Issue 15, Oct 4, 2026
The Pixel-Memory Paradox
Senior AI Engineer interview at Google DeepMind, and the interviewer asks:
“Our computer-use agent appends every screenshot and action to its context. It’s solid on 15-step web tasks. But on a 300-step desktop workflow, around step 80 it starts redoing actions it already took. The team wants to switch to a 2M-token context model. Why won’t that fix it?”
Don’t say: “Use a longer context window so the agent can see its full history.”
Why 80 look-alike screenshots silently break model attention inside giant context windows, and the 5-part memory architecture that keeps 300-step workflows on track.
Share this trapShare on LinkedIn
The full answer, with the mechanism and the arithmetic, is for paid subscribers on Substack.