DataSift – UK startup focused on making it easy to build your own tools on top of massive social media streams like the Twitter firehose. Seems a bit like Yahoo Pipes for social data, without the visual interface, and could open up the area to a much wider audience of developers.
The Doctor vs the Computer – A thousand-character limit on descriptions in medical records is so obviously arbitrary and unneeded, it hurts. Websites that have code to complain about spaces in credit-card numbers but somehow can't strip them out are bad enough, but here the bondage-and-discipline over-specification could kill people.
Trouble in the House of Google – Google's had massive success because they realized that inelegant statistical methods of detecting things like spam, plagiarism and relevance work a lot better than more elegant traditional semantic/AI techniques. Unfortunately, the black hats have figured out that there's no statistical technique in the world that can truly rate the quality of a page. Google's relying on statistical measures that used to correlate with that quality, but as the bad guys mimic those more closely, they are tricking the search engine into believing spam is the real thing. We need more inputs, whether that's a return of some kind of manual rating system, data from social networks or click-through rates.
This isn't a post about Facebook – Mourning the rise of a service that's a closed system, instead of the openness of Google. I'm not as pessimistic as Paul, I think that Facebook is demonstrating how much people want tools that reflect their off-line social world and behaviors, and once the open world absorbs that lesson, we'll see a new wave of competition for the social network. That competition will have to be more open in a technical sense, just because that's such a tempting way to get early traction.