> That last script ran into some problems, so I revised it fully
It's still awesome that this has made computing more accessible to this user but we still have a ways to go before it can solve problems like this on its own. Maybe gpt-4 would've made fewer errors?
Likely. There's a big gap between them. I don't really use 3.5 anymore even though it costs money.
My only issue with being is that it’s extremely sensitive and will end a conversation for the most ridiculous things. But it’s still well worth it if you’re prepared to deal with its pettiness & stubbornness.
I’ve been using it to help me figure out React hooks (I’ve only used class-based react) and it’s REALLY good. It even warns you of bad practices.
It’s also pretty accurate at giving me a specific tailwind class. It’s much quicker than browsing the docs.
their current implementation isn't as like a person typing
It makes me appreciate the finishing touches on the ChatGPT UX
I wonder how much of a factor that is for other people