But what does Taleb say that Silver should do differently? If he has no constructive suggestions and is just complaining, well, that's par for Taleb. When I read Taleb's The Black Swan I noticed that he spent much of the book saying the same thing over and over in different words. On the whole, his key points were important ones, but the book wasn't designed to be very actionable.

There is no magic way around the fact that we don't know what we don't know. As Taleb himself said, you can't just shift the reported probability from 80% toward 50% to account for unknown unknowns. I would point out, though, that there is a much better way to evaluate 538's projections than simply comparing the probability of who would win with who actually won. Instead, 538 should produce a probability distribution of popular vote percentages per-race and per-state. if 538 is doing a good job then the actual popular vote will usually be near the middle of the forecast probability distribution. By comparing the popular vote with the models we can get a much clearer picture of whether the models are typically well-calibrated.