Prismatic co-founder and CEO Bradford Cross doesn’t have big dreams — he just wants to revolutionize the way that we consume media, and then after that he wants to bring his brand of data-powered artificial intelligence to every other form of consumer behavior. Right now, Prismatic is a news aggregation and recommendation engine, similar to other services such as News360 or Zite, but Cross says it is only the first stage in his plan to bring personalization to other aspects of our lives — to become a kind of smart assistant and serendipity engine for the world.
These dreams may be difficult to see when you look at the cramped office Cross and his five co-workers inhabit in San Francisco (a converted apartment that had a mattress on the floor the last time I visited) but they come to life when you talk to him. Of course, that passion also means it can be hard to get a word in edgewise sometimes, as the Prismatic founder holds forth on his vision for the future, or how current recommendation services are failing users, or the state of affairs in the Twitterverse. Like a lot of startup founders, Cross can be a bit of a whirling dervish of ideas — but listening to him is almost always worth it.
Prismatic is interesting in part because it isn’t run by media experts or anyone with journalism training; in fact, Cross has no media background of any kind. His specialty is data science and machine learning, and he says he chose to focus initially on news recommendation because it is an obvious problem — the deluge of information that we are all submerged in requires smart filters — and solving it will help lead the way to other similar problems. And one of the reasons why media players of all kinds would be wise to pay attention is that this data-powered filtering ability could be the key to whatever success the media industry has in the future, as traditional gatekeepers are replaced by crowd-powered and algorithm-driven sources.
“Brad is probably one of the leaders right now in using big data and machine learning to provide consumer services,” says Jason Freedman, his former co-founder at Flightcaster, the startup that the two worked on for several years before selling it last year. Using that knowledge of data science and machine learning, Cross says that he hopes to eventually learn so much about his users via Prismatic that he can start making smart recommendations about almost everything, from movies or books to varieties of scotch or even people:
“The idea is that we become this trusted agent that you rely on to show you things, and over time we can really start to learn a lot about you. We do care a lot about [news recommendation], but we’ve also thought through how it’s a stepping stone to something much bigger. And a lot of what we do in the background, and how we slice and dice data and so on… is relevant across a really wide range of problems.”
Lessons learned from failure
Cross started out in computer science at Virginia Polytechnic Institute, but later switched to finance after he grew bored with his computer courses. Even then, the now-32-year-old founder was interested in large-scale machine learning projects and what they were capable of — and that desire would cost him dearly, at least in the short term. After university, he and a friend started their own investment fund and did fairly well (a 30-percent return in their first year, Cross says) and he parlayed that into a job with a small hedge fund called O’Higgins Asset Management in Miami.
Given his own fund to run, he made even better returns and generated a substantial income for himself, even though he was just 23 at the time. But Cross says he was obsessed with starting his own trading firm, based on his own software. Unfortunately, he admits now that he didn’t really know what he was doing, and the job he took on — building his own data-driven trading algorithms — was far beyond what he was capable of at the time, and he says he learned a painful lesson about entrepreneurialism:
“I basically failed at it, and lost all the money I had made through actual trading, which is ironic. I had no idea what I was doing. It was a pretty hard-core problem, one of the more complicated things in the world to do. It was a very interesting experience that lasted for about 12 months. I knew it wasn’t going to work but I didn’t want to let it go. And eventually I realized I just wasn’t very good at building software.”
Humbled by that experience, the Prismatic founder says he realized that he needed a lot more than just some stock-trading skills and a computer-science background in order to do what he had in mind, so he decided to try and get a job that would allow him to broaden his skill set. Freedman says this is a classic Cross-type move: when he realizes he needs to learn certain things in order to advance, the Prismatic founder figures out a way to get that knowledge. “Brad has this thing where he realizes something is holding him back and he just goes into a 100-percent all-out focused sprint to get better at it,” says his Flightcaster co-founder.
