Defeating the Villain — LessWrong