General Intuition has raised $320 million at a $2.3 billion valuation to develop AI agents that can act across video games, simulations, and physical environments.
Khosla Ventures led the round, with participation from General Catalyst, Jeff Bezos, Eric Schmidt, former Formula One champion Nico Rosberg, and researchers from Google DeepMind and MIT. The financing brings the New York-based startup’s total disclosed funding to $454 million.
Gameplay Provides Action-Labelled Training Data
General Intuition was spun out of Medal, a platform where gamers upload and share gameplay clips. The startup uses Medal’s extensive archive to train models in spatial and temporal reasoning.
The company says the most valuable part of the dataset is not simply the video. Gameplay clips can also contain records showing which buttons players pressed and when, giving the models direct information about actions and their consequences.
Chief executive Pim de Witte argues that this data can help an AI system distinguish between itself and its environment. A model can learn how movement affects a scene, how objects respond to actions, and how to navigate unfamiliar spaces.
During a demonstration, the same underlying model operated an agent inside a video game and controlled a quadrupedal robot through a camera feed. General Intuition said it adapted the robot using eight minutes of real-world data collected outside the office where it was later tested.
World Model Serves as a Training Environment
The startup has also developed a world model that generates simulated environments frame by frame. It uses this system internally as a training ground where agents can experiment and learn before being deployed in games, simulations, or physical machines.
General Intuition does not intend to sell the world model as its main product. It plans to provide access to the agentic foundation model through an API, allowing other companies to develop systems for robotics, gaming, industrial simulations, and autonomous machines.
The company has already started working with a small number of customers. It expects to make its API more broadly available by the end of summer.
Funding Will Primarily Expand Computing Capacity
Most of the new capital will be used to increase the computing resources needed to train the company’s next model. General Intuition is working with CoreWeave to support its large-scale training workloads.
The startup is also building Nerve, a work platform where gamers can earn money through data labelling, robot teleoperation, and related assignments.
General Intuition says it will not support applications intended to harm people. De Witte has ruled out lethal autonomous systems while leaving open uses such as search and rescue in hazardous environments.
The company’s central assumption is that gameplay data can reduce the amount of expensive real-world training needed for embodied AI. Whether those capabilities will transfer reliably from simulated environments to varied physical settings remains an open technical question.
Featured image credits: Magnific.com
For more stories like it, click the +Follow button at the top of this page to follow us.