AI Vocab in 5 Minutes

You've talked to Claude, seen how it works, and watched it use tools. Now it's time to put names on what you've been doing.

This chapter gives you 8 terms for better understanding of AI.

Prompt

The message you send to AI.

Explain this report like I'm presenting it to a board.

That's a prompt. Everything else flows from it.

System Prompt

A hidden instruction the AI receives before your message. Written by whoever built the app — it controls tone, role, and rules.

You won't write these until later in the course, but every AI you use already has one.

Token

A small piece of text — roughly 4 characters in English. Both your input and the AI's output are measured in tokens. Tokens affect cost and fill up the context window.

Context Window

The total amount of text the AI can "see" at once, measured in tokens. For Claude, this is around 200,000 tokens.

When the context window fills up, older messages get dropped — the AI effectively forgets them.

Hallucination

A confident, fluent — but wrong — answer. Not a bug. A natural result of how language models work. Reduced by using tools, citations, and better prompts.

Tool Call

When Claude pauses mid-response, calls an external function — web search, a spreadsheet, a document — gets data, and continues answering. You saw this in the previous chapter.

Agent

A saved, pre-configured version of Claude with a system prompt and selected tools already set up. Instead of re-explaining context every time, you just launch it. You'll build one later in the course.

LLM — Large Language Model

• Large — billions of parameters;
• Language — works with text;
• Model — a statistical pattern, not a database.

This is what Claude is.

Quick Mental Model

You → Prompt → LLM → (Tool call if needed) → Response

Done when these terms look familiar and you can roughly explain each one.