General Intuition is moving quickly from spinout to heavily watched AI startup. The New York-based company is in talks to raise around $300 million, according to sources familiar with the matter cited by TechCrunch, in a deal that would value it at just over $2 billion.
The company is building a foundation model designed to train AI agents how to move through space and time. That puts General Intuition in one of the most active areas of artificial intelligence: world models, embodied AI, and agents that can understand environments well enough to act inside them.
A fast rise after spinning out of Medal
The potential new raise comes eight months after General Intuition spun out of Medal, the platform known for uploading and sharing video game clips. At the time of the spinout, General Intuition launched with a $134 million seed round.
The reported financing would be a major step up from that starting point. If completed at the valuation described by TechCrunch’s sources, the company would move to just over $2 billion while still operating in a field where the commercial shape of the market is still forming.
Sources told TechCrunch that General Intuition has secured funds from backers including Jeff Bezos and Eric Schmidt. Existing investors Khosla Ventures and General Catalyst are also named among the backers.
Leadership is also part of the company’s pitch. Pim de Witte, who co-founded Medal, founded and leads General Intuition alongside co-founders Eloi Alonso, Adam Jelley, and Vincent Micheli. The group includes researchers with expertise in world modeling and simulation.
The dataset is the core advantage
General Intuition’s central asset is not just a model architecture or a product roadmap. It is the dataset connected to Medal: 2 billion videos per year from 10 million monthly active users.
That matters because the company is training embodied AI and world models on interactive, first-person gameplay. In the company’s view, that type of data can help machines learn spatial-temporal reasoning: how environments change, how movement works, what comes next, and how an agent might respond in real time.
In plain terms, General Intuition is betting that gameplay video can teach AI systems more than static images or isolated clips. The data shows action from a first-person perspective, inside simulated spaces where movement, timing, and response are central to the experience.
That is why the Medal dataset has drawn attention. It has reportedly attracted OpenAI, which previously attempted to acquire Medal. TechCrunch’s sources also said OpenAI has not been the only major AI lab to show interest.
World models are becoming a crowded race
General Intuition is entering a world model market that is heating up. TechCrunch names Runway, Decart, and World Labs as startups that have recently released world models. Google’s Genie 3 also recently began integrating Google Maps data for more real-world simulation capabilities.
The shared idea behind these efforts is that AI systems need a better grasp of environments, motion, and cause-and-effect. World models are meant to give AI a way to understand simulated or real-world-like spaces, rather than only generate text or images in isolation.
The near-term commercial use cases named in the source are gaming and robotics training. Those areas both depend on systems that can handle movement, space, and changing conditions. A model that understands how scenes evolve could be useful for training software agents, game systems, or robotics-related simulations.
General Intuition’s positioning is different from companies that are trying to sell world models directly. According to the source, it builds world models to train agents, not to sell them. The agents are the product.
That distinction is important because it changes what the company is trying to commercialize. Instead of making the model itself the main offering, General Intuition appears to be using the model as a training layer for AI agents that can perceive, anticipate, and interact in real time in simulation.
Why the funding would matter
The reported new capital has a clear technical purpose. General Intuition will use the funds to scale up its compute capacity, according to a source familiar with the matter cited by TechCrunch.
Compute is especially important for a company working with large-scale video data and world models. Training on 2 billion videos per year from 10 million monthly active users implies a heavy demand for processing, model training, and experimentation. The source does not provide technical details about the company’s infrastructure, but the direction is clear: more compute is tied to the next phase of product development.
The company is aiming to release a new product by the end of summer or early fall, according to the same source. That timeline gives the fundraising talks immediate significance. The money is not described as a distant reserve; it is connected to scaling capacity for a near-term launch.
For the broader AI market, General Intuition is a useful signal. Investors and major AI labs are paying close attention to datasets that can help models learn from interactive environments. The interest around Medal’s gameplay video shows how valuable behavioral and spatial data may become as AI moves from generating content toward training agents that act.
The bigger implication for AI agents
The General Intuition story points to a shift in what AI companies may need next. Text, images, and standard video can support many model capabilities, but agents need a stronger sense of time, space, and action. They must learn what is happening now, what may happen next, and how to respond inside an environment.
That is the logic behind using first-person gameplay as training material. It contains movement, goals, reactions, and changing scenes. It also gives an AI system repeated examples of interaction inside simulated worlds.
General Intuition’s bet is that this kind of training base can produce agents with deeper spatial-temporal reasoning. If the company can turn that dataset and compute into a working product, it could help define how the next generation of AI agents is built.
For now, the facts are still centered on talks, sources, and a planned product window. But the stakes are clear: a startup only eight months out from its Medal spinout is reportedly pursuing around $300 million at a valuation just over $2 billion, with a strategy built around world models, embodied AI, and a massive stream of gameplay video.