Most people I've seen saying coding is solved are ML engineers who write throw away code for running one-off experiments. If you can effectively "archive" code after getting the results you need, I can see why you would say coding is solved. Sustainment is a non issue.

Code written by not very good coders has the possibility to be short and to the point. Even if I think the architectural decisions are wrong, the code still communicates a motive and an understanding of the problem at hand. This is all lost with LLMs.

In my experience, it's very good at following standards, so good in a structured project. Not so much in a freeform project where things get out of hand pretty quick.

I'm curious, what language do you code in? I use Go for a high scale backend service at this point it is better at writing code for a jira ticket than most of my team.

There is still questions about if that jira ticket should have been written but from ticket to mergable PR it's better than me.

maybe so but most of the code produced is low quality, so what do we have to lose?