General Intuition Raises $320M Turning 2.5B Gaming Clips Into Robot Brains

General Intuition, the AI research lab spun out of gaming clip platform Medal, has closed a $320 million Series A at a $2.3 billion valuation. The round was led by Khosla Ventures, with participation from General Catalyst, Jeff Bezos, Eric Schmidt, Nico Rosberg, and researchers at Google DeepMind and MIT. It is a striking vote of confidence in a company that barely existed eight months ago.

The raise brings General Intuition's total disclosed funding to $454 million, following a $134 million round it raised at launch. The vast majority of the new capital will go toward scaling compute capacity, with a deal already in place with CoreWeave to pre-train the next version of the model. A slice has been earmarked for making its API more broadly available by the end of summer.

Fortnite to Robots: The Same Brain

The startup was spun out of de Witte's other company, Medal, which allows gamers to upload and share video game clips. The hundreds of millions of hours of uploaded gameplay provided the initial dataset to train General Intuition's model in spatial-temporal reasoning , understanding how to move through space and time. But the secret sauce is not the footage itself.

The key ingredient is the action labels embedded in those clips: records of exactly what buttons a player pressed and when. Most competitors, de Witte says, are trying to infer actions from video alone, which he argues is insufficient. Medal's clips are first-person and include ground-truth action labels , the specific controller inputs that produced each sequence of frames. YouTube and Twitch footage is recorded from a spectator perspective and lacks this action-label information, which is necessary for training models to understand the mapping from decision to consequence.

Medal is now used by 15M+ gamers every month, making it the largest clipping platform in the world. This year alone, Medal users will capture 2.5B+ clips. That is not just a large dataset , it is a self-replenishing one. As Medal grows, the training data compounds without General Intuition having to negotiate with publishers or streaming platforms.

One Model, Two Very Different Worlds

The most striking demo at General Intuition's New York office was not the AI playing Fortnite for 100 hours straight. It was what happened next: a quadrupedal robot walked in, powered by the exact same model. The capability General Intuition has demonstrated is notable: the company has built models that can understand environments they were not trained on and correctly predict what happens behind occlusions , objects or areas temporarily hidden from view , demonstrating genuine object permanence.

It took just eight minutes of real-world robotics data to fine-tune the model for the quadruped, collected on a street rather than in the office where it was navigating. General Intuition is positioning its foundation models as infrastructure, and the agents trained on those models as the actual product , a distinction that aims squarely at where enterprise buyers spend money, on deployable capabilities rather than raw model access.

The company's target customers span three verticals today:

• Gaming: Intelligent bots and NPCs that dynamically adjust difficulty, targeting a 50% win rate to maximize player engagement
• Robotics: Quadrupeds, drones, and other devices controllable via game controller or keyboard
• Simulation: Digital twins of factory floors and hazardous environments for testing before real-world deployment

Why This Data Moat Is Hard to Copy

The Verge reported that General Intuition was spun out after AI labs showed interest in Medal's data, with OpenAI reported to have offered $500 million for the startup. De Witte turned it down. The amount and quality of proprietary data General Intuition has via Medal is one of the reasons Khosla is convinced the startup is a generational bet, not an M&A target. "At this point, it would be a data acquisition, which is sort of uninteresting," Khosla said.

Vinod Khosla, whose firm led the round, says he was drawn to de Witte's vision and the company's proprietary data position. "If you look at LLMs, when reasoning emerged, it was a quantum leap," Khosla said. "In world models, I think the quantum leap is the emergence of intuition in the AI, a human intuition-like capability."

Although the office and founder are based in New York, the company is legally registered in the Netherlands , a deliberate choice. "They can theoretically stop us in America, but they cannot completely stop the model," de Witte told NRC.

A Category Exploding With Capital

General Intuition is not raising into a vacuum. The world model category has become one of the hottest in AI, with billions flowing in over the past few months alone:

• Decart.ai raised $300 million at a nearly $4 billion valuation to develop world models named Lucy and Oasis.
• Yann LeCun's AMI Labs raised $1.03 billion to train world models.
• Odyssey, an AI lab founded by self-driving car veterans, raised a $310 million Series B at a $1.45 billion valuation.
• World Labs raised $1 billion, and Runway raised $315 million.

What separates General Intuition from this field is the data angle. The broader shift in the AI market is from models that produce answers to models that understand environments, cause and effect, motion, physics, and interaction. Most competitors are building toward that goal with synthetic data or real-world camera rigs. General Intuition is building it with 2.5 billion human-generated, action-labeled clips per year , a flywheel that gets richer every time someone boots up a game and hits record.

Who Wins, Who Watches Nervously

The clearest winners are robotics and simulation companies that will eventually build on General Intuition's API. The approach translates naturally to physical systems like robotic arms, drones, and autonomous vehicles, which are often manipulated by humans using game controllers. If the sim-to-real transfer holds at scale, the cost of training physical AI agents drops dramatically.

The clearest losers are companies trying to build similar capabilities from scratch. Whoever controls 2 billion first-person, action-labeled game clips per year has an asset that is genuinely hard to replicate, because building that volume of labeled data through any other means is slow and expensive. Labs that relied on scraping YouTube or Twitch are working with fundamentally inferior signal.

Medal's 17 million monthly users also stand to benefit directly. General Intuition recently launched Nerve, a jobs marketplace that lets gamers earn money using their existing setups , starting with data labeling and eventually moving toward robot teleoperation. De Witte, who spent years working with Doctors Without Borders, is explicit about why: Medal's user base is precisely the generation most exposed to AI-driven displacement, and he wants them to have a stake in what's being built.

The Open Question

De Witte has drawn a clear line for how General Intuition's tech will be used: no agents will be employed to harm humans. The company operates as a public-benefit corporation, registered in the Netherlands partly as a structural hedge against regulatory or political pressure in the US.

The big unresolved question is whether the simulation-to-real-world transfer can hold at scale. Impressive demos in a New York office are one thing. Deploying generalized agents across thousands of real-world environments , factories, hospitals, disaster zones , is another. Odyssey's CEO believes the field is approaching the GPT-3 moment for world models , the point where they transition from a promising research direction into a breakthrough foundational technology. General Intuition is betting its $454 million that it gets there first, with better data than anyone else in the race.