You write the test → AI writes the implementation … who is doing this?

I am wondering if anyone is using this dev flow in any way … where you write the tests and pair program with the AI to do the implementation. Basically telling the AI what to do until the code looks fine and the tests pass, instead of writing the implementation code yourself. If so, what tools are you using and what is your workflow?

Thought – I was kinda doing this today, and it felt like writing the implementation code is becoming meaningless, I was not really caring how it looked like, I was too lazy. And I normally -really- care about clean code. It felt like this was the next abstraction, as in 99% of the cases nobody cares anymore about how your assembler looks, it felt the same. I did not care about the implementation, I just wanted the AI to give me working code asap.