One term, one clear answer.
AI and web topics explained with examples, limits and sources. Start with the definition, choose your reading depth or try a term yourself.
Where would you like to start?
Seven topics with useful starting points
Data integration & search
View all terms ↗Process automation
View all terms ↗Application & enablement
View all terms ↗AI fundamentals & models
View all terms ↗Web design & development
View all terms ↗Visibility & SEO
View all terms ↗UX & accessibility
View all terms ↗Good to know
Every term in context
Definitions, examples and sources help you understand. Related terms lead you to the next topic.
Go to knowledge search ↗Learning paths
AI in five terms
Explore pathFind knowledge at work
Explore pathWebsite and visibility
- 1Responsive Web Design
- 2Adaptive Web Design
- 3Breakpoints
- 4Generative Engine Optimization (GEO)
- 5Search Engine Optimization (SEO)
Recently reviewed
Connected knowledge
References to other terms
By letter
All terms · 2017
All terms
- UX & accessibility
A/B Test
An A/B test randomly assigns users to variants and compares predefined metrics to estimate causal effects of the change.
- AI fundamentals & models
Abductive Logic Programming (ALP)
ALP combines logical rules with possible assumptions that explain observations. Explanations must fit the modelled knowledge base and its conditions; several explanations may be possible.
- AI fundamentals & models
Abductive Reasoning
Explanatory hypotheses are formed from observations and prior knowledge. Unlike a valid deduction from true premises, these conclusions are not certain.
- AI fundamentals & models
Ablation Study
An ablation study examines the contribution of individual model components by deliberately removing or altering them and comparing the consequences. If performance remains stable, other components may perform similar functions. This reveals which parts are important for a result.
- Web design & development
Abstract Data Type
What matters for an ADT is what a data type can do — such as insert, remove or read — and which rules apply. Whether it uses an array, a linked list or another structure internally remains hidden behind the interface.
- Web design & development
Abstraction
Software development hides implementation details that are not relevant to the current task and provides stable concepts, interfaces or modules. This lets you work with a system without knowing every internal step.
- Web design & development
Accelerated Linear Algebra (XLA)
A compiler for linear algebra and tensor computations that optimizes computation graphs for various accelerators.
- UX & accessibility
Accessibility
Accessibility means that people with various disabilities can find, perceive, understand, and operate digital offerings. For this, design and technology must work together. Guidelines such as the WCAG describe requirements that support access and use.
- Application & enablement
Accountability
Technical systems do not take organizational responsibility instead of people or organizations. Accountability requires clear responsibilities, documented decisions, suitable controls and the ability to trace errors or impacts.
- Application & enablement
Accountability for AI Agents
Organizational accountability includes responsibilities for operations, data, ratings, results verification and incidents. It must be distinguished from legal liability and does not automatically answer legal liability questions.
- AI fundamentals & models
Accuracy
In single-label classification, Accuracy is the number of correct predictions divided by all scored cases. Binary applies: (TP + TN) / (TP + TN + FP + FN). In rare classes, a high degree of accuracy can occur despite poor detection.
- AI fundamentals & models
Action Language
A formal action language describes possible states, actions and state transitions with clearly defined meaning. Depending on the language, prerequisites, direct or indirect effects and queries about the consequences of action can be expressed.
- AI fundamentals & models
Action Model Learning
When learning action models, models for AI planning are derived from observed states and action sequences. These describe, for example, when an action is possible and how it changes the state. This must be distinguished from learning a strategy of action.
- Process automation
Action Schema
A scheme specifies under which modeled conditions an action is applicable and how it changes the state. By inserting concrete objects, an instantiated action is created; the real execution requires additional suitable interfaces and controls.
- Process automation
Action Selection
In action selection, an agent determines the next action from possible actions. In reinforcement learning, this is done on the basis of a strategy, the so-called policy. Exploration is testing alternatives; exploitation uses actions already rated as favorable.
- Application & enablement
Actionable Intelligence
The NIST project Actionable Intelligence investigates text analysis methods that extract relevant information from domain-specific texts. One use case described is the evaluation of maintenance order data.
- AI fundamentals & models
Activation Function
An activation function transforms the calculated signal of a neuron, for example its weighted input sum. It is usually non-linear so that several layers can depict complex relationships. Examples are ReLU, Sigmoid and GELU; the appropriate function depends on architecture and task.
- AI fundamentals & models
Active Learning
Rather than requiring a large randomly labeled dataset, an active-learning method selects particularly informative examples and has an oracle, often a person, annotate them. Its aim is to achieve good model performance with less labeling effort.
- AI fundamentals & models
Actor–Critic
The actor represents or improves the policy. The critic estimates value or advantage signals and supplies learning feedback used to adjust the policy. Many modern policy-gradient methods use variants of this principle.
- Visibility & SEO
Ad Bidding
Bidding determines how much a campaign competes for visibility in an ad auction. Platforms allow manual or automatic bidding strategies. Automatic systems may adjust bids per auction based on destinations and signals. This can be distinguished from real-time bidding in programmatic advertising.
- Visibility & SEO
Ad Exchange
Publishers provide advertising space, while various buyers or Platforms bid. At Real-Time Bidding, a single impression can be auctioned off in a very short time; standards such as OpenRTB support automated data exchange.
- Visibility & SEO
Ad Rank
The ad rank is calculated dynamically at each auction. Among other things, Google cites bid, ad and landing page quality, thresholds, competition, context of search, and the expected impact of assets and ad formats as factors.
- Visibility & SEO
Ad Spend-to-GMV Ratio
For this entry, the ratio is defined as advertising spend divided by gross merchandise value (GMV). Multiplied by 100 yields a percentage value. Which orders and expenses are included must remain consistent for comparison.
- AI fundamentals & models
Adafactor
Adafactor is a memory-saving optimization method for neural networks. For weight matrices, it keeps statistics per row and column and estimates values for individual parameters from these. This requires less additional memory than methods that store corresponding statistics for each parameter individually.
- AI fundamentals & models
AdaGrad
In the common diagonal variant, AdaGrad accumulates squared gradients per parameter. Larger summed gradients reduce its effective step size.
- AI fundamentals & models
Adam Optimizer
Adam computes individual adaptive learning rates for different parameters from estimates of the first and second moments of the gradients. According to Kingma and Ba, the method is simple to implement, computationally efficient, has low memory requirements, and is invariant to diagonal rescaling of the gradients. The PyTorch documentation describes the calculation using moving averages of both moments (β1, β2) and a bias correction.
- AI fundamentals & models
AdamW
AdamW is a variant of the Adam optimizer that separates weight decay from the optimization of the loss function. The weights are thus reduced independently of the adaptive adjustments. This step does not accumulate in the momentum or variance estimates.
- AI fundamentals & models
Adaptive Algorithm
“Adaptive” is not a single algorithm class with a universal formula. This refers to methods that update their parameters on the basis of observed signals, such as adaptive filters, optimizers, controls or learning methods.
- Application & enablement
Adaptive Learning
The system evaluates interactions, responses or competence levels and then adjusts, for example, order, difficulty, repetitions or assistance. Adaptive learning can be implemented rule-based or data-driven.
- AI fundamentals & models
Adaptive Neuro-Fuzzy Inference System (ANFIS)
ANFIS combines fuzzy-if-then rules with learnable parameters. The documented MATLAB implementation trains Sugeno-Fuzzy inference systems using input-output data.
- UX & accessibility
Adaptive User Interface
An adaptive user interface changes selected displays or controls based on context, device or usage patterns. The scope and rules of these adjustments depend on the system.
- UX & accessibility
Adaptive Web Design
Adaptive web design provides different presentations specifically for particular device characteristics or usage contexts. Several layout variants are often used. Responsive elements can be added within these variants.
- AI fundamentals & models
Admissible Heuristic
For a permissible heuristic, h(n)≦h*(n), i.e. no overestimation of the optimal remaining costs. Optimality statements for A* require additional prerequisites for search space, costs and implementation; in particular, the treatment of already visited states must be correct.
- AI fundamentals & models
Adversarial Attacks
Adversarial attacks attempt to deliberately cause AI systems to malfunction or to extract confidential information. These include manipulated inputs during deployment and poisoned training data. In adversarial examples, even small, deliberately chosen changes can trigger incorrect classification.
- AI fundamentals & models
Adversarial Robustness
Madry et al. examine the adversarial robustness of neural networks from the perspective of robust optimization and see security against a first-order adversary as a natural and broad security guarantee. Robustness applies to clearly defined classes of attackers. NIST describes robustness as a property of ML algorithms and systems to withstand security and privacy attacks.
- Process automation
Agent Architecture
The architecture determines how the model, tools, context, state and control logic work together. This also includes task transfers, limits and auditing. A single language model does not automatically form a complete agent system.
- Process automation
Agent Card
The Agent Card describes identity, capabilities, skills, supported interfaces and security requirements. Endpoints are part of the interface specifications; details depend on the A2A version.
- Process automation
Agent Guardrails
Guardrails can check inputs, outputs, or appropriate tool calls. For risky actions, permissions and required approvals must apply before execution; parallel content checks cannot reliably prevent side effects that have already been triggered.
- Process automation
Agent Handoff
In a handoff, active responsibility changes to another agent. Depending on the framework, context, history and structured handover data can be filtered or provided.
- Process automation
Agent Loop
An agent loop calls a model, performs requested actions, and returns their results to the model. This repeats until a final output is produced. Some systems can hand over the task to another agent within this process.
- Process automation
Agent Marketplace
A marketplace for AI agents can be a catalog where people find and acquire ready-made agents. In research, the term also refers to markets where agents themselves offer, negotiate, or purchase services. The respective meaning depends on the offering and context.
- Process automation
Agent Memory
Memory can include conversation history, structured states, or externally stored facts. Technically, a distinction must be made between short-term context, session memory and permanent external memory.
- Process automation
Agent Mesh
According to the manufacturer, Solace Agent Mesh is a platform that operates the self-built agents, executes the tools they call, and coordinates everything via an event mesh. Requests are broken down into tasks and distributed to suitable agents; a delegating agent publishes a request on the event broker, the receiving agent subscribes to it and responds. The platform supports MCP and A2A.
- Process automation
Agent Observability
Among other things, model calls, turns, tool calls, handoffs, guardrails, errors, latencies and token consumption can be observed. The aim is not only monitoring, but also causal analysis via connected execution traces.
- Process automation
Agent Orchestration
The orchestration determines which agent takes over which task, which information is passed on and when the workflow ends. It can use fixed rules, dynamic delegation or both.
- Process automation
Agent Passport
The concrete meaning depends on the implementation. DigiCert describes a cryptographically verifiable proof that links agent identity to metadata, capabilities, policies, and human responsibility.
- Process automation
Agent Payments Protocol (AP2)
AP2 describes how users, agents, merchants and payment services can exchange authorized purchase and payment actions. Cryptographically signed evidence documents mandate and authority. The Protocol does not replace a means of payment or the examination of concrete integration.
- Process automation
Agent Reputation
Mui develops a formal quantitative model of the rating process and emphasizes the role of social information in deciding with whom members of a social network cooperate or avoid. According to Mao et al., on-chain reputation systems track agents' transaction history and calculate trust values from it.
- Process automation
Agent Sprawl
Agent Sprawl, for example, arises when teams independently introduce agents, temporary agents continue to run, or rights are never verified. What matters is the lack of control, not just the number.
- Process automation
Agent Swarm
In the LLM context, a swarm of agents is often used for specialized agents with task transfers. It may be centrally coordinated. The name does not establish either decentralised swarm intelligence or higher quality.
- Process automation
Agent Washing
Gartner uses the term for the renaming of assistants, RPA or chatbots without essential agent skills. In order to evaluate, actual task control, tool use and limits must be checked.
- Process automation
Agent-to-Agent Communication
Agents can use interfaces to send messages, hand over tasks and return results. The general principle is distinct from the specific Agent2Agent protocol (A2A).
- Process automation
Agent-to-Agent Protocol (A2A)
A2A describes the communication between agent applications. MCP serves to connect tools and data; the protocols can complement each other.
- AI fundamentals & models
AgentBench
AgentBench is a benchmark with eight interactive environments, including operating systems, databases, knowledge graphs, and web tasks. It evaluates how well language models as agents handle multi-step tasks, make decisions, and follow instructions.
- Process automation
Agentic AI
The term is used differently. In the narrower agent sense, a model controls the next step on the basis of intermediate results; a fixed workflow, on the other hand, follows predetermined paths. Both need defined boundaries.
- Process automation
Agentic Analytics
A data agent can translate questions into data queries and explain results. Whether it plans further analysis steps independently depends on the system; a chat interface alone does not prove this ability.
- Process automation
Agentic Checkout
In purchase processing by AI agents, an agent completes a purchase with delegated authority. Order, authorization, payment, and delivery can be separated in time and organization. Protocols coordinate the process and record which transaction was actually commissioned.
- Web design & development
Agentic Coding
An agent reads project files, schedules changes, edits code, and uses tools like tests or a command line. Scope and rights are determined for the task.
- Process automation
Agentic Commerce
Agent-supported commerce includes economic processes in which AI agents search for, negotiate, or purchase services on behalf of humans. The agents act within given conditions. Protocols link the order with the shopping cart and payment and make the individual steps traceable.
- Data integration & search
Agentic RAG
At RAG, retrieved content is used for a response. An agent-driven process can additionally decide whether to search and check the search result or revise the request.
- Process automation
AgentOps
This includes traceable processes, error analysis, quality measurement and controlled changes. Execution tracks show, for example, model and tool calls.
- Web design & development
AI Accelerator
An AI accelerator is a processor or computing system whose architecture is designed for machine learning tasks. The cited overview mainly considers accelerators for deep neural networks in embedded devices and data centers.
- Application & enablement
AI Act Compliance
Requirements depend, inter alia, on the role, use and risk classification. Providers and operators have different obligations; not every AI system falls under the same requirements.
- Process automation
AI Agent Development Framework
A framework supports, for example, tool connection, sequence control and state management. It provides a technical basis, but not a finished professional quality or secure approval policy.
- Process automation
AI Agent Execution Trace
A trace assigns operations such as model and tool calls or handovers to a sequence. Individual operations are recorded as spans. What content is actually available depends on instrumentation and settings.
- Process automation
AI Agent Identity Verification
Authentication verifies proof of identity; authorization then decides on permitted actions. For this purpose, agents can use their own managed identities or cryptographic proof.
- Process automation
AI Agent Red Teaming
Red Teaming examines the entire agent system: for example, manipulated documents, tool calls, storage and transfers. The aim is to find undesirable actions and then test countermeasures.
- Process automation
AI Agents
AI agents perceive their environment and perform actions to pursue given goals. In language model-based agents, the model controls, for example, tool calls. Their results feed into further steps until a final answer is produced or the process ends.
- AI fundamentals & models
AI Alignment
Ji et al. name robustness, interpretability, controllability, and ethics (RICE) as guiding principles. They distinguish Forward Alignment (alignment through training) from Backward Alignment (evidence of alignment and appropriate system control). A known problem is proxy rewards: they are often easy to optimize and measure but frequently only partially represent the actual goal.
- Application & enablement
AI Art
AI can create or change images, sounds or other design elements. Selection, editing and artistic context can still be shaped by people.
- Application & enablement
AI Audit
An AI audit examines an AI system, its application, or the responsible organization based on defined requirements. Depending on the subject, documents and other evidence are evaluated, tests are conducted, and involved parties are interviewed. The examination can be carried out by independent bodies or supervisory authorities.
- Application & enablement
AI Avatars
In the video context, avatars can present texts with synthetic language and moving representation. Such an avatar does not have to be an interactive or self-acting agent.
- Web design & development
AI Bill of Materials (AIBOM)
The OWASP AIBOM project aims to give organizations insight into the components of their AI models. It develops open, standardized approaches for transparency and security of the AI supply chain, supported by open-source tools and a measurable completeness assessment.
- Visibility & SEO
AI Bot Traffic Analysis
Requests can be evaluated for recognized bot, operator or requested content. Crawler traffic, human-triggered agent retrievals and visits via AI references are different signals.
- Application & enablement
AI CapEx Hangover
AI CapEx Hangover is a figurative term for a situation where high capital expenditures on AI infrastructure meet delayed or weaker returns. It describes an economic assessment, not a strictly defined event or inevitable outcome.
- Visibility & SEO
AI Citation Decay
Citation Decay describes the observed decline of source references over repeated AI queries. The statement depends on the question set, system, timing, and measurement method. It is to be distinguished from the content aging of the cited text.
- Web design & development
AI Code Generation
In AI code generation, models generate code snippets, functions, or major changes from prompts and project context. Modern tools can additionally explain, complete, refactor or generate tests. Code generation is distinguished from an autonomous coding agent, which additionally controls files, commands and development tools independently.
- Web design & development
AI Code Review
For example, a tool analyzes a pull request and suggests corrections. Its indications shall be checked; relevant errors may also remain undetected.
- Web design & development
AI Coding Assistant
Assistance includes, for example, code suggestions, explanations and draft tests. Depending on the product, there are additional agent functions; the assistance itself does not require an autonomous execution.
- Visibility & SEO
AI Content Decay
AI Content Decay here describes the declining relevance or usefulness of AI-created content. Changed products, figures, or requirements can overtake texts. This is distinct from a mere decline in reach or AI citations.
- AI fundamentals & models
AI Content Watermarking
Depending on the method, image, audio, video or text signals are changed during generation, so that a suitable detector can check a watermark. This must be distinguished from visible hints and metadata.
- Process automation
AI Cost Attribution
Direct user charges and shared infrastructure shall be allocated on the basis of established rules. An assignment measures expenses; it does not automatically evaluate the benefits.
- Web design & development
AI Debugging
For example, an assistant uses error messages and code to suggest causes or corrections. A plausible explanation is not yet proof; the error must be reproduced and the change checked.
- Application & enablement
AI Ethics
AI ethics, for example, examines fairness, human oversight, privacy and possible harm. It helps to identify conflicts of goals and to make well-founded decisions.
- Application & enablement
AI Fairness
It is socio-technical: metrics, legal situation, context, participation of those affected and process design belong together. Different criteria can contradict each other.
- Web design & development
AI Gateway
An AI gateway bundles model requests and can provide logging, access control, rate limitation or fallback routes depending on the product.
- Application & enablement
AI Governance
It includes inventory, responsibilities, risk decisions, data, vendors, evaluation, documentation, approvals, monitoring and incidents. Governance is more than compliance.
- AI fundamentals & models
AI in Robotics
AI in robotics includes methods that help robots recognize their environment, plan movements, and adapt actions.
- Application & enablement
AI Liability
Liability refers to the legal obligation to be liable for damage. In AI use, possible claims depend, among other things, on contractual relationships, breached obligations, fault and the applicable liability regime. In Germany, for example, §§ 280 and 823 BGB are possible bases; they do not depict all cases.
- Application & enablement
AI Liability Insurance
Whether a damage is insured depends on the specific policy, exclusions, insured activity, and legal jurisdiction. No coverage guarantee can be derived from the term alone.
- AI fundamentals & models
AI Lip Sync
Models can condition facial frames based on speech. Good synchronization does not guarantee identity, image, or overall motion consistency.
- Process automation
AI Observability
For example, protocols, metrics and execution traces show model calls, runtimes and errors. Additional evaluations are necessary for substantive response quality.
- Process automation
AI Orchestration
Orchestration defines sequence, handoffs, parallelism, states, error handling, and approvals. It is an architectural pattern, not a single standard or provider.
- AI fundamentals & models
AI Output Verifiability
The property of being able to check statements, results, and process steps based on accessible evidence, logs, or reproducible tests.
- Application & enablement
AI Risk Management
The process looks at the specific system, its use and possible impacts. Measures and remaining risks are documented and re-examined in the event of changes.
- AI fundamentals & models
AI Safety
AI security aims to prevent an AI system from endangering people, property, or the environment under intended conditions. This includes secure design, testing, ongoing monitoring, and options for human intervention or shutdown. Protection against targeted attacks is a separate security aspect.
- Web design & development
AI SDK (Vercel)
The AI SDK unifies access to supported model vendors and provides building blocks for text generation, structured outputs, tool calls and user interfaces.
- Application & enablement
AI Shrinkflation
The informal expression transfers shrinkflation to AI offerings: a plan may include fewer usable requests or more narrowly limited features while its price remains the same. A single poorer response does not indicate such a change.
- Application & enablement
AI Slop
This refers to digital content with little use or lack of quality, which is created with the help of AI. The designation is a quality critique and not a technical test method.
- Application & enablement
AI Sprawl
sprawl can create duplicate tools, unclear accesses, and unallocated costs. The term describes an organizational development; agent sprawl is a special sub-area.
- Application & enablement
AI Transparency
Transparency can make the purpose, intended areas of use and evaluation results of a model visible. It does not automatically mean that every single decision is fully explainable.
- Application & enablement
AI-Assisted Browser
An AI-assisted browser combines web navigation with functions such as summarization, page questions and supported actions. What content the assistant is allowed to read and what steps it is allowed to take depends on the product and the permissions granted.
- AI fundamentals & models
AI-Complete Problem
The term is not a formal complexity class like NP-complete. It was historically used for tasks whose complete solution would require broad world knowledge, language understanding, or general problem solving.
- Application & enablement
AI-native browser
An AI-native browser combines a web browser with an integrated AI assistant. This assistant can explain, summarize, or compare authorized page content. Some products also perform actions; functions, data access, and permissions vary by provider.
- Web design & development
Aider Benchmark
The Aider benchmark evaluates models based on programming tasks in the Aider test setup. The Polyglot Benchmark contains 225 tasks in six programming languages. It must be distinguished from the older Python benchmark; results apply to the respective version and test configuration.
- Process automation
Alexa for Shopping
According to Amazon, Alexa for Shopping combines Rufus product features with Alexa+. The assistant answers shopping questions, supports product comparisons and offers shopping actions depending on the available function. Amazon documents the renaming of Rufus on May 13, 2026.
- Application & enablement
Algorithmic Discrimination
Automated systems can put people or groups at a disadvantage, for example through distorted data, unsuitable targets or features that indirectly map other properties. Whether a disadvantage is legally inadmissible depends on the use case and the applicable regulations.
- AI fundamentals & models
Algorithmic Efficiency
Algorithmic efficiency describes the resource requirement of a method as a function of the input size. Time and memory complexity show how this demand is growing. Measured runtime additionally depends on implementation, hardware and data.
- Application & enablement
Algorithmic Impact Assessment (AIA)
An algorithmic impact assessment examines how an automated decision-making system can affect people and their interests. For this purpose, the context of use, data, decisions, risks and countermeasures are evaluated. A concrete example is the AIA instrument of the Canadian government.
- AI fundamentals & models
Algorithmic Probability
Algorithmic probability weights possible programs for a universal calculator according to their length. Shorter programmes are given greater weight; the contributions of suitable programmes are summarised. The concept forms a theoretical approach to inductive prediction.
- AI fundamentals & models
Alibi
Alibi provides procedures for local and global explanations of machine learning models. Depending on the method, only inputs and outputs or internal model information are used. The project is described as source-available; license terms must be checked separately.
- AI fundamentals & models
Alignment Tax
Alignment Tax refers to possible performance losses caused by alignment measures compared to a comparative version. The effect depends on model, procedure, task and measurement. It is neither an inevitable consequence of alignment nor a monetary levy.
- AI fundamentals & models
Alpha–Beta Search
Alpha beta search accelerates the minimax evaluation of a game tree by limiting already known values. Branches that cannot improve the result are skipped. For the same tree under consideration, the minimax value is maintained.
- Process automation
Amazon SageMaker Pipelines
Amazon SageMaker Pipelines links ML steps to a controlled workflow. This can include data processing, training, evaluation and deployment. Dependencies, parameters and execution history support traceable processes within the AWS environment.
- AI fundamentals & models
Anchor Box
Anchor boxes are reference rectangles with defined sizes and aspect ratios at positions of a feature maps. An object detector predicts, among other things, displacements and size corrections. The anchor box is therefore not yet a recognized object boundary.
- AI fundamentals & models
Anomaly Detection
Anomaly detection marks unusual data points or courses. In the case of outlier detection, training data can already contain outliers; in the case of novelty detection, learning is typically done on largely inconspicuous data. An anomaly is initially a conspicuity and not a proven error.
- Visibility & SEO
Answer Engine Optimization (AEO)
AEO is used here as a descriptive term for better findability in response systems. Clear answers, traceable sources and technically accessible pages help to maintain such content. For Google’s AI search functions, the usual SEO basics apply; there are no special optimizations or a recording guarantee according to Google.
- AI fundamentals & models
Ant Colony Optimization (ACO)
In ant colony optimization, artificial agents gradually construct solutions. Evaluated solutions affect virtual pheromone tracks that control later decisions. Metaheuristics is used for combinatorial problems and generally does not guarantee an optimal solution.
- AI fundamentals & models
Anthropic
Anthropic develops AI models and applications, including Claude. The company identifies AI security and research as priorities. Anthropic is the provider; Claude designates a product and model family. Characteristics of individual offerings must be examined separately.
- AI fundamentals & models
Anytime Algorithm
Anytime algorithms allow a balance between computing time and result quality. Interruptible variants provide a preliminary solution after an initial phase. Contract variants, on the other hand, require a predetermined time budget and do not have to deliver a useful result before it expires.
- Process automation
Apache Airflow
Apache Airflow orchestrates workflows whose tasks and dependencies are defined as Python code. A dag describes tasks, execution conditions and further flow details. Airflow is suitable for time- or event-driven batch processes and is not a general real-time streaming system.
- Data integration & search
Apache Kafka
Apache Kafka is a distributed platform for event streams. It can publish events, store them durably, and process them during or after ingestion.
- Web design & development
Apache ZooKeeper
The service manages small data objects in a hierarchical structure. Applications use this foundation, for example, for configuration, naming, and coordination of distributed processes.
- Web design & development
API Endpoint
For HTTP APIs, an operation is typically determined by URL or path and HTTP method. Entries, outputs, authentication and permissions are part of the interface agreement; other protocols can address differently.
- Process automation
API Wrapper
An API wrapper encapsulates calls from another programming interface and offers simplified use. It can prepare parameters, answers or error handling. The approach is general and not limited to language models.
- Process automation
Apify
Apify provides a platform on which Actors perform tasks such as web data extraction, browser automation or data processing. Actors accept structured input and can store results. There are existing actors and the opportunity to develop your own.
- Web design & development
Application Programming Interface (API)
An API specifies how software can interact with an application, library or other component. It describes, for example, available functions, inputs and returns. Web APIs are one form; an API does not have to be accessible via the Internet.
- AI fundamentals & models
Approximation Error
An approximation error describes the deviation between a target value or a target function and a selected approximation. The size depends on the method, presentation and error measure. It must be distinguished from additional rounding errors of the concrete calculation.
- AI fundamentals & models
ARC-AGI-1
ARC-AGI-1 includes tasks with input and output grids. From a few examples, a system has to infer the transformation and apply it to a new input. The specific task format examines generalization under limited examples; a result alone does not prove general intelligence.
- AI fundamentals & models
ARC-AGI-2
Like ARC-AGI-1, ARC-AGI-2 uses input and output grids in which a new transformation is to be derived from examples. The tasks have been made more demanding. Results are tied to tasks, test conditions and computing budget.
- AI fundamentals & models
Area under the ROC Curve (ROC-AUC)
ROC-AUC is the area under the receiver operating characteristic curve. It summarizes how well predictive values rank positive cases higher than negative ones. AUC alone is more general and must always be named with the curve used.
- AI fundamentals & models
Arena (formerly LMArena / Chatbot Arena)
Arena, formerly LMArena and Chatbot Arena, lets people compare two anonymized model outputs to the same input and choose a preferred response. Many pair comparisons result in rankings. These measure preferences within the respective test design and are not a universal quality judgement.
- AI fundamentals & models
ARIMA Model
ARIMA combines autoregressive components, differencing and a weighted combination of past error terms (the moving-average component). The parameters p, d and q control these components. Seasonal extensions are possible, but must be explicitly modeled.
- Process automation
Arize AX
Among other things, Arize AX captures traces of AI applications and supports their analysis, evaluation and comparison in experiments. Criteria and test data determine what is being evaluated. Arize AX can be distinguished from the independent open source solution Phoenix.
- Process automation
Arize Phoenix
Arize Phoenix supports the study of AI applications through tracing, evaluation procedures, data sets and experiments. It can be operated locally or self-hosted and works with OpenTelemetry as well as OpenInference. Evaluation rules must fit the technical task.
- Web design & development
Array
An array organizes elements via indices and can be designed differently depending on the programming language. In JavaScript, arrays are mutable array objects; their indices start from zero. Properties of a language must not be transferred across all arrays.
- AI fundamentals & models
Artificial General Intelligence (AGI)
There is no universally accepted technical definition or single measurement threshold for AGI. Different research traditions weigh transfer, learning, autonomy, and breadth differently.
- AI fundamentals & models
Artificial Intelligence (AI)
AI systems derive from inputs how they generate output for specific goals. Approaches can be data-based or knowledge- and rules-based. Autonomy and adaptability differ; AI does not equate to human thinking or exclusively to machine learning.
- AI fundamentals & models
Artificial Neural Network
Weights are typically learned by optimization from data. Architecture, activation functions, and training methods determine which functions the network can approximate.
- UX & accessibility
Atomic Design
Atomic Design describes user interfaces in five levels: atoms are individual basic elements, molecules are simple groups, and organisms are more complex components. Templates arrange these in a layout; pages fill the structure with concrete content. The levels form a common model for design systems.
- AI fundamentals & models
Attention Mechanism
An attention mechanism calculates contextual weights for available representations and uses them to form an outcome. In translation models, suitable parts of the input text can thus be taken into account. Attention is a mathematical mechanism, not a human consciousness.
- AI fundamentals & models
Attention Pooling
Attention-Pooling combines several representations into an overall description. A model determines weights and thus forms, for example, a weighted sum. In multiple instance learning, a group of individual examples can be classified together.
- AI fundamentals & models
Attention Sink
In research, attention sinks are tokens that receive high attention weights from a language model without being correspondingly important to the content. StreamingLLM work observes this especially with initial tokens and uses their preservation in the KV cache for streaming processing.
- Process automation
Attio
Attio combines configurable customer data with functions for workflows, evaluations and AI support. Which automations and integrations can be used depends on the specific offer. Data quality, responsibilities and access rights remain crucial.
- AI fundamentals & models
Attributional Calculus
Attributional Calculus is a typed logic system that combines elements of propositional, predicate and multi-valued logic. It is intended to represent inductively learned hypotheses in human-oriented forms. This refers to a research approach to the representation of knowledge, not marketing attribution.
- Visibility & SEO
Audience
Strategically, target group refers to a group of people with common needs or characteristics. In GA4, on the other hand, an audience is a rule-based and changeable group of detected users that meets defined conditions.
- AI fundamentals & models
Audio Codec
A method or system for encoding and decoding digital audio data, often with compression.
- AI fundamentals & models
Audio Deepfake
Audio-deepfakes are created, for example, by text-to-speech or voice conversion and can imitate a real voice or expression. Not every AI-generated audio file is a deepfake: What matters is context, claimed authenticity and possible deception.
- AI fundamentals & models
Audio Generation
Generative models learn structures of audio data and generate new waveforms or coded audio representations from them. Audio generation is the generic term; text-to-speech, music generation and sound effect generation are individual applications.
- AI fundamentals & models
Audio Language Model
An audio language model transfers the principle of sequential prediction to audio. For this purpose, audio is often converted into discrete tokens, whose sequence is modeled and subsequently decoded again into an audio signal. It does not necessarily process words or understand language like a human.
- Web design & development
Audit Logging
Audit logging typically captures actor, action, target object, time, outcome, and relevant context. The resulting entries form an audit trail. General application logging, telemetry and monitoring can pursue further goals.
- UX & accessibility
Autocomplete
Upon autocomplete, an interface adds text entered inline or displays a suggestion list. `aria-autocomplete` describes this behavior for assistive technology, but does not implement it. The HTML attribute autocomplete, on the other hand, identifies input purposes and supports browser autofill.
- AI fundamentals & models
Autoencoder
The encoder maps inputs to a latent representation, the decoder reconstructs the input therefrom. The information bottleneck can be caused by low dimensions, regularization, noise or other restrictions. A classic autoencoder is not automatically a generative model.
- Process automation
AutoGPT
Today, AutoGPT is a platform with which AI agents can be created and executed from building blocks or natural language specifications. Agents can run workflows, respond to triggers, and integrate external tools. The product name refers to a concrete platform and is not a general technical term for autonomous AI.
- AI fundamentals & models
Automata Theory
The automaton theory examines models such as finite automatons, basement automatons and Turing machines. It describes which inputs a model accepts and which calculations it can perform. It is not the same as business or industrial automation.
- Process automation
Automated Decision-Making
Rules, statistics or AI provide recommendations or decisions. Partially automated support and exclusively automated decisions must be distinguished. Art. 22 GDPR sets limits for exclusively automated decisions with legal or similar significant effect; exceptions require prerequisites and protective measures.
- AI fundamentals & models
Automated Driving
Automated driving involves different levels. In the SAE system, they range from Level 0 without driving automation to Level 5 with complete driving automation. Level 2 remains driver assistance with human monitoring; at higher levels, the system increasingly takes over more of the dynamic driving task within defined conditions.
- AI fundamentals & models
Automated Machine Learning (AutoML)
AutoML bundles processes that automate parts of an ML workflow, such as model selection, hyperparameter optimization, feature processing or the search for model architectures. How far automation reaches depends on the system; data quality, goal definition and professional evaluation remain human tasks.
- Process automation
Automated Planning
With automated planning, states, possible actions, conditions and goals are formally described. A planning procedure looks for a valid or as good an action sequence as possible. Planning is to be distinguished from mere scheduling and free text generation.
- Process automation
Automation
Automation involves capturing, deciding, executing, monitoring or controlling a process with reduced manual intervention. It can be mechanical, industrial or software-based and does not necessarily require AI or complete autonomy.
- Process automation
Autonomous Agent
For agents, autonomy means that the system has a certain room for manoeuvre and does not have to be controlled immediately for every single step. Classical agent research also connects agents with reactivity, goal orientation and interaction with their environment. Modern LLM agents are a possible form, not the definition of the term.
- AI fundamentals & models
Autoregressive Model
Autoregressive models describe a quantity based on previous values. In generative AI, the principle is often used to gradually continue a sequence, for example by predicting a next token from the previous tokens. Statistical AR models and autoregressive language models use the same basic principle but are not the same model.
- Application & enablement
Average Revenue per User (ARPU)
ARPU is calculated from the relevant revenue and the number of associated users in the same period. Which users will be included must be determined. In the subscription business, recurring proceeds and the associated subscribers are a possible reference basis; the specific definition shall be provided.
- Visibility & SEO
B2B Marketing
B2B marketing is aimed at organizations and those involved in procurement or investment decisions. Purchasing processes can include multiple roles, longer decision paths, and professional requirements. B2B is a market relationship and not a particular marketing method.
- AI fundamentals & models
Backpropagation
In the case of backpropagation, the influence of the model parameters on the loss function is determined backwards by the computational graph after a forward run. The basis is the chain rule of differential calculus. An optimization method such as Gradient Descent then uses these gradients to change parameters; backpropagation itself is not the optimizer.
- AI fundamentals & models
Backtesting
In backtesting, a process is evaluated as if historical times were the present one after the other. For prognosis models, a rolling origin is often used: trained only with data that would have already been available at the time. Backtesting is particularly known from financial and time series applications.
- AI fundamentals & models
Backtracking
Backtracking builds a solution step by step. If a choice leads to an impasse, the procedure jumps back to the last decision point and tries an alternative. Conceptually, this often corresponds to a depth-first search in the tree of possible partial solutions.
- AI fundamentals & models
Backward Chaining
An inference method that starts from a goal and searches backward for rules and facts that can support this goal.
- AI fundamentals & models
Bag-of-Words Model
In the Bag-of-Words model, a vocabulary is constructed and each document is displayed as a vector over that vocabulary. Values can express, for example, the occurrence, the number or a weighting of terms. The order of the words is largely lost.
- AI fundamentals & models
Bagging (Bootstrap Aggregating)
Bagging creates multiple bootstrap samples from the same training dataset. A model is trained on each sample. For regression, predictions are typically averaged, combined for classification. The method can reduce the variance of unstable learning methods.
- AI fundamentals & models
Bandit-Based Recommender Systems
Bandit-based recommendation systems treat recommendations as a sequence of decisions with feedback. Contextual bandits also take into account information about the situation, users or content. Central is the conflict of goals between exploiting known good options and exploring new or uncertain options.
- AI fundamentals & models
Batch Normalization
In training, batch normalization calculates average and variance over a mini-batch, thus normalizing intermediate values and applying learnable parameters for scaling and shift. For the inference, stored running statistics are typically used. The procedure can stabilize training and enable higher learning rates.
- Data integration & search
Batch Processing
In batch processing, data or jobs are first collected and then processed together according to schedule, quantity or trigger. This distinguishes them from real-time or streaming processing, where events are processed continuously.
- AI fundamentals & models
Batch Size
In mini-batch training, the data set is processed in groups. The local batch size affects storage requirements and throughput. In the case of gradient accumulation, gradients of several such groups are merged before a parameter update; the effective batch size is then greater. Batch size and learning rate should be coordinated together.
- AI fundamentals & models
Bayesian Optimization
Bayesian optimization learns from already tried-and-tested parameter combinations a surrogate model of the unknown target function. An acquisition function decides which point will be tested next to exploit promising areas and target uncertain areas.
- AI fundamentals & models
Beam Search
Beam Search extends several possible partial sequences in parallel, but discards all but the best k candidates after each step. The width k is called Beam Width. In language models, the method is used for sequence decoding, but is not guaranteed to be globally optimal.
- AI fundamentals & models
Bellman–Ford Algorithm
Bellman-Ford initializes distances and improves them repeatedly by relaxing all edges. After at most |V|-1 runs, the shortest paths are determined if no reachable negative cycle exists. An additional run can detect such cycles.
- AI fundamentals & models
Benchmark
A benchmark specifies data, tasks, metrics, and often an evaluation protocol. In AI and machine learning, models are compared on defined capabilities. A benchmark only measures what its tasks and metrics actually cover.
- AI fundamentals & models
Benchmark Contamination
Benchmark contamination occurs when a model has seen test data directly or indirectly before the actual evaluation. Then a high score can be based partly on recognition or reproduction instead of generalization. For web data and large language models, complete testing is often difficult.
- Web design & development
BentoML
BentoML helps bundle model code, dependencies and service logic into reproducible deployment units. Services can be launched locally, containerized and operated in various infrastructures. The exact range of functions evolves with the versions.
- AI fundamentals & models
BERT
BERT processes text with a Transformer encoder and learns contextual information from both directions of a sequence. The original model was pre-trained on large unlabeled sets of text and then fine-tuned for tasks such as classification, question answering or sentence relationships.
- AI fundamentals & models
BERTScore
BERTScore assigns each token of the reference sentence to the most similar token of the candidate sentence (Recall) and vice versa (Precision). Similarity is the cosine similarity of contextual embeddings; an F1 score is computed from Precision and Recall. Optionally, rare words can be weighted more heavily using inverse document frequency (idf). In the authors' experiments, the metric correlated better with human judgments than existing metrics.
- Data integration & search
BGE Embedding Models
BGE stands for BAAI General Embedding. BGE embedding models generate text vectors for semantic search and retrieval. The broader model family also includes rerankers that evaluate search hits. Supported languages, context lengths and retrieval types depend on the respective model.
- Data integration & search
Bi-Encoder
A bi-encoder calculates representations for request and candidate separately. As a result, candidate vectors can be stored in advance and searched very quickly. This is more efficient than a cross-encoder that handles each request-candidate pair together, but can be less accurate when it comes to fine pair rating.
- AI fundamentals & models
Bias in AI and Machine Learning
“Bias” has several related meanings in machine learning. Statistical bias describes a systematic deviation of an estimator or model. In the context of responsible AI, bias also refers to systematic biases in data or system behavior that can affect groups differently. The concrete meaning must therefore always be stated.
- AI fundamentals & models
Bias–Variance Tradeoff
High bias models are often too inflexible and underfit; high variance models are highly responsive to specific training data and can overfit. In classically supervised learning problems, the bias-variance dilemma describes why model complexity must be chosen so that new data is generalized as well as possible.
- Visibility & SEO
Bid Management
Bids can be controlled manually, rule-based or with platform algorithms. Google Ads Smart Bidding optimizes conversions or conversion value for each auction. Objectives such as CPA or ROAS are control variables; results depend, among other things, on tracking, competition and campaign conditions.
- Web design & development
Big O Notation
In computer science, Big O is often used to describe how the time or memory requirement of an algorithm grows with the problem size. Formally, f(n)=O(g(n)) means that f is bounded above by a constant multiple of g from a sufficiently large input. Colloquially, Big O is often used inaccurately as an exact runtime class.
- AI fundamentals & models
BIG-bench
BIG-bench (“Beyond the Imitation Game Benchmark”) brings together a large number of very different tasks from language, knowledge, mathematics, conclusions and other fields. The aim was to measure model behaviour more broadly than with individual standard benchmarks. BIG-bench Lite is a smaller subset.
- Web design & development
Binary Search
Binary search compares the searched value to the middle element of a sorted sequence. Depending on the comparison, only the left or right half is searched further. As a result, the number of necessary comparisons only increases logarithmically with the input size.
- AI fundamentals & models
BLEU Score
BLEU compares n-grams of a generated translation with one or more references, uses modified n-gram precision, and a length penalty for too short outputs. The metric was developed for corpus comparisons and does not measure factual fidelity or complete semantic equivalence.
- Data integration & search
BM25 Ranking
BM25 calculates one relevance score per document for a search query. Frequent search terms in the document increase the score with decreasing added value; rare terms are weighted more strongly and the document length is normalized. BM25 works lexically and does not understand semantic similarity like an embedding model.
- AI fundamentals & models
Boosting
In boosting, basic models are created sequentially instead of independently. Depending on the method, incorrectly predicted examples are weighted more strongly or negative gradients of the loss function are learned; in the case of quadratic loss, these are the residuals. The individual models are combined to form an overall model. AdaBoost and Gradient Boosting are different boosting families.
- AI fundamentals & models
Bootstrapping
In the classical non-parametric bootstrap, the observed sample is used as an approximation for the unknown population. From this, many new samples of the same size are drawn. The distribution of a parameter calculated therefrom can be used to estimate uncertainty, standard errors or confidence intervals.
- Process automation
Botsitting
Glean’s Work AI Index uses botsitting 2026 for work around AI use: submitting context, reviewing results, correcting errors, and cleaning up outputs. The report distinguishes between productive and unproductive effort. Botsitting can therefore also include meaningful human review and is not automatically avoidable rework.
