Skrentablog

first micro virus - arguably social software :)
monster - first user-designable MUD
usenet newsreader - usenet was a huge early net.community
DMOZ - massive community success to build a web directory
Topix - huge uptake in the local community over the past year

Greg Sterling even sagely commented on my how-to-beat-google list that "A distributed editorial staff is in there somewhere."

It does seem that, no matter what you're trying to build, adding people into the mix seems to make the software better. Software is cold and shallow, people humanize it, and the public can provide wonderful extensibility and depth to a system that your cube-bound programming staff would never be able to match. Fred Wilson had the seminal post on this with his "All software should be social."

The only caveat I would add is that it's often much harder than it looks to scale the social architecture. You can get early usage takeoff quickly by throwing the doors on your system open and letting people in. But regulating quality, rejecting spam, and keeping out the various bad actors who inevitably show up once a system gains audience is essential for a service to grow beyond its initial early adopters and have a shot at a mass audience. It's pretty common for social services with early promise to crap out after they attract enough traffic to be worthy of spamming and to draw the trolls. If quality falls while the user base grows, the size of the community served becomes self-limiting.

This is why all of the successful social sites have back rooms full of reviewers, scanning every uploaded photo, reading every user flagged post, trying kill-list keyword searches against their own services to look for bad stuff. It's why Craig Newmark describes himself as a "customer support representative" at Craigslist.

People on the outside require more people on the inside. :-)