Upon entering the bustling R&D floor of General Intuition's New York office, 31-year-old co-founder and CEO Pim de Witte immediately drew attention to a monitor. On screen, a game resembling Fortnite was being played, though notably, no human was at the controls.
“Our agent has been playing for 100 hours straight,” beamed Kent Rollins, the company’s chief product officer, highlighting the AI’s endurance.
Before the captivating display of an AI skillfully navigating a virtual environment could fully sink in, the distinct electronic footsteps of a substantial quadrupedal robot echoed through the space.
“The same brain powering the agent playing the game is powering the robot,” de Witte revealed. Josh Duplantis, a data analyst, chimed in while carrying a laptop streaming a live feed from the robot’s single camera, explaining that the bot’s default setting was “exploration.”
Guided by its singular camera-eye, the large, insect-like robot approached, circled the observer, and continued its journey through the office. Its movements, akin to a curious toddler still learning its spatial awareness, occasionally led to minor collisions with chair legs or an errant trash bin. Duplantis highlighted the remarkable efficiency of its training: only eight minutes of real-world robotics data, collected on the street rather than within the office, were needed to fine-tune the AI model for the quadruped.
General Intuition's fundamental purpose, its raison d'etre, lies in developing an agentic model capable of generalizing understanding from gameplay to simulation and ultimately to physical embodiment. This groundbreaking ability to comprehend its place in the world has garnered significant support from prominent investors.
On Thursday, General Intuition confirmed previous TechCrunch reports by announcing it had successfully raised $320 million, achieving a valuation of $2.3 billion. This latest round brings the company's total disclosed funding to $454 million, following its initial $134 million round at launch last October.
The startup originated from de Witte’s other venture, Medal, a platform where gamers upload and share video game clips. The vast repository of hundreds of millions of hours of uploaded gameplay served as the foundational dataset to train General Intuition’s model in spatial-temporal reasoning—the crucial ability to understand movement through space and time.
However, de Witte emphasizes that the critical element wasn't merely the gameplay footage itself, but the embedded action labels within those clips: precise records of every button press and its timing. He contends that most competitors, who attempt to infer actions solely from video, are working with insufficient data.
“We view this as just the next stage of future pre-training,” de Witte stated. “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 point, de Witte provided a laptop running General Intuition’s world model, a simulated environment generated frame-by-frame rather than rendered by a conventional game engine. As is often the case when testing such models, the reporter intentionally walked into a series of walls. Unlike some other demos where agents might pass through, this model accurately recognized the solid boundaries. It had, through millions of hours of gameplay, intuitively learned that walls are barriers, ladders are for climbing, and shadows naturally lengthen with the sun's movement.
For General Intuition, this sophisticated world model is not the end product; it serves as the internal "gym" or training environment. The company's ultimate aim is to commercialize the agentic model itself. De Witte posits that the embedded action data in gameplay uniquely helps the model differentiate "self" from "environment," fostering a richer understanding of causality.
While General Intuition’s technology is undeniably impressive in demonstrations, it operates within a competitive landscape where other companies are also striving to solve similar problems. Moreover, scaling such a model to reliably perform in the physical world remains an unsolved challenge. Most conventional approaches demand enormous quantities of real-world data, collected slowly and expensively. General Intuition's strategic gamble is that gameplay offers a scalable and efficient shortcut.
This strategic bet has resonated with its investors. General Intuition's latest funding round was spearheaded by Khosla Ventures, with additional participation from General Catalyst, Jeff Bezos, Eric Schmidt, Nico Rosberg, and researchers from Google DeepMind and MIT.
The substantial majority of the newly acquired funds will be directed towards significantly expanding compute capacity. General Intuition has secured an agreement with CoreWeave and plans to focus intensely on pre-training the next iteration of its model. A portion of the funding has also been allocated to making its API more widely accessible by the close of summer.
Vinod Khosla, whose firm led the investment round, expressed that he was particularly drawn to de Witte’s compelling vision and the company’s unique proprietary data position.
