Integrating the Lindy Effect — LessWrong