but I totally disagree with your second point. running code often has real problems with race conditions, error handling, unwanted memory reuse, out of bounds pointers, etc. if a new version of the compiler can prove these things for me - that's invaluable.