I would think that math aptitude has very little to do with developing great products and companies, so I don't understand this argument. It doesn't even have much to do with many types of programming.
"Small differences in the distributions of men and women don't allow you to predict much about any individual trait of a randomly selected individual. But when you limit yourself to the far tails of probability distributions, they do. "
So, he's saying that by being more selective for a field that men have greater variance in, you're more likely to find more men at the far tails.
This is a counterpoint to Eric Ries, who claims:
"Here in Silicon Valley, we’re looking for the absolute best and brightest, the people far out on the tail end of aptitude...How far apart would the two populations have to be to explain YC’s historical admission rate of 4% women? It would have to be really extreme."
The only point is to do the math and demonstrate that small differences can add up to a lot if you have highly selective filters. I chose mathematical ability as an example because a) it's fairly well studied and b) Eric Ries cites a slideshare presentation which also uses it as an example. (However, the data used in the presentation are about 20 years older than mine.)