This web page was created programmatically, to learn the article in its authentic location you may go to the hyperlink bellow:
https://techcrunch.com/2026/06/25/general-intuitions-2-3b-bet-that-video-games-can-train-ai-agents-for-the-real-world/
and if you wish to take away this text from our web site please contact us
As quickly as I entered General Intuition’s R&D flooring at its New York workplace, the corporate’s 31-year-old co-founder and CEO Pim de Witte directed my consideration to a monitor perched on a standing desk. Someone seemed to be enjoying one thing like Fortnite. It wasn’t an individual.
“Our agent has been playing for 100 hours straight,” Kent Rollins, the corporate’s chief product officer, mentioned, beaming.
Before I might get absorbed within the spectacle of an AI navigating the sport’s digital setting, I heard the digital footsteps of a big quadrupedal robotic approaching.
“The same brain powering the agent playing the game is powering the robot,” de Witte instructed me.
Josh Duplantis, an information analyst carrying a laptop computer streaming a reside feed from the robotic’s single digicam, piped as much as clarify that the bot’s default mode was “exploration.”
Relying on that digicam, its singular eye, the large buglike bot walked as much as me, circled round me, and continued into the workplace. It often clipped the legs of chairs or ran into an errant trash bin, very similar to a toddler who hasn’t but realized how her physique pertains to the world round it. Duplantis mentioned it took simply eight minutes of real-world robotics information to fine-tune an AI mannequin for the quadruped. What’s extra, that information was collected on the road, not contained in the workplace the place the bot was at present navigating itself.
An agentic mannequin that may generalize from gameplay to simulation to embodiment is General Intuition’s raison d’être. And that mannequin’s capability to determine its place on this planet has secured the backing of some heavy hitters.
On Thursday, General Intuition mentioned it raised $320 million at a $2.3 billion valuation, confirming TechCrunch’s earlier reporting. The spherical brings General Intuition’s complete disclosed funding to $454 million, after the $134 million spherical it raised at launch final October.
The startup was spun out of de Witte’s different firm, Medal, which permits avid gamers to add and share online game clips. The lots of of hundreds of thousands of hours of uploaded gameplay offered the preliminary dataset to coach General Intuition’s mannequin in spatial-temporal reasoning — or understanding learn how to transfer by area and time.
But the important thing ingredient wasn’t the gameplay footage; it was the motion labels embedded in these clips: information of precisely what buttons a participant pressed and when. Most opponents, de Witte says, try to deduce actions from video alone, which he argues is inadequate.
“We view this as just the next stage of future pre-training,” de Witte mentioned. “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.”
At one level, de Witte set me up with a laptop computer operating General Intuition’s world mannequin, a simulated setting generated frame-by-frame fairly than rendered by a standard sport engine. As I typically do when testing world fashions, I walked straight right into a collection of partitions. In different demos I’ve tried, the brokers you management generally cross proper by, however this one didn’t. From the hundreds of thousands of hours of gameplay, it in some way realized that partitions are partitions, ladders are for scaling, and shadows lengthen because the solar strikes.
For General Intuition, this world mannequin isn’t the product; it’s the coaching setting (known as “the gym” internally). The firm finally needs to promote the agentic mannequin itself, and de Witte argues that the motion information embedded in gameplay helps the mannequin discern the “self” from the “environment” in a approach that offers it a richer understanding of causality.
Impressive although General Intuition’s expertise seems in demos, the corporate isn’t the one one attempting to crack this drawback. Moreover, getting such a mannequin to carry up within the bodily world, at scale, hasn’t but been completed. Most approaches of this sort require huge quantities of real-world information that’s gathered slowly and expensively. General Intuition’s wager is that gameplay is a scalable shortcut.
Its traders are okay with that wager, too. General Intuition’s newest spherical was led by Khosla Ventures, with participation from General Catalyst, Jeff Bezos, Eric Schmidt, Nico Rosberg, and researchers at Google DeepMind and MIT.
The overwhelming majority of the spherical will go towards scaling compute capability. General Intuition has a take care of CoreWeave and plans to deal with pre-training the subsequent model of the mannequin. A slice has been earmarked for making its API extra broadly obtainable by the tip of summer season.
Vinod Khosla, whose agency led the spherical, says he was drawn to de Witte’s imaginative and prescient and the corporate’s proprietary information place.
“If you look at LLMs, when reasoning emerged, it was a quantum leap,” Khosla instructed me in a cellphone interview. “In world models, I think the quantum leap is the emergence of intuition in the AI, a human intuition-like capability. The human action data and reaction data you have in games is the key part to the emergence of intuition.”
