Problem

General Intuition is addressing a limitation in today’s AI: foundation models learn primarily from language, images, and video, but lack the grounded intuition that comes from perceiving environments, taking actions, and learning from consequences. This makes it difficult for AI agents to reason about how objects and entities move through space and time, adapt to unfamiliar environments, and act reliably in the physical world. The economic pain is the scarcity of diverse, action-labeled training data and realistic environments in which agents can learn through trial and error.

The clearest near-term use cases are interactive agents in games and autonomous systems such as search-and-rescue drones. Games provide a scalable proving ground for spatial reasoning and decision-making, while the longer-term opportunity is to transfer those capabilities to robotics and other real-world environments where mistakes are costly.

Product / Service

General Intuition builds action models that decide what to do and world models that predict the outcomes of those actions. Its training advantage comes from Medal, its gaming-video platform, whose users upload billions of gameplay clips across a large range of environments. These clips capture visual context, controller actions, and unusually informative successes and failures, allowing models to learn from what an agent sees and does rather than from static text alone.

The company is commercializing this capability through an API for partners in games, simulation, and robotics. Its agents operate from visual input and controller-like actions, and the company says its models can predict actions in environments they were not trained on. The intended benefit is more capable agentic systems that can perceive, anticipate, and improvise in dynamic environments.

Market

General Intuition competes in frontier AI, particularly the emerging categories of world models, action models, spatial-temporal reasoning, and embodied or agentic AI. The evidence names Twitch and YouTube as alternative sources of training data, but does not identify a direct model competitor by name. General Intuition’s distinctive positioning is its access to Medal’s proprietary, high-volume gameplay dataset and its focus on learning how agents act in space and time.

The company appears to be early commercial rather than purely pre-revenue: it has onboarded initial partners across games, simulation, and robotics to a commercial API, while limiting broader model availability. It raised a reported $133.7 million seed round led by Khosla Ventures and General Catalyst, with participation from Raine, and is pursuing gaming first with expansion toward search-and-rescue drones and robotics. Its reported data scale—approximately two billion clips annually from ten million monthly active users—provides substantial early traction, although the available evidence does not disclose revenue or customer-contract figures.