“If you look at LLMs, when reasoning emerged, it was a quantum leap,” Khosla remarked in a phone 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.”
General Intuition is not alone in recognizing the immense value of Medal’s human action data as a pivotal piece in constructing dynamic world models and general agents. Brianna Martin, the startup’s chief of staff, disclosed that the company’s inception was partly spurred by Medal’s decision to decline an acquisition offer from a major laboratory, with further offers having materialized since.
De Witte and his co-founders, Eloi Alonso, Adam Jelley, and Vincent Micheli, are not pursuing an acquisition, nor are their investors currently seeking an exit. The sheer volume and quality of proprietary data General Intuition possesses through Medal is a primary reason Khosla views the startup as a generational investment rather than a mere M&A target, envisioning its potential to become the foundational backbone for generalized agents and world models in both simulated and real-world environments.
“At this point, it would be a data acquisition, which is sort of uninteresting,” Khosla asserted.
A significant aspect of this long-term investment also involves a profound trust in de Witte’s core values.
De Witte dedicated three years to humanitarian work, including with Doctors Without Borders. This background has profoundly influenced his stance, leading him to draw a clear ethical line: General Intuition’s technology will never be employed to harm humans.
“We don’t want to be an escalatory part of the system,” de Witte stated, posing the rhetorical question, “Let’s say I were to come out and say, ‘We’re doing lethal autonomy.’ What do you think would happen in other countries?” This firm restriction on military applications stands in contrast to Silicon Valley’s increasing embrace of defense technologies, though de Witte confirms his models are welcome for search and rescue operations.
De Witte, being Dutch, and with a predominantly European team, reflects this identity in the company's ethos. He specifically mentioned hiring Martin partly due to her public resignation from Palantir over its involvement with the United States Immigration and Customs Enforcement.
“I don’t know why Silicon Valley does what it does,” he commented. “There’s a reason I’m not there.”
De Witte’s ethical framework extends beyond simply limiting what the models will not do; it also encompasses a deep concern for individuals who might be displaced by advancements in AI. As a gamer who earned $1.5 million by building and hosting a private RuneScape server in his teenage years, de Witte possesses a unique understanding of this societal challenge.
To address this, General Intuition recently launched Nerve, a jobs marketplace designed to enable gamers to monetize their existing setups. Participants can begin with data labeling tasks and progressively move towards more complex roles such as robot teleoperation. De Witte highlighted that Medal's extensive user base represents precisely the generation most susceptible to AI-driven job displacement, and he aims for them to have a vested interest in the unfolding technological future.
De Witte envisions General Intuition as an ecosystem enabler, much like Anthropic or OpenAI—a foundational model provider that empowers others to innovate upon its technology. Currently, the startup serves a diverse clientele across gaming, simulation, and robotics sectors.
“We’re not gonna build a self-driving car company,” de Witte affirmed. “We’re gonna make it 10 times easier for the next person to build a self-driving car company.”
The company anticipates that by making its API more broadly available to customers, it can rigorously test its capabilities across a myriad of use cases. These include deploying a robot within a digital twin of a factory floor, powering a human-like bot for a gaming studio, or dispatching a quadruped robot to navigate hazardous environments.
While the quadruped robot represents General Intuition's initial foray into physical embodiment in the real world, the company has also experimented with drones and other devices, including testing its model in driving games.
“It works on anything that you can control using a game controller or a keyboard mouse,” de Witte stated, emphasizing the model’s broad applicability.
A significant objective for the company is to establish a robust data flywheel.
“We’ll pick customers where we can diversify the embodiments that this generalized foundation model is serving as the backbone for,” de Witte explained. “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 reiterated that General Intuition’s proprietary data has been instrumental in its journey thus far, and its continued ability to collect unique data will be crucial for future success. He acknowledged that despite impressive demonstrations, the question of whether the simulation-to-real-world transfer can effectively scale remains an open challenge that no one has yet fully resolved.
The Editorial Staff at AIChief is a team of professional content writers with extensive experience in AI and marketing. Founded in 2025, AIChief has quickly grown into the largest free AI resource hub in the industry.