I'd use it for running optical character recognition on millions of pages of government documents to make public information truly free, at http://citizenaudit.

I'd use it for running optical character recognition on millions of pages of government documents to make public information truly free, at http://citizenaudit.org/ http://citizenaudit.org/. But I really just need the 26 cores with a master node feeding documents for that, not actually map-reduce.