Letting Claude and Codex work together

One of the major things that has changed for me recently is I’ve started integrating Codex and Claude to work together more.

Each model has its own strengths and weaknesses, and the token cost of each model also affects how I decide to use them.

Here’s what that looks like in practice:

I’ve been using Claude on the highest settings with Opus Max to review UI and UX work and come up with a plan for what we should do. It’s basically acting as a lead designer and lead architect.

Then it hands that plan over to Codex for implementation.

Originally I was doing this manually by having both Claude Code and Codex open and going back and forth between them.

This past week, though, I gave Codex access to the Claude CLI so it can call Claude directly. I’m no longer managing that handoff myself.

I think one of the next big skills is understanding where each model excels and designing workflows around those strengths. The more I can get the models collaborating with each other, the less time I spend acting as the middleman and the more time I spend reviewing the final result.