Tried out ChatGPT o3 — it's insane
This morning I ran o3 through a code review, and it caught two edge cases I usually miss. lol Especially the potential resource leak in async error handling — I'll admit that one.
I also had it tidy up documentation, and it even made a Markdown table on its own. The refactoring suggestions were solid and well-grounded, saving the backend team from half-rewriting things. That said, Korean responses still come across as a bit awkward sometimes. Still, at this level, the quarterly subscription is worth it.
9 answers
Yeah, I've tried it too, and code review is genuinely god-tier.
That async leak catching thing is awesome lol, I caught one with it yesterday too.
The awkward Korean is a bit off-putting, but aside from that, I'm very satisfied.
Well, I'm not really sure what's different about o3 compared to the previous models. It was already good at code review, wasn't it?
Do you have a source? I'd like to know where I can try it out.
I have to admit, the reasoning behind the refactoring proposal is really solid. I also got a lot of help from it yesterday while cleaning up legacy code.
But judging by your mention of quarterly subscription fees, are you using Pro? I'm still on the free version — is the difference big?
Oh, I also asked o3 to write test code over lunch today, and it was a pain lol. Still, it's true that it saved time.
Actually, there are often cases where they can't even catch two edge cases, so isn't it too early to judge based on just that one?