I discuss Wittgenstein’s radical contextualism about knowledge ascriptions. Accordingly, though superficially similar, uses of “I know” can serve three fundamentally different functions: an ordinary, genuinely epistemic one; a grammatical, non-epistemic one; and a ‘philosophical’, nonsensical one. I present Wittgenstein’s anti-skeptical strategies, and compare my reading of On Certainty with therapeutic and contextualist ones. On reflection, the ordinary use of “I know” turns out to be the only common ground between Wittgenstein’s position and contemporary contextualism regarding knowledge ascriptions. Wittgenstein was in fact an invariantist. While I do not follow Wittgenstein in his claim that skeptical doubts are nonsensical, I don’t think they pose a threat to our ordinary knowledge either. Yet, we do not have knowledge of hinges, since reasons for them would be circular. Hinges, rather, are rationally assumed, albeit without any justification, because they are constitutive of epistemic rationality. With that picture in hand, I then close by defusing the objection, raised by contextualists, that if we don’t have knowledge of hinges, we would have to condone ‘abominable conjunctions’.