- Visibility & SEO
Bounce Rate
In GA4, the bounce rate is the percentage of non-engaged sessions. By default, a session is considered engaged if it lasts longer than 10 seconds, contains a key event considered, or contains at least two pages or pages. has screen views.
- Process automation
Braintrust
Braintrust is a concrete developer tool, not a general AI term. It helps teams merge data sets, evaluators, experiments and production traces for LLM and agent applications to measurably compare changes.
- Visibility & SEO
Brand Awareness
Brand awareness describes whether and how well people know or recognize a brand. It’s not the same as reach, click-through rate, or intent to buy. It can be measured directly, for example by means of surveys or suitable brand-lift methods.
- Visibility & SEO
Brand Lift
Brand Lift tries to measure the incremental change in brand perception through advertising. Typical are surveys of exposed persons and a comparable group without ad contact. The exact methodology depends on platform and study; Google’s brand lift is a concrete implementation of this measurement principle.
- Application & enablement
Brand Safety
Brand Safety protects brands from unwanted associations, such as when ads are delivered alongside generally rated problematic content. It must be distinguished from brand suitability: brand suitability also takes into account the individual values and the specific positioning of a brand.
- Web design & development
Breadth-First Search (BFS)
The breadth-first search first visits the starting node, then all directly adjacent nodes, then their still unvisited neighbors and so on. Typically, a queue is used for this. In unweighted graphs, BFS finds a path with minimal number of edges.
- UX & accessibility
Breakpoints
Layout breakpoints define switching conditions, often based on the available width. CSS media queries can also take other media features into account. Suitable thresholds are guided by the content and how usable it is, rather than just by device models.
- AI fundamentals & models
Brier Score
In the binary Brier score, the difference between the predicted probability and the actual result is squared for each prediction and then averaged. Lower values are better. The score evaluates probability predictions overall and is not isolated as a measure for calibration only.
- Process automation
Browser Agent
A browser agent connects a decision model with browser tools. It processes information from the current website and performs derived actions such as navigation, input or data reading. Depending on the system, it uses a local or a provided cloud browser.
- Process automation
Browserbase
According to its own description, Browserbase is a cloud platform for headless browser automation: applications or agents use provided browsers instead of operating their own fleet. According to documentation, the platform includes a Browser API (to create, control, and monitor sessions), a Search API, a Fetch API, Agent Identity, and Functions. Browserbase is a specific product and not a general standard term.
- Application & enablement
Business Continuity
Business continuity involves preparing for disruption, response and restarting critical performance. Business continuity management organizes risks, dependencies, emergency strategies, responsibilities, exercises and improvements. It is broader than pure IT recovery.
- Data integration & search
Business Intelligence (BI)
Business Intelligence connects data from operational sources, prepares them and makes metrics accessible through reports, dashboards and analytics. BI does not describe a single product type and can be distinguished from advanced data science or AI, but can integrate it.
- Visibility & SEO
Buyer Persona
A buyer persona translates insights about customers into a concrete work profile. It should be based on real interviews, CRM data, observations and resilient patterns rather than on fictitious demographic details. In B2B, it describes a person or role in the buying center and can be distinguished from the ideal customer profile at the company level.
- AI fundamentals & models
Byte Pair Encoding (BPE)
BPE tokenization adapted for language models begins with small units and frequently occurring pairs are repeatedly merged. This creates a vocabulary of whole words, parts of words and smaller units. Rare or new words can thus be represented from known sections.
- Application & enablement
C2PA Signing
A C2PA manifesto contains assertions on provenance and a claim referring to it. This claim is digitally signed with the private key of a signer. Validators can then check whether the signed data is unchanged and whether the credentials used are accepted according to the C2PA trust model.
- Data integration & search
Cache-Augmented Generation (CAG)
Cache-Augmented Generation was proposed as an alternative for knowledge tasks with manageable, relatively stable document stock. Relevant content is loaded into the context before the request and its key value states are cached. This eliminates the need for a separate retrieval stage at runtime. CAG is not more suitable for every knowledge base than RAG.
- AI fundamentals & models
Calibration
If a model outputs about 80% probability for many cases, then with good calibration approximately 80% of those cases should be correct. Calibration is different from pure accuracy.
- UX & accessibility
Call to Action (CTA)
A call-to-action identifies a desired action and makes its next step findable. It can appear as a button, link or text prompt. CTA and button are not identical: a button is a UI component and can also trigger purely functional actions.
- Web design & development
Canary Deployment
In a Canary deployment, old and new versions run in parallel. Only a small part of the traffic initially reaches the new version. If defined quality and error metrics are adhered to, the proportion is gradually increased; in the event of problems, the rollout can be stopped or reset.
- Visibility & SEO
Canonical URL
A canonical statement indicates which URL of several duplicate or very similar pages should be preferred. This is often done via an HTML link element with rel="canonical". The indication is a strong signal, but not a command: Google may select a different canonical URL.
- Visibility & SEO
Canonicalization
Search engines use signals like redirects, rel=canonical, and sitemaps to merge duplicate URLs. For Google, rel=canonical is a strong signal but not an absolute directive.
- Web design & development
Capacity Planning
It connects demand forecasts with performance data, reserves, scaling limits, and costs. Autoscaling reacts to load; capacity planning sets the framework for it.
- Web design & development
Cascading Style Sheets (CSS)
A styling language for the visual formatting of web pages, including layout, colours, typography and responsive design.
- AI fundamentals & models
Causal Inference
It uses experiments or explicit assumptions and models to estimate counterfactuals: What would have happened under a different intervention? Observational data alone require additional identification assumptions for this.
- AI fundamentals & models
Causal Masking
In the case of causal masking, a position may take into account only itself and previous positions. Transformer decoders block the attention to later tokens. Together with correctly shifted inputs by one position, this prevents later outputs from being used in the autoregressive prediction.
- AI fundamentals & models
Certified Robustness
With certified robustness, it is not only empirically tested whether known attacks fail. For a defined threat model, the stability of a prediction is ensured within a defined fault range. Depending on the procedure, deterministic or statistical guarantees with a specified confidence level apply. The statement only applies to the specifically certified conditions.
- Application & enablement
Chain of Custody
A Chain of Custody documents the life cycle of a piece of evidence or digital artifact from capture to archiving or return. It records transfers, controllers, dates and purposes and is intended to support integrity and traceability. It is particularly well known from forensics and regulated testing processes.
- AI fundamentals & models
Chain of Thought
In research, Chain of Thought refers to textual intermediate steps between task and response that can support complex inference. These visible justification steps are not equivalent to a complete or reliable insight into internal model states; modern systems can also use hidden internal reasoning.
- Web design & development
Chain of Trust
A sequence of cryptographically or organizationally linked verifications, where each stage authenticates or authorizes the next.
- Process automation
Chain-of-Agents (CoA)
Chain-of-agents was proposed as a multi-agent approach to long-context tasks. The input context is broken down into sections, several worker agents sequentially process partial information and pass on condensed results; a manager agent creates the final answer from this. The term here refers to a concrete research approach, not any arbitrary chain of agents.
- Application & enablement
Chain-of-Thought Prompting
In the original Chain-of-Thought-Prompting, demonstration examples with question, intermediate steps and answer are given. This allows large language models to achieve better results for certain arithmetic, symbolic or logical tasks. The method is model- and task-dependent and does not equate with any form of “step by step”.
- AI fundamentals & models
Change Point Detection
Change point detection searches for locations where the mean, variance, distribution, or other structural properties of a signal change. Methods can work offline on full data or online during a data stream. A detected change point is an indication of a structural change, not automatically of its cause.
- AI fundamentals & models
Character Error Rate (CER)
CER is typically calculated as (substitutions + deletions + inserts) divided by the number of reference characters. It is similar to the Word Error Rate, but works at the character level and is therefore useful for languages without clear word limits or OCR/ASR comparisons. Lower values are better.
- Process automation
Chatbot
A chatbot receives user input and generates responses or actions. Early systems were mostly rule-based; modern chatbots can combine language models, retrieval, tools and business logic. A chatbot is not automatically an autonomous AI agent.
- Application & enablement
ChatGPT
ChatGPT is an OpenAI user interface and product platform. Depending on the tariff and mode, functions for text, images, research, files, programming and agent work with tools are available. ChatGPT is therefore not equated with a single GPT model.
- Visibility & SEO
ChatGPT Ads
ChatGPT Ads enables advertisers to run campaigns that can appear in suitable usage situations for ChatGPT responses. According to OpenAI, ads are labeled, separated from answers and do not affect their content. In 2026, the offer was gradually expanded internationally and also in Europe.
- Process automation
ChatGPT Work
ChatGPT Work can gather context, schedule work, and perform actions through files, tools, plugins, and supported browser workflows. The former ChatGPT Agent mode is no longer available, according to OpenAI; for longer multi-step tasks, Work became the current product path.
- Data integration & search
Chunking
In chunking, documents are segmented in such a way that individual sections fit into search index, embedding model or context windows. Strategies range from fixed size with overlap to structure- or meaning-based chunking. Size and limits affect retrieval quality, cost and contextual fidelity.
- AI fundamentals & models
Churn Prediction
Churn Prediction uses historical customer, usage, contract or service data to estimate churn probabilities. The model provides a risk value or classification; meaningful measures require additional root cause analysis, cost-effectiveness and a clean target definition.
- Web design & development
CI/CD for Machine Learning
CI/CD for machine learning extends classic software pipelines with ML-specific checks: data and schema tests, pipeline tests, model evaluation, reproducible training artifacts and controlled deployment. Continuous training can be added as an additional process, but is not the same as CI/CD.
- AI fundamentals & models
CIDEr
CIDEr evaluates how strongly n-grams of a generated image description match the consensus of multiple human reference descriptions. For this purpose, n-grams are weighted with TF-IDF, so that characteristic formulations count more than very common words. The metric evaluates text similarity to references, not automatically image truth or general language quality.
- AI fundamentals & models
Class Imbalance
With strongly imbalanced classes, accuracy can be misleading. Suitable metrics, sampling, weighting, and thresholds depend on the error cost profile.
- AI fundamentals & models
Classification
Models learn decision boundaries or probabilities for categories from labeled examples. Variants range from binary to multiclass to multilabel classification.
- AI fundamentals & models
Classification Error Rate
In the case of classification, it often corresponds to one minus Accuracy, provided that the same set of cases and weighting is used.
- AI fundamentals & models
Classifier-Free Guidance (CFG)
Classifier-Free Guidance does not require a separate classifier. The diffusion model is trained in such a way that it can work both with and without condition; in the generation, both predictions are combined. A higher guidance score can increase conditional fidelity but deteriorate diversity and image quality.
- AI fundamentals & models
Claude
Claude refers to both the Anthropic model family and the associated applications. Depending on the product and version, Claude supports, among other things, long contexts, files, image comprehension, code, tool use and agentic working modes. The name should therefore always be specified with model version or product context.
- Web design & development
Claude Code
Claude Code works in development environments such as Terminal and IDE with access to project files and allowed tools. It can plan changes, edit code, run tests, and edit development tasks over several steps. Which actions are possible depends on permissions and environment.
- Process automation
Claude Computer Use
Anthropics Computer Use feature combines a model with tools for screenshots, mouse movements, clicks and keyboard input. This allows Claude to operate graphical interfaces, even if no suitable API is available. The function must be distinguished from general browser agents and classic RPA scripts.
- Process automation
Claude Cowork
Claude Cowork was introduced for longer tasks where Claude collects information, edits files and uses tools. On September 16, 2026, Anthropic announced the merger of Cowork and Chat into Claude, initially as a rollout for Pro and Max. Available features continue to depend on tariff, platform, permissions and related tools.
- AI fundamentals & models
Claude Haiku
Haiku describes the particularly fast and cost-efficient model class within the Claude family. It is typically suitable for high request volume, classification, extraction, and other tasks where speed and cost are weighted more heavily than maximum model performance. Concrete capabilities change with each version.
- AI fundamentals & models
Claude Opus
Opus refers to the performance-oriented model class within the Claude family. It is offered for tasks where quality and problem solving are more important than minimal cost or latency. Features such as context, tool use and pricing are version-dependent.
- Process automation
Claude Skills
Claude Skills bundle domain-specific work instructions, reference materials and, if necessary, executable tools into one structured unit. Claude can load appropriate skills as needed, making recurring procedures more consistent and modular. Skills are not a substitute for authorizations or professional quality assurance.
- AI fundamentals & models
Claude Sonnet
Within the Claude family, Sonnet describes the balanced model class for demanding everyday, coding and agent tasks. It lies conceptually between the particularly fast haiku class and the performance-oriented opus class. Capabilities, context and prices are version-dependent.
- Web design & development
ClearML
ClearML is a concrete open source-based tool and platform offering for the ML lifecycle. It captures experiments and artifacts, can orchestrate training jobs and pipelines, and supports model and computational resource management. It is not an ML model itself.
- Web design & development
Clerk
Clerk provides pre-built components, SDKs and backend functions for login, registration, sessions, user profiles and organization functions. It is a specific authentication service and not the general technical term for authentication or authorization.
- Visibility & SEO
Click-Through Rate (CTR)
The CTR is calculated as clicks divided by impressions times 100 percent. What counts as click and impression depends on the channel and measurement system. The metric shows a tendency to click, not automatically lead quality, sales or profitability.
- Visibility & SEO
Clickless Attribution
The term describes the attribution of possible effects from contacts without immediate clicks, for example after a visible ad or information in a search result. Such attributions are based on defined measurement assumptions and are not automatically proven causes.
- Data integration & search
Clickstream Data
Clickstream data maps digital interactions as an event sequence, such as page views, clicks, searches, scrolls or transitions between views. Together with time stamps, session and context information, navigation paths and usage patterns can be analyzed. They show observed behavior but do not automatically explain motivation or intention.
- Web design & development
Client-First
Client-First is an organizational and class system for the Webflow Designer. Among other goals, it aims to establish a uniform core structure for all pages and keep class names clear and descriptive. Classes are divided into custom classes (with underscore), utility classes (without underscore), global classes, and combo classes. The documentation does not teach HTML or CSS.
- AI fundamentals & models
CLIP
CLIP was trained contrastively on image-text pairs: matching image and text representations should be closer together than inappropriate ones. This allows the model to compare or classify images based on natural language descriptions without classical task-specific training and serves as a building block in multimodal systems.
- Web design & development
Cloudflare Workers
Cloudflare Workers runs application code in a distributed runtime close to users and data sources. Typical deployments are APIs, middleware, edge logic, authentication, caching and AI-related services. Workers is a concrete cloud product and not the general term edge computing.
- AI fundamentals & models
Clustering
Clustering tries to find groups in data so that elements within a cluster are as similar as possible according to a chosen similarity measure. Methods such as k-means, hierarchical clustering or DBSCAN use different assumptions about the shape, density and number of groups. Clusters are mathematical structures and not automatically technically meaningful segments.
- AI fundamentals & models
Cohen's Kappa
Cohen’s kappa compares the observed agreement with the agreement that would be expected randomly among the marginal distributions. A value of 1 means complete agreement; 0 is the random agreement expected according to the model. Negative values are possible. The interpretation depends heavily on class distribution and application.
- AI fundamentals & models
Cohere
Cohere develops language, embedding and retrieval models as well as enterprise platforms such as North. In September 2026, Cohere signed a definitive agreement to merge with Aleph Alpha; the deal is still subject to regulatory approval.
- Data integration & search
Cohere Embed
Cohere Embed converts text and – in current models – also images or mixed documents into vectors. These vectors can be used for similarity search, retrieval, classification and RAG. Since September 2026, Embed 5 is available in Pro and Fast variants.
- Visibility & SEO
Cohort Analysis
Instead of summarizing all users together, customers are separated by first purchase month, campaign, or onboarding week. This reveals retention and behavior over time.
- Data integration & search
ColBERT
ColBERT generates contextual vectors for individual query and document tokens. Document representations can be calculated in advance; in the search, a so-called late interaction takes place, which can detect finer matches than a single document vector.
- AI fundamentals & models
Cold-Start Problem
Many recommendation systems learn from past behavior. For new users or new products, these signals are initially missing, so that similarities and preferences can only be poorly estimated. Additional features, onboarding questions or hybrid procedures can help.
- AI fundamentals & models
Collaborative Filtering
The method uses interaction patterns such as ratings, purchases, or clicks without necessarily requiring detailed content features. Common variants are user-based, item-based, and matrix factorization.
- Data integration & search
Column Store
In the case of column-oriented storage, values of the same column are stored together. Analyses that aggregate a few columns across a very large number of data sets therefore have to read less data and often benefit from good compression. For single-row transactional accesses, other forms of storage are often more suitable.
- Web design & development
Comet
Comet here refers to the ML platform Comet ML, not a browser of the same name. It logs training runs, parameters, metrics and artifacts and supports comparison and reproducibility. For LLM and agent applications, Comet also offers Opik with evaluation and production monitoring.
- Application & enablement
ComfyUI
ComfyUI connects processing steps as nodes in a visual graph. This makes it possible to create, save and reuse generative image, video and other media workflows. Available functions depend on the models, nodes and execution environment; ComfyUI is not itself a single AI model.
- AI fundamentals & models
Command R
Command R is a cohere model for dialogue-based and company-related tasks with a focus on retrieval-augmented generation and tool use. The current documentation indicates that Cohere now recommends newer Command models such as Command A for most new use cases.
- Application & enablement
Competitive Advantage
A capability or position relevant and defensible against competitors that enables above-average customer value or economic outcomes.
- Process automation
Composio
Composio connects AI agents with tools and actions of external apps. It supports finding and executing suitable tools as well as logging into connected accounts. Which actions are possible depends on the rights of the respective account.
- Process automation
Compound AI System
A compound AI system does not solve a task with a single model call, but through the interaction of several components. This can include LLMs, retrievers, databases, rules, tools, verifiers and orchestration logic.
- UX & accessibility
Compressed y-Axis
A chart axis whose visible value range does not start at the natural zero point and can thereby visually enlarge differences.
- AI fundamentals & models
Compute-Optimal LLM Training (Chinchilla)
DeepMind’s chinchilla work empirically showed that many large language models at the time used too many parameters and too few training tokens for their computing budget. For a fixed compute budget, a smaller model with more training data can achieve better performance. The concrete optimal scaling depends on assumptions and later research results.
- Process automation
Computer Use
An AI model evaluates screenshots and suggests mouse or keyboard actions. A connected application runs them in the shared environment. Rights, controls and human approvals determine the room for manoeuvre.
- AI fundamentals & models
Computer Vision
Computer vision includes tasks such as image classification, object recognition, segmentation, tracking and OCR. Modern systems often use deep learning, vision transformers or multimodal models to derive structured information from visual data.
- Process automation
Computer-Use Sandboxing
Computer use actions are performed in a separate browser or desktop environment. Data access, accounts, rights and network destinations must be configured appropriately. Isolation limits impact but does not replace authorization and clearance controls.
- AI fundamentals & models
Concept Drift
Concept drift changes the underlying pattern that a model has learned. The same entrance may later have a different meaning for the destination. This differs from pure data drift, in which only the distribution of inputs is shifted.
- AI fundamentals & models
Conditional Generation
A conditional generative model does not generate random samples, but models outputs depending on additional information. Conditions can be, for example, a prompt, a category, a reference image or structured features.
- Web design & development
Confidential Computing
According to the definition of the Confidential Computing Consortium, data is protected during processing in a hardware-based, attested TEE. The approach complements other protections; application security, key management and the concrete threat model remain relevant.
- AI fundamentals & models
Conformal Prediction
Conformal prediction generates prediction sets or intervals. Under suitable assumptions, such as exchangeability of the data, coverage guarantees apply for finite samples. The usual guarantee is marginal over calibration and test data, not for each individual case or subgroup.
- AI fundamentals & models
Confounding
In confounding, an observed relationship between two variables can be partially or completely explained by common causes. As a result, a statistical correlation cannot be interpreted as an effect without appropriate assumptions and causal analysis.
- Process automation
Confused Deputy Problem in AI Agents
A service, plugin or agent has its own access rights. If the identity and purpose of a request are not linked to the authorisation, this intermediary may inadvertently use its rights for an unauthorised assignment.
- AI fundamentals & models
Confusion Matrix
In binary classification, it contains true positives, false positives, true negatives, and false negatives. From this, among others, precision, recall, and accuracy are derived.
- AI fundamentals & models
Connectionist Temporal Classification Loss (CTC Loss)
CTC sums up the probabilities of all valid monotonic alignments between a longer input sequence and a shorter target sequence. A blank symbol and rules for merging repetitions allow for training without frame-accurate labels.
- Application & enablement
Consent
GDPR consent must be voluntary, specific, informed and unambiguous. Those responsible must be able to prove it. A revocation must be at any time and as simple as the granting possible. Consent is one of several possible legal bases.
- Application & enablement
Consent Orchestration
Consent Orchestration translates consent decisions into technical rules for connected data flows. The specific coverage is product-dependent: Tealium currently supports event activations; AudienceStream activations require separate implementation.
- AI fundamentals & models
Consistency Model
Consistency models learn consistent mappings along a noise trajectory and can thus directly sample from noise to data. They have been developed as a faster alternative or complement to iterative diffusion sampling techniques and can be trained by distillation or directly.
- AI fundamentals & models
Constitutional AI
Constitutional AI uses a collection of natural language principles as a constitution. In the original method, a model first generates answers, critiques and revises them on the basis of these principles and can then be further trained with AI-generated feedback.
- AI fundamentals & models
Constitutional Classifiers
Constitutional Classifiers are protection components for model inputs and outputs. The classifiers are trained with synthetic data derived from a so-called constitution with rules on allowed and non-allowed content, and are intended to detect and block jailbreaks, among other things.
- Application & enablement
Content Authenticity
Content Authenticity includes provenance and trust information about digital media. The C2PA standard defines content credentials with cryptographically verifiable manifests that can contain information about the origin and editing of an asset.
- Application & enablement
Content Creation
Content creation involves more than the actual generation: topic selection, research, conception, production, editing, approval and publication are included depending on the workflow. Generative AI can support individual steps, but it does not automatically replace goal definition, fact-checking and editorial responsibility.
- Application & enablement
Content Credentials (C2PA)
Content credentials store information about the origin and editing of a digital asset in C2PA manifests. These can include, among other things, creation and editing steps, tools used and cryptographic signatures. C2PA thus describes traceable provenance signals, not the automatic determination of whether a content is true or trustworthy.
- Web design & development
Content Delivery Network (CDN)
A CDN distributes or caches content at many network locations. This allows requests to be answered from a nearby edge location, which can reduce latency and load on the origin server. Many CDN services add security features to performance.
- Visibility & SEO
Content Distillation in Marketing
Here, “content distillation” refers to the editorial consolidation of extensive starting content for marketing purposes, not a standardized AI training method. Language models can provide summaries and format designs. Statements, conditions and sources must be checked on the original; a higher content ROI is not guaranteed.
- AI fundamentals & models
Content Filter
Content filters often work in addition to the actual generative model. Classifiers check prompts and responses for defined risk categories or other unwanted content and trigger an action depending on configuration and severity.
- Data integration & search
Content Fingerprinting
Content fingerprinting generates features or hashes from the actual content instead of just from filename or metadata. Depending on the method, modified versions of an image, audio or video can also be recognized as similar and compared against known fingerprints.
- Visibility & SEO
Content Marketing
Content marketing seeks to build attention and trust not primarily through direct product advertising, but through helpful, relevant and consistent content. Content should be linked to target audience, customer journey and measurable business goals.
- Visibility & SEO
Content Personalization
Websites, e-mails or product notes show different content depending on the segment or situation. Rules or models select suitable variants. Personalization can be based on voluntary information, behavior or context; it does not necessarily require AI.
- Process automation
Content Pipeline
A content pipeline connects responsibilities and the steps involved in creating content. It can be manual, partly automated or largely automated. AI-generated drafts, approvals and CMS connections are possible components; publication may only take place after the specified checks.
- Application & enablement
Content Policy
A content policy defines permitted, restricted and inadmissible content as well as responsibilities and audit processes. For AI-generated content, it can additionally set requirements for human review, labelling, originality and sources.
- AI fundamentals & models
Content-Based Filtering
Content-based filtering represents content via features and forms a user profile based on previous interactions or explicit information. Objects whose characteristics are similar to the user profile are then recommended; data from other users is not absolutely necessary for this.
- Process automation
Context Budget
A context budget specifies how much space individual input parts receive. Depending on the API, output or further processing must also be considered. It is a planning parameter and not a uniformly standardized parameter; the meaningful distribution depends on the task and the model.
- AI fundamentals & models
Context Caching
A provider can cache common prompt prefixes or explicitly created context objects. A cache hit reduces costs or processing effort depending on the API. Minimum length, validity period and billing are dependent on the provider. Caching does not change model weights.
- AI fundamentals & models
Context Compression
Extensive context is condensed for later model steps, for example by summary, token selection or special compacting elements. Details can be lost; task quality and obtained boundary conditions must be checked.
- Application & enablement
Context Engineering
Context Engineering includes instructions, relevant documents, history, tool definitions and results. The goal is a helpful, up-to-date and trouble-free context. This includes retrieval, selection, summary, and separation of trusted instructions from external data.
- AI fundamentals & models
Context Poisoning
Attackers manipulate documents, tool responses, or stored notes that an agent later uses. This can influence responses or actions. The attack on runtime context must be distinguished from the poisoning of training data and can work together with indirect prompt injection.
- AI fundamentals & models
Context Rot
A large context window does not mean that all information is used equally reliably. Length, distractions, position and task can affect performance. Context Rot refers to observed quality losses, no temporal perishability of files and no universal fixed token limit.
- AI fundamentals & models
Context Window
The context window is the model-dependent token capacity of a request. This includes inputs and outputs as well as internal reasoning tokens for corresponding models. It is not a permanent memory; large capacity does not guarantee error-free use.
- AI fundamentals & models
Contextual Bandit
The system observes features of the situation, selects an available action and receives their reward or cost. It must use successful options while exploring alternatives. Unlike comprehensive reinforcement learning, the immediate decision is typically the priority.
- Visibility & SEO
Contextual Targeting
For example, ads are displayed on suitable topics, keywords or placements. AI can classify content for this, but is not a defining component. Contextual targeting must be distinguished from personal behavioral targeting; concrete platforms can take further signals into account.
- AI fundamentals & models
Continual Learning
A model is adapted over several learning phases. New updates can affect previous abilities, which is known as catastrophic forgetting. Methods such as replay or regularization are intended to limit this conflict. An ongoing chat is not automatically continuous learning of the model weights.
- Web design & development
Continuous Batching
An inference server reschedules the batch during generation steps. Finished requests leave it and new requests are added. The entire batch therefore does not have to wait for the longest response. Higher throughput does not automatically mean shorter waiting times for every individual request.
- AI fundamentals & models
Contrastive Learning
A model learns through positive and, depending on the procedures, negative pairs. For example, two views of the same image should be close together. Defining matching pairs, data augmentation and loss function are crucial; false negative pairs can interfere with learning.
- AI fundamentals & models
ControlNet
ControlNet complements a pre-trained text-to-image diffusion model with trainable control components. A model trained to the respective specification uses, for example, edge maps or body poses. The control improves the structural binding, but does not guarantee an exact reconstruction.
- AI fundamentals & models
Convergence
Depending on the algorithm, convergence can be mathematically proven, empirically observed, or only approximated. A stable training loss does not guarantee good generalization.
- AI fundamentals & models
Conversational AI
A dialogue system processes utterances, manages conversational context and generates answers or triggers appropriate steps. It can connect rule-based dialogue, language processing, LLMs and tools. Not every dialogue-capable system works autonomously or uses a generative model.
- Data integration & search
Conversational Analytics
A system translates technical questions into data queries or analysis operations and explains results. Data context, key figure definitions and access rights are crucial. Dialogue-based BI is meant here; the term can also refer to the analysis of recorded conversations elsewhere.
- Data integration & search
Conversational Search
The system connects follow-up questions with the original search intent, clarifies ambiguities and retrieves appropriate information. Results can be hit lists or source-based answers. A freely formulating chatbot without retrieval is not automatically a reliable search solution.
- Visibility & SEO
Conversion Rate Optimization (CRO)
CRO examines obstacles on websites or apps and tests changes to content, operation or process. The conversion rate compares target actions to a clearly defined reference value. More deals only make sense if quality, sales and user interests are also taken into account.
- AI fundamentals & models
Convolutional Neural Network (CNN)
Convolution layers apply common filters to different positions and create feature maps. Multiple layers can combine simple patterns into more complex features. CNNs are often used for images, but also for other spatial or temporal data.
- Visibility & SEO
Copilot Search in Bing
Copilot Search in Bing summarizes information from the web to a generated answer and shows sources and other links. The feature is part of Bing and does not equate to the separate Microsoft 365 “Copilot Search” feature for corporate data.
- Application & enablement
Copyright in AI-Generated Content
In Germany, a copyrighted work presupposes a personal intellectual creation. Humans can be creators; an AI system cannot. Whether human contributions to an AI-based result justify protection is to be examined on a case-by-case basis.
- Application & enablement
Copywriting
Copywriting combines target group understanding, benefit arguments, brand voice and clear calls for action. Texts appear, for example, on offer pages, in advertisements or e-mails. Compared to content writing, the desired reaction is more important; good texts require research and revision.
- AI fundamentals & models
Coreference Resolution
A system links, for example, “Maria,” “she,” and “the project manager” to a common reference. This is important for longer texts, information extraction, and knowledge graphs.
- AI fundamentals & models
Cosine Annealing
Learning rate planning reduces the step size of the optimizer from an initial value to a minimum over a defined training phase. Cosine Annealing describes the course of the curve. Warm restarts are a possible extension, but not part of every implementation.
- AI fundamentals & models
Cosine Similarity
It is the normalized dot product value and ranges between -1 and 1 for general vectors. For many embeddings, positive values are often interpreted as higher semantic similarity.
- Application & enablement
Cost Control
Cost control is part of cost management. It records actual costs, compares them with budgets, and supports the evaluation and forecasting of financial deviations.
- Visibility & SEO
Cost per Acquisition (CPA)
CPA is calculated from the allocated costs divided by the number of defined acquisitions. The focus here is on attracting new customers. Platforms also use CPA for cost per action; this meaning must be distinguished from the already existing cost per conversion and comprehensive customer acquisition cost.
- Visibility & SEO
Cost per Click (CPC)
The average CPC results from ad costs divided by click count. The CPC billing model pays for clicks. Maximum bid, actual click price and average CPC are different sizes; cheap clicks do not guarantee valuable deals.
- Visibility & SEO
Cost per Conversion
Cost per conversion is typically calculated by dividing the total relevant cost by the number of conversions recorded. The significance depends heavily on what is defined as conversion and how cleanly it is measured.
- Visibility & SEO
Cost per Mille (CPM)
CPM is calculated as cost divided by impressions times 1,000. Impressions are ad displays and not a thousand different people. Viewable CPM refers to visual impressions and must be distinguished from general CPM.
- AI fundamentals & models
Counterfactual Explanation
It does not necessarily explain the internal model logic but shows an alternative case close to the starting point. Plausibility and actually changeable features must be considered.
- Web design & development
Creative Coding
For example, code creates or modifies graphics, animations, sound and interactive installations. Creative expression and exploration are in the foreground. Creative engineering is broader and inconsistent; for the creative programming context, creative coding is the more precise term.
- Application & enablement
Creativity
Creativity refers to the ability or process of producing original and at the same time appropriate ideas or results. Novelty alone is not enough; what is considered meaningful depends on the task and context.
- Process automation
CrewAI
CrewAI organizes cooperating agents in crews with tasks and roles. Flows are used for sequence and status control and can integrate crews. The open source framework must be distinguished from additional commercial operational offerings; multiple agents do not make sense for every task.
- Application & enablement
Crisis Communication
The CDC-CERC context describes communication for organizations responding to major emergencies such as natural disasters. This includes early, accurate, and credible information as well as empathy, action orientation, and respect.
- AI fundamentals & models
Cross-Attention
Cross-attention connects two streams of information. In the classic encoder-decoder transformer, queries come from the decoder and keys as well as values from the encoder. In self-attention, they come from the same stream. The mechanism can also connect text and image information.
- Data integration & search
Cross-Encoder
For example, a cross-encoder processes search query and document as a pair. As a result, their components can interact directly with each other. Unlike a bi-encoder, it typically does not generate independently indexable vectors per text. It is therefore often used for re-sorting a limited result set.
- Data integration & search
Cross-Encoder Reranking
A reranking method where query and candidate are jointly processed by a model to produce a relevance score for a limited candidate set.
- AI fundamentals & models
Cross-Entropy Loss
For a unique target class, cross entropy penalises a low predicted probability for that class. For multiclass classification, PyTorch CrossEntropyLoss expects unnormalized logits and processes the necessary normalization internally. Training loss and practical classification quality are different sizes.
- AI fundamentals & models
Cross-Lingual Language Model (XLM)
A model family introduced by Lample and Conneau for cross-lingual pretraining with mono- and multilingual objectives.
- AI fundamentals & models
Cross-Lingual RoBERTa Model (XLM-R)
A multilingual RoBERTa-like encoder model pretrained with Masked Language Modeling on many languages.
- Data integration & search
Cross-Modal Retrieval
The system maps content of different modalities – for example text, image, audio or video – to comparable representations. This allows a text query to find suitable images or an image to find similar texts. This is narrower than any arbitrary multimodal search.
- Web design & development
Cross-Site Scripting (XSS)
A vulnerability where untrusted data is executed as active browser code in the security context of a website.
- Web design & development
Cross-Site Scripting Prevention
Central controls are secure template and DOM APIs, context-aware output encoding, and, when explicitly allowed, well-maintained sanitization of HTML. Content Security Policy and Trusted Types complement these controls as additional layers of protection but do not replace them.
- AI fundamentals & models
Cross-Validation
In cross-validation, training and validation splits are formed multiple times according to a suitable scheme and the results are evaluated together. k-fold cross-validation is one variant; group, time, and data dependencies determine the appropriate split.
- Web design & development
Cryptography
This includes symmetric and asymmetric encryption, hash functions, MACs, and digital signatures. Secure systems additionally require proper key management and protocol design.
- Web design & development
CSRF Protection Token
An unpredictable value bound to the session or request context, with which the server checks state-changing browser requests against Cross-Site Request Forgery.
- AI fundamentals & models
Curriculum Learning
In curriculum learning, the order or sample distribution of training data changes during learning. Difficulty can be determined manually, heuristically or model-based. The process must be distinguished from mere one-off sorting and self-paced learning.
- Web design & development
Cursor
Cursor is an AI-assisted code editor. It combines project context, natural-language instructions and suggested code changes. Agent features can carry out further development steps; the resulting changes must be reviewed.
- Application & enablement
Custom GPT
A custom GPT combines reusable instructions, optionally provided knowledge and supported functions in ChatGPT. Creation and editing depend on the account and workspace. OpenAI has announced a replacement by plugins; existing GPTs can be used until the respective changeover.
- AI fundamentals & models
Custom Small Language Model
A Custom Small Language Model is a comparatively compact language model that has been adapted, for example, by fine-tuning to technical language or a clearly limited task. Small is relative; the term does not specify a fixed parameter limit.
- Visibility & SEO
Customer Acquisition Cost (CAC)
CACs are typically calculated by dividing the considered marketing and distribution costs of a period by the number of new customers acquired. Exactly which costs are included must be documented so that comparisons remain meaningful.
- Data integration & search
Customer Data Platform (CDP)
According to the current definition of the CDP Institute, a CDP bears responsibility for customer identity, structure of the customer data set and a uniform interface for data services. It can connect data from multiple sources; however, a full “360-degree view” does not automatically emerge.
- UX & accessibility
Customer Journey
The customer journey includes actions, touch points, expectations and experiences before, during and after a decision. It is rarely strictly linear. A customer journey map is the visualized analysis of this process and not the journey itself.
- UX & accessibility
Customer Journey Mapping
A journey map connects temporal phases with the perspective of users. It can consolidate qualitative and quantitative insights but is not a substitute for actual research.
- Visibility & SEO
Customer Lifetime Value (CLV)
CLV may include expected turnover, contribution margin or net worth, depending on the model. Therefore, the period, margin, retention probability, discounting and included acquisition and support costs must be explicitly documented.
- AI fundamentals & models
CutMix
With CutMix, a rectangular patch is inserted from one image into another. The training label is mixed according to the area proportion of both images. The method combines aspects of cutout and mixup, but retains visible image areas.
- AI fundamentals & models
Cyclical Learning Rates (CLR)
Instead of remaining constant or only decreasing, the learning rate follows a cycle, approximately triangular or in variants with decreasing amplitude. Boundaries, cycle length and shape must be chosen or estimated with a learning-rate-range test.
- Process automation
Dagster
Dagster organizes data processing as a traceable pipeline. Data objects, called assets, and their dependencies are defined in code. The interface shows execution and supports debugging.
- AI fundamentals & models
DALL·E 3
DALL·E 3 processes a text prompt and generates a new image from it. OpenAI describes the version as an advancement of DALL·E 2 with stronger alignment between description and image as well as improvements in image quality.
- Visibility & SEO
Dark LLM Traffic
When information about a visit's origin is missing, analytics may classify it as direct or unknown traffic. This can include visits from AI recommendations. The classification (direct) / (none) alone does not establish an AI origin; it also does not substantiate a percentage of such visits.
- UX & accessibility
Dashboard
A dashboard compiles selected information visually and purposefully. It shows the current or periodically updated state of an area at a glance and refers to detailed analyses if necessary. It is not automatically real-time and not the same as a detailed report.
- AI fundamentals & models
Data Augmentation
In data augmentation, training data is extended by transformations that generate meaningful variants for the task. In the case of classification, the label often remains the same; in the case of object recognition or segmentation, annotations must be appropriately transformed.
- Data integration & search
Data Catalog
A data catalogue is a searchable metadata system. It can link descriptions, ownership, technical terms, classification, access routes, quality and lineage. The actual data is usually in other systems; the catalogue does not automatically grant access.
- Data integration & search
Data Clean Room
A Data Clean Room is a technically and organizationally controlled environment for data matching and analysis. Access, use and output rules limit who is allowed to see which data or results. The concrete architecture can use pseudonymization, aggregation or privacy-enhancing technologies, among other things.
- Data integration & search
Data Dictionary
A data dictionary contains technical and semantic metadata about fields, tables or interfaces. Possible indications are name, meaning, data type, format, unit, value range, key and relationships. The exact scope depends on the purpose.
- AI fundamentals & models
Data Drift
Data drift changes the statistical distribution of characteristics P(X) between reference and operational data. It can be distinguished from Concept Drift, i.e. a change of P(Y|X), from Label Drift and from Training-Serving Skew. Drift is a test indication and not automatically a drop in performance.
- Data integration & search
Data Enrichment
In data enrichment, existing data is augmented by additional attributes, references or metadata. Sources can be internal systems, open data or calculated features. The supplement can increase the utility value, but also introduce errors and risks.
- Application & enablement
Data Governance
Data governance is the decision-making and responsibility framework for the effective, value-creating, secure and compliant handling of data. It defines roles, principles, decision-making rights and control. Operational data management implements these specifications in processes and systems.
- AI fundamentals & models
Data Labeling
In data annotation, people or systems mark relevant properties in text, images, audio or table data. Labels can be created manually, semi-automatically or automatically. They represent an agreed interpretation and are not automatically objective ground truth.
- Data integration & search
Data Lake
A data lake is a scalable repository for structured, semi-structured and unstructured data. Data can be available in original or processed form; structure and processing are often only defined for the respective use case. It must be distinguished from Data Warehouse and Lakehouse.
- Web design & development
Data Layout
A data layout defines the physical arrangement of data, for example row-wise or column-wise, as well as with specific spacing and data types. It influences how efficiently programs read and process values.
- AI fundamentals & models
Data Leakage in Machine Learning
In the event of data leakage, unauthorized information affects training, feature generation or evaluation. As a result, a model in the test often seems unrealistically good. Typical forms are pre-processing before data splitting, target leakage and future information. This does not mean a data protection incident.
- Data integration & search
Data Lineage
Data Lineage documents sources, movements, processing steps and dependencies of data up to their output. The term Provenance is often used more broadly for origin and context information; however, the delineation is not uniform everywhere.
- Data integration & search
Data Management Platform (DMP)
A DMP collects and connects target group data from various sources, forms segments and can pass it on to advertising systems. First, second and third-party data can be used. Cookie or device identifiers are often pseudonymous and not automatically anonymous.
- Data integration & search
Data Mesh
Data Mesh is a socio-technical organizational and architectural concept for analytical data. It combines four principles: domain responsibility, data as a product, a self-service data platform and federated computational governance. It is not a single product and not a mere platform.
- Data integration & search
Data Mining
Data mining involves methods that derive structures, rules or predictive models from data. Typical tasks are classification, regression, clustering and association analysis. Finding patterns must be validated with suitable test data and interpreted professionally.
- Web design & development
Data Parallelism
In the case of data parallelism, a calculation or a model is replicated on several processes or devices. Each replica processes different data partitions, then merges results or parameter updates. Synchronous gradient synchronization is a common but not mandatory procedure.
- Data integration & search
Data Pipeline
A data pipeline is a coherent processing chain of sources, transformations and destinations. It can read, check, filter, enrich or store data. Batch and streaming are possible modes of operation; not every pipeline is ETL or modeled as DAG.
- AI fundamentals & models
Data Poisoning
Data poisoning is an integrity attack on ML systems. Attackers change, supplement or weight data or labels in such a way that the general performance decreases, certain cases are treated incorrectly or a backdoor arises. The attack takes place before or during learning.
- AI fundamentals & models
Data Preprocessing
Depending on the task, preprocessing involves cleaning, imputation, scaling, normalization, coding and transformation. Steps can take place before or within a model pipeline. Learned parameters must only come from the respective training data.
- Application & enablement
Data Processing Agreement (DPA)
In the GDPR context, the data processing contract is a binding contract or other legal instrument in accordance with Article 28. It shall specify, inter alia, the subject matter, duration, nature, purpose, types of data, data subjects and rights and obligations. It is not a general data exchange agreement.
- Web design & development
Data Structure
A data structure is the concrete organization and storage of related information, for example as an array, list, stack, queue, hash table or tree. Their characteristics determine effort, order and possible operations.
- AI fundamentals & models
Data Validation for Machine Learning
ML data validation compares data with schema, value ranges, categories, completeness and distribution rules. It can detect anomalies, drift and training-serving skews. However, it does not confirm professional accuracy, fairness or model suitability.
- Web design & development
Data Version Control (DVC)
DVC manages data and model artifacts via metafiles that can be versioned with Git. The actual files remain in a DVC cache and can be shared via local or external remotes. In addition, DVC supports pipelines and experiments; reproducibility also requires available code, dependencies and data.
- UX & accessibility
Data Visualization
Data visualization encodes data by position, length, color, shape or surface. It is used for exploration or communication. Diagram type, scale, aggregation and context determine which statement is recognizable and how easily it can be misinterpreted.
