Opposite for me: Claude has been schizophrenic and just wasting my time with red herrings while Codex has been saving time.
I use both just for code/logic review, for 2D Godot games, never for generating or editing code.
After asking Claude Opus 4.6 to review a single file in a simple platformer game, it goes:
> Claude: Coyote jump fires in the wrong direction (falling UP with inverted gravity)
var fallVelocity: float = body.velocity.y \* body.up_direction.y
> Claude: I owe you a correction: after re-analyzing the math more carefully, the lines are actually correct — my original review point was wrong. Let me walk through why.
It's had several other gaffes like this where it reports a finding then immediately backtracks when asked to explain, and the UI/UX is still crap (fonts don't get applied, it doesn't catch up with the updated working state after editing files etc.)
I have lots of other examples and could post screenshots of asking the same thing from Codex vs Claude, and Claude is consistently wonkier, or it just tries too hard.
Either way it seems we're entering a new frontier of fanboyism to rival the Mac vs PC wars :)