Is alignment reducible to becoming more coherent? — LessWrong