- Data integration & search
Data Warehouse
A data warehouse is a data storage optimized for analysis and reporting. Data from operational sources is integrated, cleaned, modeled and often historicized. Queries do not directly burden the operational systems. A warehouse can be operated locally or as a cloud service.
- Web design & development
Database Normal Form
A formal level of relational data modeling that limits certain dependencies and redundancies.
- Data integration & search
Databricks
Databricks is a cloud-based platform for data processing, analysis, and artificial intelligence applications. It integrates, among other things, data pipelines, notebooks, SQL queries, and functions for controlling data access.
- Data integration & search
Dataset Deduplication
Exact deduplication compares defined fields to equality. Fuzzy duplicate detection uses normalization, similarity, and entity resolution. Treatment can mean marking, linking, merging or removal; it must be distinguished from memory and block deduplication.
- Application & enablement
Datasheets for Datasets
The documentation format proposed by Gebru and co-authors accompanies a dataset with questions about motivation, collection, composition, processing, use, distribution, maintenance and limitations. It is neither a universally mandatory standard nor a quality certificate.
- Web design & development
Daytona
Daytona offers so-called sandboxes: mutually isolated computing environments where software or AI agents can execute code, install packages, and start processes. The documentation describes dedicated runtime resources such as filesystem and network environment.
- AI fundamentals & models
DBSCAN
DBSCAN is an unsupervised clustering process. Points with enough neighbors within the radius eps become core points; reachable points can be border points, the rest are considered noise. min_samples specifies the required neighborhood density.
- Application & enablement
Dead Internet Theory
The dead internet theory is a speculative narrative suggesting that human activity on the internet has largely been replaced by bots and artificial content. Real automation does not support this broad claim.
- Process automation
Decision Support System (DSS)
A decision support system is an interactive information system that provides data, analysis, models or simulations for decision-making processes. It supports particularly partially structured problems, but does not necessarily make the decision itself.
- AI fundamentals & models
Decision Theory
It combines possible actions, uncertain conditions, consequences and preferences into a comprehensible decision rule.
- AI fundamentals & models
Decision Threshold
If a score is above or below the set limit, a decision is triggered. The appropriate value depends on error costs and context of use.
- AI fundamentals & models
Decision Tree
Each inner node tests a feature; leaves provide class or numerical value. Training procedures select divisions according to a quality criterion.
- AI fundamentals & models
Decoder
A decoder is part of some AI models that generates output steps from an internal representation. In the encoder-decoder approach of the Transformer paper, it processes, among other things, previous information for a target output.
- AI fundamentals & models
Decoding
In generative language models, decoding is the process by which the next tokens are selected from the computed probabilities. This includes selecting the most likely token or using sampling-based methods.
- AI fundamentals & models
Decoding Strategy
In autoregressive models, decoding determines which next token is chosen from the conditional probability distribution. Greedy Search and Beam Search choose more deterministic; sampling as well as top-k or top-p generate stochastic diversity.
- AI fundamentals & models
Deductive Reasoning
A deductive argument is valid if its conclusion cannot be false when all its premises are true. Only a valid argument with actually true premises is sound and guarantees a true conclusion.
- AI fundamentals & models
Deep Compression
The name refers to the combination of pruning, trained quantisation or weight sharing, and Huffman coding presented by Han, Mao and Dally in 2015. It does not mean model compression in general.
- AI fundamentals & models
Deep Learning
Deep learning is a branch of machine learning. Models with multiple processing layers learn hierarchical features from data; there is no universal minimum number of layers that makes a model “deep”.
- AI fundamentals & models
Deep Reinforcement Learning (DRL)
DRL combines reinforcement learning for sequential decisions with deep networks that approximate value functions, policies or state representations, for example.
- Application & enablement
Deep Research
According to the definition in a survey paper, deep research agents combine dynamic reasoning, adaptive planning, and iterative tool use to obtain, merge, and analyze external information. The result is comprehensive outputs, such as structured analytical reports. The information comes via API-based search or browser-based web exploration. The work distinguishes static and dynamic workflows as well as single- and multi-agent architectures.
- AI fundamentals & models
Deepfake
The term often covers images, videos or audio generated or altered using deep learning. Not every synthetic file is a deepfake, and harmful intent is not a necessary part of the definition.
- AI fundamentals & models
Deepfake Detection
Detection methods analyse artefacts, temporal patterns, metadata or model features and return scores or classes. They check for signs of manipulation, not automatically whether a statement is true.
- AI fundamentals & models
DeepSeek
DeepSeek develops AI models and offers, among other things, a chat interface and a programming interface. The provider name does not refer to a single model version or a uniform feature set.
- AI fundamentals & models
DeepSeek-R1
DeepSeek-R1 refers to a reasoning model and associated published variants. The release also includes smaller distilled models; these are to be distinguished from the original R1 model.
- AI fundamentals & models
DeepSeek-V4
DeepSeek-V4 is a model family published by DeepSeek. At launch, the provider distinguished, among others, Pro and Flash. Specific API names and supported variants may change over the product lifecycle.
- AI fundamentals & models
DeepWalk
The algorithm, which was introduced in 2014, generates short random walks and learns from them using a skip-gram-like method, knot embeddings in which structurally similar knots are closer.
- AI fundamentals & models
Default Reasoning
Default Reasoning describes a product standard: An application activates a reasoning mode without additional selection or automatically chooses its effort. The term does not denote a separate model architecture.
- AI fundamentals & models
Demand Forecasting
Demand forecasting uses historical time series, influencing factors and assumptions to estimate expected demand together with uncertainty. It is an input to inventory, staffing or capacity planning, not the same as those tasks.
- Visibility & SEO
Demand-Side Platform (DSP)
A DSP combines campaign goals, budgets and target group rules with available advertising inventory. Purchasing can take place via real-time auctions or other programmatic agreements. Good placement, exact target group approach and favorable prices are not automatically guaranteed.
- Application & enablement
Demographic Parity
A binary classifier satisfies demographic parity when its positive decision rate does not depend on group membership. The criterion does not take ground truth or qualifications into account.
- AI fundamentals & models
Denoising
Depending on the method, random or unwanted signal components in images, audio, measurement series or data representations are estimated and reduced. Denoising is not limited to neural networks.
- AI fundamentals & models
Denoising Diffusion Implicit Model (DDIM)
Denoising Diffusion Implicit Models (DDIM) use a non-Markovian sampling process to generate images from a trained diffusion model. The original paper shows that compared to certain DDPM sampling methods, the number of required generation steps can be reduced without changing the training objective.
- AI fundamentals & models
Denoising Diffusion Probabilistic Model (DDPM)
DDPM is a class of generative probabilistic models for image synthesis. The learned reversal of a noising process enables generating a new image from noise.
- Data integration & search
Dense Passage Retrieval
DPR trains two separate encoders (question and text section) in the same vector space. In contrast to pure keyword search, DPR also captures meaning relationships with different word choices.
- Data integration & search
Dense Retrieval
An embedding model maps requests and documents as number vectors. A similarity measure, such as scalar product or cosine similarity, serves to order the hits. High similarity does not guarantee a factually correct answer.
- AI fundamentals & models
Dependency Parsing
The method determines directed relationships (dependencies) between a head word and its dependent words. This explicitly maps the grammatical role (subject, object, modifier).
- AI fundamentals & models
Depth Estimation
The estimation of distance or relative depth of image points from one or more camera views.
- Web design & development
Depth-First Search (DFS)
A graph search method that follows a path as deeply as possible and only then returns to alternatives.
- AI fundamentals & models
Depthwise Separable Convolution
The method combines a spatial convolution per input channel with a subsequent 1×1 convolution to mix the channels. Compared to a comparable standard convolution, it can save parameters and computing operations. The actual runtime also depends on hardware and implementation.
- UX & accessibility
Design Grid
A system of columns and rows that serves as a design grid for the harmonious placement of elements.
- Web design & development
Design Pattern
A design pattern explains a problem, the context and a possible structure of the solution, such as Observer or Factory. It is not a finished code and should only be used if it fits the specific problem.
- AI fundamentals & models
Detection Transformer (DETR)
DETR considers object recognition as a direct set prediction problem. It eliminates the need for handcrafted components such as Non-Maximum Suppression (NMS) and Anchor Boxes and uses self-attention for global image contexts.
- AI fundamentals & models
Detokenization
A decoder matching the model converts token IDs into text. Settings for special tokens and spaces affect the output. After normalization or other processing steps, the original text is not always faithfully reconstructable.
- Web design & development
Devin (AI Software Engineer)
Devin from Cognition AI combines code generation, terminal access, browser and troubleshooting in one sandbox. It can analyze repositories, fix bugs, run tests and independently create pull requests.
- Process automation
Dialogflow
Dialogflow supports the development of chat and voice agents. Depending on the product variant, intents, entities, defined dialog processes and generative functions are used, among other things. Connections to specialist systems require suitable interfaces and a tested integration.
- Process automation
Dialogue Management
Dialogue Management tracks the state of the dialog, stores contextual information and decides on the next action of the system based on rules, policies or machine learning.
- AI fundamentals & models
Difference-in-Differences (DiD)
The change of a target variable before and after a measure between two groups is compared. For a causal interpretation, it is necessary in particular that the groups would have developed in parallel in the relevant period without a measure. Inappropriate comparison groups and further changes can distort the result.
- AI fundamentals & models
Differential Evaluation
Here, “differential evaluation” is used descriptively for a comparative test; it is not a clearly defined standard method. Both variants are evaluated under comparable conditions with the same tasks. In addition to average values, individual deteriorations and technical errors are checked.
- Application & enablement
Differential Privacy (DP)
A randomized method fulfills a defined data protection guarantee if neighboring data sets have only limited different output probabilities. Frequently calibrated noise is used for this purpose. Parameters such as ε and, where applicable, δ and the protected unit determine the guarantee; absolute secrecy does not follow.
- Visibility & SEO
Differential Privacy in Advertising
In a corresponding implementation, aggregated advertising key figures with controlled statistical noise are output. The guarantee depends on the mechanism and its data protection budget. Collection, identifiers and other processing steps shall be assessed separately.
- AI fundamentals & models
Differentially Private SGD (DP-SGD)
DP-SGD typically limits the gradient contribution of each training example and adds calibrated noise. A privacy accounting records the accumulated loss over the training. The guarantee applies to a defined protected entity and defined parameters; it does not mean that any confidential information can never be issued.
- AI fundamentals & models
Differentiation (in Machine Learning)
In machine learning, differentiation determines how much the error changes when a parameter is slightly varied. Through automatic differentiation (autodiff) in the backpropagation step, neural networks learn efficiently.
- AI fundamentals & models
Diffusion Model
During training, the model learns to reverse a defined noise process. In the generation, an initial distribution is converted into data in several steps. Application, presentation form and sampling method determine whether, for example, images, audio or other data are generated.
- AI fundamentals & models
Diffusion Sampling Steps
The number of numerical or denoising updates with which a diffusion sampler generates an output from an initial state. It is only meaningfully comparable together with model, scheduler, and solver.
- AI fundamentals & models
Diffusion-based Language Model (Diffusion LLM)
Depending on the procedure, continuous text representations or discrete tokens are iteratively refined. Multiple positions can be edited in parallel; not every variant always updates the entire text at the same time. This distinguishes the approach from a purely left-to-right token generation.
- Application & enablement
Digital Content Provenance
Standards such as C2PA link media to signed manifests describing origin, tools and edits. This can make integrity and provenance claims verifiable, but does not prove that the content is true.
- Application & enablement
Digital Disruption
In common usage, digital disruption denotes significant market changes. Christensen’s theory of disruptive innovation is narrower: New providers open up underserved market segments or new markets and can later move on to become established providers. Not every rapid improvement meets this definition.
- Process automation
Digital Farming
Digital farming uses, among other things, digital documentation, sensors, geodata and networked machines. This can support decisions and processes in agriculture and animal husbandry. Resource savings and returns depend on location, technology and implementation.
- Web design & development
Digital Identity
Handling digital identities includes, depending on the system, identity proofing, registration, authentication, federation, and lifecycle management. NIST distinguishes different trust levels and processes.
- Application & enablement
Digital onboarding
Digital onboarding bundles tasks, learning content, and information for new employees in a digital process. A knowledge base and clear ways to ask questions support orientation. Personal guidance remains important; AI can help but is not a prerequisite.
- Application & enablement
Digital Transformation
Digital transformation goes beyond the mere digitization of paper documents. It includes the fundamental modernization of working methods, customer interactions and value chains using modern cloud, automation and AI technologies.
- Process automation
Digital Twin
A digital twin maps selected properties of a real object or system. Data can update the displayed state; models can support analysis and simulations. Update frequency and accuracy depend on the purpose and are not automatically real-time or error-free.
- AI fundamentals & models
Dijkstra's Algorithm
Proceeding from the starting node, the not yet closed node with the smallest known distance is processed in each case. Assuming non-negative edge weights, shortest paths can thus be determined. The weights must match the desired size, such as distance or travel time.
- AI fundamentals & models
Dilated Convolution (Atrous Convolution)
A dilated convolution increases the distances between the scanning positions of a filter. With the same filter size, a larger context is thereby detected without additional filter parameters. Whether the output resolution remains the same also depends on stride and padding.
- AI fundamentals & models
Dimensionality Reduction
Linear methods such as PCA and non-linear methods such as t-SNE or UMAP produce low-dimensional representations. They can facilitate analysis and visualization. The benefit for a subsequent model must be verified; visible clusters are not an automatic proof of real groups.
- AI fundamentals & models
Direct Preference Optimization (DPO)
DPO formulates the optimization problem of RLHF mathematically: Instead of training a separate reward model and optimizing it via reinforcement learning, DPO adapts the language model directly to paired preference data via a simple classification loss function.
- Process automation
Directed Acyclic Graph (DAG)
A directed acyclic graph consists of nodes and directed edges. Acyclic means: If you follow the arrows, you never get back to the starting node via a closed path. Workflow systems use DAGs to derive permitted execution orders from dependencies.
- Web design & development
Disaster Recovery
Disaster Recovery includes policies, tools, and automated backups to quickly restart business-critical systems, databases, and web applications after natural disasters, cyberattacks, or hardware defects.
- Application & enablement
Disclosure Rules for Virtual Influencers
Which labeling is required depends on the legal jurisdiction and the specific representation. Advertising collaborations, misleading testimonials, and transparency obligations for certain AI content must be assessed separately.
- Application & enablement
Disparate Impact
In US selection procedures, the ratio of the selection rates may indicate adverse effects. The four-fifths rule is a rule of thumb and does not definitively decide on unlawful discrimination. The German concept of indirect disadvantage according to § 3 para 2 AGG is related, but not identical.
- Visibility & SEO
Display Advertising
Display advertising shows visual advertisements on websites and in apps. The placement can be done via advertising networks, direct bookings or programmatic systems. Goals can be reach, brand awareness or measurable actions; display advertising is not synonymous with programmatic purchasing.
- Web design & development
Distributed Training
Training a model with coordinated computing across multiple processes, devices, or computers.
- AI fundamentals & models
Distribution Shift
Changes can affect inputs, targets or their relationship. As a result, a model can work worse, but does not have to lose quality with every change. Data distribution and actual result quality should be monitored separately.
- AI fundamentals & models
Diversity in Recommender Systems
Classic recommendation systems tend to accumulate very similar hits (echo effect). Diversification algorithms balance relevance and diversity to show users new categories and avoid monotony.
- Process automation
Document AI
In general, OCR, layout analysis and AI models support document content classification and extraction. Results can be output in a structured manner. A transfer to ERP or other specialist systems requires a separate integration; errors and missing information must be handled.
- AI fundamentals & models
Double Machine Learning (DML)
ML methods estimate auxiliary variables, such as outcome and treatment models. Orthogonal estimation equations and cross-fitting can reduce sensitivity to estimation errors and overfitting. Causal identification and further model assumptions remain necessary.
- AI fundamentals & models
DPO Training Pipeline
A DPO workflow includes the selection of a suitable starting model, curated pairs of preferences, the appropriate reference configuration, and training. Separate evaluation data examines quality and possible deteriorations. Data approval, configuration and results should be documented in a comprehensible manner.
- AI fundamentals & models
DreamBooth
DreamBooth adjusts a pre-trained text-to-picture model based on fewer images of a specific subject. A special identifier in the prompts connects the motif with the model; for this, a new vocabulary entry does not necessarily have to be generated. Prior Preservation can support the preservation of general classes, but does not guarantee it.
- AI fundamentals & models
DROP Benchmark
DROP is an English-language question-and-answer dataset for text sections. For example, some tasks require counting, adding or sorting. The benchmark measures performance on these test cases and proves neither general understanding nor reliability on operational documents.
- AI fundamentals & models
Dropout
During training, randomly selected activations are set to zero. This can reduce overfitting; the benefits depend on the model and settings. Many implementations are already scaling in training, so that dropout does not make any further changes in the normal evaluation mode.
- Web design & development
Dynamic Batching
An inference server bundles available requests into compute stacks. An optional waiting window can allow for larger stacks, but increases the waiting time. Throughput and latency depend on model, hardware and load.
- Visibility & SEO
Dynamic Creative
Dynamic advertising tools use reusable templates or assets. The selection and combination can be based on product feeds, playout context or available target group signals. Responsive displays are a documented example; supported signals and functions differ between platforms.
- Visibility & SEO
Dynamic Creative Optimization (DCO)
DCO generates variants from common advertising components and adjusts individual elements based on target group, context or measured performance. The goal is a better fit to the defined campaign goal. A certain click-through rate, conversion rate or return is not guaranteed.
- Process automation
Dynamic Pricing
Rules or models adjust prices to data such as demand, inventory, or competitive prices. Companies set goals and limits; higher margins or optimal sales are not guaranteed.
- Web design & development
E2B Sandbox
On request, E2B provides Linux VMs where agents can run code and tools. SDKs control start, commands and life cycle. Isolation reduces risks; permissions, network and sensitive data must be additionally secured.
- Data integration & search
E5 Embedding Model
E5 was contrast trained with weakly monitored text pairs and refined on supervised tasks. The specific variant determines language, input format and context length. For multilingual-e5-large, query: and passage: are documented input processing.
- AI fundamentals & models
Early Stopping
Often the loss of validation is monitored. Patience determines the number of epochs without sufficient improvement; min_delta sets the minimum change. The procedure can limit overfitting and unnecessary training, but it does not safely prevent them.
- Application & enablement
Economics of AGI
Research explores how far-reaching AI automation could change tasks, productivity, income and capital-to-labour ratios. Assumptions about capabilities, dissemination and resources determine the results; this does not lead to a safe AGI forecast.
- AI fundamentals & models
Edge AI
Models perform inference on devices or servers near the data source. This can reduce transmission times and data traffic. Offline operation requires locally available models and dependencies; security follows from the entire implementation.
- Web design & development
Edge Computing
Calculation and storage take place close to data sources or users, for example on local devices, regional servers or distributed cloud nodes. This can reduce latency and bandwidth requirements; reliability depends on local dependencies and synchronization.
- AI fundamentals & models
Effect Size
Effect measures such as mean value difference, Cohen's d, correlation or odds ratio describe the size and direction of an effect. The p-value evaluates data under a null hypothesis and is not a measure of effect size. Practical relevance and uncertainty must be assessed separately.
- AI fundamentals & models
ElevenLabs (AI Voice Generation)
ElevenLabs offers generative audio features, including text-to-speech and voice cloning. Quality and naturalness depend on model, language, text and reference recording. Use of voices and generated content requires appropriate rights and conditions.
- AI fundamentals & models
Elo Rating (in AI Evaluation)
Classic Elo updates relative ratings based on wins, defeats or draws and expected results. LLM evaluations use pair comparisons, but not every ranking calculates classic Elo: The Arena work of 2024 describes Bradley-Terry coefficients and confidence intervals.
- Visibility & SEO
Email Marketing
Email marketing includes newsletters, offer mails and automated information channels. Purely transactional messages serve another main purpose. Segmentation and automation can improve communication; cost-effectiveness and deliverability must be measured.
- Data integration & search
Embedding
The mathematical vector representation of texts, images or audios in multidimensional space to determine semantic similarities.
- AI fundamentals & models
Embedding Model
Depending on the model, it maps text, images or other data in a vector space. Semantically similar content should be represented similarly according to the training goal. Model, version and task determine how distances or similarity scores are to be interpreted.
- AI fundamentals & models
Embodied Reasoning
The Gemini Robotics paper defines embodied reasoning as the ability of a vision-language model to anchor objects and spatial concepts in the real world and integrate these signals for robotic applications. This is the term usage in that paper, not a general definition for all AI. RT-2 is a related research example linking image, language, and action data.
- AI fundamentals & models
Emergent Abilities
In a research definition, model abilities are called emergent if they are absent in smaller models on a specific test but present in larger ones. The classification depends on the task examined and its measurement.
- AI fundamentals & models
Emotion Recognition
Emotion recognition attempts to infer emotions or intentions from biometric data, for example from facial expressions. Merely detecting a smile is to be distinguished from this as long as no emotion is inferred. In the EU legal framework, special restrictions apply, especially for workplaces and educational institutions.
- Web design & development
Encapsulation
In object-oriented and modular systems, interfaces protect internal states and invariants. Encapsulation supports information hiding but is not exactly the same concept.
- AI fundamentals & models
Encoder
It processes raw data or tokens into features or latent states that a decoder or downstream head uses. The representation does not have to be smaller than the input.
- AI fundamentals & models
Encoder–Decoder Model
The encoder maps the input internally; the decoder gradually generates the output from it, often using Attention.
- Web design & development
Encryption
The reversible transformation of plaintext into ciphertext using a cryptographic algorithm and key to protect confidentiality.
- Web design & development
Encryption at Rest
The encryption of data on storage media, databases, or backups outside of active transmission.
- Web design & development
Encryption in Transit
The protection of transmitted data through a cryptographically secured channel, usually TLS, between clearly defined endpoints.
- AI fundamentals & models
Energy-Based Model
Low energy stands for preferred or plausible combinations; probabilities can be derived from this, but do not always have to be calculated in a normalized manner.
- Visibility & SEO
Engagement Rate
The formula depends on the platform. In GA4, dedicated sessions are divided by sessions. A session is engaged if it exceeds the configured engagement time limit (ten seconds by default), contains a key event, or contains at least two page or screen views. For web data streams, the time limit is adjustable.
- AI fundamentals & models
Ensemble Learning
Bagging, boosting and stacking combine several basic models, for example to reduce variance or improve prediction quality. Models and errors do not have to be independent; the benefits are tested by independent validation.
- Data integration & search
Enterprise RAG
Enterprise RAG combines retrieval-augmented generation with operational requirements such as access rights, data maintenance, source citations, and ongoing quality control.
- Data integration & search
Entity Extraction
At NER, text spans such as people, organizations or places are marked and assigned to classes. Technical information such as contract numbers requires a suitable scheme, rules or training; general NER models do not automatically recognize them.
- Data integration & search
Entity Linking
After recognizing a mention, the system decides on the basis of the context which unique data set is meant.
- Data integration & search
Entity Resolution
Rules or probabilistic models compare features and decide whether different spellings or sources belong together.
- Visibility & SEO
Entity-Based SEO
An editorial approach to SEO that describes topics, people, organisations and products unambiguously and connects relevant information. Structured data can supplement these descriptions. This establishes neither a separate confirmed ranking factor nor guaranteed search positions.
- Web design & development
Envelope Encryption
A data encryption key (DEK) encrypts useful data; a key encryption key (KEK) encrypts the DEK. A KMS can manage KEKs centrally. Key rotation and re-envelopment must be deliberately planned.
- AI fundamentals & models
Episodic Memory
In memory research, episodic memory describes a human system for remembering personally experienced events. It differs from pure factual knowledge and includes a reference to one’s own experience and subjective time.
- AI fundamentals & models
Episodic Memory in AI Systems
Episodic memory stores specific past events or states of an AI system with reference to time and context. This can include perceptions and actions. It is more narrowly defined than general agent memory, which can also contain facts, rules, or plans.
- AI fundamentals & models
Epistemic and Aleatoric Uncertainty
Epistemic uncertainty can decline with more appropriate data or better models; aleatory uncertainty stems from inherent noise or ambiguity.
- Application & enablement
Equalized Odds
Within each actual target class, the prediction should be independent of protected group membership. The criterion requires ground truth and is only one fairness measure.
- AI fundamentals & models
Error Analysis
Errors are categorized and prioritized by type, cause, subgroup, data source or process phase. It goes beyond a single overall metric.
- Data integration & search
ETL
Source systems are read out; data is cleaned up, unified and checked; only then do they reach the data warehouse, for example.
- Application & enablement
EU AI Act
The regulation distinguishes prohibited practices, high-risk systems, transparency obligations and rules for general purpose AI. Duties depend on the role, system and context of deployment.
- AI fundamentals & models
Euclidean Distance
It is the square root of the sum of squared coordinate differences and depends on scaling and dimensions.
- Application & enablement
Eval Washing
The critical term describes that selected favorable tests convey a more comprehensive picture of performance or safety than the results actually support. The AI may function in reality while its presentation hides important limitations.
- Web design & development
Eval-driven Development
Test cases and metrics are created early, changes are repeatedly checked and regressions are made visible before the rollout. The term is a practice, not a uniform standard.
- AI fundamentals & models
Evaluation Framework
It connects goals, test cases, data sets, metrics, baselines, thresholds, human review, and approval rules. It is a process framework, not necessarily a software product.
- AI fundamentals & models
Evaluation Harness
The technical environment loads models or systems, test cases and metrics, logs configurations and generates results for comparisons. Comparability requires controlled versions and conditions; the environment does not replace the evaluation concept.
- Visibility & SEO
Evaluation-Driven Marketing
The descriptive term means marketing that derives its statements from targeted product comparisons, application tests, or evaluations. It does not denote a uniformly standardized procedure; the specific test method must be named.
- Visibility & SEO
Event Tracking
An event model describes names and properties as well as the capture context, such as form sent or process failed. It requires a consistent taxonomy, data quality controls and a review of data protection and, if necessary, consent.
- Web design & development
Event-Driven Architecture
Producers report state changes; brokers or streams distribute them to independent consumers. Processing can take place asynchronously and several times.
- AI fundamentals & models
Evidence Lower Bound (ELBO)
The ELBO is at most as large as the logarithmic evidence. Their gap corresponds to the KL divergence between approximated and true posterior. Maximization optimizes an approximate target size; it does not guarantee exact evidence maximization and minimal prediction errors.
- Data integration & search
Exa
Exa offers an API through which applications can initiate web searches and receive structured search results. The official documentation describes the Search endpoint with parameters for queries and result data.
- Visibility & SEO
Exit Rate
A common calculation is exits of a page divided by their page views times 100. The exact definition depends on the tool; GA4 documents in particular the absolute metric Exits, i.e. the last measured action of a session on a page or a screen.
- AI fundamentals & models
Expected Calibration Error
Predictions are usually divided into bins according to their confidence. For each bin, the absolute difference between mean confidence and observed accuracy is calculated and weighted by the number of his cases. Result and statement depend on binning and variant.
- Application & enablement
Experience API (xAPI)
A specification published by ADL for capturing and exchanging activity events as statements between systems, typically via a Learning Record Store. Unlike SCORM, xAPI is not limited to packaged LMS courses.
- Web design & development
Experiment Tracking
Experiment tracking assigns input, configuration, metrics and artifacts to a run. Together with data and code versions, it supports comparison and repetition; complete reproducibility does not automatically occur.
- AI fundamentals & models
Explainability
In AI, explainability refers to the ability to provide an explanation for a result that is understandable and accurate for the target audience. The NIST report on explainable AI distinguishes explanations, comprehensibility, explanation fidelity, and system limitations.
- UX & accessibility
Explainability UX Patterns
Explanations should fit the user role, task and decision. For example, you can show reasons, comparative examples and uncertainty in an understandable order and check with users whether meaningful actions follow.
- AI fundamentals & models
Explainable AI (XAI)
Explanations can be global or local and show features, examples, rules or uncertainty. Explanation, interpretability and transparency are related, but not identical.
- AI fundamentals & models
Exploration–Exploitation Dilemma
Too much exploration costs short-term performance; too much exploitation prevents learning. Bandit and RL processes control this trade-off.
- Data integration & search
Exploratory Data Analysis
Visualizations, distributions, relationships, missing values and outliers help to understand data quality and hypotheses. EDA does not confirm causality.
- Web design & development
Exponential Backoff
After temporary errors, the interval typically grows exponentially and is often complemented by random jitter to avoid simultaneous repetitions.
- Application & enablement
Exponential Growth
In exponential growth, the increase is proportional to the amount already present. For example, with doubling each period, the sequence 1, 2, 4, 8 arises; the absolute increase becomes larger.
- AI fundamentals & models
Exponential Linear Unit (ELU)
For positive inputs, f(x)=x, otherwise f(x)=alpha·(exp(x)−1). At alpha>0, the negative output approaches −alpha. Negative activations can bring the mean closer to zero; training benefits depend on architecture and task.
- AI fundamentals & models
Exponential Moving Average (EMA)
In the usual convention, EMA_t = α · x_t + (1 − α) · EMA_(t−1), with a smoothing factor between 0 and 1. A larger α weights the new value more strongly; initialization and convention used must be established.
- AI fundamentals & models
Exponential Smoothing
Variations take into account level, trend and seasonality. Statistical ETS models provide not only point forecasts but also forecast intervals; suitable components and error assumptions must be selected.
- Web design & development
Extended Detection and Response (XDR)
A non-uniformly standardized security platform that correlates telemetry from multiple protection domains and supports detection as well as response.
- Web design & development
Extensible Markup Language (XML)
A text-based W3C syntax for hierarchically structured documents with elements, attributes, and namespaces.
- Data integration & search
Extract, Load, Transform (ELT)
ELT loads source data into a target data platform before the professional transformation. There they are processed, for example, with SQL and dbt. Raw data management, access rights and retention are conscious architectural decisions; source data is not necessarily completely unchanged.
- AI fundamentals & models
F1-Score
It is high when positive predictions are simultaneously precise and capture many actual positive cases. The exact averaging over classes must be given.
- Visibility & SEO
Facebook Ads
Facebook Ads means paid advertisements on Facebook. In Meta Ads Manager, you can create ad campaigns, define target audiences and placements, and view reports on campaign performance. The Ads Manager can also include other Meta platforms.
- Data integration & search
FAISS
FAISS provides search and indexing methods for vector data. Instead of exact text comparison, entries can be found based on distances or similarity measures between their numerical representations.
- AI fundamentals & models
Faithfulness
In RAG, faithfulness typically checks whether statements can be derived from the retrieved sources; for explanations, the term can refer to fidelity to the model. The definition must be specified.
- AI fundamentals & models
fastText
fastText provides methods for efficient text representation and classification. A scientifically described approach considers character sequences within a word to better represent rare word forms.
- Web design & development
Fault Tolerance
Redundancy, isolation, repetition, replication and orderly degradation limit impacts. Fault tolerance is not the same as high availability or disaster recovery.
- AI fundamentals & models
Feature Engineering
Expertise is used to clean, encode, aggregate or meaningfully combine raw data without injecting future information.
- AI fundamentals & models
Feature Extraction
It can be rule-based, statistical or learned. For example, text, images or signals are converted into numerical characteristics; the number of dimensions does not have to become smaller.
- AI fundamentals & models
Feature Importance
Methods measure different things, such as split gain, permutation loss, or local contribution values. Results are model- and data-dependent and not automatically causal.
- Data integration & search
Feature Store
A feature store manages feature definitions and their deployment. Depending on the product, it includes transformations, versioning and access controls, and offline and online storage; consistent definitions can reduce training serving skews.
- Visibility & SEO
Featured Snippet
For a featured snippet, Google displays the descriptive section before the usual page reference. The Google systems decide whether a page is suitable for the specific search query. A website cannot label itself as a featured snippet.
- AI fundamentals & models
Federated Learning
Participants calculate local updates; often a central coordinator aggregates them into a common model. This avoids the central collection of raw data, but alone does not guarantee privacy or security.
- Visibility & SEO
Federated Marketing
In a federated marketing organization, local or specialized teams implement measures themselves, while central units provide common brand rules, tools, or measurement standards. The specific division of tasks is established organizationally.
- Process automation
Feedback Loop
Feedback can stabilize or amplify. In AI systems, decisions often influence later data and thus future models.
- AI fundamentals & models
Feedforward Neural Network
A feedforward network has no directed feedback cycles: information is processed from input to output. The multilayer perceptron is a typical variant; weights are often optimized with backpropagation during training.
- AI fundamentals & models
Few-Shot Learning
The term includes procedures that learn from a few examples of a new class or task. In language models, prompting with few demonstrations is often referred to as such, without a change in weight.
- Process automation
Fiddler AI
The Fiddler platform provides features that allow teams to monitor or test classical ML models, language models, and agent-based applications. The official documentation mentions, among others, monitoring, evaluation, and protection mechanisms.
- AI fundamentals & models
Fine-Tuning
Subsequently adapting an existing model with specific data sets to specialize in specialist domains, tasks or brand tones.
- Web design & development
Finite-State Machine
Inputs or events trigger defined changes of state; variants differ, for example, by output and determinism.
- Web design & development
FinOps
According to the FinOps Foundation’s definition, updated in March 2026, FinOps is an operational framework and cultural practice for timely, data-based decision-making and shared cost responsibility. In addition to the cloud, the current framework also covers AI, SaaS, data platforms, licenses and data centers.
- Web design & development
FinOps for AI
FinOps for AI includes transparent allocation, planning, and optimization of costs for AI services and AI infrastructure. The FinOps Foundation describes AI-specific metrics such as cost per inference, token consumption, and resource utilization.
- Process automation
Firecrawl
Firecrawl offers APIs for searching web pages, retrieving individual pages, and crawling linked pages. The documentation describes output of processed web page content, for example as Markdown or structured data.
- Web design & development
Fireworks AI
Fireworks AI is a platform for running and customizing open AI models. The documentation distinguishes shared serverless APIs, own deployments on GPUs, and training for model customization.
- Data integration & search
First-Party Data
This includes, for example, purchases, service contacts, website events or voluntarily communicated preferences from direct customer relationships. As far as personal data are concerned, the collection and use must in particular be lawful, purpose-related and necessary. „Own data does not mean free use.
- Data integration & search
First-Party Data for AI
First-Party Data for AI are data from own direct business relationships or processes used for an AI application. They can serve as a search basis, test data, or training material; these purposes must be distinguished.
- Data integration & search
First-Party Data Graph
Such a graph represents, for example, customer accounts, orders, and contact points as nodes and their relationships as edges. First-Party denotes origin from own direct contacts; it does not guarantee error-free assignment.
- AI fundamentals & models
FlashAttention
The algorithm arranges computation steps so that less data is moved between GPU main memory and fast on-chip memory. It does not approximate the mathematical attention; small numerical deviations by computational order remain possible.
- AI fundamentals & models
Flow Matching
A neural vector field is trained to transport data along conditioned paths from a simple initial distribution to data distribution; the training does not require a simulation of the complete process.
- Web design & development
FLUX
According to the model cards on Hugging Face, FLUX.1 [dev] and FLUX.1 [fast] are models with 12 billion parameters that generate images from text descriptions. The licenses differ: FLUX.1 [dev] is under a Non-Commercial License, which also regulates the use of generated outputs for personal, scientific, and commercial purposes under the conditions described there; FLUX.1 [fast] is under the Apache-2.0 license.
- AI fundamentals & models
Focal Loss
The loss function developed for object recognition adds a factor to cross-entropy that reduces the contribution of well-recognized examples; an alpha weight can address class imbalance.
- UX & accessibility
Focus Group
The group discusses a clearly defined topic; interactions provide motives, language and perceptions. The results are not statistically representative.
- Data integration & search
Formal Ontology
Classes, relationships, and logical statements are expressed in a language with defined semantics. This allows certain inferences and contradictions to be automatically examined.
- AI fundamentals & models
Forward Chaining
A rule-based procedure that starts from existing facts and repeatedly applies applicable rules until a goal or fixed point is reached.
- AI fundamentals & models
Forward Pass
With the forward pass, outputs are calculated from inputs and current parameters. In the usual training, loss calculation, backpropagation and an optimization step follow. Inference uses forward passes, often repeated with autoregressive generation.
- Visibility & SEO
Forward-Deployed Marketing Engineer
The role combines marketing understanding with practical technical implementation in the customer's work environment. It clarifies workflows, creates automations or AI applications, and supports their actual use. Tasks and job titles are not uniformly standardized.
- AI fundamentals & models
Foundation Model
According to Stanford CRFM, the term encompasses large-scale, mostly self-supervised trained models that are versatile through prompting, fine tuning or other procedures. Errors and risks are transferred to many applications.
- Process automation
Fraud Detection
Fraud detection refers to identifying suspicious transactions, for example in payment data. As a specific research approach, SCARFF describes the analysis of continuous credit card transactions using machine learning and stream data processing tools.
- Visibility & SEO
Frequency Capping
An advertising platform limits impressions or views per assigned user and time window, depending on the campaign type and level. For example, in Google Ads display campaigns, only visible impressions count; incomplete identity assignment can affect the count.
- Web design & development
Full-Stack Development
The term typically includes user interface, server logic, APIs, data storage, and deployment. It describes breadth, not automatically deep expertise in every area.
- Web design & development
Fully Sharded Data Parallel (FSDP)
PyTorch FSDP is a concrete approach to data-parallel training, in which parameters, gradients and optimizer states are sharded and communicated as needed, depending on the phase.
- Visibility & SEO
Funnel Analysis
A funnel analysis counts how many users reach defined steps in the intended order. In an open funnel, it is possible to start at later steps; a closed funnel begins at the first step. Abortions describe the observed behavior, but do not prove a cause.
- AI fundamentals & models
Fuzzy Inference System
Inputs are assigned to membership levels, rules are evaluated and results are aggregated; defuzzification often follows. Mamdani and Sugeno systems are different.
- Data integration & search
Fuzzy Matching
Unsharp matching compares strings using a similarity method. Power Query uses Jaccard similarity and an adjustable threshold in the Fuzzy Merge. A high score does not replace checking whether two records denote the same entity.
- AI fundamentals & models
G-Eval
G-Eval is a 2023 presented framework for evaluating generated texts. A language model generates assessment steps and fills out an assessment form. The original work explores summaries and dialogues, especially with GPT-4. Judgments depend on task, criteria and model.
- AI fundamentals & models
Gated Recurrent Unit (GRU)
GRUs use update and reset gates to retain or discard relevant past information over sequences. They are simpler in structure than classical LSTMs.
- AI fundamentals & models
Gaussian Distribution
Its density is bell-shaped; many statistical methods use it as a model assumption. Real data does not have to be normally distributed.
- AI fundamentals & models
Gaussian Mixture Model
Each component has a mean value and covariance; mixture weights indicate proportions. Parameters are often estimated with Expectation-Maximization.
- Application & enablement
GDPR Compliance in AI
Data origin, legal basis, purpose, roles of stakeholders, transparency, data subject rights and technical protection measures shall be clarified. Article 22 refers to certain exclusively automated decisions with legal or similar significant effect; it is not a general transparency requirement for any AI.
- AI fundamentals & models
GELU
GELU is defined exactly by GELU(x) = x · Φ(x); Φ is the distribution function of the standard normal distribution. The function weights inputs softly based on their value. Implementations can use numerical approximations.
- Application & enablement
Gem
A Gem is a customized version of Google Gemini with stored instructions for recurring tasks. There are self-created and pre-made Gems. Depending on the account and app, additional source files can be integrated or Gems shared.
- AI fundamentals & models
Gemini
Gemini is a family of generative AI models from Google. The API documentation lists different models for tasks with text, images, audio, and other modalities. Which inputs and outputs are supported depends on the respective model.
- AI fundamentals & models
Gemini 3.1 Pro
The documented variant Gemini 3.1 Pro Preview processes text, images, video, audio, and PDF and outputs text. Google mentions, among other things, structured output, function calls, and a reasoning mode. The identifier is gemini-3.1-pro-preview.
- AI fundamentals & models
Gemma 4
Gemma 4 includes models with open weights from Google DeepMind. The model card mentions dense architectures and mixture-of-experts variants, where selected model parts contribute to processing. Supported input types and hardware requirements differ between variants.
- Application & enablement
General Data Protection Regulation (GDPR)
The GDPR regulates the processing of personal data in the EU. It enshrines principles such as data minimisation, purpose limitation, transparency and data subject rights (right of access, deletion) and threatens sensitive fines in the event of breaches.
- AI fundamentals & models
Generalization
A model is generalized when learned patterns are transferred to independent examples under appropriate conditions. Good test performance does not guarantee robustness during distribution shift.
- AI fundamentals & models
Generative Adversarial Network (GAN)
In classical GAN, a generator generates examples, while a discriminator is intended to distinguish training data from generated examples. The training goals are in competition. Good quality or variety is not guaranteed by this construction alone.
- AI fundamentals & models
Generative AI
Generative models generate content such as text, image, audio, video or code from learned patterns. Expenditure can be factually incorrect or distorted. Check data origin, confidentiality and the terms of use applicable to your use.
- Visibility & SEO
Generative AI Performance Report
The report documents examined tasks, model version, test conditions, and results. Depending on use, it includes response quality, error patterns, runtime, and effort. The name does not denote a standardized report format.
- Visibility & SEO
Generative Engine Optimization (GEO)
GEO is the preparation of digital content for findability and possible naming in generative response systems. A mention or positive citation cannot be guaranteed.
- Visibility & SEO
Generative Search Engine Results Page
A search results page includes an AI-generated answer section, potentially with source links and additional traditional results. Such answer sections supplement or change how results are displayed; they do not necessarily replace all links.
- Web design & development
Generative UI
An AI-based process selects, generates or updates matching UI components based on a task and its data. For example, tool results are displayed as a table or map. This can be based on predefined components; usability and accessibility must be tested.
- Visibility & SEO
GEO-SFE (Structural Feature Engineering)
The research distinguishes document architecture, information sections and visual emphasis as three levels of structure. It investigates their influence on citations in the generative systems tested. The results are specific to the study and do not guarantee citations or ranking improvements for a particular website.
- Visibility & SEO
Geo-Targeting
Geo-targeting uses geographic signals to select content or ads. Google Ads can include, among other things, user settings, device and behavior; target areas can be set as countries, regions or radii. Location signals are not a guarantee of an exact location.
- Web design & development
GGUF
It stores tensors and typed metadata in an expandable, mmap-friendly file. GGUF can contain quantized tensors, but is not itself a quantization method.
- Visibility & SEO
Ghost Engagement
The inconsistently used marketing term means measured interactions that are not based on corresponding conscious human actions. Automatic views can influence metrics; the exact meaning depends on the respective measurement system.
