What I want is something that will put me in touch with friends I've never had.
Most people have trouble making friends after college. In college, the shared experiences create a rich environment to meet and get to know new people. After college? Not so much.
How do you make friends after college? You go to conferences, community events, and bars. You make friends at work. These are all less than ideal. The shared experiences aren't necessarily as rich, except for coworkers. Having your primary social network and your coworkers as the same set of people offers a whole host of other problems.
What I want is a way to meet people who are interested in some subset of what I'm interested in. I built a very rough and specialized prototype in a couple of days when I built amb.itio.us, focused on people looking for startup founders. The essence is there, though. I list who I am, what I'm looking for, and give a location. The system finds intersections among users. It's all very straight-forward, except for one thing.
The whole viral-loop thing is hard to engineer in. If the entire point of the application is to discover and meet new people, I have no inherent reason to pressure my current friends to join.
This leads to the necessity of some sort of hybrid approach, maybe a layer on top of the existing social networks. What I'm looking for might just be a feature on top of (ugh) facebook. It might also be a layer on top of something like foursquare, with location awareness. I'm still settling in on this.
Possible revenue models? An obvious one is permission marketing. The same way I might signal I'm looking for someone to meet, I could signal I'm looking for something to buy. In any event, revenue models would require a pretty substantial ramp-up in users before any money actually came in the door.