Seize the Maximal Probability Moment — LessWrong