You can use these models to learn everything and to be an accelerant for what you can already do, or you can use it to learn nothing and become a dependent vegetable.

I use the crap out of LLMs as a programmer. It can do all the annoying refactors and point out collisions with new features I'm working on, and all I have to do is look at the diff and say "wow that piece there looks like shit and has an input flag for no reason" then go remove it. Or I say here's a list of things I want to do, what is an order to achieve them that brings the most value in the least amount of time? And it reflects back to me basically what I already value but with the benefit of some context awareness.

Now if you ask it to do your homework or ask it how many raccoons you should be masturbating every day it will of course give you some garbage, but that's not what it's for.