The suggested remedy of “understand the code the AI wrote” doesn’t work for me. At least not in practice. It’s simply exhausting to read, understand, think critically about a few pages of code, then adjust the prompt to address some defect, and in 10 seconds having to do it all again. I can only do a few iterations of this before fatigue sets in.