I haven’t been in development for nearly 20 years now, but I assumed it worked like that:

You generate unit tests for a very specific function of rather limited magnitude, then you let AI generate the function. How could this work otherwise?

Bonus points if you let the AI divide your overall problem into smaller problems of manageable magnitudes. That wouldn’t involve code generation as such…

Am I wrong with this approach?

We don’t know if there’s much of a difference between an AI and a brain. If we look “into” an artificial neural net we only see a lot of “weights” that don’t make any sense. If we look into a brain, we also can’t make any more sense of it. The difference is smaller than we want it to be because it seems to either give AIs a lot of credit or it makes us less than we want to be. We are very biased, don’t forget that.

So, how do art students learn? They are doing the exact same things. Only they do a lot less, because natural neural networks (aka brains) are not capable of processing training data as quickly. It’s not as if every artist has to reinvent the wheel and generative AIs don’t and as such have an unfair advantage.

Look at inventions like the printing press! Did everybody like it? The catholic church certainly didn’t! Is it a a phantastic peace of technology anyway? Sure is!

As far as I remember the secret is to log in as admin and change the ownership of the files to yourself, then change permissions and then do whatever the f you want with the files.