At fourteen, Jack Dent was selling apps for 99 cents. He had built them with friends, and the tiny payments felt wonderfully large. “I’m set for life,” he remembers thinking. It was a teenager’s estimate of a balance sheet, but a surprisingly durable introduction to software: make something, put it into other people’s hands, watch what they do with it. The sums changed. The habit stayed.
Dent would later cold-email Stripe and join while still a teenager. At Harvard, he met Joshua Meier on the first day of classes. They became study partners, then took different routes through technology. Dent made products at Stripe; Meier followed the emerging overlap of artificial intelligence and biology. For years, they kept talking. The company they eventually founded, Chai Discovery, can look sudden from the outside. Its beginnings were slow enough to include an entire other career.
That long interval is the interesting part of Dent’s story. He was close to an idea before it was ready, close to a possible co-founder before they were colleagues, and close to a new field while working on payment software. By 2024, he and Meier decided the pieces had finally moved into place. Two years later, Chai had research models, major commercial partners and a very different kind of inbox.
A friendship with a long memory
Dent was born in London. As a teenager he built apps and games, acquiring the sort of practical education that begins with making a thing run and ends with strangers paying for it. At Harvard, where he studied computer science, he met Meier in a demanding class. Dent recalls that they were often in the same courses: he was exhausting the computer-science curriculum while Meier ranged across chemistry, physics and other sciences. Their shared homework became a friendship.
They graduated in 2018, each with a bachelor’s and master’s degree. Dent went to Stripe; Meier pursued AI research at OpenAI, Meta and later Absci. Neither disappeared into the other’s contact list. They checked in roughly every three to six months. Sometimes it was at a Portuguese restaurant in San Francisco, where Dent lived; sometimes over ice cream in New York, where Meier was based. The menus were more settled than the field they were discussing.
Meier would explain what his research group was learning. Dent would listen with the experience of someone who had spent his days turning difficult technology into products. Eventually, Dent said, the work stopped sounding like a toy demonstration. The possibility became hard to ignore. Still, wanting to build a company and knowing when to start one are separate judgments. Their friendship gave them time to make both.
The invitation they postponed
There was an early nudge. After Meier left OpenAI, Sam Altman contacted Dent to ask whether his old Harvard friend might be interested in a company focused on proteins. Dent knew the person; the question was the timing. Meier thought the underlying technology had further to go. The idea stayed in conversation instead of becoming an instant startup. It was an unusually disciplined response to an offer with an unusually famous name attached.
Over the following years, research into predicting molecular structure advanced, as did the methods used to generate new designs. Meier continued in the field. Dent kept building software at Stripe. By 2024, the two thought the window had changed. They contacted Altman again, telling him that they were starting Chai. The first team, joined by Matthew McPartlon and Jacques Boitreaud, worked from space OpenAI provided in San Francisco’s Mission neighborhood.
Chai was founded in March 2024. Its early circumstances were modest by the standards of a field now accustomed to large funding rounds. The team used free cloud-computing credits at the beginning. Four founders had to decide what the new organization would actually be good at: research, software delivery, scientific collaboration and all the awkward work between them. Dent’s title, president, says less about his contribution than the career he brought into that room.
The Stripe lesson inside a research company
At Stripe, Dent became the company’s youngest staff engineer. He worked on Capital, Link and Billing, the sort of products whose complexity is hidden by a clean interface. He recalls teams on Capital and Link growing from zero to dozens of engineers. Growth brings coordination problems as surely as it brings head count. Someone has to remember the whole system while making each piece understandable to the next person who touches it.
Dent carried that concern to Chai. In a public conversation about the company’s work, he described a bug that appeared during model training. The team searched through its Git history, running more experiments to narrow the fault. He estimated that the hunt cost tens of thousands of dollars in computing time. The story ends with a prosaic answer: write tests, keep the architecture simple, and treat research software as software people will have to maintain. There is no romance in a unit test. There is plenty of romance in a breakthrough that survives one.
That outlook fits the peculiar economics of model training. A normal software bug may reveal itself the moment a user clicks the wrong button. In a research run, an error can sit quietly until the experiment finishes weeks later. Dent argues that this makes early care more valuable, not less. A cleanly separated system lets one engineer understand a small part without reconstructing the whole company in their head. It also lets a tiny team change direction without dragging a tangle of old decisions behind it. He had seen the cost of untended complexity at scale; Chai gave him a chance to begin with that lesson already learned.
“Ramping up on any new field is always just a total fight.”Jack Dent, on learning a new research discipline
He speaks candidly about crossing into a field outside his original training. Reading the papers and understanding the frontier, he says, takes concentrated work and brings waves of excitement and misery. His other answer is to build a team with specialists who know things he does not. Chai’s founders include researchers and engineers with different strengths; Dent’s argument is that the work has to travel across those boundaries and eventually reach a partner’s desk.
The night the inbox changed
Chai’s first model, Chai-1, arrived only months after the company began. Chai-2 followed in 2025, with results on designing antibodies that drew attention well beyond the founding team. Dent remembers LinkedIn messages arriving at 2 a.m. Nearly twenty pharmaceutical companies contacted the young firm after that release. His description has the exhilaration of a product launch that has escaped the building: “It was like we dropped a bomb on the field.”
Interest, however, is a starting condition. Researchers need to specify a problem, inspect results and use the models within existing work. Chai chose to sell access to its technology rather than assemble its own portfolio of drug candidates. Dent has described that choice as a challenge to an industry assumption that a company in this field had to own the assets it produced. The distinction is practical: a platform has to work repeatedly for many different partners, and it has to be legible to scientists whose time is spent elsewhere.
The announcements then accelerated. Chai reported a $130 million Series B in December 2025. In 2026 came agreements with Eli Lilly, Pfizer, Novartis, argenx and Bristol Myers Squibb, along with a $400 million Series C announced in July. Those are commitments to work together and to use or evaluate the platform. They are not the same as a finished scientific outcome. Dent’s own story is strongest when it keeps that distinction clear: an elegant model matters; making it useful in a working organization is its own test.
There is another consequence of the platform choice. A customer’s scientists keep their own questions, constraints and judgment. Chai supplies a design system that can be used across different projects. Dent has spoken about the challenge of getting internal research advances into partners’ hands promptly, a concern that sounds more like the roadmap of a software company than the diary of a laboratory. It explains why his career at Stripe matters to this particular company. A result that works once can make a paper. A tool has to endure another user, another target, another team, and a Monday morning when no founder is standing beside the screen.
A patient founder in a hurried field
It is tempting to compress Dent’s career into a tidy arc: teenage coder, Stripe engineer, AI founder. The detail that resists tidiness is the waiting. He did not abandon his friendship with Meier when their work diverged. He did not start the protein company merely because Altman suggested it. He took years to learn how larger products are built, how technical teams grow, and how easily a codebase can lose its shape. When the field moved, that experience had somewhere to go.
The 99-cent apps provide a neat first scene, but the Portuguese dinners and New York ice cream might be the better clue. Chai began with an idea that could survive a three-month interval, then another, then another. Dent and Meier kept meeting long enough to notice when their old conversation had become an engineering plan. By then, the former app seller knew something a fourteen-year-old would have liked: a good idea earns its living only once other people can use it.