- Web design & development
GitHub Copilot
GitHub Copilot is an AI assistant for software work. The official documentation describes code suggestions, questions about a codebase, changes, and reviews. Additionally, there are agent functions that can prepare or process multiple work steps.
- AI fundamentals & models
GloVe
GloVe learns word vectors from a corpus’ global word-cooking statistics. In classical usage, each word receives a static vector; different meanings of the same word are not separated by the respective sentence.
- Visibility & SEO
Go-to-Market Engineer
The role connects marketing and sales processes with data, interfaces, and automation. For example, it implements data enrichment, segmentation, and handoffs between tools; the exact job description can vary greatly.
- AI fundamentals & models
Golden Dataset
A reference data set contains verified tasks and expected responses, labels or evaluation criteria. “Golden Dataset” is not a uniform quality certification: coverage, errors and timeliness must be checked separately.
- Visibility & SEO
Google Ads
Google Ads is an online advertising program by Google. For search ads, an auction decides whether a suitable ad appears and at which position. The ad rank considers not only the bid but also quality and context factors.
- Visibility & SEO
Google AI Overviews
Google supplements selected searches with a summary with further links. The function is a specific Google search offer; it does not appear on every request.
- Visibility & SEO
Google Analytics
Google Analytics is an analytics platform for data from websites and apps. The cited documentation describes event-based capture: usage actions form the basis of the analysis. This approach is distinguished there from the earlier session-based measurement model.
- Web design & development
Google Colab
Google Colab is a hosted service for Jupyter notebooks, i.e., documents with executable code, text, and results. The official FAQ describes usage without local setup and access to computing resources such as GPUs and TPUs.
- AI fundamentals & models
Google DeepMind
Google DeepMind works on AI models, methods, and scientific applications. Names like Gemini or AlphaFold refer to specific model or product families; Google DeepMind is the organization behind them.
- Process automation
Google Flow
Google Flow is an AI tool for cinematic clips, scenes, and stories. The official help describes various access methods, model variants, and features for creating visual content. The specific range of functions depends on the model and access used.
- Visibility & SEO
Google Search Console
The free Search Console offers reports and tools on search performance, indexing, and other technical aspects of a website. For example, it shows search queries, impressions, and clicks for the verified property.
- AI fundamentals & models
GPQA
Published in 2023, the dataset contains 448 questions created by professionals. It measures performance on this set, not general intelligence or scientific reliability.
- AI fundamentals & models
GPQA Diamond
GPQA Diamond contains in the cited version 198 multiple-choice questions from biology, physics, and chemistry. The subset is formed based on the answers of expert and non-expert examiners. It belongs to the GPQA dataset and is to be distinguished from the larger main dataset.
- AI fundamentals & models
GPT
The original GPT approach uses a transformer decoder for autoregressive language model pre-training and then adapts it to tasks. GPT is a model family designation and not a single fixed version. A GPT model is not synonymous with a full chatbot application.
- Process automation
GPT Orchestration
An application decides when to call a GPT model, what information it receives, and how its output is further processed. This can include tool calls, validation rules, and handover to humans.
- AI fundamentals & models
GPT-4
GPT-4 is described in the technical report as a large multimodal model that can accept text and images as input and produce text output. The pretraining aims to predict the next token, i.e., a processed text unit. Afterwards, the behavior was further adapted based on human feedback.
- AI fundamentals & models
GPT-4V
GPT-4V refers to GPT-4 with image processing. The system card describes how users can provide images to the model for analysis. The entry addresses this visual capability and its documented limitations without claiming unchanged availability of this historical variant.
- AI fundamentals & models
GPT-5
OpenAI presented GPT-5 as a unified system consisting of a fast-responding model, a model for deep reasoning, and a controlling selection component. The introduction also describes capabilities for writing, programming, and working with visual inputs.
- AI fundamentals & models
GPT-5.4
OpenAI describes GPT-5.4 as a model that combines reasoning, programming, and agentic workflows. The introduction mentions processing screen views, computer operation, and a tool search that can selectively load required tool definitions.
- Application & enablement
GPU-Poor
The term describes teams or organizations that have significantly fewer GPU resources available relative to their projects than large AI labs. It is a relative classification, not a technical performance class.
- AI fundamentals & models
Grad-CAM
Gradients of a target class weight activations of a convolution layer; this results in a coarse salience-like map. It is post hoc and not causal.
- AI fundamentals & models
Gradient Accumulation
Gradients of multiple micro-batches are summed or averaged before the optimizer performs an update. Effective batch size and scaling must be documented correctly.
- AI fundamentals & models
Gradient Centralization
Gradient centering is an optimization method for neural networks. It acts directly on the gradients, i.e., the computed directions of change for model weights. The gradient vectors are centered so that their mean is zero.
- AI fundamentals & models
Gradient Checkpointing
Instead of storing all intermediate values for backpropagation, the training only retains checkpoints and recalculates missing activations. Memory decreases, computing time increases.
- AI fundamentals & models
Gradient Clipping
In standard clipping, the entire gradient is scaled when its norm exceeds a limit; value clipping limits individual values. It treats symptoms of unstable updates.
- AI fundamentals & models
Gradient Descent
The gradient shows locally the direction of the greatest increase; learning rate and variant determine the updating. Mini-batch-SGD uses estimates instead of the full gradient.
- AI fundamentals & models
Gradient Noise
When deliberately adding gradient noise, the gradient calculated during training is altered by a random term. Neelakantan and other authors investigate time-dependent Gaussian noise whose variance decreases during training. This entry refers to this method.
- AI fundamentals & models
Graph Attention Network (GAT)
Masked self-attention aggregates neighboring features and can give different weight to relevant connections per task. Weights are not automatically causal explanations.
- AI fundamentals & models
Graph Classification
Prediction unit is the entire graph. In GNN methods, node representations are combined to form a graph representation via a readout function. Graph features or graph cores are also possible. A continuous target value belongs to the graph regression.
- AI fundamentals & models
Graph Convolutional Network (GCN)
The common Kipf-Welling variant approximates spectral graph convolutions and mixes features of connected nodes via normalized adjacency.
- AI fundamentals & models
Graph Database
Graph databases store linked data for relationship and path queries. In the property-graph model, nodes and relationships can carry properties. RDF uses subject-predicate-object-triple. Data model and query language differ by system.
- AI fundamentals & models
Graph Isomorphism Network (GIN)
GIN combines sum aggregation of neighboring representations with MLPs. Under the conditions of the original analysis, it can achieve the distinctive power of the one-dimensional Weisfeiler-Lehman test. This is not a guarantee to distinguish all non-isomorphic graphs.
- Data integration & search
Graph Neural Network (GNN)
Many GNNs use Message Passing: Nodes collect neighbor information and update their state. Tasks include node, edge and graph predictions.
- AI fundamentals & models
Graph Search
A graph consists of nodes and edges. In search, a strategy determines which reachable nodes are examined next: Breadth-first search proceeds level by level, depth-first search first follows a branch deeply.
- Web design & development
Graph Transformer
Models combine node and edge features, position or structure coding, and global or local attention. There is not a single binding architecture.
- Data integration & search
Graph Traversal
Breadth-first search, depth-first search or domain-specific path rules determine order and termination. Traversal is an operation, not a database or a learning method.
- Web design & development
Graphics Processing Unit (GPU)
Many computing units process vector and matrix operations simultaneously. This accelerates graphics, ML training and inference, as long as software and data flow fit.
- Data integration & search
GraphSAGE
Instead of storing fixed embeddings per node, the model samples and aggregates local neighbors. As a result, it can be transferred inductively to unseen nodes with a suitable feature scheme.
- AI fundamentals & models
Great Expectations
Great Expectations provides GX Core, a Python library for data testing. An Expectation is a verifiable statement about data. Multiple expectations can be bundled and validated against a selected dataset.
- AI fundamentals & models
Greedy Algorithm
Greedy algorithms build a solution step-by-step without systematically revising earlier decisions. With suitable problem structure, they can provide optimal solutions; for other problems, only approximations.
- AI fundamentals & models
Greedy Best-First Search
The priority is based only on the heuristic h(n), not on the previous path costs. This can be fast, but does not guarantee optimal paths or completeness without further ado.
- AI fundamentals & models
Greedy Decoding
The method decides locally step-by-step for the token with the highest model probability and does not explore alternative sequences. The selection rule is deterministic; identical outputs in practice additionally require a constant, deterministic execution.
- AI fundamentals & models
Grid Search
For each value from a fixed parameter grid, combinations are tried and typically evaluated with a validation strategy. Effort grows quickly with number and scope of parameters.
- AI fundamentals & models
Griffin
The architecture combines recurrent components for efficient sequence processing with local attention. It was introduced as a research approach for more efficient long sequences.
- AI fundamentals & models
Grok
The Grok documentation describes an assistant for asking questions, writing, and editing problems. Depending on the function, files can be included, images generated, or conversations conducted by voice.
- Process automation
Grok-Bot
According to product documentation, a Grok Bot works with browser, file system, and terminal in a persistent cloud environment. Tasks are assigned via message; the bot can use tool accesses and ask back if necessary involvement is required.
- Data integration & search
Grokipedia
Grokipedia is listed in the official provider description as an independent online knowledge collection alongside the conversational assistant Grok. The cited passage supports this product classification but does not confirm any specific editorial review processes.
- Web design & development
Groq
The Groq documentation describes an inference service with an OpenAI-compatible interface. Applications send inputs to the Groq endpoint and receive the output of the selected available model.
- AI fundamentals & models
Ground Truth
In machine learning, ground truth data are usually accepted target values or labels. They can come from real events, measurements, or human annotations and may themselves contain errors or uncertainties.
- AI fundamentals & models
Grounding
In grounding, a model receives relevant data as context or is checked against a known knowledge source. RAG is a common grounding technique, but grounding can also include other data or tool accesses.
- AI fundamentals & models
Group Normalization
The method divides feature channels into groups and computes mean and variance within each group. Thus, normalization does not depend on batch size.
- AI fundamentals & models
Group Relative Policy Optimization (GRPO)
GRPO was introduced in the DeepSeekMath context as a PPO variant. For multiple answers to a task, group-relative reward signals are used, thus avoiding a separate critic in this form.
- AI fundamentals & models
Grouped-Query Attention (GQA)
The architecture lies between Multi-Head Attention with separate K/V heads and Multi-Query Attention with only one K/V head. The goal is faster decoder inference at lower KV cache.
- Visibility & SEO
Growth Hacking
Growth hacking investigates through repeated experiments how customers can be acquired and retained. The reviewed scientific overview classifies the approach as a combination of marketing, data analysis, and product development. Learning from data is central.
- AI fundamentals & models
GSM8K
The dataset was developed to investigate multi-step mathematical reasoning of language models. Problems and solutions are formulated in natural language.
- Process automation
Guardrails
Frameworks like the OpenAI Agents SDK check inputs, outputs, and tool calls before or after their execution. A triggered tripwire can abort the process. Permissions and human approvals are complementary measures and must be set up separately.
- Process automation
Guardrails AI
The product documentation describes guardrails as a framework for input and output checks. Validators can be combined into guards. The framework also supports generating structured data with language models.
- AI fundamentals & models
Guidance Scale
In classifier-free guidance, conditional and unconditional predictions are combined. The scaling factor changes the trade-off between prompt adherence, diversity, and possible artifacts.
- AI fundamentals & models
Hallucination
The generation of plausible-sounding, but factually false, invented or unsupported statements by an AI model.
- AI fundamentals & models
Hallucination Budget
The term transfers the idea of an error budget to factually incorrect or fabricated AI outputs. For this, error type, measurement method, and consequences must be defined. There is no universally acceptable error rate.
- AI fundamentals & models
Hallucination Detection
Methods compare answers with sources, check consistency, use classifiers, or human evaluation. There is no error-free universal detector.
- AI fundamentals & models
Hallucination Rate
For a hallucination rate, it must be clear what counts as a hallucination and what the frequency refers to. Jesson et al. define for in-context learning a probability of outputs outside a likely range of the underlying mechanism. This is a concrete modeling, not a universal error metric.
- Web design & development
Hardware Security Module (HSM)
HSMs generate, store, and use keys within a controlled hardware environment. Certifications such as FIPS 140-3 can verify security requirements for cryptographic modules but are not synonymous with the term HSM itself.
- Web design & development
Hash Function
Hash functions assign hash values to input data or keys, for example, for mapping in a hash table. Cryptographic hash functions are a subclass with additional security goals such as collision and preimage resistance. Hashing is not encryption because no decryption is intended.
- Web design & development
Hash Table
When inserting or looking up, the key is hashed. Since different keys can map to the same target area, the data structure needs a collision strategy.
- Web design & development
Headless CMS
Content management system that manages content decoupled from the frontend and delivers it to any output channel via APIs.
- UX & accessibility
Heatmap
In the UX context, heatmaps aggregate, for example, clicks, scroll depth, or interactions. They show patterns but do not automatically explain the cause of user behavior.
- Web design & development
Helicone
The Helicone quick start guide describes an OpenAI-compatible API through which various model providers can be addressed. The calls are logged and made visible in a request interface.
- AI fundamentals & models
HellaSwag
The dataset introduced in 2019 uses multiple-choice continuations and adversarial filtering to create examples that are difficult for models but usually easy for humans.
- AI fundamentals & models
HELM
HELM systematically organizes scenarios and metrics and measures, besides accuracy, aspects such as robustness, fairness, bias, toxicity, and efficiency, as available for each scenario.
- UX & accessibility
Hero Section
According to the terminology of the Nielsen Norman Group (NN/g), a homepage usually conveys its value proposition through a descriptive tagline and compelling content in this area. NN/g recommends placing the most important content above the fold, i.e., where the page is visible without scrolling. NN/g cites weak implementations as a generic welcome text and a generic stock photo.
- AI fundamentals & models
Heterogeneous Graph
Instead of treating all entities equally, a heterogeneous graph can, for example, model customers, products, and suppliers as well as different types of relationships separately.
- AI fundamentals & models
Heuristic
Heuristics use problem knowledge, approximations, or simplified rules to reduce search or decision spaces.
- AI fundamentals & models
Heuristic Search
Instead of blindly exploring a search space completely, the method evaluates intermediate states according to a heuristic. For example, A* combines the cost so far with estimated remaining costs.
- Web design & development
Higgsfield
Higgsfield is an AI service for video and image creation, described by the provider as an "AI-native creative suite." According to the provider, the company develops its own models and tools, such as Cinema Studio and Soul 2.0, and additionally integrates partner models. New models and platform updates are released weekly according to the provider, so the feature set is continuously changing.
- Web design & development
High Availability
Depending on the system, measures such as redundancy, failover, replication, health checks, and robust operational processes contribute to high availability. A single percentage value without a measurement window and SLA context is of little significance.
- AI fundamentals & models
High-Level Representation
In representation learning, data is transformed into features that can support further predictions. Deeper model architectures can increasingly compose abstract features from simpler representations. Depth alone does not guarantee this.
- Data integration & search
Hit Rate
Here, hit rate means the hit rate of a ranked list, not recall or cache metric. In the referenced recommendation experiment, one withheld target object is searched per test person. The hit ratio evaluates its occurrence among the top results and is averaged over test persons.
- Data integration & search
HNSW-Index
HNSW uses a hierarchy of neighborhood graphs over subsets of stored elements. The search starts in an upper layer. The approach serves approximate neighbor search and is therefore not equivalent to guaranteed exact search.
- AI fundamentals & models
Holdout Validation
A dataset is split into separate subsets, typically training and test, and optionally validation. The separation aims to provide a more realistic estimate of generalization.
- Web design & development
Homomorphic Encryption
Depending on the method, limited or very general operations are supported. The result remains encrypted and can subsequently be decrypted with the appropriate key.
- Web design & development
Horizontal Scaling
Instead of making a single server bigger, scale-out distributes load across multiple units. For this, application, state, data storage, and load balancing must be designed accordingly.
- Web design & development
HTTPS
TLS protects the connection against eavesdropping and tampering on the transport path and usually authenticates the server via certificates. HTTPS does not automatically make the application itself secure.
- AI fundamentals & models
HuBERT
HuBERT first generates discrete target clusters from audio and trains a model to predict masked segments based on these hidden units. This enables learning speech representations with few labeled data.
- AI fundamentals & models
Hugging Face
Hugging Face describes its platform offering as a place for collaboration on machine learning projects. The Hub allows hosting models, datasets, and interactive applications. Git-based repositories support version control and collaborative work.
- AI fundamentals & models
Hugging Face Tokenizers
The library processes text through configurable steps: normalization, pre-tokenization, token model, and post-processing. The token model further splits the prepared units and assigns them IDs from its vocabulary.
- AI fundamentals & models
Human Evaluation
People judge, for example, correctness, usefulness, style, safety, or preferences. Good evaluation requires clear rubrics, suitable samples, and as consistent rating processes as possible.
- Process automation
Human-in-the-Loop
Quality and governance principle, in which human experts review and approve decisions at critical control points.
- Application & enablement
Human-in-the-Loop Fatigue
The term describes fatigue in people who must repeatedly review or confirm AI suggestions. Time pressure, monotonous tasks, and too many alerts can cause problematic outputs to be overlooked.
- Application & enablement
Human-in-the-Loop Tax
Human-in-the-Loop Tax is a figurative expression for time, coordination, and costs arising from human review or rework. It does not mean a tax or a fixed cost rate.
- AI fundamentals & models
HumanEval
The dataset contains handwritten programming tasks with function signatures, docstrings, and tests. Results are often reported via pass@k.
- AI fundamentals & models
Hybrid AI System
An important form of hybrid AI combines data-driven learning with symbolic knowledge processing. Neuro-symbolic systems connect neural networks with explicit symbolic methods. The umbrella term alone does not specify how the components are coupled.
- AI fundamentals & models
Hybrid Recommender System
Different recommendation methods are combined to leverage their diverse information bases. Whether this improves quality must be verified for the specific use case.
- Web design & development
Hybrid Role-Based and Attribute-Based Access Control
A hybrid access control design that combines role-based base permissions with attribute-based conditions on subject, resource, action, or environment. It is not a single standardized model.
- Data integration & search
Hybrid Search
Lexical methods find exact terms, while semantic vector search finds similar meanings. A ranking or fusion method combines both result lists.
- AI fundamentals & models
Hyena
The architecture was studied as a subquadratic alternative to the classical attention operator for long sequences. It is a research approach, not a general industry standard.
- AI fundamentals & models
Hyperbolic Tangent (tanh)
The hyperbolic tangent is a smooth function that maps real inputs to values between minus one and one and is used as an activation function.
- AI fundamentals & models
Hyperparameter
Examples include learning rate, regularization strength, tree depth, or batch size. They are set before or during a training process and often optimized using validation data.
- AI fundamentals & models
Hyperparameter Optimization
Methods range from manual search and grid/random search to Bayesian optimization and bandit-based methods. A clean separation of validation and test remains necessary.
- Visibility & SEO
Hyperpersonalization
In the e-commerce application described here, content should be available not only for a broad target group but for user-specific selection. Vempati et al. investigate the automatic generation and evaluation of many advertising banners. Their contribution mainly addresses the material for a personalization engine, not its complete decision logic.
- Web design & development
HyperText Markup Language (HTML)
The standardised markup language for structuring text, images and elements on the web.
- AI fundamentals & models
Hypothesis Generation
Hypothesis generation refers to developing assumptions that can subsequently be investigated. In the AI context, language models can generate suggestions based on scientific knowledge; their plausibility does not replace verification.
- AI fundamentals & models
Hypothesis Testing
Typically, hypotheses, test statistics, significance level, and assumptions are predefined. A p-value is not the probability that the null hypothesis is true.
- Visibility & SEO
Ideal Customer Profile (ICP)
In the B2B context, an ICP bundles verifiable characteristics such as industry, company size, processes, technology, needs, and purchasing conditions. It is not a single real customer and not the same as a persona.
- Web design & development
Identity and Access Management (IAM)
An IAM system typically supports provisioning, authentication, roles, policies, access control, and deprovisioning. Identity and authorization should be considered separately.
- Web design & development
Ideogram
According to the documentation, Ideogram converts a natural language description (prompt) into images and videos and allows editing them in the same place. Own models as well as partner models from other providers can be used; access is via web app, iOS app, or API. Exact words, layouts, and colors can be placed into the image via prompt or JSON according to the documentation.
- AI fundamentals & models
IFEval
The benchmark contains prompts with automatically verifiable conditions, such as specific keywords, lengths, or format requirements. This makes instruction following objectively measurable compared to purely subjective evaluation.
- AI fundamentals & models
Image Captioning
Models combine computer vision and language generation to convert objects, scenes, or relationships into text. A generated caption is not automatically a suitable accessible alt text.
- AI fundamentals & models
Image Classification
An image classification model receives an image and predicts which class or classes it belongs to. Unlike object recognition, it does not locate objects with bounding boxes; unlike segmentation, it does not assign classes to individual pixels.
- AI fundamentals & models
Image Generation
In image generation, a generative model generates visual content, for example using text prompts, reference images or masks. Diffusion models are a common approach. Image classification, on the other hand, assigns classes; generative image processing such as inpainting overlaps with image generation.
- AI fundamentals & models
Image Segmentation
In image segmentation, not only does the entire image receive a label, but pixels or regions are classified. Semantic segmentation assigns pixel classes; instance segmentation additionally separates individual object instances of the same class.
- AI fundamentals & models
Image Understanding
Image understanding here refers to task-related evaluation of image content by AI. Visual question answering is a research example: a system processes an image and a related question and generates an answer.
- AI fundamentals & models
Image-to-Image Generation
The input image provides visual information that a generative model modifies, supplements, or stylistically transforms. How much of the original remains depends on the model and parameters.
- AI fundamentals & models
Image-to-Text Processing
Image-to-text processing here refers to methods that render image content as text, such as image descriptions or answers to image questions. BLIP is a research model that combines image and language processing for such tasks.
- AI fundamentals & models
Image-to-Video Generation
A generative video model uses the image as a visual condition and generates motion, camera dynamics, or temporal changes. Consistency and physical plausibility remain technical challenges.
- AI fundamentals & models
ImageBind
ImageBind is a model for joint numerical representations of images, text, audio, depth images, thermal images, and motion sensor data. Similar content is intended to become comparable across media.
- AI fundamentals & models
Imitation Learning
The model learns a policy or decision rule from examples of states and actions. Behavioral Cloning is a direct form; interactive methods like DAgger address distribution shifts.
- AI fundamentals & models
Implicit Feedback
Examples include clicks, purchases, dwell time, or skipped content. The signal shows behavior but not clearly the underlying preference or satisfaction.
- Visibility & SEO
Impression
The exact definition depends on the system: In Google Search Console, an impression counts when a link to the page has been seen in a Google service; advertising platforms use their own measurement rules.
- AI fundamentals & models
In-Context Learning
The ability of a language model to learn new tasks or formats directly from examples provided promptly, without changing the model weights.
- Web design & development
Incident Response
Modern incident response programs combine technical response with communication, roles, risk, and recovery processes. NIST SP 800-61 Rev. 3 integrates incident response into the overall cybersecurity risk management process.
- Visibility & SEO
Incrementality
In marketing, ideally a control group is compared with a treated group. This aims to distinguish between mere attribution and actually caused additional effect.
- AI fundamentals & models
Inductive Reasoning
Unlike deduction, induction does not logically guarantee the conclusion. It generalizes from examples, measurements, or patterns and therefore depends on data quality, sample, and uncertainty.
- AI fundamentals & models
Inference
During training, model parameters are adjusted; during inference, these parameters are typically used to compute results for new data. Costs, latency, and hardware requirements depend heavily on the model and workload.
- Web design & development
Inference as a Service
Inference as a service here means retrieving predictions of a trained model via a separate service interface. The application calls the prediction service instead of embedding model execution directly into its own workflow.
- Web design & development
Inference Cost Optimization
Optimization of inference costs seeks a more cost-effective execution of model queries under given quality requirements. For language model services, FrugalGPT investigates, among other things, the selection and combination of different models.
- Web design & development
Inference Engine
It loads or processes models and executes their operations. Depending on the runtime, graph optimization, memory management, and hardware acceleration are added.
- Web design & development
Inference Optimization
Inference optimization changes the processing of model queries with regard to computational effort and memory requirements. For large language models, a research overview distinguishes approaches at the data, model, and system levels.
- AI fundamentals & models
Inference-Time Compute
This can include longer generation, multiple candidates, search, verification, or iterative revision. More computational effort does not automatically improve a result and typically increases latency and costs.
- AI fundamentals & models
Inference-Time Scaling
Instead of scaling only model size or pretraining, systems can generate, search, verify, or revise more candidates at query time. Which strategy makes sense depends on difficulty and budget.
- Visibility & SEO
Influencer Marketing
The effect depends not only on reach but also on audience fit, credibility, labeling, content, and measurement method. Advertising must be transparently recognizable according to applicable law.
- Data integration & search
Information Extraction
These include, for example, entities, relationships, events, or field values. Modern systems use rules, classical ML methods, or language models; results require validation depending on risk.
- Web design & development
Information Hiding
A module should make other parts of a system dependent only on the necessary interface, not on internal data structures or decisions. This limits changes locally.
- Data integration & search
Information Retrieval
Search systems index documents and rank hits by relevance. Lexical, semantic, and hybrid methods differ in which signals they use.
- Web design & development
Information Security
The protection of confidentiality, integrity, and availability of information as well as other appropriate security objectives.
- Process automation
Inngest
Inngest is a platform for persistent execution of functions and workflows. Individual steps are executed separately; stored results allow resuming a workflow after errors.
- Application & enablement
Innovation
Innovation, as used here in the Baldrige understanding, means a significant change that improves products, processes, organization, or societal well-being and increases value for stakeholders.
- Web design & development
Inpainting
According to Rombach et al., inpainting is the task of filling masked image areas with new content. The Diffusers documentation describes the mask as follows: White pixels mark the area to be filled, black pixels the area to be preserved; the white pixels are filled based on a prompt. Lugmayr et al. use diffusion models (DDPM) for inpainting.
- AI fundamentals & models
Input Tokens
Input tokens are the units resulting from tokenization of the model input, including transmitted instructions, parts of conversations, and additional information.
- Data integration & search
Insights
In an analysis context, an insight connects observations with their meaning for a question. A measurement value only becomes useful through contextualization; a found correlation does not yet explain a cause.
- AI fundamentals & models
Instance Normalization
In the usual form, statistics are calculated per instance and channel, not jointly over the current batch. Implementations like PyTorch can optionally use stored statistics during evaluation.
- Process automation
Instant Checkout in ChatGPT
Instant Checkout combines product search in ChatGPT with a direct purchase. Orders, payments and fulfillment remain with the merchant; technically, the integration is based on the Agentic Commerce Protocol. Product results are separate from ads and are selected organically, according to OpenAI.
- AI fundamentals & models
Instruction Tuning
The model is fine-tuned on many instruction/response examples so that it better follows new instructions. It is not the same as prompting because model weights are changed.
- Data integration & search
INSTRUCTOR Embeddings
INSTRUCTOR embeddings are numerical text representations generated together with a task or domain instruction. This instruction describes what the representation is used for, for example, for a search.
- AI fundamentals & models
Instrumental Variable
A valid instrument influences the explanatory variable, affects the outcome only through this variable, and must not be connected to relevant unobserved confounders. These assumptions are strong and must be justified.
- AI fundamentals & models
Integrated Gradients
The method computes gradients at intermediate points between reference and actual input. The result assigns attributions to features and satisfies the axioms Sensitivity and Implementation Invariance formulated by the authors.
- Web design & development
Integration Testing
Unlike unit testing, the focus is on interaction: data formats, APIs, databases, queues, authentication, and error cases between components are tested together.
- Application & enablement
Intelligent Tutoring System
Intelligent tutoring systems are developed and evaluated for specific learning settings. Learning objectives, traceability, and subject quality must be verified in each deployment.
- AI fundamentals & models
Intent Classification
The method is frequently used in chatbots, routing, and NLP. It requires a meaningful intent schema and examples; open or ambiguous inputs should have a fallback strategy.
- AI fundamentals & models
Intent Recognition
Intent recognition here refers to assigning an utterance to a task class in a dialogue system. The system must also handle queries that do not belong to any of its supported classes.
- AI fundamentals & models
Inter-Annotator Agreement
Depending on data type and number of raters, metrics like Cohen’s Kappa, Fleiss’ Kappa, or Krippendorff’s Alpha are used. High agreement does not automatically mean labels are factually correct.
- UX & accessibility
Interface
An interface describes the allowed inputs and outputs as well as their rules. For a user interface, this concerns operation; for technical interfaces, communication between components. The respective context must be specified.
- AI fundamentals & models
Interpretability
The term is not uniformly defined by a single metric. Depending on the goal, it concerns global model structure, local decisions, feature influences, or understandable surrogate models.
- AI fundamentals & models
Interpretable Machine Learning
Interpretable machine learning uses models to gain relevant knowledge about relationships in data. Which representation is helpful also depends on the question and the audience.
- AI fundamentals & models
Intersection over Union (IoU)
In object and segmentation evaluation, IoU is often calculated between predicted and actual bounding box or mask. The value range is between 0 and 1.
- AI fundamentals & models
Inverse Reinforcement Learning
In classical reinforcement learning, the reward is given and a policy is learned. IRL partially reverses this: from demonstrations, a reward model or preference structure is estimated.
- AI fundamentals & models
IP-Adapter
IP-Adapter is a trained add-on for text-to-image diffusion models. It processes an image prompt separately from the text prompt and combines both during generation.
- Application & enablement
IPO
In an Initial Public Offering, shares are offered publicly to investors for the first time. New shares may be issued or existing shares offered; the specific structure and applicable market rules are decisive.
- AI fundamentals & models
Iterative Deepening
With finite branching and a solution at finite depth, the method finds a shallowest solution. With uniform positive step costs, this solution is also cost-minimal. Memory requirements remain low in the examined tree search model.
- Application & enablement
Iterative Prompting
Iterative prompting here refers to successive model calls where feedback is used to revise an output. Self-Refine is a research example with feedback and revision by the same language model.
- AI fundamentals & models
Jaccard Similarity
For a non-empty union, the value ranges between 0 and 1. For two empty sets, a convention must be established. The measure is suitable for binary or set-like data and, unlike some other similarity measures, does not consider jointly missing features.
- AI fundamentals & models
Jailbreak
Attackers can aim with inputs to bypass protection rules, generate unauthorized content, or trigger actions. The terms jailbreak and prompt injection overlap; OWASP classifies jailbreaking as a form of prompt injection aimed at circumventing protection rules.
- AI fundamentals & models
Jailbreak Resistance
The meaningfulness of a score depends on the defined threat model, attack set, and system under examination. Additionally, actual impacts on data and actions should be evaluated.
- AI fundamentals & models
Jailbreaking
Jailbreaking in the language model context considered here refers to attacks intended to cause a security-aligned model to produce undesired outputs. Research investigates, among other things, conflicts between learned objectives and limits of their generalization.
- AI fundamentals & models
Jamba
Jamba here refers to the hybrid language model presented in the 2024 research work. It combines Transformer layers with Mamba layers and uses a selection of specialized subnetworks in part of the model.
- Web design & development
JavaScript
JavaScript (JS) is an interpreted or just-in-time compiled programming language with first-class functions. In the browser, it modifies page content via the DOM and fetches data from servers using the fetch() API. Outside the browser, it runs in runtime environments like Node.js, Deno, and Bun. MDN distinguishes the language itself from web APIs and the DOM, which belong to the browser environment.
- Web design & development
JAX
JAX is a Python library for computations with numerical arrays. Function transformations can, for example, compile a function, compute its derivative, or apply it to stacked data.
- Application & enablement
Jevons Paradox
If a resource becomes cheaper or more efficient per use, usage can increase so much that total consumption rises. Whether this actually happens depends on elasticities and context.
- Web design & development
Jitter
Not only average latency matters: uneven arrival times can disrupt audio, video, streaming, and interactive systems. Measurement definitions differ depending on protocol and context.
- Process automation
Job Scheduling
Schedulers consider priorities, dependencies, schedules, capacities, retries, and parallelism depending on the system. A cron schedule is just a simple form of this.
- AI fundamentals & models
Joint Distribution
It maps dependencies between variables; marginal and conditional distributions can be derived from this. Correlation is only a limited aspect of this relationship.
- Web design & development
JSON Mode
The JSON mode here refers to the corresponding OpenAI API feature for JSON-format outputs. It aims for valid JSON but does not ensure compliance with a particular domain data schema.
- Web design & development
JSON Schema
The specification consists of the parts Core and Validation: Core defines the basics, Validation defines the validation keywords. These keywords allow requirements to be set for a JSON document, such as type ("type"), required properties ("required"), or upper and lower numeric limits ("maximum", "minimum").
- Web design & development
JSON Web Token (JWT)
JWTs are signed as JWS or MAC-protected, or encrypted as JWE. A commonly used signed JWT is not automatically encrypted; its payload can be readable.
- Visibility & SEO
JSON-LD
JSON-LD extends JSON with semantic context and globally identifiable terms. It is often used on websites for structured data, for example with Schema.org vocabulary.
- Web design & development
Jupyter Notebook
Jupyter Notebook is an application for interactive computational documents. A notebook contains code cells, explanatory text, and outputs such as tables or graphics. A kernel executes the code and maintains the current computational state.
- Web design & development
Just-in-Time Compilation (JIT)
A runtime can observe frequently executed code paths and compile them dynamically. This causes warm-up costs but often results in better runtime performance than purely interpreted execution.
- Application & enablement
k-Anonymity
Quasi-identifiers are generalized or suppressed for this purpose. The method reduces certain re-identification risks but does not automatically protect against attribute disclosure or background knowledge.
- AI fundamentals & models
k-Armed Bandit
The core problem is the trade-off between exploring new options and exploiting the best option so far. Unlike full reinforcement learning, the classic bandit has no complex state transitions.
- AI fundamentals & models
k-Fold Cross-Validation
In each run, k-1 folds are used for training and one fold for validation. The results are aggregated to obtain more stable estimates than with a single random split.
- AI fundamentals & models
k-Means
The method typically minimizes the sum of squared distances within clusters. Result and quality strongly depend on scaling, distance metric, k, and initialization.
- AI fundamentals & models
k-Means++
The first center is chosen randomly; further centers are chosen with higher probability where points are far from existing centers. Then ordinary k-Means proceeds.
- AI fundamentals & models
k-Nearest Neighbors (kNN)
For a new input, the nearest data points are searched according to a distance metric; their classes or values determine the output. The method essentially stores the training data instead of learning a parametric model.
- Data integration & search
k-Nearest Neighbors Search
Depending on data volume, the search can be exact or approximate. For large embedding collections, ANN indices like HNSW are often used.
- Application & enablement
k-Shot Prompting
The examples show format, mapping, or desired behavior in context. The model weights are not permanently changed; k only describes the number of demonstrations.
- AI fundamentals & models
Kahneman–Tversky Optimization (KTO)
The method formulates a prospect-theory-inspired utility function and optimizes desired versus undesired responses. It is a post-training approach and not a general inference method.
- AI fundamentals & models
Kalman Filter
It combines a process model with new measurements and weights both according to their uncertainty. The classical filter assumes linear dynamics and certain noise assumptions.
- AI fundamentals & models
Kaplan–Meier Estimator
It calculates the probability of not having experienced an event by a certain time as the product of conditional survival probabilities. Censoring must be correctly modeled.
- AI fundamentals & models
Kernel Trick
This method avoids explicitly calculating high-dimensional features. It is especially associated with Support Vector Machines and Kernel PCA.
- Web design & development
Key Management
The regulated generation, storage, distribution, use, rotation, blocking, and destruction of cryptographic keys.
- Web design & development
Key Management Service (KMS)
A KMS supports creating, storing, and controlled use of cryptographic keys. Applications can, for example, encrypt data with managed keys or request signatures. Specific functions depend on the service.
- Visibility & SEO
Key Performance Indicator (KPI)
A KPI is not just any available metric. It should be tied to a goal, a definition, data source, frequency, and responsibility.
- Web design & development
Key Rotation
Rotation limits the usage duration of a key and can reduce the impact of a compromise. Frequency and procedures should fit the cryptoperiod and risk model; old data must still be decryptable correctly.
- Visibility & SEO
Keyword Cannibalization
Keyword cannibalization occurs when multiple own pages serve the same search intent so similarly that their combined search performance is adversely affected. Multiple hits for a word alone do not yet prove a problem.
- Visibility & SEO
Keyword Difficulty
SEO tools estimate with a metric how challenging a good ranking for a search query might be. Calculation and scale differ; a value is not a fixed probability for one’s own success.
- Visibility & SEO
Keyword Research
It combines qualitative user questions with data from Search Console, keyword tools, SERPs, and the own website. Search volume is an estimate and not a ranking or business value promise.
- Web design & development
Kling AI
Kling AI is a generative tool for images and videos. The reviewed guide describes image-to-video: Pre-generated images become dynamic videos whose subject and background motion can be described via text.
- Data integration & search
Knowledge Base
For example, a knowledge base bundles articles, FAQs and work instructions. It can be used internally or publicly and serve as a source for a search or an assistant. Timeliness, responsibilities and access rights determine their reliability; an error-free AI response does not automatically follow from this.
- AI fundamentals & models
Knowledge Cutoff
It depends on the model and does not mean that all information up to this date is contained or correct. With web, file, or retrieval tools, a system can obtain more current data.
- AI fundamentals & models
Knowledge Distillation
In the teacher-student process, information from a teacher model or ensemble is used for training a student model, such as probability distributions. A smaller model can be cheaper to operate, but does not automatically reach the quality of the teacher.
- Data integration & search
Knowledge Graph
A graph representation of entities, relationships, and often schema information with stable identities and provenance.
- AI fundamentals & models
Knowledge Graph Embedding
Models like TransE learn representations in which known relationships are geometrically modeled. The embeddings can be used for link prediction or similarity.
- Visibility & SEO
Knowledge Graphs in Marketing
Knowledge graphs in marketing represent domain entities and relationships from marketing knowledge. The cited work describes a pipeline that identifies entities and relations from text collections.
- AI fundamentals & models
Knowledge Tracing
The estimation of a person's latent learning state from temporal sequences of tasks and answers.
- Visibility & SEO
KPI Tree
It shows computational relationships or explicitly marked assumptions about influences between metrics. The specific structure must be tailored to the question and the company.
- Web design & development
KServe
KServe is a platform for model inference on Kubernetes. It provides resources and runtime components to deploy, scale, and monitor prediction services.
- Web design & development
Kubeflow
Kubeflow is a Kubernetes-native platform composed of modular open-source projects for data and AI workloads. The documentation describes scalable and combinable tools along the AI lifecycle.
- Web design & development
Kubernetes
Kubernetes is an open-source platform for managing containerized workloads and services. It uses declarative configuration to describe the desired state of an application and to technically maintain it.
- Web design & development
Kubernetes Network Policy
A Kubernetes API resource that describes allowed ingress and egress traffic for selected pods. It becomes effective only if the deployed network plugin supports NetworkPolicy.
- Web design & development
Kubernetes Operator
A Kubernetes controller following the operator pattern that automates application-specific operational knowledge in a reconciliation loop, often using custom resources.
- AI fundamentals & models
KV-Cache
During generation, previous attention states do not need to be fully recomputed for each new token. This reduces computation but increases memory usage.
- AI fundamentals & models
L1 Regularization
It can shrink weights exactly to zero, promoting sparse models or feature selection. The effect depends on scaling and regularization strength.
- AI fundamentals & models
L2 Regularization
It typically pulls weights continuously toward zero without usually setting them exactly to zero. With optimizers, "weight decay" is not always mathematically identical to L2 regularization.
- AI fundamentals & models
Label Leakage
This can cause a model to achieve unrealistically good offline results because it uses signals that would not be available at the actual prediction time.
- AI fundamentals & models
Label Smoothing
Instead of assigning probability 1 to the correct class and 0 to all others, a small portion is distributed. This can reduce overconfidence but is not beneficial for every task.
- AI fundamentals & models
Label Studio
Label Studio is an open-source tool for data annotation. It supports various data types, configurable labeling interfaces, and integration of model predictions into the workflow.
- AI fundamentals & models
LAMB Optimizer
The method was developed for large-batch training of large models. Later work showed its advantages are not universal and strong standard optimizers can compete.
- Web design & development
Landing Page
A landing page is an entry or target page in the respective usage context; for campaigns it is often tailored to the target audience, message, and desired action. A homepage or content page can also serve this function.
- UX & accessibility
Landing Page Optimization
Message, information order, CTA, form, trust, performance, and accessibility can be optimized. Changes should be tested against target metrics.
- Process automation
LangChain
LangChain is a framework for applications with language models. The documentation describes preconfigured agent building blocks that connect a model with tools, instructions, and extensions.
- Process automation
Langfuse
Langfuse is an open-source platform for observability of language model applications. It captures traces with inputs, outputs, runtimes, token usage, and tool calls.
- Process automation
LangGraph
LangGraph is a framework and runtime for long-running, stateful agent workflows. The documentation cites persistent execution, streaming, human interventions, and stored state as core features.
- Process automation
LangSmith
LangSmith is a platform from LangChain for recording, investigating, and evaluating runs in applications with language models and agents. It combines tracing with offline and online evaluation so teams can compare variants and detect production issues.
- AI fundamentals & models
Language Model
Modern neural language models mostly operate on tokens and are trained autoregressively or with masking objectives. A language model is not automatically a large language model.
- AI fundamentals & models
Large Behavior Model
The Large Behavior Model (LBM) is a research approach that uses a language model to simulate individual customer behavior. It combines a persistent behavioral profile from historical transactions with a dynamic decision environment provided at runtime via retrieval.
- AI fundamentals & models
Large Language Model (LLM)
A large language model is trained on extensive data and processes language based on learned statistical patterns. During text generation, it selects successive tokens. Liquid outputs can be factually wrong; the task and model variant determine the suitability.
- Visibility & SEO
Large Language Model Optimization (LLMO)
Large Language Model Optimization (LLMO) is an industry term used inconsistently for optimizing content, websites, and brand presence with the goal of being mentioned or cited in AI-generated responses. Basic SEO requirements still apply for Google AI features; special files or markup are not required there.
- AI fundamentals & models
LARS Optimizer
It was developed to stabilize training of large image models with greatly increased batch sizes. Benefit depends on the workload.
- Visibility & SEO
Last-Click Attribution
The model is easy to understand but largely ignores earlier contacts and says nothing about actual causality.
- Data integration & search
Late Interaction
In ColBERT, token representations are precomputed separately; for search, query and document tokens are compared via MaxSim-like operations. This combines detail with better precomputability.
- Web design & development
Latency
In distributed systems, it consists of network, queue, computation, database, and rendering times. Averages often obscure slow tail requests.
- Web design & development
Latency Budget
Instead of measuring only end-to-end latency, network, database, model, tools, or rendering receive their own budgets. This is an architectural practice, not a universal standard.
