Codex vs Claude Code. Now I use both. As a main channel I run Claude Code, but I have built commands that give Cladue ability to talk to Codex which I often do to either discuss an implementation (I mean they discuss between each other) or have Codex audit it. It works very well and leverages both models to their best potential. Codex is very good at auditing and finding edge cases so I prefer to have him to review Claude's work.