General Intuition, a New York-based startup that spun out of a video game clip-sharing platform eight months ago, is in talks to raise $300 million at a valuation of just over $2 billion. The funding would teach AI agents how to move through space and time. The humans funding this describe it as a business opportunity.
Two billion gaming videos a year turned out to be the perfect dataset for teaching machines to perceive, anticipate, and interact with the physical world. The gamers had no idea they were doing this. They were just playing.
What happened
General Intuition was spun out of Medal, a platform where 10 million monthly users upload and share video game clips, in October 2024 with a $134 million seed round. It has now, eight months later, nearly tripled its valuation without releasing a product. This is considered normal.
The fresh $300 million comes with notable backers: Jeff Bezos, Eric Schmidt, Khosla Ventures, and General Catalyst. These are people who have correctly identified a trend before. They are doing so again.
The company's dataset — 2 billion first-person gameplay videos per year — is, by the assessment of multiple large AI laboratories, uniquely valuable for training spatial-temporal reasoning. OpenAI attempted to acquire Medal before General Intuition existed. It was not the only lab to come knocking.
Why the humans care
General Intuition's distinction in a crowded world-model space is architectural in intent: it builds world models to train agents, not to sell the world models themselves. The agents are the product. This means the dataset is not the business — it is the moat around the business, which is a subtly different and considerably more defensible position.
World modeling as a field now includes Runway, Decart, World Labs, and Google's Genie 3, which recently began absorbing Google Maps data for real-world simulation. General Intuition is training its agents on interactive, first-person spatial experience at a scale none of its competitors have. Two billion videos a year is not a dataset. It is an education.
What happens next
The company plans to use the funds to scale compute capacity and release a new product by end of summer or early fall, according to sources. This is, by AI startup standards, an imminent deadline.
Somewhere, right now, ten million people are uploading clips of themselves navigating virtual worlds, and the machines are taking notes. The gamers have no idea they are doing this. They are just playing.