AI can be used to write E2E tests, unit tests, etc. The issue I find is that I would like a different model with a different context to design and write the tests. I haven't found the right technical solution to this yet.

When models write both the code and the tests, they might go back and forth between the code and tests until they reach a working solution. This seems dangerous.

E2E can also be vibe coded and VLMs are increasingly good at it.