The thing is, any reasoning that assumes Silicon Valley is where it's at ignores all the TechCrunch articles about "the crisis in venture capital" and the inability of the VCs to find companies that need their money to lift off at all in the commodity-hardware/open-source-software era. It's very, very unlikely that the Valley-centric monoculture of the past will continue to characterize the startup scene. NY has a startup scene, Seattle, Portland, Phoenix, Pittsburgh, Los Angeles, Edinburugh, London, every place I've been to or met people from in the last ten years has a startup scene. Even Santa Fe, New Mexico, an isolated mountain city of only 60,000 people, counting daytime commuters, has a startup scene.
The radical reduction in startup cost for technology companies is THE technology-business event of the last twenty years. It absolutely decentralizes tech startups by making VCs much less important. Yes, the Valley still has an edge in terms of people, but if you're used to being the center of the universe, having an edge isn't the same.
This explains local startup scenes, and makes it obvious they'll become more important in future, not less. Paul Graham isn't just wrong for the obvious reasons - because anybody who says NY is about any one thing is smoking crack, for instance, or because of the Valley's echo chamber, or because there must be something in Boston's relative non-entity-ness despite its incredible programmer population, etc., etc., etc. - he's also wrong because he's ignoring the ginormous sea change in startup capital requirements that his own business revolves around, and which he himself first identified in one of his essays years before anyone in the Valley ever blogged about it.
He's ignored his own discoveries.