San Francisco - Field notes from a game still loading
Ege Kaan Duman has an origin story for his life in games, and it begins with the minor agony familiar to every younger cousin: watching someone else hold the controller. The machines were a PlayStation 1 and a SEGA Dreamcast. The player was his cousin. Duman was captivated. By 2013 he was trying MonoGame, a framework with rather less instant gratification than watching Sonic sprint across a television. The important thing was that he had crossed the glass. Games were no longer objects delivered by mysterious adults in faraway studios. They were things a person could make.
More than a decade later, Duman is trying to make that crossing shorter for everyone else. Orca, the San Francisco company he co-founded with Ali Kavoosi, lets people describe Minecraft mods in ordinary language. Ask for a rideable creature, a peculiar item or a voice-responsive wolf; the software writes and assembles the pieces. It can place the result on a managed server where friends may join. The grand technology is artificial intelligence. The humbler product promise is relief from launchers, version mismatches, mod folders and configuration files - the chores that arrive before the fun.
The education of a useful outsider
This is not the résumé one expects behind a Minecraft client. Duman grew up in Istanbul, then joined Duke Kunshan University’s inaugural undergraduate class in China. He studied global health on a biology track. He worked across research groups, co-authored papers with scientists and philosophers, helped start a student newsletter, and led interviewers for a student podcast. In 2021, he captained the team that won the Emory Morningside International Global Health Case Competition.
That same year brought a milestone that followed him into almost every subsequent biography: Duman became Duke Kunshan’s first Rhodes Scholar. He was also the first person from Turkey selected through the Rhodes Global Constituency. The distinction sent him to Oxford, where he completed master’s degrees in Medical Anthropology and Translational Health Sciences. It also gave him language for the kind of attention he was already practicing - how people navigate systems, where institutions create friction, and what behavior looks like outside a spreadsheet.
At the time of his Rhodes selection, Duman spoke about a future he deliberately refused to over-specify. “Only my goal to help the world is concrete,” he said. The sentence is broad enough to accommodate an astonishing number of LinkedIn pivots, but it also reveals a useful willingness to let experience redraw the plan. Oxford did. There he met Kavoosi, another person with a taste for games. Their shared interest eventually became the company that joined Y Combinator’s Summer 2024 batch.
The company first appeared as Simplifine, an AI workspace for researchers. Later, it shifted its attention toward games and became Orca. The change looks sharp from a distance: research software one day, undead real-time strategy factions the next. Up close, the method is recognizably continuous. Duman keeps choosing complicated creative systems, studying where people get stuck, and asking how software might remove the ceremony around the work.
One path, four rooms
The user is not a chart
One of Duman’s most telling startup stories includes a cross-country flight. He traveled from San Francisco to Atlanta to meet an early user, Caglar Coskunpinar, a systems engineer who had moved into full-time game development. They tested Coskunpinar’s game before release, discussed game development for hours and reviewed a game jam they had entered together. Duman’s conclusion was almost stubbornly analogue: watching a person use the product, seeing the pauses and smiles and the precise moment the app breaks, can be worth more than an analytics dashboard.
There is the anthropologist again, now conducting fieldwork beside a keyboard. Early software teams love instruments because instruments turn confusion into tidy columns. Observation returns the mess. A user hesitates over the button you thought was obvious. They ignore the feature that consumed Tuesday. Then they imagine a possibility no planning session produced. The trip to Atlanta was customer development, certainly. It was also respect: if a person is giving your unfinished software their time, the least you can do is enter the room.
The social part is the product
Orca’s current focus is Minecraft, a world whose players have treated modification as a native verb for years. Its website presents a direct loop: describe a feature, let the agent create it, then run it on a server. Every account is offered an 8 GB server and daily free usage. Orca handles plugin installation, version upgrades, crash logs and library clashes. The interesting design choice is not simply that AI can produce code. It is that creation and distribution sit together. The thing you make is already leaning toward other people.
Duman describes Orca as a platform for social play. That phrase prevents the project from becoming another lonely prompt box. A game engine traditionally helps a developer manufacture an artifact. A social creation tool has to care about the invitation: How quickly can a friend enter? What happens when a strange idea becomes a shared joke? How much technical housekeeping can disappear before the act of making begins to feel like play itself?
The answer is still under construction, sometimes with comic understatement. In one public update, Duman showed a commit from Kavoosi labeled “added physics stuff.” It changed 64 files, inserted 12,400 lines and removed 451. Duman was using the work to iterate on a new real-time strategy game with an undead faction. No pitch deck could improve the commit message. Startup life often consists of an enormous hidden diff wearing a tiny verbal hat.
64 files changed
+12,400 insertions
-451 deletions
The humor matters because Orca is built around an activity people pursue for pleasure. Duman’s public biography says, simply, “I enjoy fat cats.” His personal website describes him as a poet, ethnographer and builder and keeps a sprawling catalog of books, films and anime. These details do not prove a product thesis. They do make the thesis easier to understand. This is a founder interested in worlds, the people inside them, and the small absurdities that keep ambitious work from becoming unbearably solemn.
Reflexes, not merely replies
A recent Orca experiment gave a Minecraft dog an AI agent that could hear a voice, decide what it meant and act. Duman argued that games need more than characters that can produce plausible dialogue. They need quick action based on immediate feedback. Intelligence in a game is experienced through timing: the companion sits when asked, attacks at the right moment, or fetches before the player has forgotten the joke.
That emphasis offers a glimpse of Orca beyond automatic mod assembly. The company is exploring a world in which players can build behaviors through conversation and then meet those behaviors in motion. The maker does not have to begin by learning an engine’s ontology. They can begin with intent. The machine translates. The player tests. The friend laughs when the dog digs up diamonds instead of behaving with dignity.
Duman recently reported that Orca’s daily active users had grown 292 percent in a week after a June launch. A weekly percentage at an early startup is a snapshot, not a coronation. His own interpretation was suitably plain: there is no growth hack; build a good product. The more durable evidence will be whether people return to make a second odd creature, bring another friend or keep a world running after the novelty of the first prompt has passed.
What the detour bought
Duman’s route can be mistaken for indecision if a career is judged only by the consistency of its nouns. Biology. Philosophy. Anthropology. Research tools. Game engines. Minecraft. The verbs tell a steadier story: observe, connect, translate, build. His Duke Kunshan mentors described an unusual combination of intellectual ambition and collaborative spirit. Those qualities travel well. A founder needs appetite for a large problem and enough humility to watch somebody else misunderstand the solution.
Orca now has a narrow door: Minecraft creation without the usual pile of setup. Behind that door is a larger question about who gets to make games. Conventional tools reward patience with abstractions. Orca asks whether intent can come first and technique can arrive through the system. It will still need to survive the practical difficulties that every magical interface eventually meets: ambiguous instructions, unruly dependencies, multiplayer edge cases and players with wonderfully inconvenient imaginations.
But Duman has spent much of his life crossing between fields with incompatible vocabularies. He knows that translation is rarely perfect and often worthwhile. The child beside the Dreamcast wanted access to the world on screen. The founder in San Francisco is building a shorter route through the glass. If Orca works, the next child may not need to wait for the controller. They can describe a creature, open the server and send the link.