Originally Posted by wankawonk

I'm using grok for front-end stuff now. I like it.

I don't see a world where I'd use it for backend work. Issues:

--if the task is small, it'll always be faster to just do it myself vs engineer the right prompt
--if the task isn't self-contained, its too difficult to engineer the right prompt. The AI doesn't know how my systems work and explaining it is not realistic. Will someday it be realistic to just give it access to my whole codebase and it can figure things out on its own? Maybe. Idk.
--If the task is complicated, the AI will inevitably design its solution in a way that isn't how I would want it to be done.

AI is absolutely revolutionary for non-coders who want to build an app or website. If you're good at coding you're inevitably going to be very particular about how you want things done, you're gonna struggle to engineer the right prompt to get things done the way you want them done, and you're gonna revert to doing things by hand.

20 years from now it wouldn't surprise me if there's a job market for programmers who learned to code pre-AI in the same way there's a job market for Cobol programmers today. over-reliance on AI will lead to a shortage of truly skilled programmers and companies will realize they want to have a few humans around who actually have deep understanding of the systems and technologies being utilized, vs just being skilled at prompt engineering.