Respectfully submitted, I wholeheartedly disagree. I’ve leveraged LLM to do some pretty complex CodeGen.
The problem is not that it can't or won't do it... the problem is that it typically takes a few tries to get it right along with sometimes a lot of trouble shooting. And then depending on your level of experience you might end up with something that is hard to sort out later if you decide you need to modify it.

Do you predict AI replacing tons of coders and not eventually generating more demand for coders (i.e. managers wanting more and more data analysis) ? I hope you’re wrong, but you could be right.