Curious, I would expect a programmer of your age to remember Knuth's "beware of the bugs in above code, I have only proven it's correct but haven't actually run it".

I'm happy you know math, but my point before this thread got derailed was that we're holding (coding) AI to a higher standard than actual humans, namely to expect to write bug-free code.

Humans can actually run the code and knows what it should output. the LLM can't, and putting it in a loop against code output doesn't work well either since the LLM can't navigate that well.

I hold myself to a higher standard than AI tools are capable of, from my experience. (Maybe some people don't, and that's where the disconnect is between the apologists and the naysayers?)

> my point before this thread got derailed was that we're holding (coding) AI to a higher standard than actual humans, namely to expect to write bug-free code

This seems like a very layman attitude and I would be surprised to find many devs adhering to this idea. Comments in this thread alone suggests that many devs on HN do not agree.