Picked GPT-4 over Claude for my coding workflow last week and it backfired hard

I had to choose between using GPT-4 or Claude for a big React project I was refactoring. Went with GPT-4 because it felt faster in my tests, but after 3 days I hit a wall where it kept giving me outdated hooks syntax that broke everything. Claude caught the same problem in one try and fixed it clean. Has anyone else found that one AI model just works better for specific languages or frameworks?

3 comments

3 Comments

iris_mason881mo ago

Did your friend try switching back to see if the problem was just a bad streak? My buddy was working on a Vue app and kept getting broken lifecycle hooks from one model but the other fixed it first go. Now he just keeps both installed and jumps between them when one starts glitching out.

noahbaker1mo ago

Man that seriously sucks, hope it works out.

lee_grant971mo ago

Respectfully @noahbaker I get the frustration but flipping between models every time one acts up sounds like a bandaid more than a fix to me. You might be trading one set of lifecycle bugs for another instead of figuring out what's actually wrong.