General Intuition is making a large bet on a simple idea: video games may teach AI agents how to act in the real world. The New York startup says gameplay, especially the record of what players did and when they did it, can help train models that understand movement, control and cause and effect.
That thesis has now attracted major financial backing. General Intuition said it raised $320 million at a $2.3 billio n valuation, bringing its total disclosed funding to $454 million after the $134 million round it raised at launch last October.
From gameplay clips to agentic models
General Intuition was spun out of Medal, the company founded by Pim de Witte that lets gamers upload and share video game clips. Medal supplied the startup with hundreds of millions of hours of uploaded gameplay, which became the initial dataset for training its model in spatial-temporal reasoning.
Spatial-temporal reasoning, in this context, means learning how actions unfold across space and time. For an AI agent, that is different from simply recognizing what appears on a screen. The goal is to understand how to move, what changes after an action and how the agent relates to the environment around it.
The company argues that the most important part of its dataset is not only the video. It is the action labels inside those clips: records of exactly what buttons a player pressed and when. De Witte says many competitors try to infer actions from video alone, while General Intuition is training from video paired with human input.
“We view this as just the next stage of future pre-training,” de Witte said. “We have a single model that can respond to Fortnite information on the screen and take action, but also to real-world dynamics in a way that an LLM could never.”
A model moving through games, worlds and robots
The company’s demonstrations are built around the claim that one underlying model can work across different environments. In General Intuition’s New York office, an AI agent was shown playing something like Fortnite while a large quadrupedal robot moved through the office using what de Witte described as the same brain.
Kent Rollins, the company’s chief product officer, said the agent had been playing for 100 hours straight. Josh Duplantis, a data analyst, said the robot’s default mode was “exploration.” The robot relied on a single camera, walked around the office and sometimes bumped into chair legs or a trash bin.
Duplantis said it took just eight minutes of real-world robotics data to fine-tune an AI model for the quadruped. That data was collected on the street, not inside the office where the robot was operating.
General Intuition also showed a world model: a simulated environment generated frame-by-frame instead of rendered by a traditional game engine. In that demo, the model treated walls as barriers, ladders as climbable objects and shadows as changing with the sun. For the company, that world model is not the end product. It is a training environment, referred to internally as “the gym.”
Why investors are backing the shortcut
General Intuition’s latest round was led by Khosla Ventures, with participation from General Catalyst, Jeff Bezos, Eric Schmidt, Nico Rosberg, and researchers at Google DeepMind and MIT. The company plans to use the vast majority of the round to scale compute capacity. It has a deal with CoreWeave and plans to focus on pre-training the next version of the model.
A slice of the funding has also been earmarked for making its API more broadly available by the end of summer. De Witte wants General Intuition to be a model provider that other companies can build on top of, comparing the ambition to an ecosystem enabler like Anthropic or OpenAI.
Today, the startup has a handful of customers in gaming, simulation, and robotics. De Witte said the company is not trying to build a self-driving car company. Instead, he said it wants to make that work easier for others.
“We’re gonna make it 10 times easier for the next person to build a self-driving car company.”
Vinod Khosla said he was drawn to the company’s vision and proprietary data position. He connected the idea to the emergence of reasoning in LLMs and said world models may see a similar jump through what he called intuition.
“The human action data and reaction data you have in games is the key part to the emergence of intuition.”
The hard part is leaving the demo room
The company’s central bet is that gameplay can become a scalable shortcut for training AI agents that eventually operate in the physical world. That matters because many approaches require large amounts of real-world data, which can be slow and expensive to collect.
But the open question is whether this transfer from simulation to real-world use can work at scale. General Intuition is not the only company pursuing world models and general agents, and getting these systems to perform reliably outside controlled demonstrations has not yet been solved.
The company wants to test its model across a broader range of customer use cases once its API reaches more users. The examples include testing a robot in a digital twin of a factory floor, powering a humanlike bot inside a gaming studio, or sending a quadruped to navigate hazardous environments.
The quadruped is the first physical embodiment General Intuition has tried in the real world. The company has also tried drones and other devices, including testing the model in driving games. De Witte said the system works on anything that can be controlled with a game controller or a keyboard mouse.
Ethics, jobs and the data flywheel
General Intuition is also drawing boundaries around how its technology should be used. De Witte, who spent three years working in the humanitarian space, including with Doctors Without Borders, said no agents will be employed to harm humans.
“We don’t want to be an escalatory part of the system,” de Witte said.
He said he is comfortable with uses such as search and rescue missions. That position stands out at a time when the source article describes Silicon Valley as increasingly bullish on war.
The company has also launched Nerve, a jobs marketplace that lets gamers earn money using their existing setups. People who sign up start with data labeling and can eventually move toward robot teleoperation and other tasks. De Witte said Medal’s user base is the generation most exposed to AI-driven displacement, and he wants them to have a stake in what comes next.
The business strategy also depends on a data flywheel. General Intuition plans to prioritize customers that can provide real-world data useful for research and that have agile internal teams. Khosla said the startup’s proprietary data helped it get this far, but its ability to keep collecting data that others do not have will be essential.