Why does "deep abstraction" lose it's usefulness in the far past and future? — LessWrong