CC is more autonomous, which can be a double edged sword. In big codebases you usually don't want to make large changes and edit multiple files. And even if you do, letting the LLM decide what files to edit increases the chance for errors. I like Aider better aswell. It's a precision tool and with some /run it is pretty flexible for debugging.

Claude Code (and Cursor, for that matter) don't commit to git. Fundamentally that's just bonkers to me. Aider does, so each prompt can be /undo'd. I had a chance to use Cursor at work, and if that's how people are interacting with LLMs, it's no wonder we can't agree on this whole "are LLMs useful for programming" thing.

ChatGPT Codex is on another level for agentic workflow though. It's been released to (some?) "plus" ($20/month) subscribers. I could do the same thing manually by making a new terminal, making a new git worktree, and firing up another copy of aider, but the way codex does it is so smooth.

Claude Code is much more aggressive at doing stuff than aider is (with sonnet and gemini) in both good and bad ways. You can tell Claude to do a thing and it might churn for many minutes trying to achieve it, while aider is much more likely to do a lot less work then come back to me. Aider feels more like a small sharp tool vs Claude Code as a bulldozer.

They both can just use api credits so I’d suggest spending a few dollars trying both to see which you like.