I've been holed up, building this for the Sunlight Foundation Mashup Contest. My academically-minded friends will no doubt be horrified by the imprecision of the visualization — I'm really supposed to have done this by building a proper similarity tree, in which joined nodes become collapsed into their average. Those joined pieces are then treated as single nodes and everything is recomputed relative to them; repeat until every node has been added.

But that would've been hard! And I'd already written my algorithm. Nevertheless, thanks to son1 for pointing me in some interesting directions.

You should also head over to Tim's — he's got an entry for the Sunlight contest, too, as well as a cool 2008 Presidential aggregator that he wrote with Rich (an improved version of which is coming soon to a glossy Washington-centered magazine near you). He's also written some smart Holovaty-ish thoughts on the the overlapping future of programming and journalism. To wit: hey, reporters! It's time to learn XML, regular expressions, some statistics and WWW::Mechanize (or an equivalent web-scraping package). I'm serious. The tools will get easier, but they're already accessible and applicable enough that you could greatly benefit from learning to use them.

Speaking of Mr. Holovaty, Adrian's busy these days getting my friends into trouble, so he's on the lookout for a programmer-journalist at Washingtonpost.com to share the load. Sounds like it'd be a pretty cool job.