General Intuition, an AI research lab spun out of video game clip-sharing platform Medal, has raised $320 million at a valuation exceeding $2 billion. The round is designed to scale the company’s approach to training AI foundation models using action-labeled gameplay data rather than the static text and images that power most large language models today.
From game clips to world models
General Intuition’s secret weapon is its parent company’s data pipeline. Medal, founded by CEO Pim de Witte, boasts roughly 10 million monthly active users who collectively upload approximately 2 billion video clips per year.
The company’s leadership team includes co-founders Eloi Alonso, Adam Jelley, and Vincent Micheli, researchers with experience building models like DIAMOND, IRIS, and GAIA-2. Their focus is on what the AI field calls “spatial-temporal reasoning,” the ability to understand not just what objects are, but how they move through space over time.
The money and the backers
This $320 million round follows a $133.7 million seed round that closed around October 2025. That seed was backed by Khosla Ventures and General Catalyst, two of the most active AI-focused venture firms in Silicon Valley.
The new round reportedly includes participation from Jeff Bezos and Eric Schmidt. The combined funding of over $450 million in less than a year puts General Intuition in rare company for an AI lab that hasn’t yet shipped a widely known consumer product. The capital is expected to go toward expanding computational resources and accelerating product development.
The competitive landscape
General Intuition is entering a field that’s getting crowded fast. Runway has been building generative video models. Decart has pursued real-time world simulation. World Labs, co-founded by AI pioneer Fei-Fei Li, is working on spatial intelligence. Google’s internal Genie project explores similar territory around learning from interactive environments.
What separates General Intuition is the scale and specificity of its training data. Most competitors rely on internet video, synthetic environments, or curated datasets. General Intuition has access to billions of clips where every frame is paired with the player’s actual inputs, mouse movements, button presses, strategic decisions.
The potential applications extend well beyond gaming. Spatial-temporal reasoning is a foundational capability for robotics, autonomous navigation, and real-world simulation.