case in point, I explained the change to a CTO and their agent, got asked to review, and it missed the point - even as the CTO of the company, they lacked to nuanced understand I have as the author of the code, to notice the obvious error
If I had to say, I think steering in the AI to achieve a certain task is one of the main broad skills you need to have, but what that means exactly is a little bit hard to define. It's less about knowing exactly what to tell the AI and more about knowing how to break down a problem, give it the right context, recognize when it’s going in the wrong direction, and keep it moving toward the outcome you actually want and doing all that in the most efficient way you can.
I'll believe when I see credible evidence and evals, everyone is making claims, the evidence I see IRL points elsewhere (vibes still, but actual evidence and cases to the contrary)
A good self check is to go read Terence Tao's conversation with ChatGPT, we developers could never do something like that, and tend to extrapolate our expertise to other fields where it is not warranted, even within our own domain. Another check is all the app slop we still see from non-developers. Cool and I encourage for personal learning / usage, but not for other users and putting them at risk.