Interesting. The gist seems to be that we model market events with normal distributions, which is consistent with what I've seen in other "natural" and/or emergent processes, and this underestimates risk, because the true distribution is fat tailed.