There are surely ways to further constraint the set of programs to test, ie anything that output something not in the expected outcome should probably be an ear

There are surely ways to further constraint the set of programs to test, ie anything that output something not in the expected outcome should probably be an early return. I'd love for more work to be poured into this concept. There might be low hanging fruits and who knows an LLVM pass somewhere not to far in the distant future.