Shakespeare Social Networks
Now this is cool. Social networks like Friendster are amusing for a while, but like Clay Shirky said the other day, I already know who my friends are. Inputing metadata into social networks has the same problem of metadata input everywhere: it's really hard.
That's why I'm a believer in the value of derived metadata: textual similarity and grouping, binary similarity, full text search, PageRank. Social networking software can use observation to graph the network, like the Shakespeare example and OrgNet's email analysis and Red versus Blue book buying analyses.
What I would love to be doing right now is writing a custom CMS and network analysis program for political opposition research. I think this would be incredibly cool. You could use publically available data, Google and Lexis-Nexis searches, and intern blood, sweat, and tears to keep the data on your targets up-to-date, then use derived networks to map relationships between donors, supporters, and candidates. Would be cool.