Bringing a drug to market is a long, laborious, and expensive process—on average taking about 15 years and $2 billion in research. But Atomwise, which uses artificial intelligence to predict safe and effective molecules for drugs, can discover new medicines using far fewer resources and find new uses for existing drugs.

As a showcase of the company’s capabilities, CEO Abraham Heifets announced on stage that Atomwise recently identified two existing drugs shown to reduce the spread of infection for two Ebola strains—a process that took a week and less than $1,000 in computing costs. “We can complete a project of this caliber every week,” said Heifets. But he’s careful to note the long road ahead for this discovery, including testing the drugs on more strains of the virus, testing its effectiveness on different organs and populations, and then conducting trials in rats, primates, and ultimately humans. With a client list that includes Merck, Johnson & Johnson, and the Karolinska Institute—the body that awards the Nobel Prize in physiology or medicine—it’s clear the industry’s eager to speed up the process to find new treatments.


Self-checkout lanes are supposed to save stores money, but they’re largely a nuisance to consumers. Anti-theft measures make these systems extremely clumsy to use—and ironically enough, debugging them often require the assistance of a store employee. Mashgin believes it can simplify and speed up the process using sophisticated machine-learning and computer-vision algorithms.

The startup quickly got the attention of investors when it announced a large unnamed client representing $50 million of revenue per month. Mashgin says its technology bypasses the need to scan individual barcodes, reducing checkout time to seconds. That means one could theoretically place a cafeteria tray, for example, inside its large, brightly lit box, and the system will be able to detect and ring up the items in mere moments. Mashgin’s network of check-out machines is supposed to get smarter as it learns and recognizes more objects—to the point where it can not only differentiate between an orange and tangerine, but whether an orange is organic or not.


One of the most memorable pitches from this class came from Frederick Hutson, CEO and cofounder of Pigeonly, a company that makes it easy for people to communicate with inmates. Having served five years for distributing marijuana, Hutson began his presentation with a slide bearing 42501-08—his prison number. (His introduction resonated with the audience so much that they interrupted in applause.)

What makes Pigeonly so refreshing is that it serves a very overlooked and underserved market—prisoners and their family and friends—instantly making it more compelling than yet-another on-demand laundry, veterinarian, or marijuana service (all real pitches). Pigeonly is trying to serve a real need, and there’s major market potential too. The US has one of the highest incarceration rates in the world, and existing tools for prison communication are cumbersome to use.


“The space race is back,” declared Bagaveev CEO Nadir Bgaveyev. His monotone, pitch-perfect delivery in a thick Russian accent was enough to get the audience cracking up.

A past Y Combinator event.
A past Y Combinator event.
Image: Y Combinator

Arguably the most ambitious startup of the batch, Bagaveev aims to build a three-ton rocket to transport nano-satellites to low-earth orbit. For nano-satellites to get into space, they have to piggyback off the payloads of large rockets, but the process is long and expensive. There’s currently a two-year waiting list with a backlog of 800 nano-satellites, says Bgaveyev, and the cost of transportation usually comes to $100,000 a kilogram. Bagaveev wants its rocket, which would be powered with 3D-printed engines and is a hundredth the size of a large rocket, to launch 50 nano-satellites weighing 10 to 12 kilograms each year.


DemocracyOS summarizes itself with a very bold tagline: “We are reinventing democracy for the 21st century.” An open-source platform, DemocracyOS aims to make government information accessible to all citizens of the world, hosting government records and giving people a place to learn about and discuss political topics.

Taking its vision a step further, the Buenos Aires-based startup helped create a new political party, Partido de la Red (The Net Party), in its hometown in 2013. Representatives elected by the party agree to cast votes in alignment with the interests of DemocracyOS users. The platform relies on the Blockchain, a public ledger that’s largely used for recording bitcoin transactions, to increase transparency and accountability without exposing users’ identities. In an October 2013 election, the Net Party received about 1% of votes, and it’s hoping to elect its first representative later this year.

📬 Sign up for the Daily Brief

Our free, fast, and fun briefing on the global economy, delivered every weekday morning.