This paper critically discusses a tradition in philosophy of science—here called ‘scientific demonology’—that, going back to Laplace, considers what science would be like if it were performed by a supernatural omniscient creature instead of being performed by puny humans. This tradition has been much discussed, and criticized, especially with respect to its commitment to determinism. Setting this aside, this paper argues that the proposed philosophical methodology is fundamentally deficient for other reasons having to do with the foundations of mathematics and the nature of knowledge. Even when there are determinate facts about objects representing states of dynamical systems and determinate functions from states to states, such facts are in most cases not representable, and as a result cannot be meaningfully said to be known, even by an all-knowing demon. The lesson to be drawn is that, far from improving on the classical philosophy of science methodology consisting in drawing lessons about the nature, scope, and limits of science by examining the best human science, speculations about demonic science obfuscate scientific epistemology.