General Intuition raised $320M at a $2.3B valuation to scale AI trained on action labels from millions of hours of gameplay. The startup, spun out of Medal, argues that using button-press data rather than video alone produces agents that generalize to real-world robots.
Key facts
- $320M raised at $2.3B valuation.
- Total disclosed funding: $454M.
- Spun out of Medal, which has 100M+ hours of gameplay.
- Agent played Fortnite for 100 hours straight.
- 8 minutes of real-world data fine-tuned the robot.
General Intuition raised $320 million at a $2.3 billion valuation, confirming TechCrunch’s previous reporting. The round brings total disclosed funding to $454 million, after the $134 million round at launch last October. According to TechCrunch
The startup was spun out of CEO Pim de Witte’s other company, Medal, which lets gamers upload and share clips. The hundreds of millions of hours of uploaded gameplay provided the initial dataset. But the key ingredient wasn’t the footage; it was the action labels — records of exactly what buttons a player pressed and when. Most competitors, de Witte says, try to infer actions from video alone, which he argues is insufficient.
In a demo, a General Intuition agent played Fortnite for 100 hours straight. The same model then powered a quadruped robot that navigated an office after just eight minutes of real-world fine-tuning. “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 said.
The approach positions General Intuition against other agentic AI startups and big labs like Google and Anthropic, which rely on large language models for reasoning. By using gameplay action data, de Witte claims the model learns spatial-temporal reasoning — understanding how to move through space and time — without needing to infer actions from pixels. The company did not disclose revenue or enterprise customers.
Why action labels matter more than video
The core insight: most AI agent training uses video to infer actions, a noisy process. General Intuition has clean action labels from Medal’s 100M+ hours of gameplay. This reduces the gap between simulation and reality, a problem that has plagued robotics for years. Whether the approach scales beyond gaming environments remains unproven.
What to watch
Watch for General Intuition’s first enterprise deployment and whether the model can handle tasks outside gaming — such as warehouse robotics or autonomous navigation. A public benchmark or API release would signal readiness beyond demos.
Source: techcrunch.com
