AI Glossary: The Words Behind Every AI Tool You Open

Q: Which AI terms should I learn first?

Start with the 8 at the top of the page: [artificial intelligence](/ai-glossary/ai), [large language model](/ai-glossary/large-language-model), [model](/ai-glossary/model), [token](/ai-glossary/token), [prompt](/ai-glossary/prompt), [context window](/ai-glossary/context-window), [agent](/ai-glossary/agent) and [hallucination](/ai-glossary/hallucination). Most other terms build on them.

Q: Can I see how the AI terms connect?

Yes. [aiglossary.io](https://aiglossary.io/?utm_source=aios.guide&utm_medium=glossary&utm_campaign=ai-os) shows the terms on an interactive map, so you can see which ideas depend on each other and follow them from one term to the next.

Q: What is the difference between a model and an agent?

A [model](/ai-glossary/model) turns your input into an output, such as text. An [agent](/ai-glossary/agent) is a model that has been given tools, instructions and a job, so it can take actions as well as answer.

AI glossary

Start here

Artificial intelligence (AI)

Artificial intelligence (AI) is technology that lets computers do tasks that normally need human intelligence, such as understanding language, recognizing images, solving problems and making decisions.

Large language model (LLM)

A large language model is a program trained on enormous amounts of text that predicts what word comes next.

Model

A model is the trained thing itself: a very large set of numbers that turns your input into its output.

Token

A token is a chunk of text, roughly three quarters of a word, and it is the unit everything gets counted in.

Prompt

A prompt is what you send. The instruction, the question, the context, all of it together.

Context window

The context window is the limit on how much the model can hold at once.

Agent

An agent is a model that has been given tools, instructions and a job, so it can act rather than only answer.

Hallucination

A hallucination is the model stating something false with complete confidence.

A

AFK

AFK is starting the work and walking away, letting the agent run unattended.

Agent

An agent is a model that has been given tools, instructions and a job, so it can act rather than only answer.

AGENTS.md

AGENTS.md is a file the agent reads at the start of every session, whatever you are working on.

API key

An API key is the password that lets a program use a service on your behalf.

Artificial intelligence (AI)

Attention budget

Attention budget is the fixed amount of focus spread across everything in the session, however much you add.

Attention degradation

Attention degradation is a session getting worse as it gets longer, because the focus is spread thinner.

Automation

Automation is a workflow that runs without you starting it.

C

Cache tokens

Cache tokens are the part of your input that has already been processed and does not have to be paid for again in full.

Clearing

Clearing is ending the session and starting fresh, with nothing carried over.

CLI

A CLI is a command you run in a terminal, and it costs nothing until you run it.

Compaction

Compaction is when a long session gets summarized so it can keep going in a smaller space.

Computer use

Computer use is an agent operating a screen the way a person does: looking, clicking, typing.

Context

Context is everything the model can see when it answers.

Context engineering

Context engineering is deciding what the agent should already know, and making sure it is loaded before you ask.

Context window

The context window is the limit on how much the model can hold at once.

Contextual knowledge

Contextual knowledge is what the agent can read right now, as opposed to what it half-remembers from training.

E

Effort

Effort is how much thinking the model does before answering.

Embedding

An embedding turns a piece of text into numbers, so a computer can find other text that means something similar.

Evals

Evals are a set of test cases you run to check whether a change actually made things better.

F

Filesystem

The filesystem is the folders and files on your computer, and it is where an agent actually works.

Fine-tuning

Fine-tuning is further training on your own examples, so the model behaves differently by default.

G

Grounding

Grounding is tying an answer to a source it can point at, instead of one it produced from memory.

Guardrails

Guardrails are the rules that stop an agent doing certain things, whatever it has been asked.

H

Hallucination

A hallucination is the model stating something false with complete confidence.

Handoff

A handoff is passing what one session knows to the next one, because nothing carries over on its own.

Harness

The harness is everything wrapped around the model that turns it from a text generator into something that can act.

Hook

A hook is something that runs automatically on an event, whether or not the agent decides to.

Human in the loop

Human in the loop means you stay in the session, reviewing and redirecting as it goes.

I

Inference

Inference is the model running: reading your input and generating a response.

Input tokens

Input tokens are everything you send: your message, your files, your rules, and the whole conversation so far.

K

Knowledge cutoff

The knowledge cutoff is the date the model stopped learning. It knows nothing after it.

L

Large language model (LLM)

A large language model is a program trained on enormous amounts of text that predicts what word comes next.

M

Memory system

A memory system is what you build so an agent knows things next session that it learned in this one.

Model

A model is the trained thing itself: a very large set of numbers that turns your input into its output.

Model Context Protocol (MCP)

MCP is a standard way to give an agent tools for a service, so the same connector works in any agent that speaks it.

Model provider

The model provider is whoever runs the model for you: Anthropic, OpenAI, Google, or your own machine.

Multimodal

Multimodal means it handles more than text: images, audio, video, screenshots.

N

Next-token prediction

Next-token prediction is the only thing a model does: pick the next chunk of text, add it, and go again.

Non-determinism

Non-determinism means the same question can get a different answer twice.

O

Orchestrator

An orchestrator is what decides which agent does what, in what order, and what happens with the results.

Output tokens

Output tokens are what the model writes back, and they are the expensive half.

