General Intuition, a New York-based AI startup, is securing $300 million in funding at a $2 billion valuation to accelerate development of foundation models that teach artificial intelligence agents to perceive and interact with physical environments. The company, which emerged from Medal, a video game clip-sharing platform, plans to use the capital to expand computing infrastructure and launch a new product by late summer or early fall.

What Makes General Intuition's Approach Different?

Unlike competitors focused on licensing generative world models, General Intuition is taking a different path. The startup specializes in delivering trained AI agents as its primary product, with immediate applications in gaming and robotics. This distinction matters because it means the company isn't just building software tools; it's building intelligent systems that can actually perform tasks in simulated and real environments.

The company's competitive advantage stems from Medal's proprietary dataset, which is genuinely remarkable in scale. The platform generates approximately two billion videos annually from ten million monthly active users. These aren't random videos; they're interactive, first-person gameplay footage that captures complex spatial-temporal reasoning in action. When a player navigates a game world, anticipates obstacles, and reacts in real time, that footage becomes training data for teaching machines the same skills.

How Does General Intuition Train Its AI Agents?

The company's training methodology leverages several key components to build capable autonomous systems:

• Interactive Gameplay Data: The two billion annual videos from Medal users provide diverse examples of real-time perception, decision-making, and environmental interaction across countless scenarios.
• Spatial-Temporal Reasoning: By analyzing first-person gameplay footage, the models learn to understand how objects move, how environments change over time, and how to anticipate future states.
• Foundation Models for Embodied AI: General Intuition develops foundation models, which are large-scale AI systems trained on broad data that can be adapted for specific robotics and gaming applications.

This approach addresses a fundamental challenge in AI: teaching machines to understand and navigate three-dimensional space the way humans do. Most large language models (LLMs), which power chatbots and writing assistants, work with text. General Intuition's models work with visual, spatial information, making them suited for robots and autonomous systems that must operate in the physical world.

Who's Backing This Bet on World Models?

The funding round reflects significant institutional confidence in spatial-temporal modeling. General Intuition's investors include prominent figures and firms such as Jeff Bezos, Eric Schmidt, Khosla Ventures, and General Catalyst. The fact that major technology companies have taken notice is evident from OpenAI's previous attempt to acquire Medal, demonstrating that the gaming dataset and AI capabilities have attracted attention from leading artificial intelligence laboratories.

The company was co-founded by Pim de Witte, Medal's original creator, and is led by researchers Eloi Alonso, Adam Jelley, and Vincent Micheli. This combination of gaming platform expertise and AI research talent positions the team to bridge the gap between interactive entertainment and autonomous systems.

Where Does General Intuition Fit in the Broader Market?

The startup operates within a rapidly expanding market for generative world models, competing alongside firms such as Runway, Decart, and World Labs. Google has also entered the space, recently expanding its Genie 3 capabilities with integrated mapping data. This competitive landscape suggests that world models, which can simulate and predict how environments evolve, are becoming a foundational technology for multiple industries.

The distinction between General Intuition and its competitors is strategic. While others focus on licensing generative models as software products, General Intuition is building trained AI agents that can be deployed directly. This positions the company to capture value not just from software licensing, but from the performance and capabilities of the agents themselves.

The funding round, announced approximately eight months after General Intuition spun out of Medal with a $134 million seed investment, underscores growing institutional confidence in foundational spatial-temporal modeling and its practical applications in autonomous system training. As robotics, autonomous vehicles, and AI-driven simulation become increasingly important, the ability to train systems on realistic, diverse interactive data could prove to be a significant competitive advantage.