Over-applying rationality: Indefinite lifespans — LessWrong