The map of AI, and how people actually use it.

Start with six terms

• Context WindowThe hard ceiling on how much the model can see at once.
• AgentA model given tools, a goal and permission to keep going on its own.
• RAGLook things up, then put what you found in the prompt.
• HallucinationFluent output that is not true.
• EvaluationHow you know the system is actually working.
• Human in the LoopA person reviews or approves before the work lands.

Why this chart exists

Most confusion about AI is vocabulary confusion. Two people can mean different things by “agent”, or argue about cost without separating input tokens from output tokens. This is the shared sheet: short enough to read before you need the term, connected enough to answer what it sits next to, and specific about what people actually do with each idea.

Every entry is a plain definition, the line people really say, a worked example, an “avoid” note where a misunderstanding is worth flagging, and the primary sources behind it. The map is generated from those related-term links, so the network is the content. The sources flap lists every citation in one place.

How to use it