Not long ago, Freedman says he told Cross that if he wanted to hire engineers, he would have to become better known in the hacker community — but his blog was old and stale, and he never spent any time on Hacker News or other popular sites. “So he just went all out, and built a new profile, and wrote all these really smart blog posts, got a new blog design and in seven days had learned everything you need to know about using social media,” says Freedman, who is now running a startup called 42floors.
The building blocks of a smart recommendation engine
After blowing most of his savings on his failed attempt to build a trading system, Cross wound up getting a job with ThoughtWorks, a consulting company that he contacted when his venture was falling apart. The company’s specialty was getting parachuted in to fix things at other companies when they were coming off the rails, and one of the companies Cross spent a lot of time working with was Google. The team spent much of its time on the operating system that powered Google’s storage clusters, but he also worked on consumer products such as Books, Cross says (and met his wife, a Slovak software engineer).
“I didn’t especially enjoy it, but it taught me a lot — it taught me a lot about what not to do. You get an understanding of really gnarly technical issues [and] at the same time, you develop this sixth sense where you can see early on where something bad will lead you. That was a huge experience for me.”
After the Google stint came Flightcaster, a Y Combinator-funded startup. Freedman and his partner Evan Konwiser needed a more technical co-founder in order to get their funding from the incubator (“someone who could actually build what they were talking about,” Cross says) and so they brought Cross in to run the technical side. He designed and built the systems that sucked in terabytes worth of data from airlines, weather services and other sources and made recommendations about routes or flights. “Flightcaster was a very early permutation of this kind of thing [social recommendations], which is the whole reason why I joined,” Cross says.
Once Flightcaster was sold, Cross finally had enough time, knowledge and resources to start his own venture, and Prismatic was born in early 2011 and initially known as Woven. The idea was to use the machine-learning techniques that Flightcaster was based on to filter through and understand social data, and then use that to make recommendations about news and other content. In other words, to create a kind of customized newspaper for users based on their activity on social networks like Twitter and Facebook.
Can a mad scientist reinvent the way media works?
Facebook may have close to a billion users and vast storehouses of knowledge — as does Google — and Twitter may be focusing more on the kind of smart curation that Prismatic is talking about as the service evolves, but Cross says he is convinced that his small company (which was seed funded with $1.2 million from investors including Battery Ventures and Javelin Venture Partners) has a better chance of solving this problem than any of its much larger competitors — even though that might seem to some like an excess of hubris:
“Google won’t get this right, Twitter won’t get this right, Facebook won’t get this quite right, Amazon won’t even get this right — the company that gets it right needs to have it in its DNA. We think this is a Trojan horse into a much bigger thing… in five years time or 10 years time, AI will be all over our daily lives, everything we interact with will be intelligent, and the interfaces to it will be completely different. Backtracking from that very distant kind of vision led us to start in this place.”
There’s no question that the odds of Prismatic succeeding are almost astronomical — not only are there other older, better-funded players going after a similar goal, including News360 and Zite and even Flipboard, but Twitter and Google and Facebook are not likely to let an upstart take away what could be the future of their business.
If sheer devotion to wrestling with big problems is the secret, then Cross has a good head-start: Freedman says he remembers how Cross was always taking the engineers away for 18-hour hackathons to solve pressing issues with the code base. “One day I looked around the office and it was just my co-founder and I,” he says. “Everyone else was at Brad’s house, in his living room, working these crazy hours, but also just having fun. He loves a big problem.” Cross is definitely still a “mad-scientist type,” Freedman says, “but he’s also a lot more than that.”
When we spoke recently, Cross was busy trying to manage a hundred different aspects of being a small six-person startup: a deluge of interest based on the launch of Prismatic’s new iPhone app (an iPad app is coming soon), calls from potential acquirers, and the need to raise financing to fund the growth that he expects — or wants — to see in the future. Is it worth it, I asked? “It’s not worth it unless you have something really important that you’re trying to do,” he said. “And I think we do.”