General Intuition builds AI agents trained on 2 billion gameplay clips for embodied intelligence.
General Intuition builds AI agents trained on 2 billion gameplay clips for embodied intelligence.
General Intuition is an AI research company developing foundation models for agents that can understand, reason about, and act within dynamic environments. Spun out of Medal, the company focuses on spatial-temporal reasoning, world models, and embodied intelligence, using large-scale video data to train AI systems capable of transferring knowledge across digital and physical domains. Its research leverages Medal’s extensive archive of gameplay footage to advance video understanding, agent behavior, and predictive world modeling. While its technology extends beyond robotics, General Intuition is developing core AI capabilities that can power future humanoid robots, autonomous agents, gaming NPCs, and other embodied systems.
Platform maturity, autonomy stack, and flagship-system specifications in one view.
A concise view of platform maturity and deployment footprint.
In October 2025, General Intuition launched publicly with a $133.7 million seed round backed by Khosla Ventures, General Catalyst, and Raine. The company framed the financing as support for discovering the next generation of intelligence by training agents on Medals massive gameplay dataset.
General Intuitions site ties the companys origins to a body of prior research released from 2023 through 2025, including IRIS, Delta-IRIS, DIAMOND, and GAIA-2. That lineage matters because the company launched not as a blank-slate startup, but as a continuation of work on world models, video understanding, and sample-efficient agents.
In October 2025, General Intuition launched publicly with a $133.7 million seed round backed by Khosla Ventures, General Catalyst, and Raine. The company framed the financing as support for discovering the next generation of intelligence by training agents on Medals massive gameplay dataset.
General Intuition says Medal users upload billions of action-rich clips each year, giving the lab a uniquely diverse dataset for agents that need to reason across space and time. In Roboticos dataset, the company is notable because it explicitly connects that training substrate to transfer into robotics, autonomous drones, and other real-world agent systems.
