The longer you work with a coding agent, the more its **memory**, not its intelligence, decides whether the session stays sharp or quietly falls apart. A model that reasons brilliantly at hour one can lose the thread by hour four — not because it got dumber, but because the context window filled with clutter and the important decisions scrolled out of reach. Treating that window as a budget you spend deliberately, rather than a buffer that fills by accident, is the single biggest lever most people never pull.
A catalog job crawled for the better part of an hour, and the obvious suspect — heavy PDF parsing — took the blame. Profiling put ~40 of every ~41 seconds somewhere no one had looked: a one-line regex hunting for an email that wasn't there.
A running field guide to the traps that bite a Windows dev box day after day - the ones that are small on their own and expensive in aggregate. Each is written as a thirty-second diagnosis.
"Keep the last fourteen" is a sentence about time. A disk only understands bytes. Here is the morning that difference cost us, and the small change that fixed it for good.
The script had worked for weeks. Then, overnight, a dropdown that used to hold four options held sixteen, a rule that used to be allowed was forbidden, and a field started closing itself half a second after it opened. Nobody told us. The portal had simply changed.