I use AI to generate code sometimes but it often doesn’t work as expected. Other times I get just awfully inefficient answer. Someone who has programmed without

I use AI to generate code sometimes but it often doesn’t work as expected. Other times I get just awfully inefficient answer. Someone who has programmed without AI for a few years will spot those things. Someone who doesn’t have any “gut feeling” formed about code and just relies strictly on AI will do awful. I think your approach is good, I would introduce things that work but are terrible design decisions and have the person interviewing figure out what could be improved.