The imaginative and prescient is a generational firm

General Intuition isn’t the one firm to note that Medal’s human motion information is a key piece of the puzzle of constructing dynamic world fashions and basic brokers. Brianna Martin, the startup’s chief of employees, mentioned the corporate was born, partly, after Medal turned down an acquisition provide from a serious lab. There have been different provides since, too.
De Witte and his co-founders, Eloi Alonso, Adam Jelley, and Vincent Micheli, aren’t all for being acquired, and neither are the startup’s traders on the lookout for an exit simply but. The quantity and high quality of proprietary information General Intuition has through Medal is without doubt one of the causes Khosla is satisfied the startup is a generational wager, not an M&A goal; that it might turn into the spine for generalized brokers and world fashions in simulation and the actual world.
“At this point, it would be a data acquisition, which is sort of uninteresting,” Khosla mentioned.
Part of that wager additionally includes trusting de Witte’s values.
The entrepreneur spent three years working within the humanitarian area, together with with Doctors Without Borders. As such, he has drawn a transparent line for the way General Intuition’s tech might be used: No brokers might be employed to hurt people.
“We don’t want to be an escalatory part of the system,” de Witte mentioned. “Let’s say I were to come out and say, ‘We’re doing lethal autonomy.’ What do you think would happen in other countries?”
That restrict on army use circumstances comes as Silicon Valley is rising ever extra bullish on struggle, although de Witte says he’s comfortable for his fashions for use for search and rescue missions.
De Witte is Dutch, and far of his staff is European, which shapes the corporate’s id. He says he introduced on Martin partly attributable to her choice to publicly quit Palantir over its work with the United States Immigration and Customs Enforcement.
“I don’t know why Silicon Valley does what it does,” he mentioned. “There’s a reason I’m not there.”
De Witte’s ethics don’t merely restrict what the fashions received’t do. As a gamer who made $1.5 million by constructing and internet hosting a personal RuneScape server in his teenagers, de Witte can also be fascinated with what occurs to the individuals who get left behind by what AI fashions can do.
General Intuition not too long ago launched a platform referred to as Nerve, a jobs market that lets avid gamers earn cash utilizing their current setups. Those who join begin with information labeling and may finally transfer towards robotic teleoperation and different duties. Medal’s person base, de Witte famous, is exactly the technology most uncovered to AI-driven displacement, and he needs them to have a stake in what’s coming subsequent.
A knowledge flywheel
De Witte needs General Intuition to be an ecosystem enabler, like Anthropic or OpenAI — a mannequin supplier that permits others to construct on high of its expertise. Today, the startup has a handful of shoppers in gaming, simulation, and robotics.
“We’re not gonna build a self-driving car company,” de Witte mentioned. “We’re gonna make it 10 times easier for the next person to build a self-driving car company.”
The firm says as soon as it will get its API into extra clients’ arms, it could be capable of check its mettle with a wide range of use circumstances — like testing a robotic in a digital twin of a manufacturing unit flooring, powering a humanlike bot inside a gaming studio, or sending a quadruped to navigate hazardous environments.
While a quadruped is the primary bodily embodiment that General Intuition has tried in the actual world, it has additionally tried drones and different gadgets, together with testing the mannequin in driving video games.
“It works on anything that you can control using a game controller or a keyboard mouse,” de Witte mentioned.
The risk to construct an information flywheel is without doubt one of the targets.
“We’ll pick customers where we can diversify the embodiments that this generalized foundation model is serving as the backbone for,” de Witte mentioned. “So we’re going to prioritize picking customers on whether they can offer real-world data that’s going to be interesting and useful to move the needle on research. And if they’d have an agile internal team where we can be real embedded partners and learn from each other.”
Khosla mentioned that General Intuition’s proprietary information is what acquired it this far, and its capability to proceed gathering information that nobody else has might be important. Especially as a result of, regardless of spectacular demos, whether or not the simulation-to-real-world switch can maintain at scale is an open query that no one has totally answered but.
Correction: The headline beforehand misstated how a lot General Intuition raised on this spherical. The error has been mounted.
When you buy by hyperlinks in our articles, we might earn a small fee. This doesn’t have an effect on our editorial independence.
This web page was created programmatically, to learn the article in its authentic location you may go to the hyperlink bellow:
https://techcrunch.com/2026/06/25/general-intuitions-2-3b-bet-that-video-games-can-train-ai-agents-for-the-real-world/
and if you wish to take away this text from our web site please contact us

