AI glossary

AI tools come with a lot of vocabulary. Most of it is simpler than it sounds. Here’s a short reference you can come back to whenever a term shows up in a doc or a settings page.

I’ve grouped the terms the way you’ll meet them in this course.

Models

AI (Artificial Intelligence): software that does tasks we used to think needed a human, like reasoning, learning, and solving problems.

LLM (Large Language Model): a model trained on a huge amount of text and code that can understand and generate language. GPT, Claude, and Gemini are LLMs.

Generative AI: any AI system that creates new content, such as text, code, images, or music, based on patterns it learned.

Training data: the data a model learned from. It has a cutoff date, so the model knows nothing newer than that.

Fine-tuning: training an existing model a little more on specific data, so it gets better at a narrow job.

Multimodal: a model that can work with more than text, for example images, audio, or video.

Frontier model: one of the most capable models available right now. The label moves as new models come out.

Open-weight model: a model whose weights you can download and run yourself, on your own hardware.

Talking to a model

Prompt: what you send to the model. Your request, plus any text you include with it.

System prompt: instructions set before the conversation starts that shape how the model behaves. Tools set one for you. Rules files feed into it.

Prompt engineering: writing prompts on purpose, so you get what you asked for instead of a vague answer.

Token: the unit a model reads and writes. Roughly three quarters of a word in English.

Context window: the maximum number of tokens the model can look at in one go. Instructions, conversation, files, tool results, and the answer all share it.

Context engineering: choosing what goes into the context window so the model has what it needs and nothing else.

Chain-of-thought: asking the model to reason step by step before it answers. Reasoning models do a version of this on their own.

Hallucination: a confident answer that’s wrong or made up. It sounds plausible, and that’s the problem.

RAG (Retrieval-Augmented Generation): fetching relevant documents first, then putting them in the context so the model answers from them instead of from memory.

Agents

Agent: a model that uses tools in a loop to reach a goal, instead of answering once and stopping.

Tool (function calling): an action the model can ask the host application to run, like reading a file, running a command, or calling an API.

MCP (Model Context Protocol): an open standard for connecting AI applications to external tools and data through servers.

Skill: a folder of Markdown instructions, and sometimes scripts, that teaches an agent how to do a specific kind of work.

Rules: durable project instructions the agent reads on every task. Often kept in an AGENTS.md file.

API key: a secret string that lets your software call a model provider directly, paying per token.

Don’t try to memorize this list. Read the course, and come back here when a word trips you up.