x
Inner Alignment — LessWrong