He makes an interesting/controversial point: "It still takes ten years to become a success, web or no web." ... "Listen instead to your real customers, to your vision and make something for the long haul." I wonder how many people doing startups are prepared for success to take ten years.
I wouldn't look too far into the number "ten years" and think it applies to every single startup. Obviously, this isn't true, and I think he's using it more as just a way of saying "it takes a while."
Anyway, it's an interesting path to the ten years. Maybe you've just (http://www.paulgraham.com/bronze.html) got a bad idea that's not worth hanging on for ten years as it'll always suck. Probably an early milestone is to get (http://www.paulgraham.com/fundraising.html) ramen profitable. If you can't get there in n years, reboot (?) Maybe (http://www.paulgraham.com/bio.html) some comp sci phd angel/vc former founder will figure out the algorithm to optimize the ten years to reduce it by an order of magnitude.