My favorite numbers for startups are that they have a 1-in-20 chance of succeeding.
I don't know where I got those numbers from -- seems like I've heard PG use a 1-in-10 chance, but you get the general feel for the situation. You should work in such a way as to acknowledge those numbers. I'm not saying don't work hard -- far from it. Just work as if this is stage 1 of a 20-stage process, each stage lasting a year or two. If you're not working in such a fashion that you can go a decade or two, logic says you have mis-configured your work habits.
I had this same attitude when I moved from the commercial world into startups, get your ass in gear and buck up! That was a few years ago, and I've moderated my views quite a bit due to brutal reality. I think part of the problem for me was that I was trying to compare the startup world to either the academic world or to the commercial software world. In both of those scenarios you worked as hard as hell for a limited period of time and then you delivered. You took the test, you released the product, you passed the class.
Startups aren't like that because the work is never done. There is no huge finish line you sprint through. It just keeps going and going. The cutesy way of saying this is "it's a marathon, not a sprint" but I'm not sure that even covers it adequately. It's not a race at all. It's just a long, hard slog. Yeah sure, if you've got traction it is a hell of a roller-coaster ride, but I'd worry about then if-and-only-if that day actually arrives. Work like Facebook did to get their first million users when you actually start attracting hundreds of thousands of users and you're looking that potential in the face, not before.