Superrationality in arbitrary games — LessWrong