Hang in there as best you can. Contrary to what AI bros on HN claim, LLMs can't actually program worth a damn. In time the deficiency will become too great to ignore, and companies will be replacing their models with humans. No telling how long that will take, but this is going to get walked back just like outsourcing to cheap Indian firms got walked back in the first decade of the century.

I made lots of cash undoing the Indian outsourcing mess.

Ive also been deep in AI. Not only is AI inadequate for well engineered software, I think treating programming as a language translation task between spec and code boring. They're even worse for research, rabbit holes and bad lines of reasoning.

Im amazed people don't see how much smoke and mirrors even the frontier models and harnesses are using to mask the fact that they're reasonable sounding bullshit machines.

I really hope you are right.