I'm still fairly suspect of anyone who claims they're building full apps with AI. I ask Claude fairly basic questions that can easily be answered by looking at the Apple Developer docs & it still invents solutions more times than I'd be happy with.

That said, I'm find it useful in criticizing my own code base & then I fix it myself. It has a solid hit rate at spotting problems (it seems to lose context & doesn't get some things are just fine). It's just less reliable at offering solutions