Shift towards the hypothesis of least surprise — LessWrong