Current general LLMs are decent for prototyping or example output to jump-start you into the general direction of your destination, but their output always needs supervision and most often it needs fixing.

This.

LLMs do not produce anything that can be relied upon confidently without human review, and after the bubble pops, that’s only going to become more true.

Hell, I’m glad the first time I ever used it it gave me a bugged hallucinated and false reply. I asked it to give me a summary of the 2023 Super Bowl and learned that Patrick Mahomes kicked a field goal to win the game.