Brett Granberg Learned to Stop Adding Features

The first thing Brett Granberg and his colleagues built at Vannevar was clever, specific and almost entirely unwanted. It took Arabic text trapped inside an image and turned it into searchable English. The team made a command-line interface. They added translation. They built a user interface, named-entity extraction and search. The machine kept acquiring features while the market maintained the serene composure of a cat being offered a bath.

Granberg remembers roughly 40 or 50 sales meetings before a different question began to repeat. Could the product do Chinese? Could it do Russian? Before a trip to Washington, a contact wrote to say that Chinese aggression was the problem on their desk. The team spun up a Mandarin model over the weekend and demonstrated it the next week. Someone finally leaned forward.

Then came the sentence every founder hopes to hear, delivered by a person already working in another country: I need access now. How do I buy it? The product was still, in Granberg's description, half working. It had nevertheless discovered the indispensable feature: relevance.

The education of a product founder

Granberg tells the story without varnish. The original vision, he has said, was “half-baked.” The team mistook an absence of demand for an insufficiency of features. Six months went into a product that had received little user feedback. An existing OCR service might have answered the core question in a day. This is not the customary founder mythology, in which foresight arrives immaculate and everybody else catches up later. It is more useful than that.

The correction became one of Vannevar's principles: work on a problem that ranks among a customer's top three. The bar matters because buying new technology inside the Department of Defense is not a matter of clicking “upgrade.” A customer may have to spend six months or a year escorting a product through approvals, budgets and security requirements. Mild enthusiasm will expire somewhere around the fourth meeting.

Urgency also changed what counted as a complete product. A defense buyer was not shopping for OCR in isolation, however elegant the model. The user needed a working path from collection through processing to a decision, with as little integration labor as possible. Granberg came to distinguish between an impressive feature and a mission system. The former wins a demonstration. The latter survives contact with a secured network, an existing workflow and a buyer whose budget has its own calendar. Vannevar's expansion into multiple software products and a hardware product followed that distinction, linking unique data to the tools required to make it usable.

“We make bets on problems, not tech solutions.”Brett Granberg

That formulation now organizes how the company works. Product managers and engineers travel to military bases. They watch users at their desks rather than receiving a tidied description at a conference. When hardware is involved, they observe work in the field, including on islands in Asia. Teams of two to eight engineers can own an entire product line. The technology comes after the encounter with the problem.

One of Granberg's favorite examples compresses the philosophy into three days. A forward-deployed engineer traveled to Europe, heard about a problem and made a first version from a coffee shop using its Wi-Fi. The engineer iterated with the customer for the rest of the trip. Granberg later said the prototype became Vannevar's third most-used product. Coffee-shop networks have supported less consequential browsing.

Two views of the same broken tool

The instinct to get closer to users predates the company. Granberg studied mathematics and economics at the University of Georgia, then worked at McKinsey and In-Q-Tel, the venture organization that helps bring technology into the U.S. intelligence community. In the mid-2010s, he saw intelligence personnel working punishing weeks after a contractor-built tool failed. The contract behind it was worth close to $100 million a year. The gap between the importance of the mission and the quality of the engineering, he said, began to drive him “a little bit crazy.”

At Stanford Graduate School of Business in 2018, classmates kept telling him to meet Nini Hamrick. She had spent seven years as a counterterrorism officer and had experienced the same gap from the user's chair. During their first week, the two walked around campus and talked. Granberg understood investment and technology supply; Hamrick understood the daily work and the procurement machinery around it. They founded Vannevar in 2019 and left the safer, more legible path through business school.

The motivation was also personal. Granberg has spoken about his grandfather losing an older brother in World War II and his Scottish grandmother learning, during the Blitz, to run to school in zigzags. He grew up in comfortable suburbs of Chicago and Atlanta. Public service, in his telling, is partly an obligation to people who gave something so that his life could be ordinary.

The small contract that carried weight

Vannevar's first real deal was worth about $25,000 and served one user. The work traveled upward to the commander of a group focused on Asia. Within a few months, it informed an operational decision. Granberg remembers that as the point when eventual success felt less like hope and more like a matter of execution.

The product that emerged, Decrypt, collects hard-to-reach information and helps analysts process text, imagery, video and audio. Vannevar followed users from collection into analysis and other operations. By late 2024, Granberg said the company was deployed with roughly 40 Defense Department groups. In November of that year, the Defense Innovation Unit moved its Real-Time Information and Effects program into production with a ceiling of $99 million. The agency said Vannevar's tools had increased the speed of translating unstructured foreign-language data and helped produce insights in hours rather than days.

The scale continued. Vannevar reported $80 million in 2024 annual contract value, a third consecutive profitable year, 2,848 percent growth over three years and a $1.5 billion valuation. Its current site says the company powers 125 deployments spanning every service branch and combatant command. These are company figures, useful mainly because they describe the distance traveled from the room where nobody wanted the demo.

At the 2025 Talisman Sabre exercise in Australia, Vannevar said more than a dozen AI-powered products supported daily work. Its systems modeled scenarios, examined maritime traffic, analyzed foreign-language narratives and surfaced online threats. The work also marked the company's first deployment of an application on a secure classified network. Granberg's early rule persisted at larger scale: the models were tested beside planners and operators, then revised against what actually happened.

A blues guitar against the metronome

Mission companies enjoy stern nouns: deterrence, advantage, readiness, effects. Their people remain less grammatical. Granberg grew up playing guitar in blues bars around Atlanta. When he can step away from work, he plays music and takes blocks of time with nothing scheduled. Hamrick once joked that his second career will be playing guitar in Nashville. His favorite novel is Karl Marlantes's long Vietnam story Matterhorn; in 2023, his favorite film was Everything Everywhere All at Once, a title that may double as the chief executive's calendar.

He does not pretend intensity is costless. Granberg has said the company is always processing in his head - at night, at dinner, through weekends. Time-sensitive events do not check office hours. Vannevar recruits for what it calls an almost irrational focus on the mission, yet he also notes that such people can work too hard. The company has tried to make it acceptable to ease off and to back that permission with leave policies. An institution built around urgency still has to survive its own urgency.

His most important professional mentor offers another clue. Dan Gwak, Granberg's former boss at In-Q-Tel, left finance, took a steep pay cut, enlisted in the Marine Corps, served in Afghanistan, returned and attended Harvard Business School. Granberg describes him as “low ego, high output.” It is an unfashionably compact leadership doctrine. It also resembles the arc of Vannevar's first product: fewer claims, more contact with the work.

The technology worked. The priority had moved. Listening was the feature that finally shipped.

Granberg now speaks about building a durable, multi-product defense company across a long horizon. Government programs can take years to enter a budget, which means today's conversations may shape capabilities fielded in 2028. The company has expanded from software into hardware and from search into agentic systems. Yet the interesting continuity is not artificial intelligence. It is the discipline imposed by those first indifferent rooms.

A founder can polish an answer indefinitely. A customer with an urgent problem behaves differently: they ask for access before the demo is ready. Brett Granberg spent months learning to recognize the difference. Vannevar's story began when he stopped treating silence as a request for another feature.