AI Glossary

Plain-English definitions for AI terms every builder should know. Updated weekly.

ABCDE F G H IJK L MNOP Q R S TUVWXYZ

Start Here

Agentic Loop

The iterative cycle in which an AI agent perceives its environment, selects an action, executes it, and observes the result before deciding on the next step toward a goal.

Read definition →

AI Agent

An autonomous AI system that perceives its environment, makes decisions, and takes actions to complete a goal. AI agents use LLMs as their reasoning engine and have access to tools (web search, code execution, APIs) to act on the world.

Read definition →

AI Alignment

The field of AI research focused on ensuring that AI systems pursue goals and exhibit behaviors that are consistent with human values and intentions.

Read definition →

AI Orchestration

The coordination of multiple AI models, agents, tools, and data sources within a workflow to complete complex, multi-step tasks autonomously.

Read definition →

Chain-of-Thought Prompting

A prompting technique that instructs a language model to articulate intermediate reasoning steps before arriving at a final answer, improving performance on complex tasks.

Read definition →

Context Window

The maximum amount of text (in tokens) that a language model can process in one prompt-response cycle. Both input (prompt) and output (completion) count against the limit.

Read definition →

A

Agentic LoopThe iterative cycle in which an AI agent perceives its environment, selects an action, executes it, and observes the result before deciding on the next step toward a goal.

AI AgentAn autonomous AI system that perceives its environment, makes decisions, and takes actions to complete a goal. AI agents use LLMs as their reasoning engine and have access to tools (web search, code execution, APIs) to act on the world.

AI AlignmentThe field of AI research focused on ensuring that AI systems pursue goals and exhibit behaviors that are consistent with human values and intentions.

AI OrchestrationThe coordination of multiple AI models, agents, tools, and data sources within a workflow to complete complex, multi-step tasks autonomously.

C

Chain-of-Thought PromptingA prompting technique that instructs a language model to articulate intermediate reasoning steps before arriving at a final answer, improving performance on complex tasks.

Context WindowThe maximum amount of text (in tokens) that a language model can process in one prompt-response cycle. Both input (prompt) and output (completion) count against the limit.

E

EmbeddingsNumerical vector representations of text, images, or other data that capture semantic meaning. Embeddings allow AI systems to measure similarity between concepts by comparing positions in high-dimensional vector space.

F

Few-Shot PromptingA prompting technique that includes a small number of input-output examples in the prompt to guide a model's behavior on a new task without updating model weights.

Fine-TuningFurther training a pre-trained LLM on a domain-specific dataset to improve performance on specialized tasks. Fine-tuning adjusts model weights to better match a target behavior or domain.

Foundation ModelA large-scale AI model pre-trained on broad datasets that can be adapted to a wide range of downstream tasks with minimal additional training.

G

GroundingThe process of connecting an AI model's outputs to verifiable, real-world information sources to reduce hallucination and improve factual accuracy.

GuardrailsConstraints, filters, or rules applied to an AI system to prevent it from producing harmful, unsafe, or off-policy outputs.

H

HallucinationWhen an AI language model generates confident-sounding information that is factually incorrect, fabricated, or nonsensical. Hallucinations are a fundamental risk in any AI system that generates text.

I

InferenceThe process of running a trained AI model on new input data to generate predictions or outputs, as distinct from the training process itself.

K

Knowledge GraphA structured database that represents real-world entities and the semantic relationships between them, used to enhance AI reasoning and information retrieval.

L

Large Language ModelA type of AI trained on massive text data that can understand, generate, and manipulate human language. LLMs are the foundation of Claude, ChatGPT, Gemini, and similar tools.

LatencyThe time elapsed between submitting a request to an AI model and receiving the first token or complete response, a key metric for production AI application performance.

M

Mixture of ExpertsA neural network architecture that routes each input to a subset of specialized sub-networks (experts), enabling very large model capacity at a fraction of the computational cost of dense models.

Model Context ProtocolAn open standard developed by Anthropic that defines how AI models like Claude communicate with external tools, data sources, and services. MCP enables AI assistants to read files, query databases, and call APIs in a standardized way.

Model DistillationA model compression technique where a smaller "student" model is trained to replicate the outputs and behavior of a larger, more capable "teacher" model.

Multi-Agent SystemAn AI architecture where multiple specialized AI agents collaborate to complete complex tasks. Each agent handles a specific subtask (research, writing, review) and passes results to the next agent in the pipeline.

Multimodal AIAn AI system capable of processing and generating content across multiple data types — such as text, images, audio, and video — within a single model.

P

Prompt EngineeringThe practice of designing and optimizing input prompts to reliably elicit desired outputs from an LLM. Effective prompts combine clear instructions, relevant context, output format specification, and examples.

Q

QuantizationA model compression technique that reduces the numerical precision of a model's weights, significantly decreasing memory requirements and inference cost with minimal loss in output quality.

R

Reasoning ModelA large language model explicitly trained to perform multi-step logical reasoning before producing a final answer, improving accuracy on complex tasks.

Retrieval-Augmented GenerationA technique that enhances LLM outputs by retrieving relevant information from an external knowledge base before generating a response, combining generative power with accurate, up-to-date data.

RLHFReinforcement Learning from Human Feedback — the training technique used to align language models with human preferences. Human raters score model outputs, and the model is trained to produce outputs that humans prefer.

S

Semantic SearchA search technique that retrieves results based on the meaning and intent of a query, rather than exact keyword matches.

Structured OutputA mode of AI model generation in which the output is constrained to a specific format — such as JSON or XML — ensuring machine-parseable, predictable responses.

Synthetic DataArtificially generated data that mimics the statistical properties of real-world data, used to train or evaluate AI models when real data is scarce, expensive, or privacy-sensitive.

T

TokenThe basic unit of text that language models process. Tokens are roughly 3-4 characters or 0.75 words in English. LLM pricing and context window limits are measured in tokens.

Tool CallingA capability that allows LLMs to invoke external functions, APIs, or services during a conversation. The model decides when to call a tool, formulates the input, and incorporates the tool's output into its response.

Transformer ArchitectureThe neural network architecture underlying all modern large language models. Introduced by Google in the 2017 paper "Attention Is All You Need," the Transformer uses self-attention mechanisms to process text in parallel rather than sequentially.

V

Vector DatabaseA specialized database designed to store, index, and query high-dimensional vector embeddings efficiently. Vector databases power semantic search, RAG pipelines, and similarity-based recommendation systems.

Z

Zero-Shot PromptingA prompting approach in which a language model is asked to complete a task without any task-specific examples in the prompt, relying entirely on knowledge from pre-training.

Stay in the loop

Weekly AI tool reviews, news digests, and how-to guides.