Lesson 1158 of 1570
Context Window Discipline: What Fits in AI's Memory
Pasting a 50-page document plus your question often gets a worse answer than pasting just the relevant 2 pages.
Lesson map
What this lesson covers
Learning path
The main moves in order
- 1The big idea
- 2context window
- 3relevance
- 4context curation
Concept cluster
Terms to connect while reading
Section 1
The big idea
Models have a 'context window' — how much text they can read at once. Modern Claude and GPT can hold 200k+ tokens, but performance degrades the more junk is in there. Curating the context (only what matters) beats dumping everything in.
Some examples
- You paste a whole textbook to ask one question — Claude misses the answer that was on page 4.
- You paste only chapters 3 and 4 — Claude nails it.
- You include 10 PDFs of documentation — the model gets confused which one matters.
- You include only the 1 PDF that's relevant + a clear question — the model is sharp.
Try it!
Take a long document and a question about it. First, ask with the whole doc. Then, ask with only the section you think is relevant. Compare quality.
Key terms in this lesson
Key terms in this lesson
End-of-lesson quiz
Check what stuck
15 questions · Score saves to your progress.
Tutor
Curious about “Context Window Discipline: What Fits in AI's Memory”?
Ask anything about this lesson. I’ll answer using just what you’re reading — short, friendly, grounded.
Progress saved locally in this browser. Sign in to sync across devices.
Related lessons
Keep going
Creators · 40 min
Context Window Budgeting: What to Include, What to Cut
Long context windows tempt teams to dump everything in. Smart prompting means choosing what context actually helps — and ruthlessly cutting what doesn't.
Builders · 40 min
Negative Prompting and Constraints: Tell AI What to Skip
Sometimes the fastest way to get a good AI answer is to list what you don't want.
Creators · 40 min
Prompt Evaluation and Testing: From Vibes to Rigorous Evals, Part 1
Prompt iteration without measurement is guessing. A real evaluation harness lets you compare prompt variants on real traffic — surfacing regressions before users see them.