Probabilistic data structures for web analytics and data mining – A lot of the time we're processing massive amounts of data and producing very detailed intermediate results, only to throw almost all that detail away because all we want is a much smaller summary of the data's properties. I've got a lot of mileage out of approaches like this that cut out the middleman and produce much more manageable intermediates by throwing away parts of the data early when they are unlikely to be significant.
Twenty-four hours in a VC's life – Professional investors theoretical economic role is as discoverers of new information, finding opportunities for deploying other people's savings in useful ways. It's interesting to see some of that information transfer in action.
Haptic labs soft maps – I would have loved one of these quilt maps as a kid.
TopCoder – Using the competition model to drive software development.
American healthcare fraud and scalable investigative reporting – I'm always excited to see the new data techniques being used for more than ad targeting, and journalism is one of the most promising areas where they can make a difference.
Pete Warden's blog 