General Intuition Eyes $300M for AI World Models

General Intuition is in talks to raise $300M at a roughly $2B valuation to build world models — AI trained on 2 billion gameplay videos a year.

5 min readEAEvgenii ArsentevEvgenii Arsentev · PhD

General Intuition is in talks to raise about $300 million at a valuation near $2 billion, according to TechCrunch. The deal would arrive just eight months after the startup spun out of Medal, a platform where gamers share video clips, and only months after a $134 million seed round in October 2025 backed by Jeff Bezos, Eric Schmidt, Khosla Ventures and General Catalyst. Founder Pim de Witte leads the company alongside co-founders Eloi Alonso, Adam Jelley and Vincent Micheli.

The pitch is 'world models' — foundation models that teach AI to reason about space and time rather than just words. Instead of predicting the next token in a sentence, a world model predicts what happens next in a scene: where an object will move, how a room is laid out, what a first-person view will look like a second from now. General Intuition trains this on Medal's firehose of data — roughly 2 billion videos a year from about 10 million monthly active users, almost all of it interactive, first-person gameplay where a human is steering through a 3D environment.

Why gameplay video is the prize

That dataset is the whole story. Most video on the internet is passive — someone filmed a thing and you watch it. Gameplay clips are different: every frame is tied to a decision a player made, so the model sees cause and effect, navigation and consequence, billions of times over. That is exactly the signal an agent needs to act in a world rather than describe it. The data is valuable enough that OpenAI reportedly tried to acquire it before General Intuition was carved out as its own company. Rivals chasing the same goal include Runway, Decart and Fei-Fei Li's World Labs.

Why this matters for you

Today's chatbots are fluent and frequently clueless about the physical world — they will happily tell you a couch fits through a doorway it cannot. World models are the missing layer for anything that has to move: home robots, self-driving cars, drones, and game characters that behave like they understand the space around them. The money chasing this — a $300 million round on top of a $134 million seed in under a year — is a signal that the next race after chatbots is AI that can act, not just talk. The company says it will use the cash to scale compute and ship a first product by late summer or early fall 2026. I'd hold the applause until that product is in hands: world models demo beautifully and break on contact with the real, messy world, and a $2 billion valuation is a lot of faith in videos of people playing games.

What I'd actually do

Don't trade on the hype, but do watch the category. If you build products, the practical takeaway is that 'AI that understands physical space' is about to get cheaper and more available — useful for anything involving robotics, AR, navigation or simulation. When General Intuition's first product lands, judge it on one question: does it predict what happens next in a real scene better than today's video models, or just look impressive in a curated clip?

#ai#world-models#embodied-ai#funding#robotics

Related guides

EAEvgenii Arsentev

Author

Evgenii Arsentev

PhD · Chief Product Officer at a tech company

Want to actually build this?

Guides explain. The free course transforms — personalized, gamified, and built to get you shipping fast.

◉ Start the free course

Source: techcrunch.com