I don't hold a lot of hope for TDD if the tests are AI-generated from the AI-generated code... which I see a lot.