Sabitlenmiş Tweet

Every time you hit the context limit, your AI loses the entire architecture.
Every. Single. Session.
Ix uses 30% fewer tokens on average. Sometimes 80%+. There's a fix, see thread below.
Tanner Kocher@tanner_kocher
Our AI kept burning 30% more tokens than it needed to. Not because of bad prompts. Because every new session it had no idea what system it was working in. Losing the entire architecture. Every. Time. So we stopped trying to fix the context window. We mapped the system instead.
English


