Fundamental Oracle flaw revealed – This is a fascinating piece of detective work on a bug, but also a cautionary tale of how even the most conservative assumptions can be proved wrong as data processing speeds and volumes grow.
Extracting structured data from Common Crawl – Shows exactly why I'm so excited by the potential of Common Crawl. Even just a list of all the hcard records from five billion web pages is going to be an amazing research resource, I have plans for doing fun things with the street addresses already.
Travel itineraries with long-exposure photos – I love the way the students used analog techniques to produce a high-tech looking visualizations.
Social Graph and Needlebase are dead – Google's API for publishing unified public profile information to developers never really caught on, but it's a shame to see it vanish. Needlebase's shutdown is less surprising, it always seemed likely to be useful more internally to Google, but I'm still sorry to see it lost to the outside world, it was a great tool.
The Apple logo in unicode – It's great to see how convoluted and political something as seemingly-simple as defining an international character set can be.