ONLYEXIT FRONTIER FUND
We back founders attacking frontier problems.
Problems where the barrier to entry is permission or capability, not imagination. Conviction earned over ninety days of observed building, not a pitch meeting.
01 · THESIS
AI is resetting who gets to build.
Work that used to require a team of forty now requires a team of four. The constraint on ambitious companies is no longer engineering headcount. It is whether anyone is willing to walk into the hard markets at all.
We back founders attacking frontier problems where the barrier is either permission or capability. Those barriers kept a generation of builders out. AI just handed them the keys.
02 · WHAT WE BACK
Three frontiers.
REGULATION × AI
The barrier is permission.
Licensed, compliance-heavy markets that software avoided for twenty years. The incumbents own the permission. The founders now own the capability.
Where this shows up: AI-native operators in licensed work like insurance brokerage, clinical documentation, and municipal permitting.
CAPABILITY × AI
The barrier is technical.
Problems that were out of reach for a small team until the current model generation. The frontier moves every quarter. We want the teams standing on it.
Where this shows up: four-person teams shipping what needed a research lab in 2022, like code agents, protein design tools, and physics simulation.
PHYSICAL × AI
The barrier is the physical world.
Hardware small enough to build in the house. Sensing, actuation, and lab automation where the bill of materials fits on a workbench and the market does not.
Where this shows up: bench-scale robotics, environmental sensing, and lab automation where the prototype ships from the house itself.
03 · HOW WE SOURCE
The Hacker House.
Our companies live and build in a residency we operate. By the time we make a conviction decision, we have ninety days of operating data: what the team shipped, how they responded to a bad week, what their users actually did.
THE RULEWe do not underwrite at intake. We underwrite from observed behaviour.
Most seed decisions compress diligence into a few meetings. Ours is a ninety-day working session with the answer written in the commit history.
04 · WHY SEATTLE
The talent was always here. The reason to leave just arrived.
By SignalFire's count, 23% of US AI engineers work in the Seattle area, concentrated in two of the largest engineering organizations on earth. The founder infrastructure around them is a fraction of what the Bay Area built for a comparable talent pool.
That gap is our sourcing edge. When the talent is here and the capital is not, the first institution that shows up with a working founder engine gets the pick of the pool. But the gap itself is not new, and a gap alone starts nothing.
Founder ecosystems are built by displacement events, not talent density. Fairchild seeded Intel; the PayPal cohort formed only after eBay's acquisition scattered it. An ending, imposed from outside, converts operators into founders: in one 2022 survey of laid-off tech workers, roughly a quarter reported starting their own company.
Seattle never had its displacement event. Microsoft and Amazon were never acquired, never collapsed, never scattered a vested cohort with capital and nothing to lose. The two companies grew large enough that even employees who made real money stayed on the treadmill: no blank slate, no severed mission, no diaspora. The talent accumulated for thirty years; the catalyst never arrived. The current AI reorganization of both companies is the first genuine loss-of-mission moment in the city's software era, and we built the fund to be standing there when it happens.