I just wrote a Ruby script that does pretty much the same thing: https://gist.github.com/865841. I used a simple bloom filter for de-duplication. It's a bit overkill. A simple hash on the twitpic_id will probably work just as well if not better.
What's fun about this is it's 100% javascript. No server load :)
I see lots of photos of someone's passport or wallet and some pamphlets. Some celeb photos, emo shots, a cafeteria. No earthquake footage.