The first thing Marwan Roushdy tells me when we sit down is, “Our goal is to build a social app. Our goal isn’t to solve a problem.”
Dressed in a white T-shirt covered with what looks like Sharpie pen scribble, Roushdy is showing me his startup’s new photo-sharing app, Swipe. We’re meeting at a live-work apartment building in South Park, San Francisco’s most desirable startup neighborhood, in a sort of living room with sparse modern furniture that looks like it’s been staged by a real estate agent.
I’ve been drawn into a meeting with Roushdy because his investor Jonathan Teo wrote me an email saying, “I haven’t seen a product that’s growing this fast in a very long time.”
Photo-sharing apps have magnetic pull for tech startups and investors. Much as it’s clear that their success is a total crap shoot, people keep lining up to place their bets and hope for the next Instagram.
And that’s kind of Roushdy’s point in saying his goal is not to solve a problem. He co-founded a company called Complex Polygon with another young developer named Addison Hardy. They want to build something that people like and use and that helps them connect to one another. Their plan is to spend four months on each app, releasing it as soon as possible and then refining it. Swipe is the second one.
Here is how Swipe works. It’s an iPhone app (Android is on the way). You subscribe to other users because you know their Swipe username or have them in your phone book. When you open the app, you are presented with a stack of photos and autoplaying videos, which you can view one at a time. It’s not specified who posted them, but they are all from people you follow.
Just like Tinder, you swipe right if you like something, left if you don’t, or up to reply by writing on the photo. The people who posted the photos are notified if you’ve liked them or replied. You can only look at each thing once.
Every so often, instead of showing you a picture, the app asks you to invite one of your friends who has multiple contacts using the app. You can swipe right or left on that too.
Swipe reuses technology from Complex Polygon’s first app, Tag, which was a photo-as-a-conversation-starter app, but “wasn’t cool enough or new enough,” according to Roushdy. When his team of three sent out Swipe to testers recently, they forgot to annotate photos with the poster’s username. People seemed to like the pseudo-anonymity, so they left it. For photo apps, it’s all about finding one trick or twist that just works.
Roushdy got his start making apps during the Egyptian revolution three years ago, when he entered and won a hackathon on a whim. He now calls this period his “10 seconds of fame,” after the effort landed him starring roles in articles about tech gumption in Egypt in the New York Times, on CNN and elsewhere.
These days, Complex Polygon pays for its fancy South Park office and its minimal staff with $1.7 million raised last year in a round led by First Round Capital. Other investors include Khosla Ventures, Greylock Partners, Lowercase Capital, CrunchFund, AF Square and SV Angel. In Silicon Valley, that’s very much the in-crowd of people who invest in social apps.
As Roushdy and I are talking, I take and post some pictures of the staged coffee table, the window, and a selfie of me and Roushdy. Seconds later, I get Swipe push notifications that Teo, wherever he is, has liked the photos. The two of them are putting on a very good show, though this alpha version of the app has run my phone battery down unnervingly fast. Swipe is fun.
Roushdy wouldn’t give actual user numbers, but he said Swipe’s viral gimmicks seem to be working pretty darn well. In these very early invite-only days, the average tester uses Swipe four times per day and sends out at least one invitation per session. A version of Swipe (with the battery-hog problem addressed) is available to the public as of this week.
Meanwhile, down in Palo Alto, there’s another live-work space. It’s a rental cottage with some desks crammed into a living room that looks like grad student housing. At those desks is another small, promising mobile app team. It’s composed of two artificial intelligence researchers, Zak Stone and Nicholas Pinto, who have PhD’s in the field from Harvard and MIT, respectively.
Stone and Pinto are the co-founders of a startup called Perceptio, and they do have a problem they very much want to solve: Making phones do super advanced computations without storing their owners’ personal data in the cloud. Their technique of choice is deep learning, the particularly trendy subset of machine learning that’s based on models of how the human brain recognizes patterns.
Stone and Pinto are the kind of smart that makes you feel like you are on a game show, trying to quickly process what’s being discussed and keep up with witty banter at the same time.
And yes, they are also making a photo app. It’s called Smoothie.
Stone and Pinto have spent more than a year noodling around the best ways to run neural network algorithms computations locally, on phones. Their first focus was facial recognition.
Existing facial recognition solutions, like those offered by Facebook and Google, require uploading photos via an Internet connection to someone else’s server farms. Up in the cloud, there’s enough processing power to smartly match a whole bunch of pictures of someone’s face with their name, and in doing so train a system to recognize future captures of the same person.
While Stone and Pinto realize they are paddling upstream, they think people should have a choice to control their own stuff on their own devices.
But in order to train software to recognize people using only a phone, Perceptio needs lots of pictures of people to be on that phone. “We started capturing short videos instead of photos to provide more training data to our algorithms,” said Stone.
That’s when deep learning took a sharp turn toward photo sharing. “It turned out that those short photo-videos were surprisingly lively and fun to watch,” Stone said, “and we had even more fun remixing them by sliding back and forth on the screen DJ-style.”
Smoothie isn’t a deep-learning application. It is a byproduct of a deep-learning application, and one that’s available today, through a sort of Perceptio spinoff called Pocket Pigeon Society, run by the same two guys. The deep-learning stuff isn’t fully baked yet. Both Perceptio and Pocket Pigeon Society are self-funded.
Here is how Smoothie works. You press and hold a button to capture a video. Once you’re done, you scrub around backward and forward on the video as if you’re mixing a record. Then, a video of that remixed playback is captured and you can save it as a GIF or video. That’s it.
There’s no viral mechanism to Smoothie. This weekend, I made some sweet remixed movies of my dog jumping into a river to retrieve a stick. When I shared one on Instagram and Twitter, there was no indication of where it was made.
That’s because at this point Smoothie isn’t actually a social app — given its founders’ philosophical bent, there is currently no way to share your creations to their servers. But Stone said something like an in-app feed might be coming, to encourage people to make and share really cool GIFs.
In fact, Stone has become quite an advocate of GIFs as an art form. “So many people enjoy animated GIFs these days, but how many people create them? We want to change that ratio with Smoothie — in our ideal world, people would make as many animated GIFs as they watch,” he said.
But while Stone’s core focus is still deep learning, Swipe’s Roushdy is more serious about the value of social apps.
Roushdy is exceedingly aware that people want to write off what he’s doing as silly. He thinks it’s not.
“Social apps are used millions of times a second. It lets people get closer together in a way that’s very difficult in this fast-paced life,” Roushdy wrote me in a late-night email. “Technology makes life so much faster and people so much busier that we need constant innovation to sustain strong connections with people.”
And sure, there’s merit to both photo-sharing app aspirations. Where Smoothie is the fluky byproduct, Swipe is the latest widget from a would-be app factory. But that doesn’t mean either of them will have staying power. Much as Swipe would like to make its own viral luck, the irony for these trend-catching apps is that you have to make something your users want to keep around. And that’s not for me, or them, or the VCs to decide.
This article originally appeared on Recode.net.