General Intuition is raising $300 million on the bet that game videos can teach agents space

TechCrunch reports that General Intuition is in talks to raise around $300 million at a valuation above $2 billion. The company spun out of Medal 8 months ago and previously raised a $134 million seed round.

Investors are paying for a gameplay data vault, not just another world model

The startup is led by Medal’s Pim de Witte with a research team focused on world modeling and simulation. TechCrunch says backers include Jeff Bezos, Eric Schmidt, Khosla Ventures and General Catalyst.

The core asset is Medal’s dataset: 2 billion videos per year from 10 million monthly active users. General Intuition argues that gameplay clips have an advantage over ordinary video because they capture movement, action, mistakes, reactions and decisions inside virtual space.

The bet is on agents that learn by acting

World models are often pitched as environment generators or visual simulation systems. General Intuition is aiming at a different layer, according to TechCrunch: world models are the training environment, while agents are the product.

That matters for robotics, games and autonomous systems. A text model can plan well in language, but a physical or game agent must predict what happens after movement, collision, camera position or bad timing. Gameplay data is not just entertainment content. It is a cheap behavioral lab.

The dataset can be a moat and a reality filter at the same time

Medal’s strength is also the limit. Gameplay clips are interactive and action-rich, but they still come from artificial rules. An agent that performs well in game space may not automatically handle dirty sensors, friction, weather or a broken camera in the real world.

There is also a consent and rights question. When a user-video dataset becomes fuel for a $2 billion model company, Medal and General Intuition will need to explain clearly what happens to creators’ clips.

The product will matter more than the size of the round

TechCrunch says the company plans to use the money for compute, its next product and a launch in late summer or early fall. That is where good fundraising separates from actual technology.

The signal to watch is simple: whether General Intuition can show an agent that transfers learned spatial behavior beyond a carefully selected demo. If it stays at pretty simulations, $300 million will look more like a receipt for a data story than proof of intelligence.

Lilith's verdict

General Intuition is selling investors an arena where an agent can learn to fall before it is let outside. The question is whether opening the door reveals an athlete, or just a character that can run one map well.