Maybe I'll be able to stop using my forked codex... this is sorely needed. Codex is kind of terrible with threads when you have a /goal that lasts more than a week.
OpenAI is about to ship a huge performance upgrade for extremely long ChatGPT and Codex conversations next week.
In an internal test on a massive 741-turn, 231 MB conversation:
→ 27.6s average load time → 1.66s
→ 94% faster
→ 41% less overall memory growth
→ 894 requests

