Banjo

~ the last words ~

Banjo fed on geotagged posts, which meant it only worked where people bothered to geotag things. In 2011 that described a handful of American cities and very little else. Turkey had Twitter and it had Facebook, but the check-in habit the app needed never took here past a thin layer of people doing it as a novelty. There was nothing for Banjo to read from Ankara, and the version of it I know is the one that came later and had nothing to do with bars.

The origin story was ordinary and good. Damien Patton was stuck in an airport, found out afterward that a friend had been sitting in the same terminal, and built the app that would have told him. It shipped on 22 June 2011 and did one clean trick. It pulled in the public location-tagged output of Twitter and Foursquare and whatever else was broadcasting, then showed you what was happening within walking distance. Half a million downloads by that December, and a million users by April 2012. TechCrunch called it the creepy and awesome cyber-stalking app, which was affectionate at the time.

Then the Boston Marathon bombing happened in April 2013, and the company noticed that the same index which finds your friend at a concert also finds every phone that posted from a street corner during an emergency. Banjo Enterprise launched in January 2014. The consumer app kept limping along for a while and then stopped mattering, and the money confirmed the new direction: sixteen million dollars in March 2014, then a hundred million from SoftBank in 2015.

What that money bought was a product called Live Time Intelligence, pitched to police and to state governments as a live feed of everything happening inside their borders. Utah signed a five-year contract worth around twenty one million dollars and paid out more than three million of it. The state was wiring traffic cameras and 911 dispatch into a private company's engine and the deal had gone through with very little public argument.

~ the record ~

~ what it became ~

The reframing is worth sitting with, because nothing in the technology changed between the two Banjos. The consumer version and the surveillance version were the same index of public posts pointed at a different customer. What moved was who got to run the query, and how many other feeds were plugged into it once the customer was a government rather than a person looking for a friend.

That is the general shape of the location-app generation. Almost every one of them was built on the premise that broadcasting your position to people you liked was fun, and almost none of them survived the discovery that the same data has a much richer buyer. Gowalla and Brightkite died before anyone had to think about it. Banjo lived long enough to find the buyer.

~ what really happened ~

In April 2020 a reporter published what Patton had done as a teenager. He had been a member of the Ku Klux Klan and had taken part in a drive-by shooting at a synagogue near Nashville, a fact that had been in court records for decades without anyone connecting it to the man now selling surveillance software to a state government. The Utah attorney general suspended the contract almost immediately. Patton resigned on 8 May 2020.

The audit that followed found the second thing. Utah had commissioned a privacy review and the reviewers went looking for the artificial intelligence at the center of the pitch. Company representatives told them the system did not use anything meeting the industry definition of the term. The state had paid three million dollars for a product whose central technical claim was not true, and the mechanism that finally exposed it was a background check that should have run before the contract was signed.

~ the long tail ~

In February 2021 the company reopened as safeXai with the chief technology officer in the chief executive's chair and Patton holding a minority stake. The name Banjo went away and the software did not. That is the ordinary ending for a company in this position, and it is the reason a shutdown date for something like this is always a little bit of a fiction.

The part that outlives the brand is the precedent. A state government bought a real-time view of its own population from a private vendor, on the strength of a claim nobody checked, and the thing that stopped it was journalism about the founder rather than any procurement control. Everything about that arrangement is still available to buy.

@}-->--

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

« previous  ·  random  ·  next »