Something is a priori if it can be justified independently of experience. Dave Chalmers connects the a priori to a deep sense of "epistemic necessity". (We might roughly characterize this as what an ideal agent could recognize as being guaranteed to be true, no matter how the actual world turns out.) This connection would be problematic if there were a posteriori certainties. If I know something for certain, to deny that it is thereby "epistemically necessary" arguably stretches the phrase beyond all recognition.
But Chalmers holds (plausibly enough) that introspection is a form of experience. So the paradigmatic example of certainty -- the conclusion of Descartes' cogito ("I think, therefore I am") -- is a posteriori, and hence "epistemically contingent" on the above definition.
Externalists about justification can avoid this conclusion, however. My belief that I exist is as safe as safe can be. There is no possible world in which I am mistaken in having this belief. (Merely having the belief suffices to ensure that I exist, and thus that the belief is true!) And note that introspective experience plays no role in this externalist justification. It suffices that, as a matter of objective fact, I couldn't possibly be wrong. Since similar considerations will apply to any purported "a posteriori certainty", we see that they can be justified independently of experience -- and thus qualify as a priori -- after all.
Update: On second thought, I think I'm wrong about that. Consider the thought "I am conscious". I take this to be an a posteriori certainty. Given my introspective evidence, I couldn't possibly be wrong. But there are zombie worlds in which my counterpart is mistaken in having this belief. Of course, he'll lack the introspective evidence that I have, since he isn't conscious. But that's a fundamentally a posteriori difference. It seems that, even for externalists, there's no way to account for this epistemic necessity in purely a priori terms. We cannot vindicate a connection between the a priori and this sense of "epistemic necessity".
[Update II: Oops, I forgot that Chalmers characterizes his "deep" sense of epistemic necessity as being unrelativized to what anyone actually happens to know. It is instead supposed to underlie our more particularized or "strict" sense of epistemic necessity (which need not be tied to the a priori). So my earlier criticisms were thoroughly misguided. That doesn't much matter for the more general discussion of externalism, however.]