I also agree with your points on ethics and the knowledge shrinking. But I can’t think of a sane route. Not using AI in the current state will leave the kernel behind (at least in security matters which is obvious), and using it will degrade the quality of the software itself, and its community.