I like Vast

Writing about VAST a whole week after it launched... am I slow or what?... actually I wanted to play with it for a while before write something. Furthermore, this is obviously not a news blog so I dont feel the pressure to type up something cursory just to get on memorandum .. Working at eBay means that I track this space pretty closely so I am vested in digging around...

Vast is a classified search engine Google would have been proud to create. (wish is more like it given the mess that is googlebase) Edgeio is a classified search engine Yahoo would have been proud to create. And there lies the fundamental problem with Edgeio... its a worklow engine that relies soley on listings from the “edge” but the algorithm is still “core” focused... in the end, it is still a destination site that relies on its community to filter results, rank results, and filter spam. That is all nice a good for Craigslist who for some miracle (or first mover advantage) built a passionate community (I can say the same for eBay). Betting on that once again is hard to do... In the end, you either go edge all the way (like VAST) or focus on building core advantages and raise huge/thick walls to keep people like VAST out of your nicely manicured lawns (again like Craig). Doing it half way doesnt seem strategically consistent to me.

Vast on the other hand, truely aggregates from the edge, and its algorithm for relevancy takes into consideration data/content/link structure that is outside of its core data set. The incredible (and coolest) part is its stated ability to capture name:value attribute pairs as it crawl rather than as people submit listings like GoogleBase. Furthermore, reading Naval’s comment on VAST around the blogosphere, they seemed to have built somesort of machine learning/training scheme to use humans to train algorithms to recognize and build histograms of attribute pairs (or tag pairs). Some what like what MySimon did back in the days.

I had always wondered why Google didnt go this route like VAST, instead of relying on critical mass on its core platform to build its taxonomy universe, they should have aggregated from the edge (it from the entire web). The quick answer is that the GoogleBase team are mostly ex-eBay people who only had experience in building walled gardens... but I suspect that is not the only reason. I think GoogleBase tried to be too broad by trying to chew off the superset of problems (commerce, content, search for aliens, artificial intelligence, etc). This very hubris created an opportunity for VAST to creat a set of vertical commerce search engines that out performed GoogleBase...

I wonder what would happen if VAST was given the opportunity to crawl Gbase’s content with out the GBase specific attribute pairs exposed. Would it be smart enough to extract the majority of the tag pairs? If so... Google should be thinking about buying them just about now...