Hacker News Reader

I see. And who will be providing that assessment of correctness? Will it be humans?

Sure, that's one option. If the AI is smart enough, then it could also do its own experiments. Or if it's embodied...