If you are doing anything more complicated than gluing a pair of APIs together, it is very likely that the marginal value of work past 60 hours is negative. Reaping what you coded during a 3am espresso binge is rarely a happy experience.

But yeah, if all you are trying to do is build out Facebook before some other social network gets there maybe it makes sense to pull 100 hour weeks.