Happy one-decade anniversary, by the way. I just noticed https://www.ycombinator.com/blog/meet-the-people-taking-over... was almost exactly 10 years ago.
The trouble with your frame is that you can’t know what people actually want until you try to get them to use what you’ve built. There’s a sort of chicken and egg problem here: you have to think of an idea first, before you build anything. "Choose what people actually want" begs the question.
It’s probably true that it’s better to work on something that you personally believe in (in the https://paulgraham.com/earnest.html sense), because you’re more likely to find ways to grow it. It’s hard to be earnest about an idea that only interests you because it might grow at 5% per week. But pg chose Viaweb, and he couldn’t have known ahead of time that anybody would actually want it, since it was the first webapp. Nobody even understood how to try it out, since this was the era of desktop software. All he had was a hunch that lots of people would soon want websites for their businesses.
So which frame would you choose, if you had to decide today? Suppose you and Scott were trapped in a room, and the door unlocks only after you pick something to build together. How would you decide, other than "X seems most likely to be something that people will actually want"? (Which is similar to "has a chance of growing at 5% per week," just phrased a little differently. pg originally thought museums would want websites; if he’d asked himself whether it might morph into an idea that could grow at 5% per week, he probably would’ve realized museums didn’t actually want that.)