Home Learn Class 9
Classes 4–12 · Digital literacy and AI

What is a context window? The desk the AI works on

Think of a desk. Whatever is on the desk, the AI can use. Whatever is not on the desk does not exist to it — not what you said an hour ago, not the file you meant to attach. The desk has a size, and this is called the <b>context window</b>.

Only what is on the desk exists

The model has no memory outside the conversation in front of it. Everything it uses to answer must be in the window at that moment.

This is why it forgets what you told it earlier in a very long chat, and why it cannot use your notes unless you paste them in.

The part nobody warns you about

When the desk fills up, the oldest sheet slides onto the floor. The model does not stop. It does not say "chapter 1 is missing". It answers anyway, using what is left, and fills the gap with something plausible.

Most bad AI answers in real work are not clever failures. They are a page on the floor.

How to work with a small desk

Give it the relevant pages, not everything. A 200-page document plus the one table you care about often pushes the table off the desk.

Start a fresh chat for a new task instead of carrying a long history along.

If an answer suddenly gets vaguer, suspect the window before you suspect the model.

A four-sheet desk, five sheets given

Chapter 1 — the map is foundpushed off the desk
Chapter 2 — the journey beginson the desk
Chapter 3 — the stormon the desk
Chapter 4 — the islandon the desk
Chapter 5 — going homeon the desk
Asked: who found the map?"most likely the younger brother" — wrong

It did not say chapter 1 was missing. It said "most likely". And it was wrong.

Practice

You paste a long report and one small table, then ask about the table. The number is wrong. Why?
Show the answer
The report probably pushed the table out of the window.
Give it the table on its own. Everything else was competing for space it did not need.
Does a bigger context window fix everything?
Show the answer
No.
Bigger windows help, but attention still thins out over very long inputs, and the model still fills gaps rather than reporting them.
Why did it forget what you said twenty messages ago?
Show the answer
That message left the window.
Long conversations scroll out of the window from the oldest end. If it still matters, say it again.

Drop a page off the desk yourself

In our free course you load a four-sheet desk with five chapters, watch chapter 1 fall on the floor, then ask a question and see the answer quietly go wrong.

▶ Play Lemonade Empire free

Common questions

Is the context window the same as memory?
No. Some products add a memory feature on top, which quietly re-inserts saved notes into the window. The window itself still forgets everything the moment the conversation ends.
How big is a context window?
It varies by model and changes often, so any number here would be out of date quickly. What does not change is the behaviour when it fills: the oldest content goes, silently.

Next concepts