[Anamon]

That was my main reason for having dropped LLM coding assistance completely for now. It wasted so much of my time and energy. I got the occasional helpful response, but not before four that were almost correct, which of course is a much bigger time sink than when it's obviously wrong. Another fifteen minutes spent investigating a solution that turns out to be bogus.

I got the usual gaslighting that I'm just using the wrong models, or using them wrong, etc. But when I watched others, I found that their results were just as iffy, they just usually didn't bother to check them once they looked plausible.