Ya, he points out that a lot of fat-tailed distributions can resemble the normal distribution because the rare events may not appear for several hundred (or thousand or whatever) samples, but these rare events can be so far off the mean as to have a huge effect, making the use of the Gaussian curve a poor model of certain phenomena.

He also wrote a nice paper about how shit p-values and null hypothesis statistical testing are. Not like he's the first one to say that but he did do some nice math.

http://fooledbyrandomness.com/pvalues.pdf