There have been a handful of articles this week talking about using Twitter for algorithmic trading.  StockTwits has been successful in aggregating over 80,000 stock traders who want to share stock trading ideas.  Tweets which have a ticker symbol preceded by a $ sign ($AAPL, for example) go into the StockTwits stream for users to interpret as they like.  For algo trading, the entire Twitter stream might be used as a source of news that might affect a market (report of an earthquake) or to decipher sentiment about a major new product release – the herd is Twitter
There are a number of challenges with this; all of which will likely be overcome, but still its hard to imagine consistently extracting alpha from the Twitter stream right now.  First problem is that service is just not that reliable, and you're not sure whether or not it's gone down unless it's just stops.  Low-latency Twitter is some time away.  Second, although the messages are only 140 characters long there are plenty of them and more than 99% of them at any moment will be noise.  Lastly, as a source of real-time news Twitter has already proven itself but as a source of real-time sentiment I'm not certain there's any proof.  But as I said upfront, these challenges will likely be overcome – 10 years ago Bloomberg and many others scoffed at the internet as a channel over which to deliver market data.

From Wall Street & Technology: Algo Traders Connect to Twitter

From the Telegraph: Hedge Fund Managers Betting Twitter Will Give Them An Edge In Rapid Trading

From the StreamBase Event Processing Blog: Trading on Twitter: Opportunity, Danger, or Folly?