I’m usually not the type to complain. I know there’s only so much compute to go around right now, especially with the insane amount of API traffic these systems handle.
But wow... sometimes Codex feels really slow. 💤
Anyone else noticing a degradation in Codex GPT 5.5 recently?
I used to be a pretty happy user for more than a year. Watched the chaos around Claude’s tiny limits and honestly laughed a bit. But now Codex seems to hit similar problems.
After compacting, the model often loses coherence completely. It suddenly acts like the repo and all the existing files don’t exist anymore. Instead of following established patterns, it starts reinventing things from scratch, and often in the wrong way.
At this point I basically have to force it after every compact to reread bootstrap and governance files again and again.
Forget it once?
Then there is a good chance the next tiny feature gets implemented in a completely different style than the whole rest of the project.