Twine

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

~ the last words ~

twine belonged to the branch of web 2.0 that believed the semantic web was going to happen. tim berners-lee had been saying so for years. radar networks, nova spivack's startup, was one of the few teams that tried to build something normal users could touch on top of it. what they built was a bookmarking site with an rdf brain, and it survived two and a half years before the market decided it had asked the wrong question.

the invite-only version opened on 19 october 2007. the public version opened on 21 october 2008 with the framing "wikipedia meets facebook". you saved links and articles into a "twine" (a shared workspace built around a topic), and behind the scenes the system tagged everything semantically. it read the text, extracted concepts, tied them to a taxonomy derived from wikipedia. the same information could be filtered by concept rather than by folder.

at peak in mid-2008 twine had around 34,000 active users. this was smaller than delicious in the same window and much smaller than digg. the product was expensive to run because every save triggered nlp processing on server infrastructure that predated cheap cloud compute. the semantic layer was invisible to the casual bookmarker, and the technical audience that would have appreciated it was mostly happy with delicious tags.

i was in the last stretch of engineering university in 2008 and my bookmark habit was firefox's built-in bar plus delicious. twine was one of those products i read about on techcrunch and never signed up for, partly because "semantic web" was a phrase that already meant "conference paper" more than "product". by the time it seemed like it might be useful, evri had already bought and closed it.

~ the record ~

~ the pitch ~

radar networks pitched twine as a semantic desktop for the web. instead of tagging manually, you saved a page and the system inferred the tags. instead of folders, you had twines, each one a small graph of related items.

the rdf triples underneath were the point. in principle, everything a user saved could be queried like a database. in practice, most users treated the site as a slower, weirder version of delicious. the concepts got in the way of the workflow more often than they helped.

~ the fall ~

by early 2009 the direction was clear. facebook had passed 200 million users and twitter was in its early growth curve. the "web 2.0 collaboration platform" category was being flattened by simpler tools. radar networks had raised about $24 million in venture funding for a technology that produced small user numbers and an expensive stack.

evri, a seattle semantic-content startup backed by paul allen's vulcan capital, acquired radar networks and twine in march 2010. the announcement said the technology would be folded into evri's content platform. what actually happened was that twine.com became a redirect to evri.com on 14 may 2010, and evri itself was mostly shut down by october 2012 after laying off its staff.

~ the long tail ~

the semantic web thesis did not die with twine. it moved. schema.org rolled out in 2011 as a joint search-engine effort to standardize the same rdf-flavored metadata twine had been trying to build user interfaces for. google's knowledge graph launched in 2012 on top of it. everything twine had been trying to do at the user layer got absorbed into search infrastructure instead. the visible bookmarking product died, and the invisible metadata infrastructure that motivated it became a load-bearing part of the modern search engine.

@}-->--

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

« previous  ·  random  ·  next »