Hofstadter's Superrationality — LessWrong