I think one of the frustrating things about the "are AI tools useful" debate is that people just assume you don't think the AI can generate code or just post stories about "I needed this one simple thing but I didnt know python so I had the AI write all my stuff".
Because it seems like validation towards AI being useful, but in actuality it doesn't deviate at all from the expectations people against using AI actually have. We know it can do things like this
The problem is writing some short thing that you need and don't really need to understand is both something the AI will have plenty of ready made examples to go through, and doesn't get anywhere close to the amount of complexity required to start to see the problems.
We know it can generate things, the problem is maintaining those things and actually understanding how they work. You cannot reasonably be a kernel maintainer for example without understanding the code you're maintaining
You can go into great detail about how the AI has no understanding of what it's generating and the maintenance issues that will come from that only to be squashed by random Joe Smith who successfully created a small project for one specific thing with the AI.
"I see code, the AI made code, there's no problem"
@Lyude just make the AI the maintainer :P