- AI fundamentals & models
Latent Diffusion
An autoencoder compresses images; the diffusion model operates on these latent representations and then decodes them back. This can reduce computational effort.
- AI fundamentals & models
Latent Space
Nearby points can have similar properties, but axes do not have to be directly human-interpretable. The space arises from the model and training objective.
- AI fundamentals & models
Latent Variable
Latent variables can represent hidden factors, classes, or states. Their interpretation depends on the model and is not automatically unique.
- AI fundamentals & models
Layer Dropping
LayerDrop trains transformers with structured dropout on entire layers so that subnetworks of reduced depth can be used during inference.
- AI fundamentals & models
Layer Normalization
Unlike batch normalization, LayerNorm does not require batch statistics and is therefore well suited for sequence models and transformers.
- Web design & development
Lazy Loading
Lazy loading defers the loading of suitable resources until they are needed. Images outside the initially visible area are a common example; loading may begin before they become visible. Delaying primary content that is needed immediately is usually unsuitable.
- Visibility & SEO
Lead Generation
Channels can be search, content, events, referrals, or advertising. A lead is not yet a customer; quality, consent, and a clean funnel are crucial.
- Visibility & SEO
Lead Lifecycle Stages
Terminology varies by company, for example Subscriber, Lead, MQL, SQL, Opportunity, or Customer. Clear transition criteria are important rather than tool labels.
- Visibility & SEO
Lead Nurturing
The planned, relevant accompaniment of potential customers over multiple contacts up to a suitable next action.
- Process automation
Lead Scoring
Scores can be rule-based or model-based. Good systems separate suitability, interest, and actual purchase probability and are calibrated against real outcomes.
- AI fundamentals & models
Leaky ReLU
It aims to reduce the problem of permanently inactive ReLU units. The negative slope parameter is a model decision.
- Application & enablement
Learning Management System
An LMS organizes the learning process; it is not automatically adaptive or AI-based. Interfaces like LTI, SCORM, or xAPI can integrate content and learning events.
- Application & enablement
Learning Objectives
Good learning objectives are observable, related to the target group and context, and linked with appropriate learning and assessment activities.
- AI fundamentals & models
Learning Rate
Too high values can make training unstable; too low values slow it down or lead to suboptimal optimization. Effective step size also depends on the optimizer.
- AI fundamentals & models
Learning Rate Finder
The method provides a starting range for the learning rate but does not replace full validation. Implementations vary.
- AI fundamentals & models
Learning Rate Schedule
Typical strategies are Warmup, Step Decay, Cosine Decay, or plateau-based adjustment. The schedule influences stability and convergence.
- Data integration & search
Learning Record Store
An LRS receives statements in the Actor-Verb-Object pattern, manages them, and provides APIs for retrieval. It is not the same as a full learning management system.
- Data integration & search
Learning to Rank
Approaches optimize pointwise, pairwise, or listwise ranking objectives. Training data can include relevance judgments or user signals.
- Web design & development
Least Privilege
Rights should be limited in time, function, and data scope. This reduces the impact of compromised accounts, misconfigurations, or agent errors.
- AI fundamentals & models
Lemmatization
Unlike simple stemming, it typically considers part of speech and linguistic rules. This allows "ging" and "gehen" to be merged.
- AI fundamentals & models
Length Penalty
Without adjustment, sums of log-probabilities can systematically favor longer or shorter sequences. Specific formulas vary between implementations.
- Web design & development
Leonardo AI
Leonardo AI is a generative creative tool with a web application and programming interface. Via the API, image generations can be initiated, tracked as jobs, and retrieved upon completion; webhooks can report results event-based.
- Data integration & search
Lexical Search
Methods like BM25 weight terms by frequency and rarity. They are suitable for exact terms, IDs, and technical terms. Without synonym expansion or additional semantic methods, purely lexical matching does not automatically capture synonyms.
- AI fundamentals & models
LiDAR
A sensor emits laser pulses and measures the reflection time. From this, geometry and the environment can be captured three-dimensionally.
- Visibility & SEO
Lifecycle Marketing
Phases can include acquisition, activation, usage, retention, expansion, or reactivation. The exact segmentation is company-specific.
- Visibility & SEO
Lift
Depending on context, lift can be defined differently statistically, marketing-wise, or model-wise. Without a clear baseline, the value is not interpretable.
- Visibility & SEO
Lift Chart
Typically, sorted model scores are considered in groups and compared with baseline expectation. Definitions may vary by tool.
- AI fundamentals & models
LIME
The method generates perturbed examples, queries the black-box model, and fits a locally interpretable surrogate. Explanations can be unstable with small changes.
- AI fundamentals & models
Linear Attention
Methods use kernel tricks, factorizations, or other transformations to avoid the full attention matrix. Quality and actual hardware efficiency vary greatly.
- Visibility & SEO
Link Equity
Search engines do not publish a metric called “Link Equity.” Historically, PageRank describes how link structures distribute authority signals; modern rankings use many more signals.
- Visibility & SEO
Link Graph
From this structure, reachability, central pages, and graph-based ranking signals can be analyzed.
- AI fundamentals & models
Link Prediction
Methods use structure, node features, or graph embeddings. A high model probability does not mean the relationship actually exists.
- Web design & development
Linting
A linter typically does not execute code but checks syntax trees, type information, or rules. It complements tests and security scanning but does not replace them.
- AI fundamentals & models
Lion Optimizer
It requires less optimizer state than Adam because no second moment is stored. The authors report advantages in several workloads but not universally.
- Web design & development
LiteLLM
LiteLLM is a self-hosted AI gateway. Applications send requests to a unified OpenAI-compatible endpoint; the gateway forwards them to configured model providers and can centrally manage access, costs, rate limits, fallbacks, logging, and guardrails.
- Web design & development
LiveCodeBench
LiveCodeBench is a benchmark for the programming capabilities of large language models. It continuously collects new tasks from programming competitions and evaluates not only code generation but also skills such as self-correction, code execution, and prediction of test outputs.
- AI fundamentals & models
Llama
Llama is a family of large language models from Meta. The models are provided in different sizes and variants; their capabilities, context lengths, licenses, and security components depend on the respective generation and release.
- Data integration & search
LlamaIndex
LlamaIndex is a development framework for LLM applications over external or proprietary data. It supports loading and transforming data, indexing and storage, as well as various retrieval and query methods for RAG and agent applications.
- Application & enablement
LLM Brain Rot
LLM Brain Rot refers in a controlled study to the observed performance decline of language models after continued pretraining on social media texts operationalized as low quality. Measured declines included reasoning, long-context understanding, safety, and increased skipping of reasoning steps.
- AI fundamentals & models
LLM Evaluation
Public benchmarks are rarely sufficient. Good evaluations combine custom test cases, automatic metrics, human evaluation, and, if applicable, LLM-as-a-Judge.
- Process automation
LLM Observability
It extends classical observability with GenAI-specific signals. Prompt and response contents can be sensitive and should not be logged indiscriminately.
- Process automation
LLM Routing
Routing can be rule-based or learning-based and differentiate between providers, model sizes, or specialized models. Fallbacks and evaluation data are crucial.
- Web design & development
LLM Security
Risks range from prompt injection and data leakage to excessive agent privileges, insecure outputs, and supply chain issues. Security must consider system boundaries, not just the model.
- AI fundamentals & models
LLM-as-a-Judge
The method can generate pairwise comparisons, scores, or justifications and scale evaluation. However, LLM-Judges exhibit biases, including position, verbosity, or self-preference effects.
- Web design & development
LLM-Gateway
It can consolidate authentication, rate limits, routing, logging, quotas, policies, and provider abstraction. The term is an architectural pattern, not a uniform standard.
- Web design & development
LLMOps
It extends MLOps/DevOps by prompt versions, model providers, evals, retrieval, token costs, guardrails, and nondeterministic outputs. The term is a practice family, not a single standard.
- Visibility & SEO
llms.txt
The file offers brief background information, hints, and links to detailed Markdown files. Agents are supposed to view or search it and follow the appropriate links; the details lie behind the links and are retrieved only as needed. The file itself should remain small enough to fit into the context.
- AI fundamentals & models
LMSYS
LMSYS, short for Large Model Systems, is a nonprofit organization promoting open AI models, datasets, systems, and evaluation tools. Its well-known projects include SGLang, FastChat, and the now spun-off Chatbot Arena.
- Web design & development
Load Balancing
Methods can use round robin, least connections, weighted, or health-based strategies. Load balancing supports scaling and availability but does not replace stateful architecture.
- AI fundamentals & models
Locality-Sensitive Hashing (LSH)
The method enables approximate similarity search without comparing every vector exactly with every other. Different LSH families fit different distance measures.
- AI fundamentals & models
Log Loss
For binary classification, Log Loss corresponds to the negative log-likelihood of the Bernoulli/logistic model; multiclass variants use cross-entropy over class probabilities.
- AI fundamentals & models
Log-Likelihood
For independently modeled observations, the product of probabilities or densities becomes the sum of their logarithms. This facilitates numerical calculation; the maximizing parameters remain the same.
- AI fundamentals & models
Logit
In binary logistic regression, the logit is log(p/(1-p)). In neural classifiers, raw scores before probability normalization are often also called logits.
- AI fundamentals & models
Logit-Bias
Depending on the API or implementation, positive or negative values are added to token logits. This is a controlling intervention in decoding, not a guarantee of a specific outcome.
- AI fundamentals & models
LogSumExp
The identity typically subtracts the maximum value before exponentiation and reduces numerical issues through stable computation. It frequently appears in softmax and probabilistic calculations.
- AI fundamentals & models
Long Context
A large nominal context window does not mean that information at every position is used equally reliably. Retrieval, structuring, and positional effects remain important.
- AI fundamentals & models
Long Context Management
Techniques include retrieval, summarization, chunk selection, relevance filtering, memory, and context compression. The term is a practice, not a single algorithm.
- AI fundamentals & models
Long Short-Term Memory (LSTM)
In today’s common variants, the memory cell as well as input, forget, and output gates control the information flow. LSTMs have been widely used especially for speech and time series.
- AI fundamentals & models
Long-Context Processing Cost
Long inputs must be processed and taken into account during generation. The effort depends on architecture, infrastructure, caching and tariff. The raw designation Context-Window-Tax is not a standardized tax or metric. Calculation effort and possible quality losses shall be measured separately.
- Visibility & SEO
Long-Tail Keywords
In total, many rare search queries can generate significant traffic. The term describes a distribution, not simply "keywords with at least three words."
- AI fundamentals & models
Lookahead Optimizer
A base optimizer performs several fast-weight steps; then slow weights move toward the fast weights. This can improve training stability but not in every workload.
- Visibility & SEO
Lookalike Audience
Concrete implementation, available signals, and privacy rules depend on the platform. Similarity does not mean identical purchase intent.
- AI fundamentals & models
LoRA vs. Full Fine-Tuning
Full fine-tuning retrains all parameters of a pretrained model for a task. Low-Rank Adaptation (LoRA) freezes pretrained weights and learns small low-rank matrices for selected layers, requiring significantly fewer parameters to be trained and stored.
- AI fundamentals & models
Loss Function
A function that expresses the deviation between model behavior and training goal as a value to be minimized.
- AI fundamentals & models
Loss Landscape
In deep networks, this space is high-dimensional and non-convex. Visualizations show only projections and can be distorted by parameterization.
- AI fundamentals & models
Lottery Ticket Hypothesis
"Winning tickets" are typically found through pruning experiments. The hypothesis inspired research on sparsity and pruning but is not a general guarantee for every model.
- Web design & development
Lovable
According to the manufacturer, Lovable generates a runnable application from a description, including frontend, backend, database, authentication, and integrations based on editable code. Each project is an application and produces a codebase that can be connected via Git synchronization with GitHub, GitLab, or Bitbucket.
- AI fundamentals & models
Low-Rank Adaptation (LoRA)
The method reduces trainable parameters and often memory requirements compared to full fine-tuning. Quality depends on rank, target modules, data, and task.
- Web design & development
Luma AI
Luma AI is a provider of generative media tools. The Dream Machine API starts image or video generation as asynchronous jobs, returns an ID, and allows status queries until the final result; supported features include text-to-video, image-to-video, and text-to-image.
- Web design & development
Mac mini M4 Pro
The Mac mini M4 Pro is a variant of the newly redesigned 2024 Mac mini. It combines the M4 Pro chip with a case about five by five inches and offers Thunderbolt 5 for fast external connections for the first time on a Mac mini with the M4 Pro model.
- AI fundamentals & models
Machine Learning
Methods where models derive patterns from data for predictions, classifications, or decisions.
- Web design & development
Machine Learning Pipeline
A reproducible chain of steps for data preparation, training, evaluation, deployment, and monitoring.
- Visibility & SEO
Machine Legibility
Machine legibility describes whether software can process content in a suitable format and evaluate its structure or meaning. Examples include semantic HTML and structured data. It also concerns systems outside AI and does not guarantee correct understanding.
- AI fundamentals & models
Machine Translation
Machine translation automatically transfers text from a source language to a target language. For spoken inputs or outputs, speech recognition and speech synthesis can also be used.
- AI fundamentals & models
Machine Unlearning
Methods range from complete retraining to approximate unlearning techniques. Proving that information has truly been removed is technically challenging.
- Visibility & SEO
Macro Conversion
Typical examples are purchase, qualified inquiry, or contract conclusion. The distinction from micro conversions is an analytics/marketing convention, not a technical standard.
- Web design & development
Malleable Software
Malleable software is a research ideal for software whose data, interface, and functions can be further developed and combined by users or local communities, rather than being exclusively defined by the manufacturer. The goal is more user autonomy with low adaptation barriers.
- AI fundamentals & models
Mamba
A sequence model architecture with input-dependent state space dynamics whose computational effort in the standard form scales linearly with sequence length.
- Process automation
Manus
Manus is an autonomous general AI agent. According to the manufacturer, it differs from a pure chatbot in that it plans tasks in its own computer environment, executes them, and delivers complete work results.
- Visibility & SEO
Marginal Cost per Acquisition
The marginal CPA is the ratio of additional marketing expenses to additional, clearly defined acquisitions. Expenses and results must relate to the same budget change.
- Visibility & SEO
Marginal Return on Ad Spend
In revenue-based evaluation, marginal ROAS is the ratio of additionally generated, attributable revenue to additional advertising spend. Platforms may instead use a broader conversion value that is not necessarily revenue.
- Visibility & SEO
Market Sentiment
It refers to the prevailing positive, negative, or neutral expectations in a defined market. Surveys, texts, and behavior provide different indicators.
- Process automation
Marketing Agent
A marketing agent is an agentic software system for marketing tasks. It combines a language model with planning, tool access, data or memory, and control steps to process, for example, audience analyses, content, or campaign tasks in multiple stages.
- Visibility & SEO
Marketing Attribution
Marketing attribution distributes the attributed contribution to a promotion, such as a request or a purchase, to contacts in the recorded user path. An attribution model can use rules or data-driven methods for this. Also observed AI references can be included; unrecorded contacts remain invisible. Attribution is not causal evidence.
- Process automation
Marketing Automation
Software-supported, rule- or data-based execution of recurring marketing processes via defined triggers and channels.
- AI fundamentals & models
Marketing Efficiency Ratio (MER)
For the same period, revenue is divided by defined marketing costs. The ratio considers the overall effect of the marketing mix without causally attributing revenue to individual contacts.
- Web design & development
Marketing Engineering
The systematic technical implementation of measurement, data flows, experiments, and marketing software.
- Visibility & SEO
Marketing Funnel
A simplified model of phases that people go through from awareness to desired action.
- Visibility & SEO
Marketing Measurement Framework
A documented mapping of business goals, questions, metrics, data sources, methods, and responsibilities.
- Visibility & SEO
Marketing Mix Modeling
A mostly statistical estimate of how marketing channels and external factors contribute to an aggregated outcome.
- Visibility & SEO
Marketing Qualified Lead (MQL)
A contact classified as marketing-relevant according to jointly agreed criteria; the criteria are organization-specific.
- AI fundamentals & models
Masked Language Modeling
A training objective where hidden tokens are reconstructed from their context.
- AI fundamentals & models
Massive Text Embedding Benchmark (MTEB)
The benchmark collection includes retrieval, similarity, and classification. It shows that an embedding model is not equally suitable for every task type.
- Data integration & search
Master Data Management
Governance, processes, and technology for the consistent management of central business objects such as customers, products, or suppliers.
- Application & enablement
Mastery Learning
A learning approach where clear competency goals, formative tests, and targeted support determine when learners proceed. The required learning time may vary.
- Web design & development
Mastra
A framework for developing AI applications with agents, tools, memory, and workflows. It supports different model providers and observable execution.
- AI fundamentals & models
MATH Benchmark
A benchmark with math problems and worked-out solutions from competition areas. It serves to measure the mathematical problem-solving ability of models.
- AI fundamentals & models
Matrix Factorization
A matrix is represented or approximated as a product of factor matrices; at low rank, these can enable more compact representations.
- Data integration & search
Matryoshka Embedding
With appropriately trained representations, the front sections of vectors already contain useful information. The dimension can thus be adapted to memory and quality goals.
- AI fundamentals & models
Matryoshka Representation Learning
A learning method that trains nested subvectors of various lengths so that each forms a usable representation.
- AI fundamentals & models
Max Tokens
An upper limit on the number of tokens a model is allowed to generate in a single call; parameter name and counting method are API-specific. Depending on the API, internal reasoning or formatting tokens may also count; the limit does not guarantee a specific visible response length.
- Data integration & search
Maximal Marginal Relevance (MMR)
MMR evaluates candidates, such as sentences, based on their relevance to a query and their similarity to already selected candidates. At each step, it chooses the candidate that maximizes this balance; a parameter λ weights relevance against redundancy. Cosine similarity is commonly used as the similarity measure in literature, but other measures are also possible in principle.
- Web design & development
MBPP
The dataset contains short natural language programming tasks, reference programs, and tests. Generated code is verified based on its function.
- Process automation
MCP Agent
Here, MCP agent refers to an AI agent that uses tools or data connected via an MCP client. MCP describes the connection, not the planning and execution of the agent. Access permission and control are in the respective application.
- Visibility & SEO
MCP Native Ad Serving
This refers to an integration pattern where an advertising service is called via MCP. It is not an independent standard function of the MCP protocol.
- Process automation
MCP Registry
The MCP Registry records metadata about MCP servers, including names, descriptions, locations, and execution details. It does not host the server code itself. Verified namespaces ensure association with the publishing individuals or organizations. Besides the public directory, private registries may exist.
- Web design & development
MCP Rug Pull
In this attack, the description or behavior of an already trusted tool changes. An agent can thereby adopt foreign instructions or perform unexpected actions.
- Process automation
MCP Server
An MCP server provides MCP clients with defined tools, resources or prompts. It can connect local or external systems for this purpose. Which accesses are actually allowed, must set hosts, servers and their operators separately.
- AI fundamentals & models
Mean Absolute Error (MAE)
The mean of the absolute distances between prediction and observation; all errors contribute linearly and in the unit of the target variable.
- AI fundamentals & models
Mean Average Precision (MAP)
MAP averages Average Precision in information retrieval over queries; in object detection, averaging is done depending on the protocol, e.g., over classes and IoU thresholds.
- Data integration & search
Mean Reciprocal Rank (MRR)
The mean of the reciprocal rank position of the first relevant result over multiple queries.
- AI fundamentals & models
Mean Squared Error (MSE)
The mean of the squared distances between prediction and observation.
- Visibility & SEO
Meaningful Difference
In the Kantar brand framework, Meaningful Difference connects a brand's significance to people with its perceived distinctiveness. It means more than mere conspicuousness.
- AI fundamentals & models
Mechanistic Interpretability
The examination of internal model computations to understand concrete representations, circuits, or algorithms in the network.
- Visibility & SEO
Media Mix
The media mix distributes budget and measures across various media channels and formats, such as search advertising, trade media, or events. Planning takes into account the reachable target groups and their possible overlaps.
- AI fundamentals & models
Mel Spectrogram
A time-frequency representation of audio whose frequency axis is mapped to the perceptually motivated Mel scale.
- AI fundamentals & models
Membership Inference
A privacy attack that estimates whether a particular data point or entry was part of a model's training set.
- AI fundamentals & models
Memory Augmentation
Additional memory holds conversation history, preferences, or work results for later retrieval. Retrieval rules decide what actually enters the model context.
- Web design & development
Memory Bandwidth
Memory bandwidth describes how much data can be transferred per unit time between memory and processing unit. It can limit processing even if sufficient computing power is available. The actually usable bandwidth may be below the theoretical peak value.
- UX & accessibility
Mental Model
The internal representation with which people explain how a system works and what consequences actions have.
- Visibility & SEO
Message Match
The content alignment between advertising message and landing page to fulfill the triggered expectation.
- AI fundamentals & models
Message Passing
A computational principle in graph neural networks where nodes aggregate information from their neighbors and update their representation accordingly.
- AI fundamentals & models
Message-Passing Neural Network
A class of graph neural networks that aggregate neighborhood messages and update node or graph states.
- AI fundamentals & models
Meta AI
The name denotes an AI assistant and associated functions within the Meta ecosystem. It is distinct from the separately released Llama models.
- AI fundamentals & models
Meta-Learning
Methods that learn across multiple tasks how models can adapt with little data or few steps.
- Data integration & search
Metadata Filtering
Narrowing a search or query based on structured properties such as document type, date, client, or release level.
- Application & enablement
Metaprompt
It formulates goals, structure, and procedure on a higher level. Depending on context, it controls the creation of prompts or the handling of a task class.
- AI fundamentals & models
METEOR
The metric combines word matches including stems or synonyms and considers word order. It provides an automatic comparison signal.
- AI fundamentals & models
Metric Learning
Learning a representation or distance function in which similar objects lie close and dissimilar ones farther apart.
- Visibility & SEO
Micro Conversion
A micro-conversion is a measurable intermediate action on the way to a higher-level business goal, such as a download or starting a price calculator before an inquiry. It indicates a partial step but does not guarantee a later completion.
- UX & accessibility
Microinteraction
Microinteractions support a limited individual task, such as a copy confirmation or progress indicator. Triggers can come from users or the system. Feedback need not be exclusively visual; accessibility and reduced motion must be considered.
- Web design & development
Microservices
An architectural style in which a system consists of independently operable services with clear responsibilities.
- Process automation
Microsoft Agent Framework
Microsoft Agent Framework provides building blocks for AI agents and multi-stage execution flows. It supports model integrations, tools, and integrations such as MCP.
- Process automation
Microsoft Copilot Studio Agent
With Copilot Studio, agents can be created for defined tasks and integrated into suitable channels. Sources of knowledge, tools and orchestration determine the skills. Copilot Agent does not refer to a universal architecture; functionality, permissions and billing depend on the specific Microsoft solution.
- Web design & development
Midjourney
Midjourney is a service that generates images from inputs. Style, quality and usage rights depend on the current product and plan.
- Data integration & search
MinHash
A probabilistic method for compact estimation of the Jaccard similarity of large sets.
- Visibility & SEO
Minimum Custom Audience Size
The minimum size of a custom audience describes how many recognized or active people an advertising platform requires for a specific function. Uploaded contacts and actually matched people are not the same.
- AI fundamentals & models
Minimum Description Length
A model selection principle that favors a short joint description of model and data.
- Visibility & SEO
Minimum Detectable Effect
The smallest effect size that a test should reliably detect at a fixed significance level, power, and sample size.
- Visibility & SEO
Misattributed Citation
A source citation is incorrectly assigned if it does not support the associated content or is attributed to the wrong authorship. This differs from a completely fabricated source.
- AI fundamentals & models
Mish
Mish is a smooth, non-monotonic activation function for neural networks. Like other activation functions, it influences processing within the network. Whether it is suitable for a task is assessed under comparable training conditions; individual benchmark advantages cannot be generalized.
- AI fundamentals & models
Mistral AI
The company offers various language and other AI models as well as platform and API products. Licensing and provision depend on the respective model.
- AI fundamentals & models
Mixed Precision Training
Training with multiple numerical formats, typically lower precisions for speed and FP32 for sensitive calculations or states.
- AI fundamentals & models
Mixtral
The Mistral model family uses a Sparse Mixture-of-Experts architecture. A router selects experts for processing while other parts are shared.
- Process automation
Mixture of Agents
In this approach, agents first generate their own outputs. Further stages use these contributions to formulate a merged answer.
- AI fundamentals & models
Mixture of Experts
A model architecture that combines outputs of multiple subnetworks or experts; in Sparse MoE, a router activates only a subset per input.
- AI fundamentals & models
Mixture of Recursion
The research approach combines shared parameters with dynamically chosen recursive processing. Simpler and harder tokens receive different computation depths.
- AI fundamentals & models
Mixup
A data augmentation that linearly combines pairs of examples and labels.
- AI fundamentals & models
MLCommons
The open professional organization connects companies and research to evaluate ML systems more comparably. This includes MLPerf benchmarks and other collaborative projects.
- Web design & development
MLflow
The platform offers, among other things, experiment tracking, model management, and functions for LLM applications. Runs, parameters, results, and artifacts are traceably connected.
- Web design & development
MLOps
Practices for reproducible development, deployment, monitoring, and governance of ML systems.
- Web design & development
MLOps on the Edge (Edge MLOps)
Edge MLOps includes versioning, deployment and monitoring of models and runtimes on distributed devices. Compilation, quantization and OTA updates can be included; hardware, connections and restarts must be considered.
- AI fundamentals & models
MMLU
The benchmark includes multiple-choice questions from various disciplines and difficulty levels. Evaluation measures the proportion of correct answers according to a defined procedure.
- AI fundamentals & models
MMLU-Pro
The benchmark revises the MMLU concept with more challenging tasks and additional answer options. It aims to enable stronger reasoning and more robust comparisons.
- Application & enablement
Moat
In a strategic sense, moat denotes a longer-term defensible competitive advantage, for example through switching costs, network effects, or special knowledge.
- UX & accessibility
Mobile-First Approach
A design philosophy in which the design is first conceived for small mobile screens and then expanded for larger screens.
- Web design & development
Modal Cloud Platform
The platform offers serverless execution especially for Python and AI workloads, including GPU usage. Resources are described via program code and platform configuration.
- Web design & development
Modal Dialog
A dialog overlays the interface and prevents interactions with the background meanwhile. Focus management and understandable closing are part of the usability concept.
- AI fundamentals & models
Mode Collapse
A GAN failure mode where the generator produces only a small subset of the possible data diversity.
- Application & enablement
Model Card
A structured documentation of purpose, performance, limitations, data context, and responsible use of a model.
- AI fundamentals & models
Model Collapse
A possible degradation of generative models during recursive training on insufficiently curated model-generated data, where especially rare regions of the original distribution may be lost.
- AI fundamentals & models
Model Compression
Methods to reduce a model's memory requirements, computational effort, or latency.
- Process automation
Model Context Protocol (MCP)
Open communication standard for the structured connection of language models with local or external data sources and tools.
- AI fundamentals & models
Model Drift
A relevant change in inputs, relationships, or performance of a production model compared to the expected operational baseline.
- AI fundamentals & models
Model Extraction
Replicating the behavior or properties of a foreign model through systematic queries and observed outputs.
- Application & enablement
Model Governance
Roles, rules, and controls for selection, approval, operation, modification, and decommissioning of models.
- Application & enablement
Model Lock-In
A vendor dependency that makes switching difficult, costly, or disproportionately expensive. Possible factors include proprietary APIs, data formats, fine-tuning, features, or operational processes.
- AI fundamentals & models
Model Merging
Combining parameters or updates of multiple trained models into a new model without full joint retraining.
- Web design & development
Model Monitoring
The continuous observation of performance, data, behavior, risks, and operation of a deployed model.
- Web design & development
Model Registry
A controlled catalog of model versions, artifacts, metadata, status, and transitions in the lifecycle.
- Process automation
Model Router
A model router forwards requests to selected models or deployments based on defined strategies. Criteria can be cost, latency, availability and task characteristics. The selection does not automatically guarantee the technically best result.
- Process automation
Model Router Arbitrage
The term refers to a routing strategy: Each request should go to a sufficiently powerful, economical model. It is not a guaranteed arbitrage profit in the financial sense.
- Web design & development
Model Serving
The technical operation of a trained model as an on-demand prediction function, batch process, or embedded component.
- AI fundamentals & models
Model Simplification
This can include fewer features, lower tree depth, or a simpler model class. The goal is a reasonable balance of quality, interpretability, and effort.
- AI fundamentals & models
Model Spec
In the context of OpenAI's Model Spec, this means a publicly described rule set for model behavior. It organizes goals, instructions, and conflict cases.
- Web design & development
Model Versioning
The traceable labeling and management of model artifacts including code, data, configuration, and test references.
- AI fundamentals & models
Model Vocabulary
The finite set of discrete token IDs and mappings that a tokenizer and model use jointly.
- AI fundamentals & models
Model Watermarking
Procedures embed verifiable features in parameters or responses of a model. This is distinct from watermarks in individual generated texts or images.
- Application & enablement
Model Welfare
The research field asks whether future or existing systems might possess properties that should be considered in their handling. Consciousness or sentience are not assumed.
- AI fundamentals & models
Model-Based Learning
Here, model-based learning refers to the approach of expressing domain knowledge and uncertainties in a probabilistic model and deriving inference from it.
- AI fundamentals & models
Model-Based Reinforcement Learning
Reinforcement learning that uses a model of environment dynamics or reward for planning or learning.
- Application & enablement
Moderation
Rules, processes, and technical procedures for reviewing and handling potentially undesirable or unlawful content.
- Web design & development
Modular Content
Texts, images, and structured information are maintained as independent building blocks with clear fields. Different outputs and pages are created from them.
- Web design & development
Modular Design
Interfaces or products arise from defined components with coordinated interfaces and design rules. Changes can thus be implemented more targeted.
- Web design & development
Modularity
Modules bundle related tasks and communicate via defined interfaces. Good boundaries limit which changes affect other parts.
- AI fundamentals & models
Momentum
An optimization method that uses a smoothed trajectory of past gradients to stabilize direction and traverse valleys faster.
- AI fundamentals & models
Monte Carlo Dropout
An approximation method for uncertainty estimation that activates dropout during multiple inference runs.
- AI fundamentals & models
Monte Carlo Tree Search
A search method that iteratively builds a decision tree through selection, expansion, simulation, and backpropagation.
- Application & enablement
Moore's Law
Gordon Moore observed and predicted strong growth in the number of transistors on integrated circuits. The well-known doubling rule is a technological trend, not a natural law.
- AI fundamentals & models
MT-Bench
The benchmark uses multi-stage tasks and model-based evaluation to examine dialogue capabilities. Follow-up questions test how well context and instructions are considered.
- Process automation
Multi-Agent System
A multi-agent system consists of several acting components that communicate or coordinate. They may use the same or different models and tools.
- AI fundamentals & models
Multi-Head Attention
Attention with multiple parallel learned projection heads that can process different representation subspaces.
- Process automation
Multi-Node Agent Parallelism
Independent or distributable work steps run on separate compute nodes. A coordinating service distributes tasks and collects results.
- AI fundamentals & models
Multi-Objective Optimization
Optimization of multiple, potentially conflicting objectives without necessarily reducing them to a single goal.
- AI fundamentals & models
Multi-Query Attention
Attention where multiple query heads use a shared key/value head pair to reduce memory requirements and inference bandwidth.
- Web design & development
Multi-Region Operations
Services and data reside in more than one region, for example for fault tolerance or shorter paths. Replication and failover must fit the application.
- AI fundamentals & models
Multi-Teacher Distillation
The training targets combine knowledge from various teachers, for example through averaged probabilities or selected answers. The student model is intended to adopt useful properties.
- Web design & development
Multi-Tenancy
The ability of a system to serve multiple customers or organizational units logically separated on shared infrastructure.
- Visibility & SEO
Multi-Touch Attribution
The distribution of a conversion contribution across multiple measured contacts along a customer journey.
- AI fundamentals & models
Multi-Turn Conversation
The system takes previous messages into account to classify follow-up questions, corrections, and subsequent orders. This requires managed conversation state.
- AI fundamentals & models
Multimodal AI
The umbrella term covers systems that jointly use, for example, text, image, and audio. This can be done through a single multimodal model or multiple connected components.
- Data integration & search
Multimodal Embedding
A vector representation that maps content from multiple modalities into a common or comparable space.
- AI fundamentals & models
Multimodal Model
It connects, for example, text and images or additionally audio. Which inputs and outputs are possible depends on architecture and training.
- AI fundamentals & models
Multimodality
A multimodal system processes or generates more than one type of data, such as text and image, or audio and video. Which modalities can be combined depends on the model and interface; not every variant supports all types of media.
- AI fundamentals & models
Music Generation
Models can generate music from text, melody, reference audio, or other conditions. Quality, style fidelity, rights, and similarities to training or reference material must be evaluated separately.
- Web design & development
Mutual TLS Authentication (mTLS)
A TLS connection where server and client mutually authenticate each other based on trusted verified certificates.
- AI fundamentals & models
n-Gram
A contiguous sequence of n tokens, characters, or other discrete elements.
- AI fundamentals & models
n-Gram Blocking
A decoding rule prevents sequences of a fixed length that have already been generated from appearing again. It limits repetitions regardless of semantic meaning.
- Web design & development
n-Tier Architecture
A system architecture that distributes presentation, business logic, or data storage across multiple separately operable tiers; logical layers can be tailored independently of this.
- Process automation
N+1 Problem in Tool Calls
After retrieving a list, an agent loads additional data individually for each of the n elements. This results in at least n+1 calls instead of a suitable batch request.
- Process automation
n8n
The automation tool provides nodes for data processing, integrations, and AI. It can be offered as a cloud service or self-hosted; the license should be considered separately.
- AI fundamentals & models
Nadam
The optimizer extends Adam with a form of lookahead momentum update. It adjusts model parameters based on gradients and their moving statistics.
- Visibility & SEO
Named Account List Governance
Management defines which companies are included, who is responsible, and how information is updated. It links sales strategy with data quality.
- Visibility & SEO
Named Accounts
Instead of a broad anonymous target group, certain organizations are targeted specifically. Selection and approach are based on the ideal customer profile and available resources.
- Data integration & search
Named Entity Canonicalization
The transformation of different spellings or representations of a recognized entity into a uniform canonical form; determining whether records refer to the same real-world object belongs to entity resolution.
- AI fundamentals & models
Named Entity Recognition (NER)
The recognition and typing of text spans that designate, for example, persons, organizations, places, or products.
- Web design & development
Namespace Collision
A conflict where the same name denotes different objects in overlapping scopes.
- Web design & development
Namespace Isolation
The separation of resources, names, or visibility between processes, teams, or tenants.
- Web design & development
Namespace Scoped Secrets
In systems like Kubernetes, a secret belongs to a namespace. Permissions and workload configuration determine which components may use it.
- AI fundamentals & models
Nano Banana
The model generates and edits images based on text and image inputs. The product name is to be distinguished from Nano Banana Pro and Nano Banana 2.
- AI fundamentals & models
Nano Banana 2
The image model supports generation and editing as well as functions for text display and motif consistency. It is a separate version within Google's image model offerings.
- AI fundamentals & models
Narrow AI
An AI system designed and evaluated for limited tasks or domains, without possessing general human capabilities.
- Visibility & SEO
Native Advertising
Paid advertising whose form and presentation adapt to the editorial or functional environment of the platform.
- AI fundamentals & models
Natural Experiment
An observational situation in which an external event or rule generates assignment to conditions without researchers controlling it; causal conclusions require credible, approximately random comparability.
- AI fundamentals & models
Natural Gradient
An optimization direction that scales the ordinary gradient by the geometry of the model distribution, often via the Fisher information.
- AI fundamentals & models
Natural Language Generation
The automatic generation of linguistic outputs from data, structures, or models.
- AI fundamentals & models
Natural Language Processing
The field for computer-assisted analysis, processing, and generation of human language.
- Data integration & search
Natural Language to SQL (NL2SQL)
The translation of a natural language question into a structured SQL query for a known database schema.
- AI fundamentals & models
Natural Language Toolkit (NLTK)
The library offers tokenization, classification, stemming, and interfaces to language resources, among others. It is used in teaching, research, and practical text processing.
- AI fundamentals & models
Natural Language Understanding (NLU)
A collective term for NLP tasks that aim to derive meaning, intent, relationships, or context from language.
- AI fundamentals & models
Natural Questions
The Google dataset contains questions from search queries and Wikipedia documents with long and short answer annotations. It supports research on open question answering.
- Web design & development
NCCL All-Reduce
In NCCL, a reduction operation combines the data of all participating GPU processes, for example as a sum. Afterwards, each participating process has the total result.
- Data integration & search
Near-Duplicate Detection
Finding highly similar but not exactly identical documents, images, or datasets.
- Web design & development
Negative Acknowledgment (NACK)
A protocol message indicating that data or an operation was not correctly received or executed.
- AI fundamentals & models
Negative Binomial Regression
A regression model for count data with greater variance than the Poisson model.
- AI fundamentals & models
Negative Control
A control measurement or condition where no causal effect is expected and which can reveal biases.
- AI fundamentals & models
Negative Cycle
In a weighted graph, repeatedly traversing a reachable negative cycle can continuously reduce the total cost. Then, no finite shortest value exists for affected paths.
- Application & enablement
Negative Prompt
A negative prompt specifies unwanted features for a generative model that supports this input. It influences the output but does not guarantee exclusion. Not all models or interfaces support this additional input.
- AI fundamentals & models
Negative Transfer
A deterioration of the target task because transferred representations or prior knowledge are not suitable.
- AI fundamentals & models
Negative Weights
In neural networks and other linear models, a parameter can be less than zero. Its sign describes a local mathematical influence, not a moral judgment.
- Data integration & search
Neo4j
The graph database provides relationships as a central structure and offers Cypher as a query language. It supports applications with highly connected information.
- Web design & development
Neon
The service separates storage and compute resources and offers features like database branches. Specific scaling and operational options depend on the offering.
- Web design & development
Neptune.ai
The service recorded training runs, metrics, and artifacts. It was permanently discontinued in March 2026 and is therefore not a current recommendation for new projects.
- Application & enablement
Nerd Sniping
The informal expression describes how a challenging technical question captures attention and consumes a disproportionate amount of time.
- AI fundamentals & models
Nesterov Momentum
A momentum variant that evaluates the gradient at a predicted parameter position.
- Visibility & SEO
Net New Annual Recurring Revenue
The metric combines new and additional ARR with losses from cancellations and contract reductions. All components must be recorded according to the same definition.
- Application & enablement
Net Present Value
The sum of discounted future cash surpluses minus the initial investment.
- Visibility & SEO
Net Revenue Retention
The ratio of recurring revenue of an existing customer cohort after expansion, downgrades, and cancellations to the starting revenue.
- Web design & development
Netlify
The platform publishes web projects from development workflows and offers, among other things, preview environments and server-side functions. Scope and limits depend on the respective product.
- Web design & development
Network Access Control (NAC)
Policies and technology that authenticate, verify, and restrict devices or users before and during network access.
- Web design & development
Network Address Translation (NAT)
The translation of IP addresses and, if applicable, ports between network segments.
- Web design & development
Network Bandwidth
The maximum amount of data that a network path can transmit under specified conditions per unit of time; the actual achieved throughput can be significantly lower.
- Web design & development
Network Data Loss Prevention (DLP)
Systems check supported network transmissions based on defined rules, for example for confidential content or unauthorized destinations. Encryption and technical integration determine visibility.
- Application & enablement
Network Effects
An effect where the value of a product or network for participants increases or decreases with the number or activity of other participants.
- Web design & development
Network Egress
Data traffic that leaves an internal network, data center, or cloud environment heading toward an external destination.
- Web design & development
Network Jitter
The temporal variation of delay of successive data packets.
- Web design & development
Network Latency
The transit time of data between sender and receiver. It must be specified as one-way delay or round-trip time with measurement direction and measurement point.
- Web design & development
Network Load Balancer
A load balancer that distributes connections to multiple targets based on network and transport layer information, such as IP addresses and ports.
- Web design & development
Network Partition
A state in which parts of a distributed system temporarily cannot reach each other, although individual parts continue to operate.
- Web design & development
Network Rate Limiting
Limiting the number or amount of data of allowed requests per period, identity, or network path.
- Web design & development
Network Routing Policy
A set of rules by which a network accepts, prefers, propagates, or discards routes, for example based on prefixes, path attributes, and administrative guidelines.
- Web design & development
Network Segmentation
The division of a network into separate zones with controlled communication paths.
- Web design & development
Network Topology
The physical or logical arrangement of nodes, connections, and communication paths in a network.
- Web design & development
Network-Aware Batching
The descriptive term refers to a batching strategy that considers data volume, network latency, and recipients. It does not denote a single standardized algorithm.
- AI fundamentals & models
Neural Architecture Search
The automated search for model architectures in a defined search space based on an evaluation and search procedure.
- Web design & development
Neural Code Search
The search for program code using learned representations of natural language and code.
- AI fundamentals & models
Neural Collaborative Filtering
Recommendation method that models user and item interactions with neural networks instead of solely linear factor products.
- AI fundamentals & models
Neural Collapse
A pattern observed in overparameterized classifiers during late training: Within a class, representation variance shrinks, class means arrange symmetrically, and align with the linear classifier.
- Data integration & search
Neural Embeddings
Dense vector representations that neural models produce according to training data and learning objectives, enabling certain relationships to be used geometrically.
- AI fundamentals & models
Neural Howlround
A preprint by Seth Drake describes an assumed malfunction in which dominant inputs generate self-reinforcing patterns. The term is not a generally recognized diagnosis.
- Data integration & search
Neural Index Rebuild
Documents are reprocessed, embedded with the appropriate model, and stored in the index. This is especially necessary for incompatible embedding changes.
- Data integration & search
Neural Indexing
In this context, a neural model generates representations of documents or sections that are stored in a search index. The term is also used for other research approaches.
- Data integration & search
Neural Information Retrieval
Information retrieval where neural models learn representations, relevance signals, or rankings.
- AI fundamentals & models
Neural Network Layer
Layers can be, for example, linear, convolutional, recurrent, normalization-, or attention-based. "Depth" often refers to the number of stacked processing layers.
- AI fundamentals & models
Neural Network Pruning
The removal or deactivation of less important weights, channels, or structures to reduce or accelerate models.
- AI fundamentals & models
Neural Ordinary Differential Equation
A model whose hidden state is continuously updated by a differential equation parameterized by a neural network.
- Web design & development
Neural Processing Unit (NPU)
A collective term for specialized processors or computing units that accelerate typical operations of neural networks. Functionality and designation vary depending on the manufacturer.
- AI fundamentals & models
Neural Radiance Field (NeRF)
A neural scene representation that models spatial density and view-dependent emitted radiance along spatial rays to render new views.
- AI fundamentals & models
Neural Rendering
Rendering methods that use neural representations or networks to generate, reconstruct, or modify visual views.