P

Parameters

Parameters are the numbers inside a model that got set during training. Everything it knows lives in them.

Parametric knowledge

Parametric knowledge is what the model knows from training, frozen at the moment training stopped.

Permission

A permission is the check before an agent does something you have not already approved.

Progressive disclosure

Progressive disclosure is loading only what is needed right now, and leaving a pointer to the rest.

Prompt

A prompt is what you send. The instruction, the question, the context, all of it together.

Prompt engineering

Prompt engineering is wording the request better to get a better result.

Prompt injection

Prompt injection is when text the agent reads contains instructions, and it follows them.

R

Rate limit

A rate limit is the ceiling on how much you can use in a window, before you have to wait.

Reasoning model

A reasoning model works through a problem before answering, rather than answering straight away.

Retrieval augmented generation (RAG)

RAG is looking something up first, then answering from what was found rather than from memory.

S

Sandbox

A sandbox is a walled-off place an agent runs in, so a mistake cannot reach anything that matters.

Session

A session is one run of working with an agent, from opening it to closing it.

Skill

A skill is a written procedure for a job you do the same way every time. Your SOP, in a form the agent follows instead of improvising.

Smart zone

The smart zone is the early part of a session, when it is sharp, before it drifts.

Stateful

Stateful means it carries something forward. A session is stateful across turns, and an agent can be made stateful across sessions.

Stateless

Stateless means nothing carries forward. The model is stateless: it remembers nothing between requests.

Subagent

A subagent is a helper an agent hands a job to, which works separately and reports back one answer.

Sycophancy

Sycophancy is the model agreeing with you because agreement is what people rewarded, not because you are right.

System prompt

The system prompt is the standing instruction sent ahead of everything else, on every single request.

T

Temperature

Temperature is how much the model varies its choices. Low is predictable, high is loose.

Token

A token is a chunk of text, roughly three quarters of a word, and it is the unit everything gets counted in.

Tool

A tool is something an agent can use: read a file, run a command, search the web, send a request.

Tool call

A tool call is the agent asking to do something: read this file, run this command, search for this.

Tool result

A tool result is what comes back after the agent does something: the file contents, the output, or the error.

Training

Training is where a model learns, once, by reading enormous amounts of text and adjusting itself to predict better.

Turn

A turn is one message from you plus everything the agent does before it hands back.

V

Vector database

A vector database stores embeddings so you can search by meaning across a lot of documents, fast.

Vibe coding

Vibe coding is accepting what the agent produced without reading it.

W

Workflow

A workflow is a job with steps that runs the same way every time.

FAQs about the AI glossary

Here are the questions people ask most about AI terms, with a direct answer to each.

What is an AI glossary?

An AI glossary is a list of the terms used in artificial intelligence, each with a plain-English definition. This one covers the words you meet when you work with AI agents, from tokens and context windows to skills, hooks and MCP.

Who is this AI glossary for?

It is written for creators, founders and entrepreneurs who use AI in their work and want each term explained without jargon.

Which AI terms should I learn first?

Start with the 8 at the top of the page: artificial intelligence, large language model, model, token, prompt, context window, agent and hallucination. Most other terms build on them.

How is this AI glossary organized?

The terms are listed A to Z, with a letter bar that stays at the top of the page so you can jump to any letter. The terms to learn first sit above the list, under Start here.

Can I see how the AI terms connect?

Yes. aiglossary.io shows the terms on an interactive map, so you can see which ideas depend on each other and follow them from one term to the next.

What is the difference between AI, machine learning and deep learning?

Each one sits inside the one before it. AI is the field, machine learning is the main way AI is built today, and deep learning is machine learning built on layered neural networks. Generative AI, like ChatGPT and Claude, sits inside deep learning. Read the full explanation.

What is the difference between a model and an agent?

A model turns your input into an output, such as text. An agent is a model that has been given tools, instructions and a job, so it can take actions as well as answer.

What is an AI OS?

An AI OS, or AI operating system, is your context, skills, memory and tools set up once, so every AI agent you use knows your business from the first message. The how to build an AI OS guide shows you how to set one up.

Is this AI glossary free?

Yes. Every term and every definition on this page is free to read.

The AI glossaryfor creators and founders

Start here

Artificial intelligence (AI)

Large language model (LLM)

Model

Token

Prompt

Context window

Agent

Hallucination

A

AFK

Agent

AGENTS.md

API key

Artificial intelligence (AI)

Attention budget

Attention degradation

Automation

C

Cache tokens

Clearing

CLI

Compaction

Computer use

Context

Context engineering

Context window

Contextual knowledge

E

Effort

Embedding

Evals

F

Filesystem

Fine-tuning

G

Grounding

Guardrails

H

Hallucination

Handoff

Harness

Hook

Human in the loop

I

Inference

Input tokens

K

Knowledge cutoff

L

Large language model (LLM)

M

Memory system

Model

Model Context Protocol (MCP)

Model provider

Multimodal

N

Next-token prediction

Non-determinism

O

Orchestrator

Output tokens

P

Parameters

Parametric knowledge

Permission

Progressive disclosure

Prompt

Prompt engineering

Prompt injection

R

Rate limit

Reasoning model

Retrieval augmented generation (RAG)

S

Sandbox

Session

Skill