General Intuition is moving fast. Just eight months after spinning out of the gaming platform Medal, the New York-based startup is in talks to raise $300 million. The deal would value the company at roughly $2 billion.

This is a massive jump. It signals that investors are betting heavily on the race to build "embodied" AI—systems that don't just chat, but actually navigate and interact with the world.

The Power of the Dataset

Most AI labs train models on static text or images. General Intuition is doing something different. It is using Medal’s massive repository of 2 billion gameplay videos per year.

This data is unique. It captures 10 million monthly active users playing games in real time. For an AI, this is a goldmine. It provides a perfect sandbox to learn spatial-temporal reasoning. The goal is to teach machines how to perceive, anticipate, and act within a simulation.

Big players have noticed. Sources say OpenAI previously attempted to acquire Medal to get its hands on this data. They aren't the only ones. The race for high-quality, interactive training data has become the new arms race in Silicon Valley.

Why Agents Are the Real Product

Many startups are building world models to generate video or art. General Intuition is taking a different path. It is building world models to train agents.

It doesn't want to sell the simulation. It wants to sell the intelligence that lives inside it. By training agents in these high-fidelity virtual environments, the company believes it can create systems capable of real-world tasks.

This approach has attracted heavy hitters. Jeff Bezos and Eric Schmidt are reportedly among the backers. Existing investors Khosla Ventures and General Catalyst are also doubling down. The capital is there. The vision is clear.

Scaling for the Next Phase

What comes next? Compute. Training these models requires massive amounts of GPU power. The new funding will go directly toward scaling that capacity.

Leadership is already looking ahead. The team, led by co-founder Pim de Witte, plans to release a new product by late summer or early fall. They are moving quickly. They have to.

Key Takeaways

  • Massive Valuation: General Intuition is seeking a $2 billion valuation, a significant leap for a company only eight months old.
  • Unique Data Advantage: The startup leverages 2 billion annual gameplay videos from Medal to train agents on spatial reasoning.
  • Strategic Focus: Unlike competitors building models for video generation, General Intuition is focused entirely on training autonomous agents.

The Road Ahead

The world model space is crowded. Runway, Decart, and World Labs are all competing for the same ground. Google is integrating mapping data into its Genie 3 model to simulate real-world physics.

General Intuition has a head start with its proprietary dataset. But the window is closing. The company’s upcoming product launch will be the first real test of whether their simulation-based training can translate into capable, real-world agents. The market is watching. We will know soon.