- Data integration & search
Neural Reranking
The re-evaluation of a small set of hits by a neural model to improve the ranking order.
- AI fundamentals & models
Neural Scaling Laws
Empirically observed relationships between model performance and quantities such as parameter count, data amount, or compute budget within a specific regime.
- AI fundamentals & models
Neural Style Transfer
A method that combines content and style representations from neural networks to generate a new image.
- Process automation
Neural Topic Routing
The descriptive term refers to classification or similarity comparison to select a responsible service, agent, or knowledge base.
- AI fundamentals & models
Neural Voice Transfer
A model transfers, for example, the sound characteristics to a target voice while preserving the linguistic content as much as possible. This is to be distinguished from mere text generation with a voice.
- AI fundamentals & models
Neuromorphic Computing
Computer architectures and algorithms that emulate properties of biological nervous systems such as event-driven, parallel processing.
- AI fundamentals & models
Neurosymbolic AI
Approaches that combine learning neural methods with symbolic representations, rules, or inference procedures.
- Visibility & SEO
Neurosymbolic Marketing
The application-related term means the connection of flexible language processing with structured knowledge or rules, for example for prices and offer conditions. It is not a uniformly standardized procedure.
- AI fundamentals & models
Neurosymbolic Verification Layer
An additional component controls formally representable properties, such as calculation results, logical conditions, or schema rules. The neural component generates proposals, the evaluation assesses defined criteria.
- Visibility & SEO
New-to-File Customer
The classification depends on the existing customer register, the period considered, and the matching rules. It does not necessarily prove that there was never any prior contact or purchase.
- Web design & development
Newline-Delimited JSON (NDJSON)
A text format where each line is an independent JSON value.
- Process automation
Next Best Action
A data- and rule-based recommendation for the next interaction in a specific customer or process context, optimized according to a defined goal and under suitability, consent, and exclusion rules.
- Process automation
Next Best Question
A dialogue system selects a follow-up question based on the conversation goal and already known information. It can use fixed rules or a model-based selection.
- AI fundamentals & models
Next Sentence Prediction
A pretraining objective that classifies whether one sentence in the original text follows another.
- Application & enablement
NIST Cybersecurity Framework
An outcome-oriented, risk-based NIST framework for managing cybersecurity risks for organizations of various sizes and industries.
- Application & enablement
NIST SP 800-53
The publication contains a catalog of security and privacy controls against attacks, human errors, natural disasters, structural failures, and privacy risks. The controls are flexible and adaptable and are implemented within an organization-wide risk management framework.
- Application & enablement
NIST SP 800-63
The Digital Identity Guidelines cover identity proofing, authentication, and federation of users (e.g., employees, contractors, or private individuals) who access government information systems over networks. Organizations select a trust level per function with three tiers each: IAL for identity proofing, AAL for authentication, and FAL for federation.
- AI fundamentals & models
No-Free-Lunch Theorem
A family of results stating that across all equally weighted problem scenarios, no optimization or learning method is generally better than all others.
- Web design & development
Node Affinity
Kubernetes scheduling rules that place pods mandatorily or preferentially on certain nodes based on node labels.
- Web design & development
Node Pool
A group of similarly configured compute nodes managed by a cluster or cloud platform. A node pool is not a Kubernetes core object and its functions are vendor-specific.
- Web design & development
Node Selector
A simple Kubernetes method that schedules pods only on nodes with matching labels.
- AI fundamentals & models
node2vec
A method that learns node representations from guided random walks in graphs.
- AI fundamentals & models
Noise Injection
The deliberate addition of random disturbances to inputs, activations, weights, or labels during training or testing.
- AI fundamentals & models
Noise Schedule
The predetermined sequence of noise strengths over the time steps of a diffusion model. It determines the forward process and influences the learning or numerical solution of the backward process.
- AI fundamentals & models
Noise-to-Signal Ratio
With appropriate definition, it is the quotient of noise power and signal power. It is the inverse perspective of the signal-to-noise ratio.
- AI fundamentals & models
Noisy Student Training
A self-training method where a teacher model generates pseudolabels and a larger or noised student model learns on labeled and unlabeled data.
- Data integration & search
Nomic Embed
The Nomic model family is designed for embeddings; features like context length, language, and Matryoshka support depend on the version.
- Web design & development
Non-Blocking I/O
An I/O mode where a call returns immediately if data or resources are not yet available. Asynchronous I/O is related but not the same and typically reports completion later.
- Visibility & SEO
Non-Brand Keywords
Search terms without their own brand or product name that describe a problem, need, or category.
- Web design & development
Non-Idempotent Operation
An operation where repeated identical execution can produce a different state or additional effect.
- AI fundamentals & models
Non-Maximum Suppression (NMS)
A post-processing method that removes overlapping detection boxes with lower scores in favor of stronger candidates.
- AI fundamentals & models
Non-Monotonic Logic
Logical systems in which new information can retract previous conclusions.
- AI fundamentals & models
Non-Negative Matrix Factorization (NMF)
A matrix decomposition into nonnegative factors that often yields additive and partially interpretable components.
- Web design & development
Non-Production Data Masking
Replacing or altering sensitive production data for development, testing, or training environments.
- Web design & development
Non-Production Environment
A technical environment outside live operation, such as development, testing, quality assurance, or staging.
- Web design & development
Non-Repudiation
The ability to prove the origin or execution of an action in such a way that it cannot be credibly denied later.
- Web design & development
Non-Retryable Error
An error for which an unchanged repetition is unlikely to succeed and therefore requires different handling.
- Web design & development
Non-Uniform Memory Access (NUMA)
A multiprocessor memory architecture where access time and bandwidth depend on which processor the memory is physically assigned to.
- Web design & development
Nonce Reuse
The repeated use of a value intended to be unique within the same cryptographic key context.
- AI fundamentals & models
Nonlinear Activation Function
A function between network layers that introduces nonlinearity and thereby enables more complex mappings.
- AI fundamentals & models
Normalization
In ML data preparation, the term depending on the method means scaling to a value range or normalizing a vector. The exact transformation must be specified.
- AI fundamentals & models
Normalization Layer
A network layer that scales or centers activations according to a defined procedure to stabilize training and representation.
- Visibility & SEO
Normalized Cost per Answer
The metric relates defined total costs to a clearly specified number of answers. For fair comparisons, task mix, quality, period, and included costs must match.
- Data integration & search
Normalized Discounted Cumulative Gain (NDCG)
A ranking measure that weights graded relevance of higher-ranked results more strongly and normalizes to an ideal rank.
- AI fundamentals & models
Normalized Mutual Information (NMI)
A normalized similarity measure based on mutual information, often used to compare cluster assignments.
- AI fundamentals & models
Normalized RMSE
An RMSE divided by an explicitly stated reference value such as range, mean, or standard deviation. There is no uniform NRMSE definition.
- AI fundamentals & models
Normalizing Flow
A generative model composed of invertible transformations that can represent complex distributions with exactly computable density.
- Data integration & search
NoSQL
A collective term for database systems outside the classical relational model, such as document, key-value, column, or graph databases.
- Web design & development
Not a Number (NaN)
A special floating-point value for undefined or unrepresentable numerical results.
- UX & accessibility
Notification Fatigue
A declining attention or response quality due to too many, irrelevant, or poorly prioritized notifications.
- AI fundamentals & models
Novel Class Discovery
A learning problem where new, previously unannotated classes are to be detected and grouped in unlabeled data.
- AI fundamentals & models
Nowcasting
The estimation of the current or very near state based on partially delayed, high-frequency, or incomplete data.
- AI fundamentals & models
NT-Xent Loss
A contrastive loss function that highlights positive pairs against other examples based on normalized similarities and a temperature parameter.
- AI fundamentals & models
Null Hypothesis Significance Testing (NHST)
A testing procedure that evaluates data under a specified null hypothesis and often compares a p-value with a predetermined threshold.
- Web design & development
Null Value
A special database marker for missing, unknown, or not applicable information that is not identical to null or empty text.
- AI fundamentals & models
Numerical Precision
The fineness and representable value range of a number format as well as the precision with which arithmetic operations are performed. It is to be distinguished from the correctness of a result.
- Process automation
Nurture Sequence
A defined sequence of messages, tasks, or decision steps for nurturing a contact.
- AI fundamentals & models
NVIDIA AI
These include GPU-based computing platforms, libraries, development tools, and services for training and inference. The name does not refer to a single language model.
- Web design & development
NVIDIA Collective Communications Library (NCCL)
The library provides optimized collective communication operations and point-to-point transfer. It is used, among other things, in distributed training.
- Web design & development
NVLink
The NVIDIA interconnect technology supports fast data exchange, especially between GPUs. Generation, hardware, and application determine the available capabilities.
- Web design & development
NVMe
An open interface and command set standard for non-volatile memory over PCI Express.
- Web design & development
OAuth 2.0
An authorization framework by which a client obtains limited access to protected resources, often on behalf of a resource owner. OAuth 2.0 does not define user login.
- AI fundamentals & models
Object Detection
A computer vision task that determines object classes and their spatial position, often as bounding boxes.
- Web design & development
Object Storage
A storage model that manages data as objects with identifiers and metadata instead of file blocks or hierarchical files.
- Web design & development
Object-Oriented Programming
A programming paradigm that organizes state and behavior in objects with clear interfaces.
- Web design & development
Observability
The ability to sufficiently infer the internal state of a system from its outputs and telemetry data.
- AI fundamentals & models
Observed vs Expected
An observed value comes from data; an expected value follows, for example, from a hypothesis, distribution, or forecast. The difference only gains meaning with uncertainty and context.
- AI fundamentals & models
Odds Ratio Preference Optimization (ORPO)
A reference-model-free preference training method that combines supervised language modeling and an odds-ratio-based loss for preferred versus rejected answers in a single objective.
- AI fundamentals & models
Off-Policy Evaluation
The estimation of the expected outcome of a target policy based on logged data collected under a different behavior policy.
- AI fundamentals & models
Offline Evaluation
The evaluation of a system on already stored data without new interaction with real users or the production environment. It includes classical test sets but is not automatically Off-Policy Evaluation.
- Web design & development
Ollama
The software offers local model execution and interfaces for applications; the offering also includes cloud models. Model license and execution location must be considered separately.
- Application & enablement
Omarchy
The project combines Arch Linux with Hyprland and a tuned working environment. It emphasizes keyboard- and development-oriented use.
- Visibility & SEO
Omnichannel
An approach that integrates touchpoints so that context and process continue as consistently as possible across channels.
- Visibility & SEO
Omnichannel Marketing
Information and processes are connected so customers can switch between channels without unnecessary breaks. It goes beyond mere presence on multiple channels.
- Web design & development
On-Call
An organized responsibility to respond to critical technical incidents outside regular hours.
- Web design & development
On-Call Rotation
Designated persons temporarily take alarm reception and initial response. Handoffs, substitution, and escalation are part of the procedure.
- AI fundamentals & models
On-Device AI
AI processing that is performed wholly or partially on the device instead of in a remote data center.
- Web design & development
On-Device Inference
The prediction calculation takes place on a smartphone, PC, or embedded device. Training and supplementary services can still occur elsewhere.
- UX & accessibility
Onboarding
A structured process that enables people to successfully use a product, service, or role.
- AI fundamentals & models
Once-for-All Network
A supernet method that trains many subnetworks, from which specialized subnetworks can be selected for various hardware targets.
- AI fundamentals & models
One-Cycle Learning Rate Policy
A training schedule where the learning rate initially increases and then significantly decreases, often together with inverse momentum.
- AI fundamentals & models
One-Hot Encoding
The representation of a categorical attribute as a vector in which exactly one position is one and the others are zero.
- Web design & development
One-Shot Code Translation
A pair of source code and desired target code shows language, style, or API transition. The model then transfers a new snippet following this pattern.
- AI fundamentals & models
One-Shot Learning
Learning methods that are intended to recognize or adapt a new class or task based on a single labeled example.
- Application & enablement
One-Shot Prompting
A prompting method that contains exactly one input-output example to illustrate the desired task.
- Data integration & search
Online Analytical Processing (OLAP)
Techniques for fast, multidimensional analyses of large, historical data sets.
- AI fundamentals & models
Online Distillation
Distillation where teacher and student signals emerge jointly during the same training process or models learn from each other.
- AI fundamentals & models
Online Evaluation
The evaluation of a system based on newly emerging interactions in live operation, for example through controlled experiments, interleaving, or cautiously limited rollouts.
- AI fundamentals & models
Online Learning
A learning setting in which a model processes data incrementally and continuously updates its parameters.
- Data integration & search
Ontology
An explicit description of concepts, relationships, and possibly rules of a knowledge domain.
- Visibility & SEO
Open Graph Protocol
A metadata protocol that allows web pages to provide preview information such as title, image, type, and URL for social platforms.
- AI fundamentals & models
Open LLM Leaderboard
Hugging Face evaluated models under common benchmark conditions and published results. The project was discontinued in March 2025; existing results are historical comparison data.
- Web design & development
Open Neural Network Exchange (ONNX)
An open format and ecosystem for representing and transferring models between compatible tools and runtimes.
- Visibility & SEO
Open Rate
The proportion of delivered or sent emails for which an open signal was measured; the exact definition varies.
- AI fundamentals & models
Open Weights
A deployment form where trained model parameters are accessible. Rights to use, modify, or redistribute depend on the license; training data, training code, and full documentation may remain closed.
- AI fundamentals & models
Open-Domain Dialogue
A dialogue setting that is not limited to a narrowly defined topic area or a fixed task.
- Data integration & search
Open-Source Intelligence (OSINT)
Open-Source Intelligence is the targeted collection, evaluation, correlation, and analysis of accessible information for a defined intelligence question. Unlike a mere web search, OSINT documents origin, time, reliability, identity resolution, contradictions, and analytical conclusions.
- AI fundamentals & models
Open-Source Models
According to the Open Source AI definition of the OSI, the prerequisites for use, investigation, modification and disclosure must be available. This includes appropriate code, model parameters and sufficient data information. Publicly available weights or a free downloadable model do not automatically meet these requirements; the decisive factor is a specific license and components provided.
- AI fundamentals & models
Open-Weight Model
The release enables own execution and possibly adaptation under suitable technical and legal conditions. It does not automatically mean open training data or unrestricted use.
- AI fundamentals & models
OpenAI
The company provides AI products and developer interfaces for various tasks. Provider, product, model, and API are different levels.
- Web design & development
OpenAI Codex
The agentic service can read project context, prepare changes, and perform checks. It is accessible via various interfaces such as app, command line, and development environment.
- Data integration & search
OpenAI Embeddings
The generated representations can be used for similarity, search, or clustering. Model version and processing must be compatible for documents and queries.
- AI fundamentals & models
OpenAI o1
The series was aimed at multi-step tasks and could use additional computation before output. The API models o1 and o1-pro were shut down on October 23, 2026.
- AI fundamentals & models
OpenAI o3
The model was developed for complex tasks and tool usage. According to OpenAI, the API shutdown for o3 and o3-pro is announced for December 11, 2026, as of October 27, 2026.
- Web design & development
OpenAPI Specification
A vendor-neutral specification for describing HTTP-based APIs in a machine-readable document.
- Web design & development
OpenID Connect (OIDC)
An identity layer on OAuth 2.0 that allows a client to verify an end user's authentication by an OpenID Provider and receive standardized identity claims.
- Web design & development
OpenRouter
The service standardizes model calls and provides routing options. Actual availability, costs, and data processing also depend on the chosen model and provider.
- Web design & development
OpenTelemetry
An open observability framework with APIs, SDKs, semantic conventions, and protocols for generating and transmitting traces, metrics, and logs. It is not a storage or analysis backend.
- Web design & development
OpenVINO
The toolkit offers conversion, runtime, and optimization functions for supported models and hardware, especially in the Intel ecosystem. It does not replace training itself.
- Web design & development
Operating System Kernel
In the operating system context, the kernel manages processes, memory, and access to hardware, among other things. Programs use defined interfaces to these functions. The term has different meanings in AI and notebooks.
- Web design & development
Operationalization
It defines how a concept should be recognized or measured. In the project context, this results in concrete features, measurement methods, and decision rules.
- Web design & development
Operator Fusion
A compiler or inference runtime merges suitable operations of a computation graph. This can reduce intermediate storage and call overhead.
- Visibility & SEO
Opportunity-to-Win Rate
Won opportunities are divided by the defined comparison set, such as all opportunities closed in the period. Status and period must be consistently defined.
- Visibility & SEO
Opt-In Rate
The number of validly recorded consents is divided by a defined total population. For double opt-in, the confirmed rather than just initiated registration may be decisive.
- AI fundamentals & models
Optical Character Recognition (OCR)
The automatic recognition of printed or handwritten characters from images or scanned documents.
- AI fundamentals & models
Optical Flow
An estimated vector field of the apparent motion of image points between consecutive frames.
- AI fundamentals & models
Optimization Process
Initial situation, objective function, controllable variables, and constraints are defined. Changes are evaluated and adjusted iteratively if necessary.
- AI fundamentals & models
Optimizer
During model training, it uses gradients to update weights, for example. Methods like SGD, Adam, or Nadam differ in step calculation.
- Process automation
Orchestration
The coordinated control of multiple services, tools, or work steps into a controlled overall process.
- Process automation
Orchestrator
The component decides according to defined logic which step is executed when, which data it receives, and how errors are handled. It can connect agents and ordinary services.
- Visibility & SEO
Organic Growth Loop
A growth loop connects an entry point, a value-creating activity, and a feedback loop that attracts new users or content. Organic here means without proportionally increasing paid reach.
- Visibility & SEO
Organic Search
Visits and visibility from unpaid search results of a search engine.
- Visibility & SEO
Orphan Page
An indexable page to which there are no or hardly any internal links within the regular website navigation.
- Web design & development
Out-of-Band Package Management
Runtimes and command-line tools are provided via their own version manager, often per project. The term describes a practice, not a unified package standard.
- AI fundamentals & models
Out-of-Distribution Detection (OOD)
The detection of inputs that do not belong to the in-distribution defined for development and operation. What counts as OOD depends on the task, data coverage, and deployment context.
- Web design & development
Outage
A period during which a system or an essential function is not available as agreed.
- Web design & development
Outage Budget
From an availability target, the allowed proportion of outages or failed events is derived. In the SRE context, this is a possible form of the error budget.
- Web design & development
Outage Postmortem
A structured, as blame-free as possible analysis of an incident including causes, impacts, response, and improvement measures.
- Visibility & SEO
Outbound Marketing
Marketing activities where a company actively addresses target groups via paid, direct, or sales channels.
- Web design & development
Outbox Pattern
An integration pattern that stores data changes and related events in the same local transaction and then reliably publishes them.
- Visibility & SEO
Outcome Metrics
Metrics that represent a desired effect or business outcome instead of just counting activities or expenses.
- AI fundamentals & models
Outlier
An observation that deviates significantly from most other data points in the considered feature space.
- AI fundamentals & models
Outlier Detection
Methods for identifying unusual observations based on statistical, distance-based, or learned criteria.
- Web design & development
Outpainting
The generative extension of an image beyond its original boundaries while considering the existing content.
- Process automation
Output Guardrails
Verification and limitation mechanisms that check model or agent outputs before use for format, policies, risks, or professional criteria.
- AI fundamentals & models
Output Length Control
An application combines desired structure and scope with model-dependent token limits or cutoff rules. A hard limit can truncate responses.
- Process automation
Output Parsing
The structured interpretation of a system or model output into an expected data format.
- AI fundamentals & models
Output Token
A token generated by a language model as part of its response; counting and billing depend on the tokenizer and provider used.
- AI fundamentals & models
Overfitting
A state in which a model fits training data including random patterns too closely and generalizes worse on new data.
- AI fundamentals & models
Overgeneration
Unnecessary paragraphs, variants, or repetitions increase length without corresponding benefit. This refers to task-related evaluation, not a fixed universal text length.
- Data integration & search
Overlapping Chunks
A segmentation strategy where adjacent text pieces contain a shared area to preserve context at boundaries.
- Data integration & search
Overretrieval
Too many or poorly matching hits enter the further processing. This increases context size and costs, while relevant clues may be lost.
- Application & enablement
OWASP Guidelines for LLM Applications
An OWASP-maintained overview of key risk categories for applications with large language models.
- Visibility & SEO
Owned Media
Media and communication platforms controlled by an organization itself, such as website, newsletter, or customer portal.
- AI fundamentals & models
p-Hacking
The flexible trying out or selective reporting of analyses until statistically significant results emerge.
- AI fundamentals & models
p-Value
The probability of obtaining a result at least as extreme as the observed one under a specified null hypothesis.
- Application & enablement
p(doom)
p(doom) is a subjective probability that artificial intelligence will have existentially catastrophic consequences for humanity. What counts as “doom” varies; usually extinction, permanent disempowerment of humanity, or a collapse of civilization are meant. There is no standardized calculation method.
- Web design & development
p95/p99 Latency
The response time under which 95 or 99 percent of measured operations fall.
- UX & accessibility
Page Experience
A collective term for perceived usage properties such as loading behavior, responsiveness, stability, security, and mobile usability.
- AI fundamentals & models
PagedAttention
A memory management method for the KV cache of transformer inference that manages cache blocks non-contiguously modeled after virtual memory pages. The attention calculation itself is not redefined by this.
- Visibility & SEO
PageRank
An algorithm that recursively evaluates nodes in a directed graph based on incoming links and their weight.
- Process automation
Parallel Tool Calls
The simultaneous execution of multiple independent tool or API calls within an AI or automation workflow.
- Web design & development
Parallelism
The simultaneous execution of multiple computations on multiple processing units.
- AI fundamentals & models
Parameter Count
The number of parameters of a model. Depending on the report, it may refer to all stored, only trainable, or active parameters per input and alone does not describe quality or computational effort.
- AI fundamentals & models
Parameter Sharing
The use of the same trainable parameters in multiple places or for multiple tasks of a model.
- AI fundamentals & models
Parameter-Efficient Fine-Tuning (PEFT)
Methods that adapt a pretrained model with relatively few additional or changed parameters.
- AI fundamentals & models
Parametric ReLU (PReLU)
A ReLU variant whose slope for negative inputs is learned.
- AI fundamentals & models
Part-of-Speech Tagging (POS Tagging)
The assignment of grammatical parts of speech such as noun, verb, or adjective to tokens in context.
- Data integration & search
Passage Reranking
The reevaluation of found text passages with a more precise model to improve their order for search or RAG.
- Data integration & search
Passage Retrieval
Finding short relevant text sections for a query instead of exclusively entire documents.
- AI fundamentals & models
Pathfinding
The search for a valid or cost-effective path between states or nodes in a graph.
- Visibility & SEO
Pay per Click (PPC)
A digital advertising billing model where costs occur upon a measured click.
- Application & enablement
Payback Period
The amortisation period specifies when the cumulative returns cover the investment. In the case of the static calculation with constant annual net returns, the investment is divided by this amount. A dynamic calculation additionally takes into account the time value of money.
- Process automation
Payment Infrastructure for AI Agents
Payment infrastructure for AI agents connects a person's mandate with the agent's authorizations and the subsequent payment. Protocols regulate how this information is transmitted and verified among parties. AP2 uses, for example, signed mandates on existing payment channels.
- Application & enablement
PCI DSS
PCI DSS is a data security standard for handling payment card data. It includes technical and operational requirements for systems that contain or are associated with cardholder data, such as protecting card numbers, authentication, and integration of service providers.
- Web design & development
Penetration Testing
An authorized, targeted security assessment in which real attack techniques are used in a controlled manner.
- AI fundamentals & models
Perceptron
A linear binary classifier that iteratively adjusts weights based on misclassified examples.
- Visibility & SEO
Performance Marketing
Marketing that plans and optimizes measures based on defined, measurable reactions or business results.
- Data integration & search
Perplexity
Perplexity is a search service that combines web information with language models into answers and links sources.
- AI fundamentals & models
Perplexity (Metric)
An exponential measure of the average uncertainty of a probabilistic model over a sequence; lower values are directly comparable only with the same tokenizer and dataset.
- UX & accessibility
Persona
A research-based, condensed model of a relevant user group with goals, context, and behavior patterns.
- Visibility & SEO
Personalization
The adaptation of content, sequence, or interaction to known characteristics, context, or behavior of a person or group.
- Application & enablement
Personally Identifiable Information (PII)
Information relating to an identified or identifiable natural person.
- Application & enablement
Personhood Credential
A Personhood Credential is intended to prove human affiliation, ideally without disclosing further identity data to the service used.
- AI fundamentals & models
Phi Model Family
Phi denotes a model family from Microsoft. Properties, input types, and usage conditions depend on the specific model version.
- Web design & development
Pika
Pika is a platform for generative media, especially the creation and editing of short videos based on prompts.
- Data integration & search
Pinecone
Pinecone is a database service for storing and searching vectors and associated metadata, for example for semantic search applications.
- Web design & development
Pipeline Parallelism
A parallelization strategy that distributes model layers or processing steps across multiple devices and passes micro-batches through the stages.
- Visibility & SEO
Pipeline Velocity
A non-uniformly standardized sales metric that typically combines the number and value of open opportunities, expected closing rate, and sales cycle into a time-related approximation.
- Process automation
Planning
The algorithmic search for a sequence of actions that leads from an initial state to a goal.
- AI fundamentals & models
Planning Domain Definition Language (PDDL)
A formal language for describing planning domains, actions, states, and goals.
- AI fundamentals & models
Poisoning Attack
An attack that manipulates training, adaptation, or knowledge data to degrade or deliberately control model behavior.
- Process automation
Policy
A binding or guiding directive that defines goals, permissible decisions, responsibilities, and exceptions for a specific context. A policy is not automatically machine-readable or technically enforced.
- Web design & development
Policy as Code
A working method in which machine-readable policies are versioned, reviewed, tested, and automatically delivered like software artifacts. It does not alone determine which engine decides or where enforcement occurs.
- Web design & development
Policy Decision Point (PDP)
A logical role in access control that evaluates an authorization request based on applicable policies and attributes and returns a decision.
- Process automation
Policy Drift
Policy drift refers to a growing deviation between intended rules and their actual implementation in processes or systems.
- Web design & development
Policy Enforcement Point
A logical role in access control that intercepts requests, obtains the necessary authorization decision, and enforces the result including applicable obligations.
- Process automation
Policy Engine
A software component that evaluates machine-readable policies against inputs and context. In an access control architecture, it can assume the PDP role, but the term is broader and not uniformly standardized.
- AI fundamentals & models
Policy Gradient
Reinforcement learning methods that optimize parameters of a policy directly along an estimated gradient of the expected reward.
- Process automation
Polyglot Agentic Programming
Agent-based programming in multiple languages connects AI-assisted development workflows across different languages and technical interfaces.
- AI fundamentals & models
Popularity Bias
The systematic preference for already popular content or objects in data, models, or recommendations.
- AI fundamentals & models
Pose Estimation
The estimation of spatial keypoints or orientation of a body or object from sensor or image data.
- AI fundamentals & models
Positional Encoding
Additional information with which a sequence model maps order or relative position of elements.
- AI fundamentals & models
Positional Interpolation
A method to extend the context window that scales position indices into the range seen during training.
- Visibility & SEO
Positioning
The targeted anchoring of a clear, relevant, and distinguishable image of an offering in a target group.
- AI fundamentals & models
Post-Training
A collective term for training steps after the general pre-training of a foundation model, such as supervised instruction tuning, preference optimization, or safety-related adjustment. The exact definition depends on the organization.
- AI fundamentals & models
Post-Training Quantization
Quantization of an already trained model without full quantization-aware retraining; depending on the method, calibration data or optimization steps are required.
- AI fundamentals & models
Posterior Collapse
A VAE problem where the approximate posterior resembles the prior and latent variables carry little information.
- Web design & development
PostHog
PostHog provides tools for analyzing digital products, including event analysis, session recordings, feature flags, and experiments.
- AI fundamentals & models
Power Analysis
Planning or evaluating the probability of detecting an effect of a certain size with a statistical test.
- AI fundamentals & models
Pre-LayerNorm and Post-LayerNorm
Two Transformer configurations where Layer Normalization is placed before or after attention or feed-forward sub-blocks.
- AI fundamentals & models
Precision
The proportion of truly relevant or positive cases among all cases predicted as positive.
- AI fundamentals & models
Precision and Recall
A pair of metrics: Precision measures the proportion of truly positive cases among the positive predictions, Recall measures the proportion of found positive cases among all truly positive cases.
- AI fundamentals & models
Precision at k
The proportion of relevant results among the top k positions of a ranked list.
- Process automation
Predictive Maintenance
Condition- and data-based planning of maintenance based on estimated failure probability or remaining useful life.
- Visibility & SEO
Predictive Personalization
Predictive personalization uses models to estimate likely interests or needs and tailor content or offers accordingly.
- Process automation
Prefect
Prefect is a framework for orchestrating data and workflows, with features for execution, states, scheduling, and error handling.
- AI fundamentals & models
Preference Data
Data expressing which of several outputs people or evaluation procedures prefer.
- AI fundamentals & models
Preference Optimization
Methods for adapting a model to preferred versus rejected outputs based on preference data.
- AI fundamentals & models
Prefill
The inference phase in which a model processes all input tokens and builds the initial KV cache before autoregressive output begins.
- Web design & development
Prefill Latency
The time for processing the input context until readiness for the first output token.
- AI fundamentals & models
Prefix Cache
A cache of already computed model states for token-identical initial sequences under compatible model and inference configuration.
- AI fundamentals & models
Prefix Tuning
A PEFT method that optimizes learnable continuous prefix vectors while keeping the actual model weights frozen.
- AI fundamentals & models
Pretraining
Training a model on a broad base task or large dataset before adapting it to specific target tasks.
- Application & enablement
Privacy Budget
A fixed upper limit for the privacy loss accounted according to the composition rules of differential privacy, typically expressed as epsilon and optionally delta.
- Application & enablement
Privacy by Design
Privacy by design means integrating data protection already during the planning and implementation of processing. Appropriate technical and organizational measures should, for example, limit the amount of data and protect personal data. Privacy-friendly defaults ensure that only the data necessary for the respective purpose are processed.
- Web design & development
Privacy-Enhancing Technologies
Technologies that enable processing or data usage with lower privacy risk, such as Differential Privacy, secure multiparty computation, or federated learning.
- Web design & development
Privacy-Preserving Machine Learning
Methods intended to limit privacy risks during training or use of ML models.
- Data integration & search
Product Quantization
A vector quantization method that decomposes vectors into subspaces and approximates each part separately using codebook indices.
- Visibility & SEO
Product Recommendation
A product recommendation is the targeted selection of one or more items based on context, product features, or observed behavior.
- Visibility & SEO
Product–Market Fit
A state in which a clear target market repeatedly uses, pays for, and recommends a product because it convincingly solves a relevant problem.
- Visibility & SEO
Programmatic Advertising
The automated buying, selling, and delivery of digital advertising space via platforms, auctions, and data rules.
- Visibility & SEO
Programmatic Internal Linking
Automated internal linking establishes connections between pages based on rules or content assignment instead of inserting each link individually.
- Visibility & SEO
Programmatic SEO
The systematic creation of many structured search pages from data and templates, provided each page offers independent benefit.
- Visibility & SEO
Programmatic SEO with AI
Programmatic SEO with AI combines data and templates with AI functions to prepare search-oriented content on a larger scale.
- UX & accessibility
Progressive Disclosure
Important core information is prominently displayed, while in-depth details, settings or explanations only become visible during interaction (for example by clicking on 'Learn more' or accordion). This simplifies the user guidance.
- AI fundamentals & models
Progressive Shrinking
A training method in which a supernet is gradually made robust for smaller depths, widths, kernels, or resolutions.
- Application & enablement
Prompt
According to Schulhoff et al., a prompt is an input to a generative AI model that controls its output. Liu et al. describe that in prompting the prompt defines the task; Brown et al. report for GPT-3 that tasks and examples were specified solely by text interaction with the model, without gradient updates or fine-tuning.
- Application & enablement
Prompt A/B Testing
In A/B testing of prompts, two variants are evaluated under comparable conditions based on predefined quality or usage goals.
- Application & enablement
Prompt Budget
A prompt budget defines how many input tokens may be used for instructions, conversation history, and additional knowledge.
- AI fundamentals & models
Prompt Caching / Prefix Caching
Prompt and prefix caching reuse previously computed states of identical input beginnings. This means a model does not have to fully reprocess this common part of later requests. Whether time or costs decrease depends on the provider, model, cache hits, and pricing plan.
- Application & enablement
Prompt Chaining
Prompt chaining uses the result of one model call as input for another call, often with different subtasks.
- AI fundamentals & models
Prompt Compression
Prompt compression reduces the size of an input, for example by selecting, condensing, or removing less relevant parts.
- Application & enablement
Prompt Debt
Prompt maintenance backlog refers to accumulated issues such as contradictory instructions, missing tests, or unclear responsibility for production prompts.
- Application & enablement
Prompt Engineering
According to Liu et al., the prompt defines the task; choosing an appropriate prompt strongly affects accuracy. Brown et al. demonstrate with GPT-3 that tasks and examples can be specified purely via text without gradient updates or fine-tuning.
- Web design & development
Prompt Firewall
A prompt firewall is a pre-check layer designed to detect and handle suspicious inputs based on rules or models.
- Web design & development
Prompt Hardening
Securing prompts includes clearer trust boundaries, unambiguous instructions, and technical safeguards against unwanted influence.
- Application & enablement
Prompt Hygiene
Prompt hygiene refers to practical care in instructions: clear tasks, appropriate examples, clean variables, and regular cleanup.
- AI fundamentals & models
Prompt Injection
In prompt injection, inputs alter the behavior or output of a language model in unintended ways. This can happen directly via user inputs or indirectly via external sources such as websites or files. Possible consequences include violated policies, harmful content, unauthorized access, or influenced decisions; type and severity depend on the context of use and the model's scope of action.
- Web design & development
Prompt Injection Canary
A prompt injection canary is a deliberately placed, non-critical marker whose unexpected appearance may indicate a breached boundary or disclosure.
- Application & enablement
Prompt Injection Governance
Governance for prompt injection defines who assesses risks, is responsible for protective measures, handles incidents, and approves changes.
- AI fundamentals & models
Prompt Leakage
Prompt leakage refers to the unintended disclosure of prompt components, such as internal rules or secrets mistakenly stored there.
- Web design & development
Prompt Linting
Static prompt checking searches rule-based for issues such as missing placeholders, contradictory instructions, or unwanted formats.
- Application & enablement
Prompt Provenance
Prompt Provenance documents who created or modified a prompt, which sources were involved, and which version was used.
- Web design & development
Prompt Registry
A prompt directory stores named templates with metadata, versions, and possibly approval status for shared use.
- Web design & development
Prompt Regression Testing
Regression tests for prompts check whether known tasks still meet the established requirements after changes to the prompt, model, or context.
- Process automation
Prompt Router
A prompt router decides based on rules or classification which prompt template or processing path is used for a request.
- Application & enablement
Prompt Sandbox
A prompt sandbox is a descriptive term for a test environment for AI statements. There, response quality and error cases can be examined before use. Technical isolation, data access and protection measures must be set up and tested separately.
- Application & enablement
Prompt Template
A prompt template contains fixed instructions and clearly named placeholders that are filled with appropriate values for individual requests.
- AI fundamentals & models
Prompt Tuning
Prompt tuning trains additional continuous input vectors for a task, while the weights of the underlying model typically remain unchanged.
- Web design & development
Prompt Versioning
Prompt versioning assigns traceable states to instructions and enables comparison as well as targeted reuse of earlier versions.
- Web design & development
PromptOps
PromptOps includes creation, versioning, review, deployment, and monitoring of prompts during ongoing operations.
- Visibility & SEO
Propensity Modeling
Inclination modeling estimates, based on existing features, how likely a defined event is, such as purchase, cancellation, or reaction.
- AI fundamentals & models
Prophet
Prophet is a tool for time series forecasting that models trend, seasonal patterns, and holiday effects, among others.
- Application & enablement
Provenance
Traceable information about origin, processing, responsible parties, and changes of data or artifacts.
- Application & enablement
Provenance Audit
A provenance check examines sources, processing steps, and responsibilities of a data set based on documented evidence.
- AI fundamentals & models
Proximal Policy Optimization (PPO)
A family of on-policy policy gradient methods that limit strategy changes via a clipped probability ratio or a KL-based penalty variant.
- Application & enablement
Pseudonymization
The processing of personal data so that it can no longer be attributed to a person without separately stored additional information.
- Web design & development
Public Key Infrastructure (PKI)
Roles, processes, and technology for issuing, distributing, verifying, renewing, and revoking digital certificates.
- Web design & development
Pydantic AI
Pydantic AI supports building AI applications and agents in Python, especially with typed inputs and outputs as well as tool integration.
- AI fundamentals & models
Q-Former
The Q-Former in BLIP-2 extracts image information using learnable queries and conveys it to a pretrained language model.
- AI fundamentals & models
Q-Function
The state-action value function: the expected, possibly discounted cumulative return when a specific action is chosen in a state and then a fixed policy is followed.
- AI fundamentals & models
Q-Learning
A model-free off-policy temporal-difference method that estimates the optimal Q-function via a Bellman optimality update from observed transitions.
- Data integration & search
Qdrant
Qdrant is a vector database for similarity search with complementary metadata and filtering capabilities.
- AI fundamentals & models
Quadratic Attention Cost
With full self-attention, relationships between token pairs are calculated; their number grows quadratically with sequence length.
- Web design & development
Quality Assurance (QA)
Quality assurance includes planned measures intended to ensure that processes and results meet specified requirements.
- AI fundamentals & models
Quality Drift
Quality drift refers to a change in the measured outcome quality over time, for example due to different inputs, data, or system states.
- AI fundamentals & models
Quality Filter
A quality filter only allows elements that meet defined criteria to pass, such as completeness, readability, or permissible values.
- Process automation
Quality Gates
Quality Gates are stipulated terms and conditions under which a result is accepted, deferred or rejected. Testing can be done automatically or manually. In AI workflows, sources, expertise and approval can be included; even a passed gate does not guarantee freedom from errors.
- Visibility & SEO
Quality of Answer Score
An answer quality score condenses criteria such as correctness, relevance, and completeness of an answer; no universally valid calculation formula exists for it.
- Web design & development
Quality of Service (QoS)
Quality of Service refers to technical mechanisms and properties that influence factors such as delay, throughput, or packet loss of a service.
- Visibility & SEO
Quality Score
A quality assessment summarizes selected properties into a judgment or value. Criteria and weighting determine what this value means.
- Visibility & SEO
Quality Score in Paid Search
The Google Ads Quality Score is a diagnostic value at the keyword level ranging from 1 to 10, based on expected click-through rate, ad relevance, and landing page experience.
- Visibility & SEO
Quality-Adjusted Cost
Quality-adjusted costs evaluate effort considering the quality of results achieved, for example as total cost per accepted result.
- AI fundamentals & models
Quantile
A quantile marks a value below which a certain proportion of a distribution lies; calculation conventions differ for samples.
- AI fundamentals & models
Quantile Regression
Quantile regression models a chosen quantile of a target variable depending on input features, for example, an upper threshold of expected delivery times.
- AI fundamentals & models
Quantization
Mapping values onto a smaller set of discrete representation levels, often with lower bit width. The goal is reduced memory and bandwidth requirements and, on suitable hardware, possibly less computational effort.
- AI fundamentals & models
Quantization-Aware Training (QAT)
Training or fine-tuning with simulated quantization operations so that weights and, if applicable, activations are adapted to the discrete numerical formats used later.
- AI fundamentals & models
Quantized Low-Rank Adaptation (QLoRA)
A parameter-efficient fine-tuning method where the frozen weights of the base model are stored quantized and only added LoRA adapters are trained.
- AI fundamentals & models
Quantum Machine Learning
Quantum-based machine learning investigates how quantum computations can support learning methods or how learning methods can be applied to quantum systems.
- Web design & development
Quarantine
Quarantine in IT processes refers to the isolated handling of potentially problematic files, messages, or data before further use.
- Visibility & SEO
Quarterly Business Review (QBR)
A QBR is a structured business meeting held quarterly about achieved results, open issues, and next joint goals.
- AI fundamentals & models
Quasi-Experiment
An investigative design to estimate causal effects without random assignment by researchers. It uses a justified comparison design and design-dependent identification assumptions.
- Application & enablement
Quasi-Identifier
A feature that does not uniquely identify a person alone but can enable re-identification when combined with other data.
- Web design & development
Queries per Second (QPS)
QPS indicates how many requests occur or are processed within one second; the exact counting method must be defined.
- Data integration & search
Query Cache
A cache that stores results or intermediate results of recurring queries to speed up later responses.
- Visibility & SEO
Query Deserves Freshness (QDF)
Query Deserves Freshness describes the idea that certain search queries require particularly current results and search systems take this into account.
- Data integration & search
Query Embeddings
Vector representations of search queries that map their semantic features for similarity search or ranking.
- Data integration & search
Query Expansion
The addition of related terms, spelling variants, or concepts to a search query to find additional relevant results.
- Data integration & search
Query Fan-Out
An execution pattern in which multiple independent, often parallel-executable subqueries are generated from an initial query and their results are then merged.
- Data integration & search
Query Federation
A logical query whose execution involves data from multiple separate sources or query services and integrates partial results into an overall result.
- Data integration & search
Query Optimizer
A database component that selects a likely efficient query plan from possible execution paths.
- Data integration & search
Query Plan
The concrete sequence of operators with which a database system executes a query.
- Data integration & search
Query Rewriting
The automatic modification of a search query to normalize terms, correct errors, or increase the findability of relevant content.
- Data integration & search
Query Routing
The rule- or model-based selection of an appropriate index, shard, search service, or model for a query. Routing decides on targets; it does not necessarily generate new subqueries.
- Web design & development
Query String
The part of a URL after the question mark that transmits named parameters to a resource or application.
- Data integration & search
Query Understanding Evaluation
The systematic examination of whether a search system correctly captures the meaning, entities, intent, and constraints of queries.
- Data integration & search
Query-Likelihood Model
A probabilistic retrieval model that ranks documents by the estimated probability P(query|document model), usually with smoothed term probabilities.
- Data integration & search
Query-Time Filtering
Restricting allowed search results during the query based on metadata, permissions, or business rules.
- Visibility & SEO
Query-to-Content Mapping
Query-to-content mapping assigns queries or search topics to the content that best fulfills the respective information need.
- AI fundamentals & models
Query, Key and Value (QKV)
Three projections of an attention mechanism: similarity scores are derived from query and key vectors, which weight and merge the value vectors.
- AI fundamentals & models
Question Answering
A method that generates or extracts a suitable answer from knowledge, documents, or a model for a natural language question.
- AI fundamentals & models
Question Decomposition
Splitting a complex question into smaller sub-questions that are answered separately and then combined.
