Yebol

» see yebol.com as it lived, on the wayback machine

~ the goodbye ~

Yebol was a search engine that did not want to return a list. Type a term and it gave you back something closer to a built page about that term: clusters of sites alongside related keywords, arranged on one screen instead of paginated. The company called it a decision engine and described the goal as a unique homepage for every possible search term. The public beta opened on 27 July 2009 and claimed to cover ten million terms already.

Behind it was Hongfeng Yin, who holds a doctorate and had spent years on artificial intelligence and pattern recognition work, including a stretch at Yahoo. The company sat in San Jose and ran on angel money. The method, as the patents and the press described it, combined clustering and classification over crawled pages with human labeling on top, so that the categories a query produced would look like something a person had organized rather than something a ranking function had sorted.

In Turkey none of this was a live question. By 2009 Google in Turkish was the default to the point of being a verb, and alternative engines were a thing you read about on technology blogs rather than something you typed into. I was working as an engineer by then and my relationship to semantic search was entirely academic, in the literal sense: papers, not a search box.

~ the chronicle ~

~ the pitch ~

The pitch arrived at the right moment and in the wrong decade. Bing had launched six weeks earlier, Cuil had already burned through its own launch cycle, and every trade outlet wanted a story about somebody taking a run at Google. Yebol got written up as exactly that, and the company helped by framing search's future as something other than a pay-per-click layout.

Then it promised a schedule. Coverage of every permutable search term by early in the first quarter of 2010, which for a small team with a crawler and a cluster of classifiers is not a roadmap so much as a wish. Nothing in the record suggests it got there.

~ the last two weeks ~

There is no announcement, no farewell post, no acquisition. What exists is the capture record of the front page, and it is unusually tight. On 26 January 2012 the archive has the full site: search box, suggestion script, the nav row reading Yebol News and About Yebol and Yebol API, a copyright line for 2010. On 8 February 2012 the same URL returns a body with nothing in it. Stub responses continue through September 2012 and then the crawler finds nothing worth keeping for a year.

That window is worth stating against the only other published verdict on Yebol's death. In August 2015 Stephen E. Arnold wrote it up as a goner and placed its disappearance in the 2009 to 2010 period, reading the end of the press coverage as the end of the service. The captures say the engine was answering queries for two years after the noise stopped.

~ still being served ~

Here is the part I did not expect when I started checking. yebol.com answers today. It serves the 2010 design at about 16 kilobytes, the About and News pages come back too, and the footer still reads 2010. The search form is there in the markup, pointing at /a.jsp, with its autocomplete calling /suggest.jsp. Both of those return 404.

So the thing you can visit is the cabinet with the contents removed. A search engine is an index and a ranking function, and what survives of this one is the furniture that was wrapped around them: a logo, a nav bar, a privacy policy, and a form that submits into nothing.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »