Thanks a lot for the insights. I like the Machiavelli thing.
reply> Then I ask GPT 5.5 Pro on the web with extensive reasoning set on. It will take sometimes 30 minutes or more to reply.
Any reason why Codex can't do that?
If Pro is the same model (hard to tell, I'm not sure) it has a token budget to think (test time scaling) which is huge compared to the Codex endpoint.
reply