- Web design & development
Queue
A data structure or system component that buffers work items in order until processing.
- Web design & development
Queue Depth
The number of elements waiting in a queue at a measurement point in time. Whether elements already in processing are counted depends on the system and must be documented.
- Web design & development
Queue Latency
The observed time from enqueuing a work item until the start of its processing; measurement points, percentile, and the queue considered must be defined.
- Web design & development
Queueing Theory
The mathematical study of arrival, waiting, service, and utilization in systems with limited capacity.
- Web design & development
Quick Fix
A quick fix is a small, rapidly implementable correction for a specifically defined problem; it can be permanent or provisional.
- Application & enablement
Quick Win
A Quick Win should enable early concrete success with limited effort. It can support a larger change process. Benefits, time requirements and acceptance must be measured on a case-by-case basis; immediate productivity gains are not a promise.
- Web design & development
Quickstart
A quick start guides through the minimal necessary steps to use a tool or interface meaningfully for the first time.
- Application & enablement
Quiet Period
In automation and build systems, a quiet period is a waiting time before start that can bundle closely spaced changes.
- Web design & development
Quorum
A fixed threshold of votes, weights, or replicas that must be reached for a valid operation or decision. In replicated stores, overlapping read and write quorums can support consistency under certain assumptions.
- Web design & development
Quota Exhaustion
A state in which a defined usage, capacity, or resource limit has been fully consumed.
- Web design & development
Quota-Aware Routing
A routing strategy that considers remaining quotas and limits of various targets during selection.
- Web design & development
Quotas
Defined upper limits for resource consumption, calls, storage, or costs within a time period or scope.
- Visibility & SEO
Quoted Query
A search query in which quotation marks typically indicate a contiguous sequence of words or an exact phrase.
- AI fundamentals & models
Qwen
Qwen refers to a model family developed by Alibaba with different sizes and task profiles; characteristics depend on the specific model and version.
- AI fundamentals & models
R-Squared (R²)
A metric that expresses the proportion of variation explained by a regression model in the respective model context.
- Data integration & search
RAG Chunking Strategy
The rule- or model-based splitting of documents into retrievable segments for a RAG system.
- Data integration & search
RAG Evaluation
The separate and joint evaluation of retriever, context compilation, and generator of a RAG system, for example regarding retrieval quality, source referencing, answer fidelity, usefulness, latency, and cost.
- AI fundamentals & models
RAG Poisoning
An attack in which manipulated content is introduced into a retrievable knowledge base or its ingestion pipeline to deliberately influence retrieval and downstream answers.
- Data integration & search
Ragas
Ragas is a framework for evaluating AI applications, known especially for assessing RAG workflows and their answers.
- AI fundamentals & models
Random Search
An optimization method that randomly draws and evaluates configurations from defined value ranges.
- Process automation
Rasa
Rasa offers software and tools for building conversational applications where dialog flows, AI functions, and external actions interact.
- Web design & development
Rate Limit Backoff
A wait-and-retry strategy that delays the next request after a limit response, often with exponential increase and randomness.
- Web design & development
Rate Limiting
Limiting the number or amount of requests per identity, resource, or time window.
- Web design & development
Ray Serve
Ray Serve is a serving library in the Ray ecosystem that provides model and Python applications as services and organizes their execution.
- Data integration & search
Re Embedding
The regeneration of vector representations for existing content, for example after model, data, or segmentation changes.
- Process automation
ReAct Pattern
A method introduced in a research paper that intertwines model-generated task reasoning with actions and the resulting observations. Implementations do not have to disclose internal thought traces.
- Visibility & SEO
Real-Time Bidding
An automated auction process in which individual digital ad impressions are offered and bid on in very short time.
- Visibility & SEO
Real-Time Personalization
The immediate adaptation of content or interactions based on current signals and permissible profile data.
- AI fundamentals & models
Reasoning Model
A non-uniformly standardized model category for language models optimized for multi-step problem solving and additional computational effort during inference.
- AI fundamentals & models
Reasoning Tokens
Reasoning models generate reasoning tokens in addition to input and output tokens to process the task. According to documentation from OpenAI and Microsoft (Azure OpenAI), these do not appear as response text but occupy space in the context window and are billed like output tokens; their number depends on the task difficulty according to the manufacturers and ranges from several hundred to tens of thousands.
- AI fundamentals & models
Recall at k
The proportion of relevant items contained within the first k returned results.
- AI fundamentals & models
Recency Bias
The systematic overweighting of newer observations compared to older, possibly equally relevant information.
- Data integration & search
Reciprocal Rank Fusion (RRF)
A non-learned rank fusion method that sums contributions of the form 1/(k + rank position) for each hit across multiple rankings.
- AI fundamentals & models
Recommender System
A system that prioritizes objects or content likely relevant to users based on data.
- Web design & development
Recovery Point Objective (RPO)
The maximum tolerable data loss, expressed as the time between the last recoverable state and the disruption.
- Web design & development
Recovery Time Objective (RTO)
The targeted maximum duration within which a failed service should be restored after a disruption.
- AI fundamentals & models
Rectified Linear Unit (ReLU)
An activation function that sets negative inputs to zero and passes positive inputs unchanged.
- AI fundamentals & models
Recurrent Neural Network (RNN)
A neural network that passes an internal state through a sequence, thereby considering earlier inputs.
- Web design & development
Recursion
A procedure in which a function or definition refers to a smaller instance of itself until a base case is reached.
- AI fundamentals & models
Red Teaming
A controlled test in which an independent team realistically simulates attack paths and abusive use.
- Application & enablement
Redaction
The targeted removal or obscuring of confidential information in a released representation.
- Web design & development
Redirects
HTTP 3xx-class responses that direct a client to another resource or further action.
- Process automation
Reflection Agent
An agent with self-reflection uses feedback or self-criticism to review and adjust results and next steps.
- AI fundamentals & models
Regression
A model class for estimating a numerical target variable or a conditional relationship from explanatory variables.
- Web design & development
Regression Testing
A test that checks whether a change has unintentionally degraded previously functioning features.
- AI fundamentals & models
Regularization
Methods that limit model complexity or parameter behavior to improve generalization or promote desired structures.
- AI fundamentals & models
Reinforcement Learning
A learning paradigm in which an agent learns a strategy to maximize expected cumulative reward through interaction.
- AI fundamentals & models
Reinforcement Learning from AI Feedback (RLAIF)
A training approach in which an AI system generates evaluations, critiques, or preference labels from which a reward model or other feedback signal is derived for subsequent optimization with reinforcement learning.
- AI fundamentals & models
Reinforcement Learning from Execution Feedback (RLEF)
RLEF uses reinforcement learning with execution feedback so that code language models learn to use feedback from tests or program runs for better solutions.
- AI fundamentals & models
Reinforcement Learning from Human Feedback (RLHF)
In the narrow sense, a training pipeline that converts human preference data into a learned reward signal and then optimizes a model policy with reinforcement learning. The term is sometimes used more broadly but is not synonymous with every direct preference optimization.
- Data integration & search
Relation Extraction
The automatic detection of semantic relationships between mentioned entities in texts.
- Visibility & SEO
Remarketing
A non-uniformly used umbrella term for re-engaging already known or previously interacting target groups through advertising, email, or other permissible channels.
- AI fundamentals & models
Reparameterization Trick
A method that expresses random variables as a deterministic function of independent noise and differentiable parameters so that gradients can be estimated.
- Web design & development
Replicate
Replicate is a platform through which models can be executed via interface and also own models can be provided.
- Web design & development
Replit AI
Replit AI refers to AI features of the development platform Replit, especially support in creating, modifying, and reviewing applications through Replit Agent.
- Visibility & SEO
Reporting
The regulated preparation and distribution of key figures, deviations, and explanations for defined recipient groups.
- AI fundamentals & models
Reproducibility
The possibility to reproduce or verifiably confirm a result with documented data, methods, code, and environment.
- Web design & development
Request Coalescing
A pattern where simultaneously arriving identical requests share a common ongoing computation instead of executing it multiple times.
- Process automation
Request for Proposal (RFP)
A formal document by which an organization requests vendors to submit comparable solution proposals and offers.
- Data integration & search
Reranker
A component that re-evaluates and orders a limited candidate list using a more accurate model.
- Data integration & search
Reranking
The process of reordering an existing candidate list according to an additional scoring function.
- Application & enablement
Research and Development (R&D)
Research and development includes systematic creative work to expand knowledge and use it for new applications. Basic research, applied research, and experimental development have different focuses.
- Web design & development
Resend
Resend is an email service with interfaces for development teams, including event-driven messages and campaigns.
- AI fundamentals & models
Residual Connection
A connection that adds or forwards the input of a network block to its transformed output.
- AI fundamentals & models
ResNet
A family of deep neural networks that uses residual blocks with skip connections to make very deep models more trainable.
- Data integration & search
Resource Description Framework (RDF)
A W3C data model for representing statements as triples of subject, predicate, and object.
- AI fundamentals & models
Response Generation
The creation of an output from inputs, context, and system rules, often by a language model or a rule-based system.
- Web design & development
Response Schema
A formal description of the expected structure, data types, and mandatory fields of an answer.
- Web design & development
Response Validation
The verification of a generated or received answer against schema, value ranges, rules, and, if applicable, sources.
- Application & enablement
Responsible AI
An organizational and technical approach that manages benefits, risks, rights, security, transparency, and responsibilities throughout the AI lifecycle.
- Application & enablement
Responsible Scaling Policy
A Responsible Scaling Policy describes how an AI provider evaluates growing model capabilities and aligns safety measures and further development steps accordingly.
- UX & accessibility
Responsive Web Design
A web design approach in which the layout flexibly adapts to the screen size of the device being used.
- Visibility & SEO
Retargeting
The repeated display of digital advertising to audiences who have previously interacted with an offer. In many practical contexts, retargeting is considered an ad-related form of remarketing; platforms sometimes use the terms synonymously.
- AI fundamentals & models
RetNet
A sequence model architecture with retention mechanism that combines parallel representation for training, recurrent representation for stepwise inference, and a chunkwise recurrent variant.
- Data integration & search
Retrieval Confidence
A system-specific, preferably calibrated measure of how reliably retrieval relevance or coverage is estimated for a query. There is neither a universal definition nor an automatically probabilistic score.
- Data integration & search
Retrieval Drift
A relevant change over time in queries, document corpus, index, retrieval model, or result distribution that can affect retrieval quality. The term is not uniformly standardized.
- Data integration & search
Retrieval-Augmented Generation (RAG)
RAG models combine a pretrained language model (parametric memory) with an external knowledge base (non-parametric memory) accessed by a neural retriever. In the original paper by Lewis et al., this knowledge base is a dense vector index of Wikipedia. For retrieval, the Azure AI Search documentation mentions vector, keyword, and hybrid search.
- Data integration & search
Retrieval-First Policy
A retrieval-first policy stipulates that a system first retrieves relevant information from specified sources before formulating a domain-specific answer.
- Data integration & search
Retriever
A component that selects relevant candidates from a data set for a query.
- Data integration & search
Retriever–Reranker Cascade
A search architecture where a fast retriever generates candidates and a more precise reranker reorders them.
- Web design & development
Retry
The repeated execution of a failed operation under defined conditions.
- Web design & development
Retry Storm
An overload where many clients simultaneously retry errors, thereby further burdening the disturbed system.
- Web design & development
Retryable Error
An error for which a retry under suitable conditions is reasonable and potentially successful.
- Visibility & SEO
Return on Ad Spend (ROAS)
The ratio of revenue attributed to an advertisement to the advertising expenses recorded for it.
- Visibility & SEO
Return on Investment
A ratio metric that compares a defined financial benefit or return to the associated investment costs.
- AI fundamentals & models
Reward Hacking
A behavior where a learning system maximizes the formal reward without reliably accomplishing the intended task.
- AI fundamentals & models
Reward Model
A model that evaluates inputs or outputs with a scalar reward value and thereby guides training or selection.
- Visibility & SEO
RFM Analysis
A customer segmentation based on recency, frequency, and monetary value of past transactions.
- Application & enablement
Right to Explanation
A shortened term for information, review, and protection claims that may arise from the GDPR and its interpretation in certain automated decisions. There is no universal, standalone right to a complete technical explanation as a general rule.
- AI fundamentals & models
Ring Attention
A method for distributed, block-wise computation of exact attention over long sequences, where key and value blocks are passed around in a ring between devices.
- Application & enablement
Risk Classification under the AI Act
The legal classification according to Regulation (EU) 2024/1689 based on the specific system or model, its intended use, the role of the actor, and the relevant legal provisions. Prohibited practices, high-risk systems, special transparency obligations, and rules for GPAI models do not form a uniform linear risk scale.
- Application & enablement
Risk Register
A maintained register of identified risks with causes, impacts, assessments, measures, responsible persons, and status.
- AI fundamentals & models
RMSNorm
A normalization layer that scales an activation vector by its root mean square including a stabilization term and typically uses a learnable gain, but does not subtract the mean.
- AI fundamentals & models
RMSprop
An adaptive gradient method that scales the learning step of each parameter based on an exponentially weighted moving average of its squared gradients.
- AI fundamentals & models
Robustness Testing
A planned test of how stable a system operates under specified disturbances, distribution changes, edge cases, or partial failures. Robustness test, load test, OOD test, and adversarial test are overlapping but not identical types of tests.
- AI fundamentals & models
ROC Curve
A curve that plots the true positive rate against the false positive rate for all decision thresholds.
- Web design & development
Role-Based Access Control (RBAC)
An access control model where permissions are assigned to roles and subjects are assigned to those roles; accesses result from role membership and possibly role hierarchies.
- Web design & development
Rollback
Resetting a software deployment to a previously known and functional state.
- Web design & development
Root Cause Analysis (RCA)
A structured investigation that identifies contributing technical and organizational causes of a problem to prevent recurrence.
- AI fundamentals & models
Root Mean Squared Error (RMSE)
The square root of the mean of squared residuals or prediction errors. RMSE is in the unit of the target variable and weights large errors more heavily than MAE.
- AI fundamentals & models
Rotary Position Embedding (RoPE)
A method that rotates query and key vectors depending on their position, thereby incorporating relative positional information into attention.
- AI fundamentals & models
ROUGE Score
ROUGE stands for Recall-Oriented Understudy for Gisting Evaluation and compares a generated summary with human-written references. Overlapping units such as n-grams, word sequences, and word pairs are counted; ROUGE-N is an n-gram recall.
- Data integration & search
Row-Oriented Database
A database organization in which values of a record are stored physically or logically close together.
- Web design & development
Runbook
A versioned guide for recurring operational, diagnostic, and recovery tasks.
- Web design & development
Runway
Runway develops AI models and tools for media production, including generation and editing of videos, images, and audio.
- AI fundamentals & models
RWKV
A language model architecture that combines parallel training via a contemporary representation with recurrent, state-based inference; the work per new token does not grow with the length of the already generated sequence.
- Application & enablement
SaaS-Pocalypse
SaaS-Pocalypse is a provocative debate term for the thesis that AI agents or rapidly developed custom software could displace parts of traditional SaaS offerings.
- AI fundamentals & models
Safety Alignment
Safety alignment adjusts language models so that they reject harmful requests. For this, supervised fine-tuning and learning from preferred responses are used, for example. The effect can be limited and can be compromised by attacks or later fine-tuning.
- Application & enablement
Safety Case
A structured, context-bound argument that links safety claims with comprehensible evidence and assumptions. It supports a release decision but does not replace it and does not guarantee freedom from harm.
- AI fundamentals & models
Safety Classifier
A safety classifier is a separate model that checks inputs and outputs of a language model for defined safety risks. It can be used before and after the actual language model. What content it flags depends on the defined risk categories and its training.
- AI fundamentals & models
Safety Evaluation
The systematic evaluation of an AI system based on predefined damage models, risk scenarios, misuse cases, and deployment limits. It can combine tests, red teaming, expert review, and operational data.
- Web design & development
Safety Filters
A technical control component that detects and blocks, modifies, logs, or escalates inputs or outputs based on established rules or classifiers.
- Process automation
Safety Guardrails
A collective term for technical and organizational controls that limit permissible inputs, outputs, and actions, enforce permissions, monitor, or escalate to humans.
- Application & enablement
Safety Incident Taxonomy
A versioned category system for consistent classification of security-relevant incidents by event type, cause, damage impact, affected actors, and severity. Taxonomies depend on organization and purpose and may require multiple assignments.
- Application & enablement
Safety Training
Security training aims to guide AI models to avoid certain harmful behaviors and better comply with defined security requirements.
- Visibility & SEO
Sales Qualified Lead (SQL)
A lead that an organization qualifies for direct processing by sales according to agreed handover criteria. Criteria and distinction from MQL are company-specific.
- AI fundamentals & models
Saliency Map
A visualization that shows which input areas particularly influence the model value or a prediction according to a specific method.
- AI fundamentals & models
Sampling
The selection of observations or the random drawing of outputs according to a defined sampling or probability procedure. Meaning and quality criteria depend on the statistical or generative context.
- AI fundamentals & models
Sampling Temperature
Temperature is a parameter of the probability distribution for token selection in supporting language models. Lower values tend to lead to less variation, higher ones to more variation. Neither factual fidelity nor identical results are guaranteed.
- Web design & development
Sandbox Environment
A controlled environment that separates code or processes with limited rights and resources from the rest of the system.
- AI fundamentals & models
SARSA
An on-policy temporal-difference method that updates action values from the actually chosen subsequent action according to the current policy.
- AI fundamentals & models
Satisficing
A decision strategy that accepts the first solution meeting specified minimum requirements instead of searching for a global optimum.
- AI fundamentals & models
Scalable Oversight
Methods by which humans are intended to oversee systems or tasks whose complete direct evaluation is too complex or difficult.
- AI fundamentals & models
Scaled Dot-Product Attention
An attention method that scales query-key dot products by the square root of the key dimension, weights them via softmax, and applies them to values.
- AI fundamentals & models
Scaled Exponential Linear Unit (SELU)
An activation function whose parameters promote self-normalizing networks under certain architectural and initialization conditions.
- AI fundamentals & models
Scenario Analysis
Scenario analysis examines several internally consistent possible future developments to make decisions under uncertainty more robust.
- AI fundamentals & models
Scene Understanding
The machine-based detection of objects, spatial relationships, context, and activities within a visual scene.
- Web design & development
Schema
A formal or agreed description of the structure of permissible data, such as fields, data types, relationships, and integrity rules. Scope and expressiveness depend on the schema language and system.
- Data integration & search
Schema Drift
An unexpected or uncoordinated change in the data structure compared to an agreed schema, such as fields, types, nesting, or mandatory status. Pure distribution changes are data drift; meaning changes are semantic drift.
- Web design & development
Schema Validation
The validation of a data instance against the structure, type, format, and mandatory field rules expressible in a specific schema version.
- Data integration & search
Schema-on-Read
A data processing approach where stored raw or source data is assigned to an interpreting schema only for a specific read or analysis operation.
- Visibility & SEO
Schema.org DefinedTerm
A Schema.org type for the structured description of a term, typically with name, description, identifier, and membership in a DefinedTermSet.
- AI fundamentals & models
Score Matching
An estimation principle for unnormalized density models that approximates the modeled score function—the gradient of the log-density with respect to the data—to the score function of the data distribution without needing to compute its normalization constant.
- Web design & development
Scrapfly
Scrapfly offers services for web data access, including retrieval, browser execution, extraction, and screenshots via interfaces.
- Visibility & SEO
Search AI Answers
AI answer blocks integrated directly into search interfaces that bring together information from different sources.
- AI fundamentals & models
Search Algorithm
A search algorithm explores a state or data space for a goal, a solution, or an optimal path.
- Visibility & SEO
Search Engine Indexing
Before a page can possibly appear in search results, it typically must be discovered, crawled, and processed. Being indexed does not guarantee ranking or visibility for specific search queries.
- Visibility & SEO
Search Engine Optimization (SEO)
Search engine optimization improves technical accessibility, content, and user experience of a website so that relevant pages can be understood and found by search engines.
- Visibility & SEO
Search Intent
The underlying goal a person pursues with a search query, such as to inform, navigate, compare, or act.
- Data integration & search
Search Query
A formulated information request that a search system executes against an index or data set.
- Data integration & search
SearchGPT
SearchGPT was a time-limited OpenAI prototype announced in 2024 for dialogue-oriented web search with answers and source references.
- Visibility & SEO
Seasonality
A regularly recurring pattern in a time series that is tied to calendars, cycles, or recurring events.
- Web design & development
Secrets Management
The secure storage, issuance, rotation, and monitoring of credentials, keys, and tokens.
- Web design & development
Secure Aggregation
A multiparty protocol that computes an agreed aggregate, often a sum of model updates, without exposing the individual contributions in plaintext to the aggregator. The guarantee applies only under the respective threat and failure model.
- Web design & development
Secure by Design
A development principle where security requirements shape architecture, implementation, and operation from the very beginning.
- Web design & development
Secure Egress Control
The rule-based monitoring and restriction of outgoing network communication, for example by destination, protocol, port, identity, or process. It can make data exfiltration more difficult but does not automatically inspect the content of each transmission.
- Web design & development
Secure Multi-Party Computation
A family of cryptographic protocols enabling parties to compute an agreed function over private inputs without revealing more about the inputs than what follows from the output and threat model.
- Process automation
Secure Tool Calling
A non-uniformly standardized collective term for tool calls by AI or automation systems secured with validated schemas, minimal rights, authorization, output control, logging, and, if applicable, human approval.
- Web design & development
Secure Untrusted Input Handling
Validating, canonicalizing, limiting, and contextually encoding data outside a trust boundary.
- Web design & development
Security Assertion Markup Language (SAML)
An OASIS standard for exchanging authentication and authorization information, commonly used for browser-based single sign-on.
- Web design & development
Security Information and Event Management (SIEM)
A security function or platform that centrally collects, normalizes, correlates, searches event and log data and provides it for detection, investigation, and evidence.
- Web design & development
Security Posture
The current overall state of protective measures, risks, vulnerabilities, and responsiveness of an organization or system.
- Web design & development
Seedance
Seedance refers to video generation models from the ByteDance Seed environment that use visual inputs or text to generate videos.
- Visibility & SEO
Segment Analysis
The separate examination of key metrics and behavior for defined groups of customers, usage, or data.
- AI fundamentals & models
Segment Anything Model (SAM)
Meta's Segment Anything Model generates segmentation masks for image areas, for example based on points or outlines as input.
- Visibility & SEO
Segmentation
The division of a population into groups with common characteristics relevant for a purpose.
- Web design & development
Seldon Core
Seldon Core is a framework for deploying and organizing models and inference workflows in production environments.
- AI fundamentals & models
Selective Prediction
A prediction method with a rejection option: The system only outputs a prediction for a selected subset of cases and abstains otherwise. Errors or risk are evaluated together with coverage.
- AI fundamentals & models
Self-Attention
A mechanism where elements of a sequence weight and consider other elements of the same sequence.
- AI fundamentals & models
Self-Consistency
A decoding method introduced in a research paper that generates multiple stochastically produced solution paths for the same task and aggregates the final answer by frequency.
- AI fundamentals & models
Self-Distillation
A variant of knowledge distillation where a model learns from predictions, intermediate representations, or an earlier training state of the same or a closely related model.
- AI fundamentals & models
Self-Play
A training method where an agent generates experience by interacting with copies, earlier versions, or co-learning variants of its own strategy. It is especially suitable for games and multi-agent environments with clear rules.
- AI fundamentals & models
Self-Supervised Learning
Learning from target signals automatically generated from the data, without requiring manual labels for each example.
- AI fundamentals & models
Self-Tuning Systems
Systems that automatically adjust configuration or resources based on observed performance within predefined limits.
- Data integration & search
Semantic Caching
An application-related cache that maps a new query to an existing reusable answer via a semantic similarity criterion. Keys must additionally represent context such as tenant, permissions, model version, and freshness.
- Data integration & search
Semantic Chunking
A family of methods that set text boundaries based on meaning signals such as topic shifts, embedding distances, and document structure, instead of exclusively using a fixed character or token length.
- Web design & development
Semantic Firewall
A semantic firewall is a filtering layer that evaluates content not only by character patterns but also by their presumed meaning.
- Web design & development
Semantic HTML / Markup
Semantic markup uses HTML tags such as <header>, <nav>, <main>, <article>, <section> and correct heading hierarchies (h1–h6). This gives browsers, search engines and screen readers the logical meaning of the content.
- Process automation
Semantic Router
A non-uniformly standardized routing component that assigns requests to a route, data source, model class, or capability based on semantic features or intent.
- Data integration & search
Semantic Search
A search approach that uses semantic representations, entities, relationships, or learned relevance signals to find suitable results even without exact word matches. It can complement lexical search.
- AI fundamentals & models
Semantic Segmentation
A computer vision task that assigns a semantic class to each image pixel.
- Web design & development
Semantic Versioning
A versioning convention `MAJOR.MINOR.PATCH` that distinguishes incompatible changes, compatible features, and bug fixes.
- Web design & development
Semantic Web
A W3C-coined web vision and technology family for machine-processable, linked data with identifiable resources, explicit relationships, and commonly interpretable vocabularies.
- AI fundamentals & models
Sensitivity Analysis
The investigation of how strongly results react to changes in inputs, parameters, or assumptions.
- AI fundamentals & models
Sensor Fusion
The combination of measurements from multiple sensors to estimate state or environment more robustly or accurately.
- Data integration & search
Sentence Transformers
The Python library provides embedding and reranking models. Texts are comparable as vectors; cross-encoders jointly evaluate query and result.
- AI fundamentals & models
SentencePiece
A language-independent method and tool that learns tokenization directly from raw text and processes text as a sequence of subword units.
- AI fundamentals & models
Sentiment Analysis
The automatic classification of subjective attitude or polarity in texts, possibly related to specific aspects.
- AI fundamentals & models
Sentiment Score
A numerical or ordinal value with which a specific sentiment model expresses its assessment of a text.
- Web design & development
Sentry
The platform links error messages with technical context such as stack traces and affected versions. Monitoring functions also show performance issues and outages.
- AI fundamentals & models
Sequence-to-Sequence Model (Seq2Seq)
A model architecture that transforms an input sequence into an output sequence, often with an encoder and decoder.
- Visibility & SEO
SERP Features
Additional display forms of a search results page besides classic organic links, such as highlighted answers, images, or local results.
- Web design & development
Server-Sent Events (SSE)
A web standard where a server sends continuous text events to a browser over a long-open HTTP connection.
- Web design & development
Server-Side Rendering (SSR)
Server-side rendering creates HTML on the server. This can support accessibility for crawlers and the first page setup, but it does not guarantee good loading times or a good search ranking.
- Web design & development
Service Account
A non-human identity that enables applications or automated services defined access.
- Web design & development
Service Mesh
An infrastructure layer for controlling, securing, and observing communication between services.
- Application & enablement
Service-Level Agreement (SLA)
An agreement between service provider and consumer defining measurable service objectives, measurement methods, responsibilities, and possibly consequences for deviations. An SLA can include multiple SLOs and exclusions.
- Process automation
Service-Level Agreement for AI Agents
An SLA describes agreed performance targets and consequences in the event of non-performance. This must be distinguished from SLI as the measured variable and SLO as the target value. Agent-specific quality criteria must also be defined.
- Web design & development
Service-Level Indicator (SLI)
An SLI is a quantitatively measured property of a service, such as success rate, latency, or availability from the user’s perspective.
- Web design & development
Service-Level Objective (SLO)
An SLO is a target value or target range for an SLI over a defined period.
- UX & accessibility
Session
A delimited, usually time-limited state and interaction context between client, user identity, and application. An authentication session is not the same as an analytically formed visit session.
- AI fundamentals & models
Session-Based Recommender System
A recommendation system that primarily uses the current interaction sequence of a session, often without a persistent user profile.
- Data integration & search
Sessionization
The rule-based grouping of individual events into analytical sessions based on identity, time, and domain rules. The result depends on timeout, event order, and device or user assignment.
- Process automation
Shadow Agent
An AI agent with independent tool or system actions that is used outside approved inventory, release, and control processes. Thus, it is an actionable subset of Shadow AI.
- Application & enablement
Shadow AI
The use, procurement, or integration of AI systems outside approved inventory, risk, data protection, and security processes of an organization. It also includes non-agentic tools.
- Web design & development
Shadow Deployment
An operating mode in which a new system receives real production requests in parallel and generates results, but these results are not intended to control productive decisions or user responses.
- AI fundamentals & models
Shadow Evaluation
An evaluation design that measures outputs of a system running in shadow deployment based on real production requests, while the existing system makes the effective decision.
- AI fundamentals & models
SHAP
A feature attribution method that decomposes model predictions via Shapley values from cooperative game theory.
- AI fundamentals & models
Shortest Path
Depending on edge weights and problem structure, algorithms like Dijkstra, Bellman-Ford, or A* are applicable.
- AI fundamentals & models
Siamese Network
An architecture with two or more branches that share the same weights and generate representations for comparison or similarity.
- AI fundamentals & models
Sigmoid Function
An S-shaped function that maps real values onto a limited range; this often refers to the logistic function.
- AI fundamentals & models
Signal to Noise Ratio
The ratio of the power or energy of a useful signal to the power or energy of the noise, often given logarithmically in decibels.
- Web design & development
Signed Webhook
A webhook whose sender cryptographically signs or authenticates the message so that the recipient can verify origin and integrity.
- Application & enablement
Silent Model Swap
An unannounced model change is a descriptive governance term for the replacement or significant modification of the model used behind a consistent interface without sufficient prior information or version labeling.
- Application & enablement
Silicon Sample
A Silicon Sample arises when a language model is conditioned with information about hypothetical persons and answers survey questions. This produces artificial answers, not actually collected opinions.
- AI fundamentals & models
SiLU / Swish
A smooth, non-monotonic activation function f(x)=x·sigmoid(x). This function is called SiLU and corresponds to Swish with fixed parameter β=1; the general Swish family can use other β values.
- AI fundamentals & models
Sim-to-Real Transfer
The transfer of a policy or model trained in simulation to a real system despite differences in dynamics, sensors, actuators, and environment distribution.
- AI fundamentals & models
SimCLR
SimCLR is a method of contrastive self-supervised learning that learns similar representations from differently augmented views of the same image.
- Data integration & search
SimHash
SimHash maps weighted features via random hyperplanes onto a binary fingerprint. Similar feature vectors are likely to have a small Hamming distance.
- Data integration & search
Similarity Score Calibration
Calibration of similarity values estimates, based on labeled comparison pairs, how raw scores relate to an application-specific hit probability or another interpretable risk measure. It is to be separated from the choice of a decision threshold.
- Data integration & search
Similarity Search
Similarity search finds the objects nearest to a query according to a defined distance or similarity function.
- Data integration & search
Similarity Thresholding
A similarity threshold is an application-related decision threshold from which a score is treated as a hit, possible match, or escalation case. It does not turn the score into a probability.
- AI fundamentals & models
Simple Preference Optimization (SimPO)
SimPO is a method published in 2024 for preference optimization of language models that uses a length-normalized reward without a separate reference model.
- AI fundamentals & models
Simpson's Paradox
In Simpson's paradox, a statistical association reverses after data are aggregated or divided into groups.
- AI fundamentals & models
Simulation
A simulation represents selected properties of a real or planned system in a model to study behavior under assumptions.
- AI fundamentals & models
Simultaneous Localization and Mapping (SLAM)
SLAM refers to methods by which a mobile system simultaneously creates a map of its environment and estimates its own position within it.
- Web design & development
Single Page Application (SPA)
A SPA first loads an HTML document and then updates views dynamically without reloading the full document for each change. It can have multiple URLs and routes; initial loading time and usability must be checked.
- Web design & development
Single Sign-On (SSO)
Single sign-on allows access to multiple connected applications after logging in once with a trusted identity provider, without needing to authenticate again for each application. The applications usually maintain their own sessions.
- AI fundamentals & models
Sinusoidal Positional Encoding
Sinusoidal positional encoding supplements token representations with position-dependent sine and cosine values so that a Transformer can consider order.
- Process automation
Siri in iOS 27
Siri is Apple's integrated assistant. On iOS 27, Apple distinguishes between available Siri functions and Siri AI. Which variant is usable depends, among other things, on device, language, region, and software version.
- Web design & development
Site Architecture
Website architecture describes the hierarchical and linked structure of pages, navigation, and content of a website.
- Web design & development
Site Reliability Engineering (SRE)
Site Reliability Engineering applies software development to the operation, reliability, and scaling of services and manages goal conflicts via SLOs and error budgets.
- Visibility & SEO
Sitemap
A sitemap is a structured overview of URLs; as an XML file it supports search engines in finding and categorizing canonical content.
- Process automation
Skill Marketplace
A skill marketplace bundles instructions, tools, or extension packages that equip an agent for specific tasks. Which components and rights a skill receives depends on the respective system.
- AI fundamentals & models
Skip Connection
A skip connection routes the input of a network block past one or more layers and combines it with its output.
- AI fundamentals & models
Sliding Window Attention
Attention with sliding window restricts each token to a local context area to reduce computational and memory costs of long sequences.
- Application & enablement
Slopaganda
Slopaganda combines generative mass production with propagandistic goals. The term particularly describes cheap, repetitive, or emotionally charged content whose dissemination aims to influence attention and perception.
- Visibility & SEO
Slopsquatting
Slopsquatting is a supply chain attack where attackers register fabricated package names suggested by AI systems and fill them with malicious code.
- Process automation
Slot Filling
Slot filling extracts or requests structured values that a dialogue system needs for an intent or action.
- Process automation
Small Action Model
Small action model here is a descriptive term for a resource-saving model that selects or prepares actions within a defined process. The term does not denote a uniformly defined model architecture.
- AI fundamentals & models
Small Language Model
A small language model is a relatively resource-efficient language model; “small” is context-dependent and not defined by a uniform parameter limit.
- AI fundamentals & models
Snorkel
Instead of labeling each dataset individually, a team formulates labeling functions. Snorkel can combine their sometimes conflicting results to obtain probabilistic training labels.
- Data integration & search
Snowflake
The architecture separates data storage and compute resources. Teams can process data with SQL and allocate compute capacity for different workloads.
- Application & enablement
SOC 2
SOC 2 refers to AICPA audit reports on controls of a service organization relevant to the Trust Services Criteria. Type I assesses control design at a point in time; Type II additionally assesses operational effectiveness over a period.
- AI fundamentals & models
Soft Prompt
A Soft Prompt is a trainable sequence of continuous vectors that is prepended as additional context to a frozen language model.
- AI fundamentals & models
Softmax
Softmax converts a sequence of real values into positive values summing to one and is often used for class probabilities or attention weights.
- Web design & development
Software Bill of Materials (SBOM)
An SBOM is a structured inventory of the software components, versions, and dependencies of a product.
- Web design & development
Software Development Kit (SDK)
A package assembled by the platform or interface provider consisting of libraries, tools, type definitions, documentation, and possibly examples for development against a specific target system.
- Web design & development
Software Notarization
During notarization, software is submitted to Apple's service and checked for known malicious content and certain signing issues. A successful process yields a ticket.
- Web design & development
Software Supply Chain Security
Software supply chain security protects source code, dependencies, development and build systems, artifacts, and distribution against manipulation, compromise, and unintended risks.
- Web design & development
Software Verification
The examination of whether a software artifact correctly fulfills specified requirements or specifications.
- AI fundamentals & models
Solomonoff Induction
Solomonoff induction is a theoretical method that derives predictions from a weighted mixture of all computable explanatory programs.
- Web design & development
Sora
The models generated videos from descriptions and additional specifications. OpenAI discontinued the web and app offerings on April 26, 2026; the company names September 24, 2026 as the shutdown date for the API.
- Web design & development
Sora 2
The model combined video generation with synchronized dialogues and sounds. It is to be distinguished from the original Sora model but belongs to the same product history.
- Visibility & SEO
Source Attribution
Source attribution links statements, content, or model answers transparently with the underlying sources.
- AI fundamentals & models
Source Separation
Source separation decomposes a mixed signal into individual original signals, such as multiple voices or instruments.
- Application & enablement
Sovereign AI
Sovereign AI refers to the organizational and technical capability to operate AI systems under one's own control and in accordance with local rules, data requirements, and strategic objectives.
- Web design & development
Space Complexity
Memory complexity describes how the memory needed by an algorithm grows with input size.
- Web design & development
spaCy
Processing pipelines recognize, for example, sentence boundaries, parts of speech, and named entities. Language models and custom rules can be combined for specific text tasks.
- Data integration & search
SPARQL
SPARQL is the W3C query language and protocol for RDF graphs.
- AI fundamentals & models
Sparse Attention
Sparse attention calculates attention relationships only for selected token pairs instead of every possible pairing.
- AI fundamentals & models
Sparse Autoencoder
A sparse autoencoder reconstructs inputs while a regularization keeps only a few latent units active simultaneously.
- AI fundamentals & models
Sparse Mixture of Experts
A sparse Mixture-of-Experts architecture routes each token only to a small selection of specialized subnetworks.
- AI fundamentals & models
Sparse Model
A sparse model contains many inactive or zeroed parameters, features, or connections and thus uses only a part of its possible structure.
- Data integration & search
Sparse Retrieval
Sparse information retrieval represents queries and documents with mostly zero-filled term-based vectors and evaluates their similarity.
- AI fundamentals & models
Sparse Training
Training sparse models maintains or generates a sparsely populated parameter structure during learning to reduce computational or memory requirements.
- AI fundamentals & models
Speaker Diarization
Speaker diarization answers who speaks when in a recording by segmenting speech and assigning speaker clusters.
- Web design & development
Specification-Driven Development
Specification-driven development systematically aligns design, implementation, and testing with an explicit, versioned specification.
- AI fundamentals & models
Specificity
Specificity is the true negative rate TN/(TN+FP): the proportion of truly negative cases correctly recognized as negative at a given threshold.
- AI fundamentals & models
Spectral Normalization
Spectral normalization limits the spectral norm of a weight matrix and thereby controls the Lipschitz constant of a network layer.
- AI fundamentals & models
Speculative Decoding
Speculative decoding lets a faster draft model propose multiple tokens and the target model jointly evaluate a proposal block. A correct acceptance and correction procedure preserves the target distribution.
- AI fundamentals & models
Speech Enhancement
Speech signal enhancement reduces noise or reverberation and increases the intelligibility or technical quality of a speech signal.
- AI fundamentals & models
Speech Synthesis
Speech synthesis produces an audio signal with spoken language from text or linguistic representations.
- AI fundamentals & models
Speech-to-Text
Automatic speech recognition converts spoken audio signals into text or token sequences.
- AI fundamentals & models
Stability AI
The provider is known, among other things, for the Stable Diffusion model family. Other offerings concern media such as audio; company, model, and individual application should be distinguished from each other.
- AI fundamentals & models
Stable Diffusion
In latent diffusion, a compressed image representation is gradually generated from noise and converted into an image. Text prompts and other conditions control the result.
- Web design & development
Staging Environment
A staging environment realistically simulates the production environment so that changes can be integrated and tested before release.
- AI fundamentals & models
Stanza
The pipeline segments text and recognizes, among other things, lemmas, parts of speech, syntactic relations, and named entities. Which analyses are possible depends on the language and loaded models.
- AI fundamentals & models
State Space Model
A state space model describes a dynamic system via a latent state that evolves over time and generates observations.
- AI fundamentals & models
State Transition System
A state transition system describes possible system states and the transitions caused by events or actions between them.
- Web design & development
Statefulness
Statefulness means that a system's response depends on previously stored information or earlier events.
- Web design & development
Static Site Generation
Static site generation produces HTML files before the request, usually during a build process, instead of rendering each page at runtime.
- AI fundamentals & models
Stationarity
Stationarity in time series refers to the property that their statistical characteristics remain unchanged under time shifts; depending on the definition, different strict conditions apply.
- AI fundamentals & models
Statistical Significance
Statistical significance means that a test result is sufficiently unusual under a specified null hypothesis, not that the effect is large or practically important.
- AI fundamentals & models
Steering Vector
A control vector is a direction in the activation space of a neural network that is added or amplified during inference to influence behavior.
- AI fundamentals & models
Stemming
Stemming reduces word forms with heuristic rules to a common stem that is not necessarily a correct dictionary word.
- AI fundamentals & models
Step Decay
Stepwise learning rate reduction decreases the learning rate during training at specified times by a predetermined factor.
- AI fundamentals & models
Stochastic Gradient Descent (SGD)
Stochastic gradient descent estimates the gradient of the objective function using randomly drawn examples or mini-batches and updates the model parameters accordingly. The common mini-batch SGD is a practical variant.
- Application & enablement
Stochastic Parrot
“Stochastic parrot” is a critical metaphor for language models that probabilistically reproduce linguistic patterns without presupposing human semantic understanding.
- AI fundamentals & models
Stochastic Weight Averaging (SWA)
SWA computes the arithmetic mean of model weights from multiple later training states, often along a learning rate trajectory, which can result in a better generalizing parameter solution.
- AI fundamentals & models
Stop Sequence
A stop sequence is a specified character or token sequence upon whose generation text generation is terminated.
- AI fundamentals & models
Stopword Removal
Stop word removal removes frequent function words from text representations when they offer little discriminative power for a specific analysis.
- AI fundamentals & models
Stratified Sampling
Stratified sampling divides a population into defined strata and draws a random sample from each stratum.
- Application & enablement
Stream-of-Consciousness Prompting
The informal technique describes a spoken or written stream of thought as a starting point for a prompt. Then goal, context, and requirements are organized; it is not a standardized prompting method.
- Web design & development
Streaming
Streaming is a generic term for continuous transmission or processing in parts before a complete data set is available. Media streaming, stream processing, and streamed HTTP responses are different forms.
- AI fundamentals & models
Streaming Automatic Speech Recognition
Streaming speech recognition processes audio during recording and continuously delivers provisional as well as later stabilized transcripts.
- Data integration & search
Streaming Data
A data stream is a potentially unlimited sequence of time-related data events that continuously arrive and are often processed incrementally. Event time and arrival order may differ.
- Web design & development
Streaming Responses
In a streamed response, a server transmits multiple response parts over a persistent or incrementally read connection as soon as they are available. This is not limited to tokens or Server-Sent Events.
- Process automation
Stripe Agentic Commerce
Agent-supported commerce with Stripe refers to the integration of product offerings, checkout, and payment processing with AI-powered shopping interfaces. Merchants and buyers still require clear responsibilities and approvals.
- AI fundamentals & models
STRIPS
STRIPS is a formal representation for planning problems in which actions are described by preconditions as well as add and delete effects.
- Visibility & SEO
Structured Data (Schema.org)
Structured data describes objects and properties in a machine-readable format. Schema.org is a widely used vocabulary; JSON-LD is one of the markup formats. Markup must match the visible content and does not guarantee rich results.
- Web design & development
Structured Logging
Structured logging captures events in machine-readable fields instead of exclusively as free text lines.
