General Intuition, spun out of clip sharing app Medal, is in talks to raise about $300 million at roughly $2 billion by pitching first person gameplay footage as a training ground for AI that acts in the real world.
The moment worth looking at is not the kill shot. It is the half-second before it: the player sliding behind cover, the camera tucked against a wall, the split-second calculation about where an opponent will appear next. A first-person game records that calculation as it happens, from the player's own point of view, and the New York startup General Intuition has built its business on the idea that those clips are exactly what AI needs to learn how to act in physical space.
That proposition is now attracting serious money. General Intuition is in talks to raise roughly $300 million at a valuation of just over $2 billion, according to TechCrunch, as reported by SiliconANGLE, citing people familiar with the talks. The company has not confirmed the round, and the terms are not final.
Eight months is a short time to reach a $2 billion paper valuation. General Intuition launched in October 2025 after spinning out of Medal B.V., the New York company behind a clip-sharing platform used by gamers to record and post their best moments. Its founding seed round was $134 million; the new figure is roughly four times that, a markup that reflects how quickly investors have decided that embodied AI, the still-nascent effort to build systems that navigate and act in the real world, needs a different kind of training data than the polished spectator footage already saturating YouTube and Twitch.
The pitch hinges on what a gameplay clip actually captures. Medal's network pulls in about two billion clips a year from more than ten million monthly users across thousands of games, General Intuition has said. Unlike a Twitch stream or an esports highlight reel, those clips are recorded from the player's own viewpoint, at the moment a decision is being made, in interactive environments where the camera moves because the player moves. To a researcher training an agent to navigate a warehouse or a sidewalk, that footage is closer to what the agent will see in the real world than anything scraped from broadcast video.
The framing the company leans on is that YouTube and Twitch already hold millions of hours of polished spectator gameplay, but those videos are recorded from a fixed camera watching a player act. A Medal clip is recorded from inside the player's head. The difference is the difference between watching someone navigate a room and learning to navigate one yourself.
There are reasons to slow down before treating the round as proof that the thesis is right. General Intuition has no public revenue numbers to anchor its valuation, the new funding has not closed, and the field of companies pursuing similar world-model and embodied-AI approaches is crowded. The 4x markup in eight months also means a lot of expectations are now baked in.
Then there is the consent question that any honest look at this model has to raise. Most Medal users upload clips to share highlights, not to train AI systems that may eventually power robots or autonomous vehicles. The terms of service cover a wide range of uses; what individual gamers understood they were signing up for is a different matter. General Intuition is betting that the dataset is valuable enough to make those questions manageable. It has not yet made them go away.
What the round does signal is that a category once treated as a curiosity is now being treated as infrastructure. Gameplay clips, the kind that have lived on social feeds and Discord servers for years, are now part of the supply chain for the next wave of physical AI. The check from investors is a bet that the rest of the industry will agree.