Context window
A context window is the amount of text a model can consider at once; anything outside it, including past sessions, is invisible to the model unless re-supplied.
In detail
The context window is bounded and ephemeral: when a session ends, everything in it is gone. This is why teams re-brief their AI repeatedly. A memory layer works around the limit by storing context durably outside the window and retrieving only the relevant parts into each new session, instead of trying to cram everything back in.