Asymptotic Logical Uncertainty: A Modification to the Demski Prior — LessWrong