Break your habits: be more empirical — LessWrong