Why not stuff it all in the prompt?

In chapter 1 the assistant invented a refund policy, because it had never read Larkspur’s real one. The team’s first idea: skip anything clever and paste the whole company drive into the prompt.

This chapter: Why does “just give it everything” break down, even with today’s huge context windows?

You will learn

How the size of real company data compares with what a context window holds
Why text that does fit still gets missed or goes stale
Why one shared prompt can’t respect who is allowed to see what

Words you’ll meet (4)

Corpus: All the documents the assistant could draw on: here, the company drive.
Context window: How much text a model can read in one request, counted in tokens (a token is about ¾ of a word).
Truncation: Cutting text off when it doesn’t fit. Whatever is cut, the model never sees.
Lost in the middle: Models tend to miss facts buried in the middle of a long prompt.

The picture on the left is live. Each step below asks you to do one thing in it; the story continues here as you go.

Step 1 of 7

Your team tried the obvious fix: paste the company drive into the assistant’s prompt. Ask it about enterprise refunds.