A startup just raised $320 million on the idea that the path to robots which understand the physical world runs through Fortnite. According to TechCrunch AI, General Intuition closed the round on Thursday at a $2.3 billion valuation, led by Khosla Ventures with Jeff Bezos, Eric Schmidt, General Catalyst, F1 champ Nico Rosberg, and researchers from Google DeepMind and MIT all writing checks. That brings the company’s total disclosed funding to $454 million, after a $134 million launch round last October.
This is one of the bolder bets in AI right now, and it’s worth understanding why serious money is backing it.
What General Intuition is actually building
The pitch is a single agentic model that can play a video game, run inside a simulation, and pilot a physical robot, all with the same “brain.” TechCrunch AI reports that during a demo at the company’s New York office, one agent had been playing a Fortnite-style game for 100 hours straight. A few feet away, the same model was driving a quadruped robot that walked up, circled the reporter, and bumped into chairs like a curious toddler.
The number that stands out: it took just eight minutes of real-world robotics data to fine-tune the model for that quadruped. And that data was collected on the street, not in the office the robot was navigating.
Why gameplay is the secret ingredient
General Intuition spun out of Medal, CEO Pim de Witte’s other company, where gamers upload and share clips. That gave it hundreds of millions of hours of gameplay to train on. But de Witte told TechCrunch AI the footage itself isn’t the prize. The real value is the action labels baked into those clips: exactly which buttons a player pressed, and when.
Most rivals try to infer actions from video alone. De Witte argues that’s not enough. His claim is that pairing what’s on screen with what the player did teaches the model to tell “self” from “environment,” which gives it a richer grip on cause and effect.
The company built a world model it calls “the gym” internally. It’s a simulated space generated frame by frame, not rendered by a game engine. From all that gameplay, it learned that walls are solid, ladders are for climbing, and shadows stretch as the sun moves.
Why this matters
Teaching AI to handle the messy physical world is one of the hardest problems in the field. The usual approach needs huge piles of real-world data that’s slow and expensive to gather. General Intuition’s wager is that gameplay is a scalable shortcut to that same spatial reasoning.
Here’s the catch, and the report is honest about it. General Intuition isn’t the only one chasing this, and nobody has made a model like this hold up in the physical world at scale yet. Impressive demos are not the same as a shipping product.
Vinod Khosla, whose firm led the round, framed the upside in big terms. “If you look at LLMs, when reasoning emerged, it was a quantum leap,” he told TechCrunch AI. “In world models, I think the quantum leap is the emergence of intuition in the AI, a human intuition-like capability.”
What comes next
Where the money goes tells you the plan:
- Compute, mostly. The vast majority of the round funds scaling up, through a deal with CoreWeave, to pre-train the next version of the model.
- An API by end of summer. A slice is earmarked to make the model more broadly available, so outside developers can start building on it.
- No exit. De Witte and his co-founders, Eloi Alonso, Adam Jelley, and Vincent Micheli, have already turned down acquisition offers, including one from a major lab. Khosla called it a “generational bet,” not an M&A target.
My read: the proprietary data position is the moat everyone’s paying for here. If General Intuition pulls off the jump from gameplay to robots at scale, it could become foundational infrastructure for general-purpose agents. If it doesn’t, it’s still sitting on a dataset the big labs clearly want. Watch for that API drop later this summer as the first real test. Full details are in the original TechCrunch AI report.