> people delivering code via LLMs aren't critically evaluating the code it generates

I guess this is the flip side of "you're prompting it wrong": "you only think it works because you can't tell how much it sucks."