Strict Dominance for the Modified Demski Prior — LessWrong