On serious programming, 5-6 hours seems to be an upper limit for the average. That's not to say that you can't do a 12-hour day occasionally, but you'll tire out faster and need more light weeks. To me, the best schedule seems to be 11 3-hour blocks (contiguous, no interruptions) over 5.5 days.
For startup founders it's a bit different because there's variation in workload. Code isn't the only thing they do. They also have to meet with investors, manage people, set priorities, and network. With that additional variation, you can more easily get to a 70+ hour week and not burn out because even though you're pushing yourself to the limit of what most people can handle in terms of work and stress, you're doing different kinds of work. It's still exhausting, but it's less exhausting than doing the same kind of work for 70 hours.
Look, I enjoy the hell out of programming. It is a lot of fun. And I'm good at it. I've been doing it for long enough to know that I can't get 3,500 hours per year of quality code. There are people who claim to be different, but for every case of that where I've had the "privilege" of seeing his or her work, I've found that person not to be different.
Now, I have to say this bluntly: you're just wrong on one thing. The idea that you can't have real achievements without long hours is completely incorrect. With the right insights and exposure, you can. I'm not saying it's easy, and it typically comes only after several years of continuous hard work, but what's appealing about software is that often you can save a company millions with under a hundred hours of work. (Well, actually, that's because of the thousands of hours the person spent learning before that point, but the nice thing about programming is the acceleration of value creation. Once you learn more stuff, you get to a point where you can reliably add value to such degrees.)