One-Magisterium Bayes — LessWrong