Moore's paradox pits our intuitions about semantic oddness against the concept of truth-functional consistency. Most solutions to the problem proceed by explaining away our intuitions. But "consistency" is a theory-laden concept, having different contours in different semantic theories. Truth-functional consistency is appropriate only if the semantic theory we are using identifies meaning with truth-conditions. I argue that such a framework is not appropriate when it comes to analyzing epistemic modality. I show that a theory which accounts for a wide variety (...) of semantic data about epistemic modals (update semantics) buys us a solution to Moore's paradox as a corollary. It turns out that Moorean propositions, when looked at through the lense of an appropriate semantic theory, are inconsistent after all. (shrink)