> But I was also very skeptical about AI being able to code semi-reliably during the early stages of GPT hype, and look where I'm now: most of the code I produce is written by an AI

My impression from a couple years ago was that it was fairly decent at coding, it was just slow to go from question -> code, and the tooling around that has improved significantly so that it's all pretty quick. I think whether or not the models are fundamentally better at raw coding is a murkier question.

They still fall down at bigger architectural tasks, go off the rails, hallucinate, etc. So, it seems to me like a core problem with the current technology.

> it doesn't matter if the end result is flawed, it matters that "mission accomplished" and someone is out of a job

This is a short term problem. If the market has any sanity left, the shops that maintain the talent to execute well will out-perform the shops that were short-sighted.