General Intuition is an AI lab that trains "large action models" — AI systems that can perceive environments, predict what happens next, and decide what to do — using billions of gameplay clips from Medal, its sister gaming platform with 17M+ monthly active users. Each clip carries precise action metadata: every button press, mouse movement, and split-second decision a real human player made. That action-labeled data is the core asset. General Intuition uses it to pre-train world models (which predict how an environment will evolve) and action models (which generate the best move given what the model sees), then targets robotics, autonomous vehicles, and simulation as commercial applications. The company has already demonstrated a quadruped robot navigating a room using only 8 minutes of real-world data after pre-training on gameplay. Customers are buying API access across gaming, simulation, and robotics. The long-term goal is to be a foundation model provider — the OpenAI/Anthropic of embodied AI — licensing agent models to companies building autonomous systems in the physical world.