I think it’s pretty obvious that you haven’t used any AI tools in a while. They have gotten considerably better at not spewing complete nonsense, and at being very convincing to people who don’t know better. We’re well past the obvious days of glue pizza and telling people to eat rocks, and entirely fabricated garbage.
My job is doing a full court push of AI to make sure we “don’t get left behind” by nebulous competitors’ hypothetical productivity gains. So I’ve started picking it up. It’s useful for searching for information using normal language terminology, and on rare occasions can call out edge cases and additional considerations early, rather than when I come to those conclusions after reading 10 different documentation sites and mentally mapping everything out.
I still think the bleeding edge early adopters were/are insane, AGI/ASI isn’t going to coalesce out of the current approaches if ever, and it’s a net negative for humanity. That all said, I see the benefits and use cases more clearly now too.
Unfortunately, not checking anything and pushing to prod now can result in things that work well enough to become relied on before it hits the wall. You end up with a project five layers past the initial slop code (that the initial slopper claims is rock solid and good to go) when the problems start showing up, then you realize that you’ve been building off slop foundation and have to tear everything back down to the frame.
Makes the whole thing worse tbh.