BookLamp

» see booklamp.org as it lived, on the wayback machine

~ the eulogy ~

BookLamp was founded in Boise, Idaho in February 2007, which is roughly 700 miles from anyone who funds software for a living. Its idea was that a book could be measured the way a song had been measured by the Music Genome Project: not by what the cover said it was, but by what the text actually did. Pacing, density of dialog, how much description sat between events, where the perspective moved. Run that over a library and you can hand someone a novel that feels like the last one they loved, even when the two share no genre label.

Aaron Stanton got the company started by doing something no investor deck recommends. He traveled to Google's offices in California and waited there until somebody agreed to hear the pitch, and the stunt earned enough press that the idea had a name before it had a product. The company raised about $900,000, almost all of it from backers in Idaho. By 2013 the engine was analyzing between 40,000 and 100,000 titles a week and selling the output to publishers who wanted to know what they were actually sitting on.

Then the public project ended without a press release. The last full version of booklamp.org in the Internet Archive is from 10 February 2014. By early April the site was a single page reading Booklamp: "Thank you for everything.", and that file carries its own last-modified date of 26 March 2014, which is a better shutdown date than any of the coverage managed. The reason surfaced on 25 July 2014, when TechCrunch reported that Apple had bought the company for somewhere between $10 million and $15 million. Apple confirmed it the same evening and said nothing else.

~ the dossier ~

~ what the engine measured ~

The pitch to publishers was unsentimental. A publisher's catalog is mostly dark to the publisher itself: thousands of titles whose only machine-readable description is a genre code someone picked in 1998. BookLamp read the books and produced a profile for each one, which made it possible to answer questions like which of our own backlist titles are closest to this bestseller we do not own.

Readers got the toy version at booklamp.org, where you named a book you liked and the site proposed others with a similar shape. It was never a mass-market destination and the public site was always the smaller half of the business.

~ where it went ~

Apple bought discovery, not an app. iBooks in 2014 was a competent store attached to a bad recommendation system, up against Amazon, which had spent fifteen years learning what people read next. Buying a team that could classify a book by its contents is a direct answer to that problem. Whether any of it shipped is a separate question, and Apple has never said.

The founder's own framing held up better than the company did. A novel has structure you can measure before you have an opinion about it, and a reader looking for the next thing is usually describing structure rather than subject. Every recommendation system that has worked since runs on some version of that observation.

~ the goodbye page ~

The farewell page is 718 bytes of HTML on a lighttpd server. Open the source and there is an earlier draft of the message commented out above the one that shipped: Booklamp: "Thanks for all the Fish...". Someone wrote the Douglas Adams joke, thought better of it, replaced it with the plain version, and left the joke in the markup where it sat for years. That commented-out line is the most human thing anybody involved published about the ending, and the timestamp on the file is the only reason this grave can name the day.

@}-->--

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

« previous  ·  random  ·  next »