Using ChatGPT to generate unit tests works great almost out of the box, but I guess this system solves the remaining 5% to make it fully automated end-to-end. I

Using ChatGPT to generate unit tests works great almost out of the box, but I guess this system solves the remaining 5% to make it fully automated end-to-end. I believe this will work and help us write better software, given that I have experienced numerous cases where the generated tests (even with inferior models) catch no-so-obvious bugs.