I'm excited to see LLMs get much better at testing. They are already good at writing unit tests (as always, you have to review them carefully). But imagine an LLM that can see your code changes _and_ can generate and execute automated and manual tests based on the change.