Does Bayes Beat Goodhart? — LessWrong