So it's not a matter of believing in 80-hour engineering weeks and it's not a matter of choice for many people. It's just how things naturally develop when you are starting on your own.
As I mentioned (and as the links demonstrate) 80 hour work weeks don't double productivity in the long term. They may double the perception of productivity, and they may satisfy one's sense of urgency by never giving time to think, but I have never seen any evidence that they double productivity in the creation of software.
I believe that for certain sorts of work, 80-hour weeks might be more productive than 40-hour weeks. E.g., digging ditches (at least for the very young and very healthy). But in software the costs of error are so high, and so hard to trace back, that it will take very solid studies to convince me that the increased error rates from long hours really are counterbalanced by the large number of hours put in.
I also long ago stopped believing that 80-hour engineering weeks are much more productive than 40-hour weeks. For one week? Sure. Two? Maybe. Months? Nope! Definitely not better, probably worse.
(Why is a long answer, but for one take, this is good: http://www.lostgarden.com/2008/09/rules-of-productivity-pres... But even if none of that human-factors stuff were true, I'd still believe that because the cost of error. A tired programmer can easily be negatively productive, because small errors tonight turn into land-mine bugs that later blow up hours or days of time. 80-hour weeks value apparent productivity over actual productivity.)
All of those people should have gone home. Sense of urgency + stupid behavior + big payout is not actually proof that the stupid behavior led to the big payout. For every success like his, I can find a dozen stories of death marches that led to absolutely nothing worthwhile.