I think the important thing is finding that happy space where you can feel certain that you are the author of the ideas that are being put to text (or code).

I'm not sure what that middle ground looks like but for me its having a really clear separation between the LLM acting as a really good search engine and static analyzer and a human authoring code.

In an ideal world, professionals would have been given LLM's and told "make of this what you will, we have no expectations of output". We did not get that world unfortunately.