Stuff the context window

Paste in the full thread or document, learn what tokens and context windows are, and watch the answers improve.

Day 7 of Prompting Path, a 30-day AI course with one 10-minute lesson and one real task a day. Day 0 to Day 8 are free.

By Liam D, Founder and Principal Advisor at Futureformed · Updated 15 September 2026

The lesson

Your single biggest unlock as a prompter: paste in more context. Modern AI can read tens of thousands of words at once. Most people give it three sentences.

Things you can (and should) paste:

• The full email thread, not your summary of it.
• The whole document, not the bit you think is relevant.
• Your meeting notes, your brief, the prior version.

Why does this matter so much? Because AI has enormous knowledge and zero context. It knows everything published. It knows nothing about your client's politics, your team's history, or what your boss said in the corridor. You are the only source of that, so supply it.

A bit of jargon that pays for itself: tokens. AI doesn't read words, it reads tokens, chunks of roughly three-quarters of a word each ("unbelievable" might be un + believ + able). The context window is how many tokens a model can hold in its head at once: your prompt, everything you've pasted, and the whole conversation so far. Typical sizes right now: Claude holds around 200,000 tokens (roughly 150,000 words, a decent-sized novel), ChatGPT is in the same ballpark, and Gemini stretches to a million or more. The numbers keep growing; the point stays the same: there's far more room in there than most people ever use.

One caveat before you paste in your entire life: stack a very long chat with lots and lots of material and the AI can start to lose track of what came early on. People call it context drift or context rot. You'll only hit it at real volume, and the fix is simple: start a fresh chat and bring just the relevant bits with you.