The strategy-stealing assumption — LessWrong