General Intuition in Talks to Raise $300M at Around $2B Valuation
What if the next leap in artificial intelligence isn’t driven by theoretical breakthroughs, but by the data hidden in the pixels of video games? General Intuition, a New York-based startup, is reportedly in advanced talks to raise $300 million in funding, which would value the company at just over $2 billion. This development marks a significant shift in the world of AI agent development, where the startup is leveraging its origins in Medal — a platform for sharing and uploading video game clips — to build a unique foundation model that trains AI to understand and interact with spatial-temporal environments. Unlike many of its peers, General Intuition isn’t focused on selling world models as standalone products. Instead, it's using them to train intelligent agents capable of navigating and responding to complex, real-time scenarios.
A Dataset Unlike Any Other
At the heart of General Intuition’s approach is a dataset that is both vast and deeply interactive. The company processes 2 billion video clips per year, sourced from its 10 million monthly active users on Medal. This dataset, built from first-person gameplay, is unlike anything else in the AI space. It allows agents to learn from real-time interactions, making it an ideal training ground for embodied AI — systems that don’t just process information, but physically engage with their environments.
The data provides a naturalistic, human-centric view of spatial reasoning. It captures complex interactions that are difficult to simulate in controlled environments. The scale and diversity of the dataset offer a unique training ground for AI to understand cause and effect in dynamic systems.
This kind of data has already drawn the attention of major players in the field. Sources indicate that OpenAI once tried to acquire Medal, and other AI labs have reportedly shown interest in General Intuition’s approach. With this new funding, the startup is well positioned to expand its capabilities and bring its product to market by the end of summer.
A Strategic Differentiation in a Competitive Field
The world model space is growing rapidly. Startups like Runway, Decart, and World Labs have all released their own models in recent months, and even Google is enhancing its Genie 3 platform with real-world data integration. However, General Intuition’s focus on training AI agents rather than selling models sets it apart. The startup is building systems that can perform tasks in simulated environments — a crucial step toward real-world applications in robotics and autonomous systems.
Its co-founders — including Pim de Witte, who previously co-founded Medal — bring deep expertise in world modeling and simulation, giving the company a strong technical foundation. Backers such as Jeff Bezos, Eric Schmidt, Khosla Ventures, and General Catalyst have already committed to supporting the company, signaling confidence in its long-term vision.
A Vision Beyond the Model
General Intuition’s approach suggests a future where AI doesn’t just understand the world — it interacts with it in real time. The startup’s ability to train agents through first-person, real-time gameplay could have profound implications for fields ranging from robotics to autonomous vehicles and even virtual environments. By focusing on agent training rather than model sales, it’s positioning itself as a builder of the next generation of intelligent systems.
As the AI landscape becomes more competitive, General Intuition’s unique data pipeline and clear product focus may give it an edge. With this new funding, it’s not just about scaling up compute — it’s about accelerating the path to a future where AI doesn’t just see the world, but acts within it.
