Feedback vs Claude Code

[lucasmike]

I have been using Claude Code with Opus 4.1 and I have been mighty impressed with it. However, I think Codex moves the bar higher and it beats Claude Code in reasoning in complex codebases, implements complex new features better, drafts more concise execution plans, picks up better on code-wide conventions, is not lazy in pretending it completed tasks.

The only downside is that Codex CLI presentation seems messier than Claude Code. It's not so easy to follow, it seems to use a lot of Python scripts to edit Node and JavaScript files, which looks messy on screen so I can't really follow up what's going on (maybe Claude Code does the same but it's hiding the details behind the one line status "updating the file") - but that aside, it seems that Codex beats Claude Code on overall quality of execution. Well done OpenAI - this is awesome!

[YoloMark98]

true!

[SevenThRe]

I don't like codex's reading and writing tools either, but you can change its default behavior through MCP. Additionally, I remember that codex can be called as an MCP, which is detailed at https://github.com/openai/codex/blob/main/docs/advanced.md#codex-mcp-server-quickstart. You can use claude code's reading and writing function tools and use codex for planning.