General Intuition, a New York-based startup backed by Jeff Bezos, has raised $320 million at a $2.3 billion valuation to pursue an unconventional approach to artificial general intelligence (AGI) — training AI agents using video game footage instead of text from the internet.
Gaming Data Is the New Frontier for AI Training
The startup, founded by CEO Pim de Witte, spun out of Medal — a popular platform where gamers upload and share gameplay clips. The core insight? Those hundreds of millions of hours of gameplay contain something that text-based datasets can’t provide: real action labels showing exactly what buttons players pressed and when, teaching AI models how objects move through space and time.
“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,” de Witte told TechCrunch.
The latest funding round was led by Khosla Ventures with participation from General Catalyst, Jeff Bezos, Eric Schmidt, former F1 champion Nico Rosberg, and researchers from Google DeepMind and MIT. The round brings General Intuition’s total disclosed funding to $454 million.
From Video Games to Real-World Robots
During a visit to the company’s New York R&D floor, a quadrupedal robot powered by the same AI model that plays Fortnite demonstrated the ability to navigate office spaces and interact with people — after just eight minutes of fine-tuning with real-world data collected on the street.
The company believes that the action data embedded in gameplay helps its models understand causality in a way that large language models cannot. While LLMs like ChatGPT excel at text-based reasoning, they struggle with understanding physical space, movement, and time — essential skills for AGI that can operate in the real world.
What’s Next for General Intuition
The vast majority of the new funding will go toward scaling compute capacity through a deal with CoreWeave. The company plans to make its API more broadly available by the end of summer 2026, positioning itself as a model provider — similar to OpenAI or Anthropic — that enables others to build on its technology.
“We’re not gonna build a self-driving car company,” de Witte said. “We’re gonna make it 10 times easier for the next person to build a self-driving car company.”
The startup has also launched Nerve, a jobs marketplace allowing gamers to earn money through data labeling and robot teleoperation, creating a data flywheel while addressing AI-driven job displacement.
Ethical Boundaries
De Witte has drawn a clear line: General Intuition will not build agents designed to harm humans. “We don’t want to be an escalatory part of the system,” he said, limiting military use while remaining open to search and rescue applications.
Frequently Asked Questions
What is General Intuition?
General Intuition is a New York-based AI startup that trains AI agents using video game data to develop spatial-temporal reasoning — understanding how objects move through space and time. It was valued at $2.3 billion in its latest funding round.
How does video game data help train AI?
Video game footage contains embedded action labels showing exactly what buttons players pressed and when. This provides richer training data than text alone, teaching AI models about physical interactions, causality, and movement in 3D environments.
Who is backing General Intuition?
Investors include Jeff Bezos, Khosla Ventures, General Catalyst, Eric Schmidt, Nico Rosberg, and researchers from Google DeepMind and MIT.