just having more test doesn't mean you actually test more (or more correctly)

Having AI specialized on assisting testing is viable but in the end would be done quite different as far as I can tell even if it uses LLMs. Through maybe domain adoption and a very different scaffolding might be enough.