Harlan Crystal and the Quiet Code Behind 500 Million Downloads

The first useful thing Harlan Crystal did at Pocket Gems was decide what he did not want his young company to become: a room full of people nursing servers. It was 2009. The iPhone App Store was barely a year old, mobile games were learning to walk, and Crystal and Daniel Terry were writing their first title in their spare time. Their office was an apartment above a pizza shop. The arrangement supplied atmosphere, if not glamour. Their game, Tap Farm, supplied the surprise. Released that December, it climbed the charts.

Crystal and Terry had met while studying computer science at Cornell. Both had done artificial-intelligence research funded by the National Institutes of Health, an origin story considerably less obvious than a childhood devoted to arcade cabinets. They came to games as engineers and industry outsiders. In the App Store's unruly opening years, that distance helped. They had fewer inherited rules about what belonged on a phone.

They wrote all the code for Tap Farm themselves. Its premise borrowed the familiar pleasure of tending a virtual plot, but its natural home was the touchscreen in a player's pocket. The pair could ship, watch and adjust. Pocket Gems grew out of that loop.

A small team buys itself time

Success immediately created the sort of problem that can make a young company miserable. A popular game generates traffic without asking whether its creators are ready. Crystal and Terry had chosen Google App Engine for the back end, giving the infrastructure work to a platform designed to absorb it. Hardware, redundancy, load balancing and upgrades became somebody else's problem.

Through June 2011, Pocket Gems had only one server engineer. Its systems were handling tens of thousands of hits per second. Crystal contrasted that with companies where dozens of people maintained server rooms, air conditioners and power supplies for far less traffic. The point was not thrift for thrift's sake. A server not repaired by a game maker returned an hour to the game.

“We might not have been able to scale quickly enough.”Harlan Crystal, on the early Google App Engine decision

That choice became a quiet thesis for Crystal's career: remove the work that prevents makers from making. Cloud infrastructure supplied the conditions that let two coders keep acting like two coders even while millions of players arrived. By the first half of 2011, Pocket Gems ranked among the five leading iPhone app publishers by gross revenue. It had found an audience before it had acquired much bureaucracy.

The office with the inflatable palm tree

By 2014, the apartment had given way to a San Francisco office colorful enough for a TechCrunch tour. Every employee received $100 to add a decoration. Some pooled their allowance for a pool table. Others chose inflatable palm trees, lounge chairs or fencing equipment. The result looked less like a corporate design system than a group project with a generous definition of taste. Crystal guided the camera through it with the air of an engineer pleased that the machine had developed quirks.

The decorations revealed something useful. Pocket Gems was becoming larger, yet its identity still depended on individuals leaving visible marks. This mattered when the company later retired some of the games that had established it and bet on two very different forms of entertainment. Episode turned the phone into a branching story engine. War Dragons pursued synchronous battles, guilds and graphically rich 3D worlds.

Neither was a tidy sequel to Tap Farm. Their coexistence became the more interesting achievement. A studio known for light tap games taught itself narrative design, creator tools, a proprietary graphics engine and real-time multiplayer. Episode invited people to make and consume stories. War Dragons asked them to command armies and coordinate attacks. One company had built two canvases.

Reading the chart: These are cumulative company download milestones, not annual totals. They show scale rather than a smooth year-by-year trajectory.

A CTO in the support queue

A modest anecdote about Crystal is especially revealing. During the company's shift toward Episode and War Dragons, he often spent weekends answering player troubleshooting tickets. A founder in a support queue might look like a poor use of executive time. It can also be a fast route to the truth. A bug reported by a player has already escaped the reassuring atmosphere of a planning meeting.

Pocket Gems' current culture describes the same instinct in more polished language. Teams are encouraged to share work before it feels finished, test often, look for clarity before polish and treat feedback as more trustworthy than personal instinct. Crystal observes that people dislike failing in public or exposing imperfect work. The company pushes rough ideas into the open anyway.

“Most people don't like failing publicly or sharing things that aren't yet perfect.”Harlan Crystal, on iteration at Pocket Gems

There is a useful modesty in this method. It assumes the product knows things its maker does not. A launch becomes an instrument rather than a verdict. The player taps, leaves, pays, returns, creates a story or files a ticket. The maker receives an answer and revises the question.

Before the farm, the laboratory

Crystal's route to the App Store passed through a collection of places where software had to interpret a complicated world. At Cornell, he earned a bachelor's degree in computer science and continued into graduate study. He and Terry spent several years doing artificial-intelligence research supported by the National Institutes of Health. The surviving accounts do not turn this period into a neat prophecy about games, which is useful. Careers usually make sense only after somebody edits them backward.

In 2007, Crystal worked as a computer-vision research engineer at the Blindsight Corporation. He later served as a teaching assistant at Harvard and moved into engineering leadership at Smartleaf, a financial technology company. He also founded OurTaste in 2006. The sequence is less a straight road than a tour of software's different obligations: recognize the visual world, explain technical ideas, manage an engineering team, attempt a company. When a new pocket computer appeared with a marketplace attached, he had several kinds of practice ready.

Pocket Gems called its early approach “mobile first.” The phrase has since been sanded smooth by years of corporate use. In 2009 it carried a real constraint. A phone had a small screen, limited memory, uncertain connectivity and an interface built around fingers. Crystal and Terry could not assume that a successful web or console design would survive the trip. Tap Farm made the screen itself part of the pleasure. Later games pushed different edges of the device: Episode made reading interactive, while War Dragons asked mobile hardware to render rich 3D battles and keep distant players in sync.

This is where Crystal's combination of research and engineering leadership becomes legible. Research rewards a good question. Engineering makes the answer operate under pressure. A game company needs both, plus the humility to accept that the player may ignore the feature everyone admired in the meeting. Pocket Gems paired creative bets with analytics from its early years. Imagination suggested a world; behavior reported whether anyone wished to stay there.

The value of beginning again

In 2017, Tencent invested another $90 million in Pocket Gems after an earlier reported investment of $60 million. The company had reached 300 million downloads, and Episode users had viewed billions of episodes. War Dragons had staged billions of battles. Capital gave the studio room to work on longer, riskier projects, but the more durable resource was the ability to discard a comfortable formula.

Crystal's public biography is spare. It lists the role and lets the work carry the detail. Before Pocket Gems, he founded OurTaste, worked in computer vision at the Blindsight Corporation, taught at Harvard and led engineering at Smartleaf. His description on Pocket Gems' own site is simpler: a “guy who loves to make things.” The phrase would look affected from someone farther from the work. From the CTO who answered weekend tickets, it feels accurate.

Seventeen years at one startup can sound like constancy. Crystal's record is more restless. The first hit was a farm. Then came a storytelling network, dragons rendered through a proprietary engine, and a company whose current releases still refuse to share a single genre. Late in 2025, Pocket Gems launched Episode: Reality Stars, combining storytelling with game mechanics while keeping Episode and War Dragons running.

The numbers are now large: more than 500 million downloads and over $1 billion in product revenue, according to Pocket Gems. Yet the clearest line through the story remains small. Two people wrote a game. One infrastructure choice bought them time. A founder took support tickets home. Rough work met real players. Then the next version went out.

Technology careers are often narrated through breakthroughs, as though progress arrives with trumpets. Crystal's is better understood through friction. Find it. Remove it. Show the unfinished thing. Listen when it complains. The pizza shop downstairs eventually disappeared from the picture. The appetite to make did not.