262ventures.com

I’ve been messing around with twitter for the past couple of months.  My network is relatively small, keeping small updates on my brother and twit-stalking a few of the VC guys that are big believers in this stuff.  So for me, I use it relatively infrequently.  For others though it’s huge, completely viral, and informative.

Fred Wilson, a VC at Union Square Ventures, has been a big proponent of twit-bots.   For example you can create a twit-bot on wine recommendation (@winetweets) .   This allows everyone interested in wine recommendations to post to the twit-bot and the bot published the recommendation to everyone who follows the twit-bot.  There is some real power of the masses at work here: if you get enough people, surely the wisdom of crowds should allow you to find the cream of the crop recommendations based on number of recommendations alone.  However there is oversaturation of information here as well: what happens when you get 1000, 10000, 1 million recommendations a month from the twitbot?  There is just no way of handling this information.

This leads me to the idea of ‘data-mining’ information from twitter (I’m sure someone is working on this already) . Because of the format, twits are limited to 140 characters, necessitating brevity.  In a comparison with yelp.com, there is no opportunity to supply context for the recommendation that a full review on yelp can provide.  Thus I think it will take some sweet statistical analysis along with some meta-data (some metric for how ‘good’ the recommender is) to really extract some value from this.