AI Made Crystal Clear · chapter 4: Talking to AI: Prompts, System Prompts, and the Context Window
Training data is a sealed library the model learned from once
2026-10-15
The context window is the desk it works at right now, and everything on the desk competes for the same fixed space.
Below: the paragraph from the book that builds this idea, then the diagram itself (Figure 4.1), and a recap. About a minute of reading.
And everything counts toward the limit: the hidden system prompt, every message you sent, every document you pasted, every reply the model wrote back. The window is measured in tokens, the word-pieces from chapter 3. As of this writing, mainstream models hold from a couple of hundred thousand tokens (roughly 150,000 words, about two novels) up to around a million for the largest windows. That sounds infinite. A long working session with pasted reports gets through it faster than you would think. Figure 4.1 shows the whole arrangement as a desk.
Recap
- The idea: The context window is the desk it works at right now, and everything on the desk competes for the same fixed space.
- The picture: Figure 4.1, from chapter 4 ("Talking to AI: Prompts, System Prompts, and the Context Window") of AI Made Crystal Clear.
- Go deeper: the chapter builds this step by step, with recipes and sources at the end.
This diagram is one of many in AI Made Crystal Clear.
Every chapter opens with the gist, draws the hard ideas, and ends with recipes and sources.
Get the book