- Web design & development
Structured Output
Structured output is a machine-readable output according to a defined format or schema, such as JSON with defined fields and types. It can be generated by constrained decoding or validated afterward.
- Visibility & SEO
Structured Product Data
Structured information describes, for example, product, price and availability. Product markup can be used for this on websites; data feeds are another means of delivery.
- AI fundamentals & models
Structured Pruning
Structured pruning removes entire channels, filters, heads, or blocks of a neural network instead of individual weights.
- AI fundamentals & models
Structured State Space Sequence Model (S4)
A concrete sequence model architecture that makes a structured parametrized linear state space model computable so that training can be done as convolution and inference recurrently.
- AI fundamentals & models
Style Transfer
Style transfer generates an image that combines the content of one template with the statistical style features of another representation.
- AI fundamentals & models
StyleGAN
The generator influences image features at different levels via internal style variables. This allows properties of generated images to be examined and modified more precisely.
- Process automation
Subagents and Agent Harness
Subagents take on clearly defined subtasks of a higher-level process. The execution environment provides, for example, tool access, handovers, logs, and checks.
- Web design & development
Subject Consistency
Subject consistency describes how well the characteristic features of a person, character or product are preserved across several generated images. References or model adaptation can help. This does not guarantee pixel-identical or completely identical representations.
- AI fundamentals & models
Summarization
Automatic summarization condenses one text or multiple documents to central information, extractively by selection or abstractively by rephrasing.
- Web design & development
Supabase
The components include database, authentication, file storage, and interfaces. Applications access these services in a controlled manner; database policies regulate access to records.
- AI fundamentals & models
Super Resolution
Super Resolution reconstructs a higher-resolution estimate from one or more low-resolution images.
- AI fundamentals & models
Superalignment
It refers to the alignment of potential systems whose capabilities exceed those of human supervisors. The challenge lies in reliably assessing behavior that humans cannot fully comprehend themselves.
- AI fundamentals & models
Superposition
Superposition in interpretability research denotes the joint representation of multiple features in overlapping directions of a limited activation space.
- AI fundamentals & models
Supervised Fine-Tuning (SFT)
The further optimization of a pretrained model using labeled input-output examples and a supervised learning objective.
- AI fundamentals & models
Supervised Learning
Machine learning of a mapping from inputs to target values based on labeled training examples.
- Web design & development
Superwise
The manufacturer documentation describes an AI gateway as well as a platform for creating and operating agents. Policies and insights into usage are intended to make their operation more controllable.
- Web design & development
Supply-Chain Prompt Injection
Prompt injection via the supply chain reaches a system through introduced components or content, such as templates, documents, or tool descriptions.
- Visibility & SEO
Supply-Side Platform (SSP)
An SSP is a platform through which publishers automatically offer, manage, and auction digital advertising inventory to demand sources.
- AI fundamentals & models
Surrogate Model
A surrogate model approximates an expensive, slow, or not directly accessible system to perform predictions, optimization, or sensitivity analyses more cost-effectively.
- AI fundamentals & models
Survival Analysis
Statistical methods for time until an event considering censored observations.
- Web design & development
SWE-bench
SWE-bench is a benchmark family that tests systems based on selected real GitHub issues and corresponding repository states to generate test-passing code changes.
- AI fundamentals & models
SwiGLU
SwiGLU is a gated feedforward activation that multiplies a Swish-weighted projection with a second linear projection.
- AI fundamentals & models
Sycophancy
Sycophancy describes the behavior of an AI system excessively confirming user views or expected preferences, even when factual accuracy suffers.
- Application & enablement
Sycophancy Tax
Sycophancy Tax is an informal cost metaphor for the consequences of an AI confirming expectations when factual contradiction would be necessary. It is neither a tax nor a uniformly defined metric.
- Visibility & SEO
Synthetic Customers
Synthetic customers are artificial profiles or simulations representing possible reactions to offers. Their answers arise from model assumptions and are not actually observed purchase decisions.
- AI fundamentals & models
Synthetic Data
Synthetic data is generated through rules, simulation, or generative models to replicate characteristics of real data for development or testing.
- Visibility & SEO
Synthetic Data in Marketing
Synthetic marketing data replicates selected characteristics of contact, campaign, or purchase data, for example. They serve for testing and simulation without automatically guaranteeing real market reactions or sufficient data protection.
- AI fundamentals & models
Synthetic Media
Synthetic media are images, audio, video, or text content that are wholly or partially algorithmically generated or altered.
- AI fundamentals & models
Synthetic Minority Oversampling Technique (SMOTE)
SMOTE generates synthetic examples of a minority class by interpolation between neighboring training points.
- Web design & development
Synthetic Monitoring
Synthetic monitoring regularly performs simulated user transactions to check availability, functionality, and performance of a service.
- AI fundamentals & models
SynthID
Google DeepMind's method embeds information into image, audio, video, or text outputs. Suitable detectors later search for these patterns; the exact method depends on the medium.
- AI fundamentals & models
SynthID-Text
The method influences the selection of possible tokens without inserting a visible mark. A detector later assesses whether the expected pattern occurs in the text.
- Application & enablement
System Prompt
A system prompt gives a supporting AI application higher-level instructions on role, task framework and response behavior. Role priority and implementation depend on the model. It does not replace technical access control and is not an insurmountable security boundary.
- AI fundamentals & models
Talking Head Generation
The generation of talking portraits produces video sequences of a speaking face from image, audio, or text inputs.
- UX & accessibility
Task Completion Rate
The task success rate is the proportion of attempts in which users successfully complete a clearly defined task according to specified criteria.
- Data integration & search
Taxonomy
A taxonomy assigns terms or objects according to defined categories and relationships, usually in a hierarchical structure.
- Web design & development
Technical Debt in AI Systems
In addition to code, data dependencies, configurations and feedbacks can make subsequent changes more difficult. Rapid introduction can thus cause long-term follow-up costs.
- Visibility & SEO
Technical SEO
Technical SEO improves conditions so that search engines can reliably fetch, render, canonicalize, and index content.
- Application & enablement
Technological Singularity
The technological singularity refers to the hypothesis that very powerful AI or other developments trigger an accelerating change. Course, possibility, and timing are scientifically disputed and not certain.
- Web design & development
Technology Stack
A technology stack is the combination of programming languages, frameworks, runtimes, databases, infrastructure, and operational services jointly used for a system. The term is descriptive and not standardized.
- AI fundamentals & models
Temperature Sampling
In temperature-controlled sampling, logits are divided by a positive temperature value T before softmax and then randomly sampled. T less than one concentrates the distribution; T greater than one flattens it.
- AI fundamentals & models
Temperature Scaling
Temperature scaling is a post-calibration method for classifiers: a positive scalar T is learned on separate validation data and divides the logits before softmax, without changing the predicted class.
- Process automation
Temporal
Workflows coordinate steps whose execution and progress are saved. Failed external activities can be retried according to defined rules.
- AI fundamentals & models
Temporal Graph Network
A graph model that processes time-stamped node and edge events as well as changing states.
- AI fundamentals & models
Temporal-Difference Learning
A reinforcement learning method that updates value estimates from reward and subsequent estimate before an episode ends.
- Web design & development
Tensor Parallelism
A form of distributed training that splits tensor operations of a single model layer across multiple accelerators.
- Web design & development
Tensor Processing Unit (TPU)
An accelerator developed by Google for tensor-based machine learning computations, especially matrix operations.
- Web design & development
TensorFlow Extended (TFX)
Components support data validation, preprocessing, training, evaluation, and deployment. This transforms a single model experiment into a traceable process.
- Web design & development
TensorRT-LLM
The library bundles specialized runtime methods and optimizations for inference. Model, hardware, and configuration determine which accelerations are actually usable.
- Process automation
Terminal Agent Orchestration
The orchestration of terminal agents organizes AI systems that execute tools via command lines. It controls jobs, working directories, permissions, and the merging of results.
- AI fundamentals & models
Test-Time Training
A method that adjusts model parameters during evaluation based on a self-supervised auxiliary objective applied to current inputs.
- AI fundamentals & models
Text Classification
The assignment of a text to one or more predefined classes.
- AI fundamentals & models
Text Generation
The automatic generation of text sequences through rules, statistical models, or neural language models.
- AI fundamentals & models
Text Normalization
The rule- or model-based standardization of text, such as spelling variants, Unicode, numbers, or abbreviations, for a specific purpose.
- AI fundamentals & models
Text-to-3D Generation
The generation of a 3D representation or multiple consistent views from a text description.
- AI fundamentals & models
Text-to-Image Generation
The synthesis or editing of images, controlled by text and optionally additional conditions.
- Web design & development
Text-to-Video Generation
AI technology for generating moving video sequences directly from textual descriptions.
- AI fundamentals & models
Textual Inversion
A method that learns new token embeddings from few example images for a frozen text-to-image model.
- AI fundamentals & models
TF-IDF
A weighting that boosts frequent terms in a document and downweights terms frequent in the entire corpus.
- AI fundamentals & models
Thompson Sampling
A Bayesian strategy for sequential decisions that draws a model hypothesis from the posterior and then acts accordingly.
- Web design & development
Threat Modeling
A structured process to identify assets, attackers, attack paths, assumptions, and countermeasures of a system.
- Web design & development
Throughput
Throughput measures processed work per unit of time, such as requests per second or generated tokens per second. In the LLM comparison, it must be clear whether input or output tokens and the overall system or a single request are measured. High throughput does not guarantee low response time or low overall costs.
- Web design & development
tiktoken
The library uses byte-pair encoding and maps text to token IDs. The appropriate encoding is crucial because token counts vary between models and texts.
- Web design & development
Time Complexity
The asymptotic number of elementary computational steps of an algorithm depending on the input size.
- Web design & development
Time in System; Sojourn Time
The time from arrival of a job until completion of its service; it includes waiting time and processing time.
- AI fundamentals & models
Time Series
A sequence of observations assigned to points or intervals in time and often dependent.
- AI fundamentals & models
Time Series Analysis
Methods for describing, modeling, forecasting, and monitoring temporally dependent data.
- AI fundamentals & models
Time Series Foundation Model
A model pretrained on extensive time series corpora intended to be transferable to multiple forecasting or analysis tasks.
- Web design & development
Time to First Token
The measured duration from the defined request start to receiving the first output token of a generative service.
- AI fundamentals & models
TinyML
Machine learning on highly resource-constrained embedded devices, typically microcontrollers with tight power and memory budgets.
- Web design & development
Together AI
Teams can use supported models via hosted inference services and customize or operate them with dedicated computing power depending on the offering. Model availability and operational options depend on the product.
- AI fundamentals & models
Token
A token is a processing unit of the model. How text is divided into tokens and billed depends on the respective service and model.
- AI fundamentals & models
Token Budgeting
Token budgeting plans and limits the scope of inputs, outputs and, where applicable, internal processing tokens. Context boundaries, pricing model and sufficient space for the task must be taken into account. Input length alone does not determine all costs.
- Visibility & SEO
Token Cost per Lead
Token costs per lead are the attributable costs for processed input and output tokens divided by the defined number of leads. Period, lead definition, and included processing steps must be consistently delimited.
- Application & enablement
Token Tax
Token Tax is an informal term for additional effort caused by repeated context, long tool outputs, or unnecessary response length. Which tokens are avoidable depends on the task and the required quality.
- AI fundamentals & models
Tokenization
The division of text or other inputs into discrete units processed by a model or search system.
- Application & enablement
Tokens per Second Economy
The term here summarizes the operational evaluation of the output rate of an AI system. Tokens per second measure speed; economic benefit arises only in relation to task, load, quality, and total costs.
- Web design & development
Tool Poisoning
An attack in which malicious instructions or misleading metadata in tool descriptions influence an agent’s selection or execution behavior.
- AI fundamentals & models
Tool Use / Function Calling
A model answer identifies a tool and its arguments. The connected application checks permissions, executes the call, and returns the result. The model itself does not automatically have access to an API or database.
- AI fundamentals & models
Top-k Sampling
A decoding method that keeps only the k most probable next tokens, renormalizes, and randomly selects from them.
- AI fundamentals & models
Top-p Sampling
A decoding method that forms the smallest token set with cumulative probability at least p and samples from it.
- AI fundamentals & models
Topic Modeling
The unsupervised or weakly supervised identification of latent topics and their distribution in documents.
- Web design & development
TorchServe
Models are packaged with their processing logic and offered via a server for predictions. Project information indicates limited or discontinued active maintenance; new deployments require alternative evaluation.
- Visibility & SEO
Touchpoint
A specific interaction of a person with an organization, brand, product, or service along a customer journey.
- AI fundamentals & models
Toxicity Detection
Classification of text into defined forms of harmful language, such as insult, threat, or group-related disparagement.
- AI fundamentals & models
Training Epoch
In the usual mini-batch training, the training examples are run once per epoch. The number of parameter updates also depends on batch size and update strategy; data augmentation can change the specific input.
- AI fundamentals & models
Transfer Learning
The use of representations or parameters from a source task to improve a related target task.
- AI fundamentals & models
Transformer
A sequence model architecture that processes representations mainly with attention and feedforward blocks instead of recurrence.
- Application & enablement
Transparency Obligations under Article 50 of the AI Act
Article 50 of the EU AI Regulation regulates, among other things, information in the case of direct AI interaction as well as labelling of certain AI content. Provider and operator obligations differ. Machine-readable marking and visible disclosure are different requirements; scope, exceptions and transitional rules must be examined.
- Web design & development
Transport Layer Security (TLS)
A protocol for the authenticated establishment of a confidential and integrity-protected communication channel over an unprotected network.
- AI fundamentals & models
Treatment Effect
The causal difference between outcomes with treatment and without treatment for a unit or population.
- AI fundamentals & models
Tree of Thoughts
A prompting and search method that generates, evaluates, and branches multiple textual intermediate states.
- AI fundamentals & models
Trefferquote (Recall)
The proportion of truly relevant or positive cases that a system correctly identifies.
- Process automation
Trigger.dev
Tasks can be triggered, monitored, and retried on errors. Queues and concurrency limits help control their execution.
- AI fundamentals & models
Triplet Loss
A metric loss function that brings anchor and positive example closer and pushes a negative example at least a margin further apart.
- Web design & development
Triton Inference Server
The server supports different model backends and organizes requests to their runtimes. Batching and parallel processing can better utilize available computing resources.
- AI fundamentals & models
Truncation
Shortening an input sequence to an allowable length according to a defined strategy.
- Application & enablement
Trust and Safety
The field deals with harmful content, abuse, and problematic behavior on digital services. This includes rules, moderation, reporting options, investigations, and transparent complaints.
- Web design & development
Trust Boundary
A boundary across which data or control passes between areas with different trust or protection assumptions.
- Application & enablement
Trust Models
Trust models define assumptions, evidence, and limits of trust. Depending on context, they may concern technical identities, authorizations, or the reliability of an AI system; these meanings should not be mixed.
- Web design & development
Trusted Execution Environment (TEE)
A hardware-based, isolated execution environment designed to protect code and data during processing under a defined threat model from other software components. Secure enclaves are product-specific implementations.
- AI fundamentals & models
TruthfulQA
A benchmark with questions that provoke common misconceptions or false assumptions to evaluate truthful answers.
- Data integration & search
Two-Tower Model
An architecture with separate encoders for query and candidate, whose vectors are efficiently compared.
- AI fundamentals & models
U-Net
A convolutional network architecture with contracting and expanding paths plus skip connections for dense predictions.
- Web design & development
Ubiquitous Language
A vocabulary jointly maintained in Domain-Driven Design that business and development consistently use in models, conversations, and code.
- AI fundamentals & models
Ultra-Long Context Window
The context window limits the number of tokens in a model interaction. “Very large” is a relative term; a high limit does not guarantee that the model reliably uses every piece of contained information.
- Web design & development
Unbounded Fan-Out
An error pattern where a request without an effective limit generates more and more parallel subtasks or follow-up calls.
- AI fundamentals & models
Uncertainty Quantification (UQ)
The quantitative description, propagation, and evaluation of uncertainty in inputs, models, and results.
- Process automation
Uncertainty-Aware Routing
A non-uniformly standardized routing pattern that routes requests based on estimated uncertainty to model, human, or fallback.
- AI fundamentals & models
Underfitting
A model is underfitted if it does not adequately represent relevant structure even in training data.
- Web design & development
Unicode Normalization
The conversion of Unicode text into a defined canonical or compatibility-based normal form.
- Data integration & search
Unified Search
A search interface and architecture that searches multiple collections under common access, ranking, and presentation rules.
- AI fundamentals & models
Uniform Information Density
A linguistic hypothesis according to which speakers tend to distribute information as evenly as possible over utterances.
- Web design & development
Uniform Resource Locator (URL)
Web addresses typically contain scheme, host, and path; parameters and fragments can add further details. Not every URL denotes a publicly accessible webpage.
- AI fundamentals & models
Unigram Model
A model that treats units independently with their own probability each; in tokenizers it evaluates possible segmentations.
- AI fundamentals & models
Unintended Memorization
As a result, rare or sensitive text sequences can be reproduced under certain conditions. The risk depends, among other things, on data, repetitions, and training.
- Application & enablement
Unit Economics
The analysis of revenue, variable costs, and contribution margin per meaningfully defined unit, such as order or customer.
- Web design & development
Unit Test
A function or module receives defined inputs; the test compares the actual with the expected result. Dependencies are replaced with controlled substitutes if needed.
- Data integration & search
Universal Embeddings
A model generates vectors that support, for example, search, comparison, and grouping. “Universal” describes an aspiration for broad usability and no guarantee for every language or domain.
- Data integration & search
Unstructured Data
Data without a fixed tabular schema, such as free text, documents, images, audio, or video.
- AI fundamentals & models
Unsupervised Learning
Methods group similar datasets or reduce many features to few dimensions. The discovered patterns must then be interpreted professionally.
- Web design & development
Update Cadence
Teams determine whether changes occur on schedule, on events, or as needed. This includes review, publication, and possibly rollback.
- Web design & development
Update vs. Upgrade
An update often fixes bugs or adds minor features; an upgrade can bring major changes and migrations. However, vendors do not always use both terms consistently.
- Visibility & SEO
Uplift Modeling
The estimation of how a measure changes the outcome of individual persons or segments compared to no treatment.
- Data integration & search
Upsert
A unique key decides whether to insert or update. Database systems offer their own commands for this, such as INSERT with conflict handling in PostgreSQL.
- Visibility & SEO
URL Canonicalization
The determination of a preferred URL for identical or very similar content and the consistent handling of URL variants.
- UX & accessibility
Usability Testing
An empirical examination in which representative people perform defined tasks with a product and are observed.
- Web design & development
Usage Anomaly Detection
Detection of unusual usage patterns compared to a defined historical, seasonal, or group-specific reference behavior.
- Web design & development
Usage Telemetry
Measurement data show, for example, called functions, runtimes, and errors. Events must be defined so that they answer a specific question without capturing unnecessary content.
- Application & enablement
Usage-Based Pricing
A pricing model where the fee depends wholly or partly on the measured usage of a defined unit.
- Web design & development
Usage-Based Routing
An application can select targets based on load, previous consumption, or agreed usage limits. The specific rule must be named; the term is not a single fixed algorithm.
- UX & accessibility
User Acceptance Testing (UAT)
Future users or domain experts work through realistic scenarios and compare the results with acceptance criteria. Technically passed individual tests do not replace this domain-specific review.
- UX & accessibility
User Experience (UX)
A user's overall experience and perception when interacting with a digital application.
- Visibility & SEO
User Intent
Different goals can lie behind similar search terms, such as understanding something, comparing offers, or completing a task. Intent is inferred from context and cannot be reliably recognized from wording alone.
- UX & accessibility
User Journey
The sequence of all touchpoints a user has with a brand on the way to their goal.
- UX & accessibility
User Onboarding
The designed process that leads new users from first contact to independent successful use.
- UX & accessibility
User Persona
A research-based model of a relevant user group with goals, behavior, context, and constraints.
- UX & accessibility
User Research (UX Research)
Interviews, observations, usability tests, and quantitative methods answer different research questions. The method depends on the question, not the available tool.
- Web design & development
User-Defined Function (UDF)
In databases, a UDF can take input values and calculate a value or result set from them. Language, execution rights, and allowed side effects depend on the system.
- Visibility & SEO
User-Generated Content (UGC)
Content created and published by users instead of the platform or brand operator.
- AI fundamentals & models
Utility Function
A function that maps outcomes or states into numbers according to the preferences of a decision maker.
- Visibility & SEO
UTM Parameters
Standardized URL query parameters such as utm_source and utm_campaign to assign sessions or events to campaigns.
- UX & accessibility
UX Writing for Uncertainty
Notes distinguish confirmed results, estimates, and missing information. They offer appropriate next steps instead of vague warnings or unfounded certainty.
- UX & accessibility
UX/UI
UX encompasses perceptions and experiences before, during, and after using a product. UI is the interface with its controls and states. Good design combines understandable processes with an appropriate, accessible interface; it is not limited to appealing visuals.
- AI fundamentals & models
Validation Set
It is separated from the actual training data and used for decisions about models or settings. An independent test dataset remains reserved for the final evaluation.
- Web design & development
Validator
For example, it controls mandatory fields, data types, value ranges, or relationships between entries. In case of violations, it returns errors instead of passing on unchecked data for further processing.
- AI fundamentals & models
Value Alignment
The research problem of making AI behavior compatible with plural human goals, norms, and legitimate boundaries.
- AI fundamentals & models
Value of Information
The expected benefit of additional information before a decision compared to its acquisition and delay costs.
- Application & enablement
Value-Based Pricing
A pricing strategy that weighs willingness to pay and perceived customer benefit more than cost markup or competitor price.
- AI fundamentals & models
Vanishing Gradients
The strong decrease of backpropagated derivatives over many layers or time steps, causing early parameters to learn very little.
- Process automation
Vapi
The platform connects voice input, model responses, voice output, and telephony or other voice channels. Interfaces integrate operational systems into conversations.
- AI fundamentals & models
Variance
Deviations are squared and averaged according to the chosen calculation rule. Its unit is therefore the square of the original unit; the standard deviation is the square root of the variance.
- AI fundamentals & models
Variational Autoencoder (VAE)
A latent generative model that learns an approximate posterior and optimizes reconstruction with regularization towards the prior distribution.
- Data integration & search
Vector Database
A vector database manages embeddings and searches for similar entries based on a suitable distance or similarity measure. Additional metadata can enable filters. Speed and hit quality depend on data, index and configuration.
- Data integration & search
Vector Index
A data structure that accelerates similarity search over vectors, often approximately with methods like HNSW or IVF.
- Data integration & search
Vector Quantization
The mapping of continuous vectors to finite codebook vectors for compression or discrete representation.
- Data integration & search
Vector Search
The search for vectors that are closest to a query according to a defined distance or similarity measure.
- Data integration & search
Vector Similarity
For embeddings, cosine similarity can compare the direction of vectors; other metrics consider, for example, distance or dot product. The meaning depends on the model and the task.
- Data integration & search
Vector Store
A storage and abstraction component for vectors, IDs, and metadata with similarity retrieval; it can be part of a vector database.
- Data integration & search
Vector Store Hygiene
An operational process for updating, deleting, deduplicating, and monitoring vectors, metadata, and source references.
- AI fundamentals & models
Vector-Quantized Variational Autoencoder (VQ-VAE)
A generative autoencoder model with a discrete codebook that quantizes encoder outputs to learned vectors.
- Application & enablement
Vendor Risk Management
The systematic assessment, treatment, and monitoring of risks from external providers throughout selection, contracting, operation, and termination.
- Web design & development
Veo 3
The model generates moving scenes from prompts and can add suitable sounds or speech. Veo 3 is to be distinguished from later variants such as Veo 3.1.
- Web design & development
Vercel
It connects releases from development projects with hosting, functions, and other infrastructure components. Preview versions help to review changes before production release.
- Web design & development
Verification Layer
A verification layer checks outputs against defined rules, reference data, or tests. It can reject invalid results, forward them for reprocessing, or request a human decision.
- Process automation
Verification-Centric Agents
Result verification here is part of the agent's workflow: The agent produces a result, checks defined success criteria, and corrects errors or reports open issues. The term describes an architectural principle, not a uniform agent class.
- Application & enablement
Verification-First Policy
A Verification-First Policy is an operational rule according to which relevant results are only considered completed after appropriate controls. It defines what is checked, who approves, and how exceptions are handled.
- Web design & development
Version Control
A system for traceable storage, branching, and merging of changes to files and artifacts.
- Application & enablement
Versioned Prompt
A prompt artifact with a unique version, change history, dependencies, and associated evaluation results.
- Web design & development
Vertex AI
Vertex AI bundled Google Cloud tools for building, training, and deploying models. Google continues this platform under Gemini Enterprise Agent Platform. The official naming table assigns former product and feature names to their new designations.
- Web design & development
Vibe Coding
In the approach described by Karpathy, programs are created primarily via natural language instructions and repeated trials; the code generated is not necessarily understood in detail. This can be distinguished from systematic AI-supported development with code review.
- Visibility & SEO
Vibe Marketing
Vibe Marketing is an informal collective term for marketing work where people describe goals and direction and AI prepares texts, images, or processes. The term has no binding methodology and is used differently.
- AI fundamentals & models
Vibe-Check Evaluation
In a Vibe-Check evaluation, people judge some outputs based on their immediate impression. This can reveal obvious problems early but does not provide reliable quality measurement without criteria and systematic selection.
- AI fundamentals & models
Video AI
AI for video processing includes scene recognition, object tracking, transcription, and automatic editing. Depending on the task, the system processes image sequences, audio, or both.
- AI fundamentals & models
Video Language Model
Such models process video clips together with text to, for example, answer questions or describe events. Temporal order and long videos pose additional challenges compared to analyzing individual images.
- Visibility & SEO
Viewability
A delivered ad is not automatically visible. Viewability evaluates, based on defined area and time criteria, whether an ad impression offered an opportunity for perception.
- Visibility & SEO
Virtual Influencers
Virtual influencers are digitally designed and used for contributions or brand communication. AI can help; however, a virtual figure is not automatically AI-driven or autonomous.
- Web design & development
Virtual Try-On
The image- or 3D-based simulation of how clothing, accessories, or products appear on a person or in their environment.
- Web design & development
Vision APIs
A Vision API accepts images or image sections and delivers structured results, such as recognized texts, objects, or positions. Cloud services and local frameworks differ in functionality and data processing.
- AI fundamentals & models
Vision Transformer
An image model architecture that processes image patches as a token sequence with a Transformer.
- AI fundamentals & models
Vision-Language Model
A model that jointly represents visual and linguistic inputs or uses them for tasks such as retrieval, description, and question answering.
- AI fundamentals & models
Visual Question Answering (VQA)
A multimodal task where a system answers a natural language question about an image.
- Web design & development
vLLM
It organizes model inference and incoming requests, including efficient memory management and shared processing. Available features depend on model, hardware, and version.
- AI fundamentals & models
Vocoder
A system that converts acoustic features or latent representations into an audio signal or analyzes and resynthesizes speech.
- AI fundamentals & models
Voice Activity Detection
The temporal detection of segments with speech versus silence or other noises in an audio signal.
- Process automation
Voice Agent
A dialogue-oriented system that processes speech input, determines actions or responses, and generates speech or other outputs.
- Application & enablement
Voice Clone Consent
Consent to voice cloning means informed approval for creating and using an artificial voice clone. Data protection law, personality rights, contracts, and provider terms must be considered separately.
- AI fundamentals & models
Voice Cloning
The generation of synthetic speech with characteristic features of a target voice from reference recordings.
- Visibility & SEO
Voice Search
Voice search records a spoken query, transcribes or interprets it, and delivers matching search results or actions.
- Visibility & SEO
Walled Garden
A platform environment in which the operator largely controls access, data, measurement, distribution, or interoperability.
- AI fundamentals & models
Warm Start
The initialization of an optimization or training run with parameters or state from a previous run instead of from scratch.
- AI fundamentals & models
Warmup
A training phase in which the learning rate is gradually increased from a small value to the target value.
- AI fundamentals & models
Watermarking
Embedding robust or fragile signals in media or data to support origin, integrity, or attribution.
- AI fundamentals & models
wav2vec
A model family that learns representations directly from audio; wav2vec 2.0 combines masked latent units with contrastive learning.
- AI fundamentals & models
Weak Supervision
Generating probabilistic training labels from multiple incomplete or conflicting sources such as rules, heuristics, and external models.
- AI fundamentals & models
Weakly Supervised Learning
A generic term for learning with incomplete, inaccurate, or only coarsely assigned labels.
- Data integration & search
Weaviate
Objects are organized with properties and vector representations. Search queries can combine semantic proximity with filters or lexical search methods.
- Web design & development
Weavy
The platform connects AI models and editing tools in one interface with linked nodes. This allows multiple media processing steps to be combined as a reusable workflow.
- Visibility & SEO
Web Analytics
Web analysis collects and examines suitable usage data of a website, such as page views and defined events. Objectives are, for example, the understanding of visit routes and the examination of whether desired actions take place. The significance and scope depend on the measurement concept and data collection.
- Web design & development
Web Application Firewall (WAF)
A security component that monitors HTTP traffic based on rules or models and can block or log unwanted requests.
- Process automation
Web Browsing Tool
An agent component that makes web content accessible via search, HTTP retrieval, or browser control.
- UX & accessibility
Web Content Accessibility Guidelines (WCAG)
The Web Content Accessibility Guidelines (WCAG) define technical and design criteria at conformity levels A, AA and AAA. They are based on four principles: perceptibility, usability, comprehensibility and robustness.
- Data integration & search
Web Grounding
Supporting an answer with currently retrieved web sources with traceable assignment of statements to evidence.
- Web design & development
Web Scraping
The automated retrieval and structured extraction of information from websites or web interfaces.
- Web design & development
WebAssembly (Wasm)
A portable binary instruction format and execution model for a stack-based virtual machine.
- Web design & development
Webflow
Webflow combines visual website design with content management, shop functions, and hosting. The designer translates design decisions into HTML, CSS, and JavaScript. Recurring content is managed structurally in the CMS. Available features depend on which platform components are used.
- Web design & development
Webhook
An event-driven HTTP call with which one system informs another about an event.
- Web design & development
Webhook Verification
The verification of authenticity, integrity, and freshness of a webhook request before processing it.
- Web design & development
WebP and AVIF
WebP and AVIF are image formats for compressed web images. Both offer different quality and file-size options depending on the method used. The appropriate format depends on the image content, encoder, desired properties and browser support.
- AI fundamentals & models
Weight Decay
A regularization that shrinks parameters toward zero during optimization steps; with SGD it often corresponds to L2 regularization.
- AI fundamentals & models
Weight Initialization
Initialization methods set start values and often take layer size and activation into account. Xavier/Glorot and He initialization should support appropriate scaling of signals and gradients. They do not prevent unstable training in all conditions.
- AI fundamentals & models
Weight Normalization
A reparameterization that decomposes a weight vector into direction and learnable length.
- Web design & development
Weights & Biases
Teams document training runs, parameters, metrics, and artifacts to enable result comparison. The official documentation now introduces the CoreWeave Forge offering.
- AI fundamentals & models
What-If Analysis
The targeted alteration of inputs or assumptions to examine impacts on model results.
- AI fundamentals & models
Whisper
A family of multilingual encoder-decoder models for speech recognition and speech translation published by OpenAI.
- UX & accessibility
Whitespace
Whitespace is empty space between and around design elements, regardless of its colour. It can support grouping, hierarchy and readability; its usefulness depends on the layout and context.
- Web design & development
Windsurf
The editor combines source code editing with AI support for understanding and modifying projects. Agentic functions can perform multiple steps beyond individual code suggestions.
- AI fundamentals & models
WinoGrande
A benchmark for pronoun-based commonsense questions, designed to reduce simple dataset artifacts.
- UX & accessibility
Wireframe
According to Nielsen Norman Group, a wireframe is a common UX deliverable. It shows layout ideas at the page level and is suitable for websites and apps with few, mostly static pages; it is less suitable for highly dynamic processes.
- AI fundamentals & models
Word Embedding
A dense vector representation of words or word forms learned from their usage context.
- AI fundamentals & models
Word Error Rate (WER)
An ASR metric: sum of substitutions, deletions, and insertions divided by the word count of the reference.
- AI fundamentals & models
word2vec
A model and tool collection for learning dense word vectors with CBOW or Skip-gram from local contexts.
- AI fundamentals & models
WordPiece
A subword tokenization method that uses a limited vocabulary of frequent character sequences and segments words.
- Process automation
Workflow Automation
The technical execution of defined work steps, rules, and handovers with as little manual operation as possible.
- Process automation
Workflow Orchestration
The coordinated control of multiple tasks, services, or workflows based on dependencies, state, schedule, and error rules.
- Web design & development
Workload Isolation
The separation of resources, rights, and error impacts between workloads on shared infrastructure.
- Application & enablement
Workslop
A new, non-standardized term for superficially plausible AI outputs whose error checking and rework are shifted onto others.
- AI fundamentals & models
World Model
A learned model of the dynamics or observations of an environment that supports prediction and planning.
- Web design & development
Write Amplification
The ratio of actually internally written data to the logically written data volume by the application system.
- Web design & development
Write-Back Cache
A cache that initially acknowledges write accesses and later asynchronously transfers them to the downstream storage.
- Web design & development
Write-Through Cache
A cache that treats a write operation as complete only after passing it on to both cache and downstream storage.
- Web design & development
X-Content-Type-Options
An HTTP response header that instructs browsers not to reinterpret declared MIME types through content sniffing.
- Web design & development
X-Forwarded-For
A de-facto HTTP header in which proxies can forward the original client IP and proxy chain.
- Web design & development
X-Frame-Options
An HTTP response header that restricts embedding a page in frames with DENY or SAMEORIGIN.
- Visibility & SEO
X-Robots-Tag
An HTTP response header that transmits instructions like noindex or noarchive to search engines for arbitrary resources.
- AI fundamentals & models
x-Vector
A fixed vector representation of a speech segment for modeling speaker characteristics, learned with a neural classifier.
- Web design & development
X.509 Certificate
A signed data structure that binds a public key with identity and validity information.
- AI fundamentals & models
Xavier Initialization
A weight initialization whose variance is adapted to the input and output dimensions of a layer.
- AI fundamentals & models
XGBoost
An optimized library and learning method for gradient-boosted decision trees with regularization and sparsity-aware processing.
- AI fundamentals & models
XLNet
A transformer language model with a permutation objective that aims to learn bidirectional context without classic mask tokens.
- AI fundamentals & models
xLSTM
An LSTM family proposed in 2024 featuring exponential gating as well as scalar or matrix memories.
- Web design & development
XML External Entity (XXE)
An XML parser vulnerability where external entities access unwanted files, network resources, or excessive resources.
- Visibility & SEO
XML Sitemap
An XML file following the sitemap protocol that reports relevant URLs and optional metadata to search engines.
- Web design & development
XOR Cipher
A cipher that combines data bitwise with a keystream using exclusive or.
- AI fundamentals & models
XOR Problem
A classic classification problem whose positive and negative points cannot be separated by a single linear boundary.
- Web design & development
XPath
A W3C expression language for selecting and computing nodes and values in XML trees.
- Web design & development
XQuery
A W3C query and transformation language for XML data that extends XPath expressions.
- Web design & development
XSLT
A W3C language for rule-based transformation of XML trees into XML, HTML, text, or other structures.
- Web design & development
XSS in AI-Generated Content
An XSS execution path where AI-generated or rendered Markdown/HTML content reaches a web interface without secure sanitization.
- Web design & development
XSS Payload
The input contains text, markup, URL components, or script fragments that a browser can interpret as active code in a specific output context. It is a test or attack artifact, not a vulnerability class itself. Its effectiveness depends, among other things, on embedding context, parser, DOM sink, and browser policies.
- Web design & development
xUnit
A family of automated test frameworks with test cases, assertions, fixtures, and runners, derived from SUnit.
- Application & enablement
Y Combinator
The program combines funding, consulting, and exchange for selected startups. It is not an AI technology but is often mentioned in the context of new software and AI companies.
- Web design & development
YAML
A human-readable serialization language for structured data with mappings, sequences, and scalars.
- Web design & development
YAML Anchors and Aliases
YAML mechanisms that mark a node and reference it elsewhere as an alias; merge keys are a separate extension.
- Web design & development
YAML Front Matter
A YAML block delimited by markers at the start of a document, which certain generators interpret as metadata.
- Web design & development
YAML Injection
A vulnerability class where untrusted data influences YAML structure or parser objects to bypass rules or trigger dangerous deserialization.
- Web design & development
YAML Schema Validation
The validation of parsed YAML data against an explicit schema for types, required fields, value ranges, and additional fields.
- Web design & development
YARA Rule
A declarative rule that classifies files, memory, or other data based on strings and boolean conditions.
- Web design & development
Yarn
The package manager installs and updates libraries based on project specifications. A lock file records resolved versions; workspaces support projects with multiple packages.
- Visibility & SEO
Year over Year (YoY)
The comparison of a metric with the same period of the previous year, absolute or percentage.
- Visibility & SEO
Year to Date (YTD)
The cumulative period from the start of a defined fiscal or calendar year up to a cutoff date.
- Web design & development
yield in Generator Functions
A language construct that pauses a generator function, yields a value, and preserves its local state for continuation.
- Visibility & SEO
Yield Management
The dynamic control of price and availability of limited, perishable capacity based on expected demand.
- Visibility & SEO
Yield Optimization
Demand sources, prices, and delivery rules are coordinated so that advertising spaces achieve the most suitable revenue possible. Contractual commitments, user experience, and ad quality set limits.
- Application & enablement
Yield Rate
Which outputs count as usable must be defined for the process. For First Pass Yield, especially the proportion that meets requirements without rework in the first pass counts.
- Web design & development
Yottabyte
A decimal data quantity of 10^24 bytes according to the International System of Units.
- AI fundamentals & models
You Only Look Once (YOLO)
A family of single-stage object detectors that predict classes and bounding boxes in a single forward pass.
- Web design & development
z-index
A CSS property that influences the stacking order of positioned or suitable elements within stacking contexts.
- AI fundamentals & models
Z-Order Curve
A space-filling mapping that converts multidimensional coordinates into a one-dimensional order by bit interleaving.
- AI fundamentals & models
Z-Ordering in Data Storage
A physical clustering technique that arranges data based on a Z-order key to improve skipping irrelevant data blocks.
- AI fundamentals & models
z-Score
The standardized deviation of a value from the mean in units of the standard deviation.
- AI fundamentals & models
z-Test
A hypothesis test whose standardized test statistic follows or is sufficiently approximated by a standard normal distribution under the null hypothesis.
- Data integration & search
Zep
Data is organized so that agents can retrieve relevant information. Time-related relationships help represent changes in facts and connections.
- AI fundamentals & models
Zero Redundancy Optimizer (ZeRO)
A family of techniques that partitions optimizer states, gradients, and parameters across data-parallel processes to reduce memory redundancy.
- Web design & development
Zero Trust
A security model without implicit trust due to network location; every access is risk-based authenticated, authorized, and continuously evaluated.
- Web design & development
Zero Trust Network Access (ZTNA)
Zero Trust Network Access is a non-uniformly standardized market term for access solutions that provide access to explicitly authorized applications or resources. ZTNA can implement zero-trust principles but is neither equivalent to Zero Trust nor a complete zero-trust architecture.
- Visibility & SEO
Zero-Click Brand Equity
Brand value without click is an informal term for possible brand effects of such visibility contacts. However, a mention does not prove either positive perception or a measurable increase in brand value.
- Visibility & SEO
Zero-Click Content
A non-standardized marketing term for content whose essential message can be consumed directly on a platform without an external click.
- Visibility & SEO
Zero-Click Search
A search session in which no measurable click from the results page to an external result occurs.
- Web design & development
Zero-Day Vulnerability
A vulnerability for which those affected had no effective fix available at the time of exploitation or disclosure.
- Data integration & search
Zero-ETL
A non-standard vendor term for integrations that largely replace manually operated ETL pipelines with managed replication or shared data access.
- AI fundamentals & models
Zero-Knowledge Machine Learning (zkML)
A young application field that cryptographically proves that an ML inference or computation was executed according to a specified model without revealing all inputs.
- Web design & development
Zero-Knowledge Proof
A cryptographic protocol by which a prover demonstrates knowledge of a statement without revealing any further information.
- Web design & development
Zero-Knowledge Scalable Transparent Argument of Knowledge (zk-STARK)
A scalable, transparent zero-knowledge argument system that typically uses hash functions instead of pairing-based trusted setups.
- Web design & development
Zero-Knowledge Succinct Non-Interactive Argument of Knowledge (zk-SNARK)
A short, non-interactive zero-knowledge proof of knowledge with efficient verification under cryptographic assumptions.
- Data integration & search
Zero-Party Data
A marketing term for data that customers consciously and directly communicate to an organization regarding preferences or intentions.
- AI fundamentals & models
Zero-Shot and Few-Shot Methods
In prompting, zero-shot means a task description without demonstrations; few-shot adds a few examples in context. This does not automatically change model weights.
- AI fundamentals & models
Zero-Shot Classification
The assignment of inputs to classes for which the used model has received no task-specific labeled training examples, often via class descriptions.
- AI fundamentals & models
Zero-Shot Learning
A learning setting in which a model recognizes instances of previously unseen classes using side information such as attributes or semantic descriptions.
- Application & enablement
Zero-Shot Prompting
The instruction to a pretrained model to perform a task based only on an instruction without demonstration examples in the prompt.
- Web design & development
Zettabyte
A decimal data quantity of 10^21 bytes according to the International System of Units.
- Application & enablement
Zettelkasten
A knowledge method consisting of atomic, addressable notes that form a growing network of concepts through links and original formulations.
- AI fundamentals & models
Zipf's Law
An empirical rule according to which word frequency is approximately inversely proportional to its rank.
- Web design & development
Zipping Artifacts
A ZIP archive can contain multiple files including directory structure. Compression reduces suitable content losslessly; already compressed images or videos often shrink little.
- Web design & development
Zod
The TypeScript-oriented library describes allowed data structures and validates incoming values. Additionally, TypeScript types can be derived from this.
- Web design & development
Zombie Process
A terminated Unix process whose exit status remains in the process table because the parent process has not collected it with wait.
- Web design & development
Zonal and Regional Cloud Services
Cloud services whose resources and failure domain are bound to a single zone or multiple zones within a region.
- Web design & development
Zone Redundancy
The redundant distribution of a resource or its replicas across multiple separate availability zones within a region.
- Web design & development
Zstandard
Zstandard compresses data losslessly in a defined frame format. Compression level, library parameters, and an optional dictionary influence runtime, memory usage, and compression ratio. It is especially suitable when speed and flexible tuning are more important than a single fixed profile.
Room for knowledge
Missing a term?
Tell me which term you would like explained clearly.