KartOO

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

~ the goodbye ~

kartoo showed up around 2002 as the kind of thing turkish tech blogs would write a single excited post about and then quietly forget. i opened it once on the family pc on a 56k connection, which was the wrong combination of variables. the page loaded a flash applet, the applet asked for a query, you typed your query, the applet drew a map of the result space with blobs and lines instead of a list of links. for the demo it was striking. for actually looking something up it was slower than altavista and less useful than the google front page that had taken over by then.

the company was kartoo sa, based in lyon and founded by two cousins, laurent and nicolas baleydier, around 2001. the underlying engine was a meta-search: it queried other search engines (google and altavista among them in its earlier years) and ran its own layout on top of the merged results. the rendered page was a swf file that drew result clusters as colored blobs with red lines between related links. hovering over a blob highlighted its neighbors. clicking a blob added that term to your query and redrew the map.

the press coverage in the early 2000s was sympathetic. visual search was supposed to be the future for a while. siteshape, grokker, mooter and yippy were all chasing the same hypothesis: that a result map was a better interface than a result list. kartoo was the best looking of them by a clear margin. for a stretch of 2003-2005 it kept appearing in european search-engine-comparison lists as the example of an alternative to the page of ten blue links.

the real story of why visual search did not win is just google. by 2005 google had taken enough of the search market that a meta-search engine querying it was both an additional hop in the user's path and dependent on a competitor's api. by 2008 google had a market-share number that made the question of interface variation purely academic. kartoo kept publishing through the late 2000s and the engine quietly stopped serving in january 2010, which is the closure date the wayback machine snapshots confirm.

~ the chronicle ~

~ the visual part ~

the kartoo interface had a real idea inside it. a search result is not just a list of pages, it is a set of relationships between concepts the pages share. a layout that drew those relationships as physical proximity could in theory help a user navigate a topic faster than scrolling a list. that was the pitch. the implementation suffered from the basic problem that the relationships kartoo could compute from page text were weak signals compared to the relationships google could compute from link structure. the result maps looked beautiful and were not actually more useful than a list of titles.

the second problem was flash player itself. by 2007 the typical user response to "please wait for the swf to load" was to close the tab. by 2010 the runtime was visibly dying. a search engine that required a flash applet to display results was a museum piece at that point.

~ the long tail ~

the company tried a follow-on product called ujiko in the mid 2000s, a search interface with a similar visual flavor but a simpler ui. it did not catch on either. the kartoo domain stopped serving fresh results in january 2010. the cousins moved on to other companies. the wayback machine snapshots of kartoo.com from 2004 still load the flash, but only if you have a browser plugin willing to run swf, which is not a casual ask in 2026.

the visual-search idea kept reappearing. yippy ran until 2021 with a clustered-result format that was the descendant of the kartoo concept without the flash dependency. bing has had a few visual-search experiments along the way. none of them have meaningfully dented the ten-blue-links interface that google standardized and that everybody else copied.

~ what remains ~

kartoo is one of the cleaner examples of a search interface that was admired more than it was used. the screenshots survive in a lot of mid-2000s blog posts about alternative search engines. the engine itself stopped working sixteen years ago. the lesson the writeups tended to draw was that interface novelty does not beat index quality, which is true but less interesting than the actual artifact. the artifact was a flash applet that drew a map of your query, and for a few minutes the first time you saw it, you thought search might actually look different in the future.

@}-->--

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

« previous  ·  random  ·  next »