Post

I feel the same. Also one more thing to consider - someone is ultimately responsible for all the generated code, and I don't think people appreciate just how much you take on by speeding up on delivery with AI. Including the amount of bugs you ship.

Good point, yeah the long term affects might still take some time to really surface. Right now the answer seems to be use AI to review AI. But I notice in my own projects that it's struggling.

0 reposts1 like

I use something akin to superpowers/teams to roast my reviews, it helps if you do those on a fresh, unprompted session. But I still have to provide guidance and input on what to focus on, otherwise things can go off the rails on moderately complex changes. All of that is quite draining.