The Maturing of the Implicit Web

I’ve been fascinated with the notion of the Implicit Web since I determined that I was tired of my computer (and the Internet in general) being stupid.  I wanted it (my computer as well as the Internet) to pay attention to what I, and others, were doing.  Theoretically “my compute infrastructure” should learn, automate repeated tasks (automatically), figure out what information I actually want, and make sure I get it when I want it.

In 20 years, I expect we will snicker at the idea of having to go search for information by typing a few words into a text box on the screen.  It’s way better than 20 years ago, but when you step back and think about it, it’s pretty lame.  I mean, I’ve got this incredible computer on my desk, a gazillion servers in the cloud, this awesome social network, yet I find myself typing the same stuff into little boxes over and over again.  Ok – it’s all pretty incredible given that it wasn’t so long ago that people had to rub sticks together to get fire, but can’t it be amazing and lame at the same time?

Several companies that I’ve got a personal investment in that play in and around the implicit web recently came out with new releases that I’m pretty excited about; each addresses different problems, but does it in elegant and clever ways.

The first – OneRiot – came out with a new twist on using Twitter for search.  OneRiot’s goal is to provide a search engine for the real time web.  To that end, they’ve historically gotten their data on what people are looking at from a collection of browser-based sensors (anonymous, opt-in only).  They’ve built a unique search infrastructure that takes a variety of factors, including number of people on a specific URL in a particular time period, freshness of the content, and typical content weighting algorithms.  A little while ago they realized that people were tweeting a huge number of URLs, mostly via URL shorteners (which are loathed by some very smart people.) Twitter search addresses keywords in the tweet, but it doesn’t do anything with the URL’s, especially the shortened ones.  So, OneRiot built a pre-processor that grabs tweets from Twitter’s API that include a URL, tosses the shortened URL into OneRiot’s search corpus (which expands the URL and indexes the full page text), and then references it back to the original tweet.  It also correlates all tweets with the same URL (including re-tweets) across any URL shortened service.  Now, imagine incorporating any URL data that’s real time that has an API, such as Digg.  Aha!  It’s alpha so forgive it if it breaks – but give it a try.

The second – AdaptiveBlue – has released their newest version of Glue.  Glue is a contextual network that uses semantic technology to automatically connect people around everyday things such as books, music, movies, stars, artists, stocks, wine, and restaurants.  It uses a browser-based plugin to build this contextual network implicitly.  When you are on a site such as Amazon,, Netflix, Yahoo! Finance,, or Citysearch, the Glue bar automatically appears when it recognizes an appropriate object, categorizes it, and let’s you take specific action on it if you want.  Glue has been evolving nicely and now includes the idea of connected conversations between friends (e.g. talk about whatever you like regardless of the site you are visiting), smart recommendations (e.g. implicit recommendations), and web wide top lists of the aggregated activity of all Glue users.

In addition, we’ve finally found a company that we think is attacking a wide swath of the problem of the Implicit Web the correct way, at least given today’s technology. We hope to close the investment and start talking publicly about it early next month. 

For now, I expect the applications around the Implicit Web to continue to fall into the early adopter / you need to see it to believe it category (where it’s harder to explain than just to show).  In the near term, if you are interested in this are, try out OneRiot and Glue – they are both evolving and maturing very nicely.