But if the ai writes the tests and you don't read them, how would you know if that's 100 lines of assertTrue(true) ?

So, "red, green, refactor" has become "red, green, deploy"?

That's a change, for sure, but I wouldn't call it an evolution.