Yeah, I also thought about automating it before. Especially when I'm middlemanning noncritical plans where I don't care what the solution is.

I did make a skill /ask-codex that gets Claude to call Codex for one round of feedback which is pretty useful.

But for critical plans, I suppose being the middleman gives me a good opportunity to learn about the solution and its trade-offs as it's being incrementally refined. Especially in a domain I don't know much about.