Now, Jonathan at Fake Barn Country argues that the skeptic's dream scenario threatens the safety of our beliefs. There is a close possible world in which I am dreaming all of this, such that many of my present beliefs would be false. So even though my beliefs are all true (let us suppose), they fail to count as knowledge because I easily could have been dreaming and thus mistaken.
As I responded in comments:
the phenomenology of dreaming is quite different from that of waking life. Even though I might not realise when I'm dreaming, I certainly do realise when I'm awake. This asymmetry is important. There is no chance at all that I'm dreaming right now, because I know that I never have this quality of phenomenological experience while asleep. So I don't see dreams as skeptically dangerous.
My understanding of dreaming is influenced by Dennett's metaphor of anosognosia: sometimes when it seems we are aware of something (e.g. filling in the blind spot), it's really just that we are not aware of our deficit. I think dreams are like this: it's not that we positively think we're awake, it's more just that we aren't aware that we are asleep. Most dreams aren't convincing hallucinations, it's just that our judgment is so impaired at the time that we don't realise how different and unconvincing they are. The potential for confusion is negative, not positive; it's due to a lack, not a presence.
What I want to emphasise here is the crucial difference between my waking knowledge that I'm awake (which is in no danger of being mistaken), and my dreaming inability to tell whether I'm awake. The latter in no way threatens the former. From the fact that if I were dreaming right now I would not be able to tell the difference, it does not follow that I actually cannot tell the difference! That's the "asymmetry" idea I was trying to get at before. (I worry that I'm not expressing this very clearly, but I hope you at least get the general idea.)
As such, I consider my beliefs quite unthreatened by dream-based skepticism. If we include the appropriate background, i.e. the rich 'positive' phenomenology that grounds my beliefs, then I think they even remain 'safe'. Given my phenomenology, I could not easily be mistaken. I could have a different phenomenology and fail to realise it, e.g. if dreaming, but that is of no relevance here. It's no problem, because as it happens I'm not just overlooking a negative, I'm actually noticing a positive. I am fully aware of the richness of my present (waking) experiences. If dreaming, I would not have that awareness. Instead, I would be unaware of my deficit. Two negatives do not make a positive. We should conclude that my beliefs are safe after all.
Alternatively, if you insist that the dream scenario implies that my beliefs are technically not 'safe', then that simply shows that safety is not what matters. All that justification really requires is the sort of "asymmetrical safety" that I have described above. 'Symmetrical safety' is overly stringent in much the same way as 'sensitivity' is. So we should not be too worried if our beliefs fail to meet these tall requirements in light of various skeptical scenarios.
To demonstrate: suppose that you've been slipped a drug that screws up your logical capabilities. So you engage in wacky trains of illogical reasoning, all the while unaware of your irrationality. At some point, you reason to yourself, "I think, therefore I am", and it strikes you as utterly certain - you think you couldn't possibly be wrong. But suppose it turns out you are wrong. We can't comprehend this of course, because we're all on the wacky drug ourselves, so our logical capacities are screwed. But that, I hereby stipulate, is really how it is (in the scenario we're imagining).
Does that scenario in any way threaten our present certainty in Descartes' cogito? I should think not. This then helps reinforce my central point: The fact that if we were wrong we wouldn't know it, does not establish that we currently do not know that we're right. To think otherwise is to neglect the asymmetries I've pointed to, and Dennett's insight that neglecting absence does not create presence - it merely seems like it.