The Self-Reinforcing Binary — LessWrong