The only way to succeed in grad school is to be single-mindedly focused on mathematics. Not on being a "generalist", not about "cliques", tenure, etc. This "meta-math" reasoning is a bad substitute for doing actual work. Clearly, the guy is better off doing something else.
Just because grad schools don't currently engender generalists doesn't mean that they are a bad thing or less important. It's just harder to measure how accomplished and brilliant a generalist is than it is to measure the contribution of a researcher that has made significant progress in a specific area. Generalism is only really well rewarded when it produces an unexpected solution to or progress toward a solution of an open problem.
I succeeded in grad school, by doing actual mathematical work (my PhD thesis was in Banach Algebras). But I agree with the article, and it's a decent description of some of the reasons I left mathematics.
