Had a great chat with one of my favorite investors, @generalcatalyst General Intuition x General Catalyst - The Next Phase of Pre-training

rsvp on the thread below the post

More and more, experts are coming to the consensus that world models, or AI that can understand the physical world, are the key to human-level generalized intelligence. But in an industry that's poured billions into betting on models that can only understand language, where does the next frontier lay? For our latest @thedeepview long form, we break down the anatomy of the world model market and discuss how language models may be a stepping stone towards AGI, rather than the endgame. thedeepview.com/articles/how…

Backstory / GI lore :)

ICML 2026! 🇰🇷 We're heading to Seoul next week. Come find our research team at Booth 111 to talk world models and action models - the core of what we're building at General Intuition. We're also hosting a party on Wednesday, July 8th for a curated group of researchers working at the frontier. If that's you, the event RSVP is linked on the thread. See you there!

In all seriousness, the name @gen_intuition comes from this @demishassabis talk, as the north star of the lab is a general agent, where "Play the World" applies to scientific problems that can become games too! Another example: youtube.com/watch?v=JMAv0NS6…

Don't threaten me with a good time, Andrew

General Intuition is headed to ICML in Seoul next week, July 6–11. Find our research team at Booth 111 to talk world models, action models, and the future we’re building toward. Going to be there? DM the team, we’d love to connect.

If this is how it all ends for me I’m here for it because it’s an S tier shitpost :)

it does seem like utilizing text & video to scale up to physics understanding of the real world is likely not enough. google is doing a bit of this with street view maps data simulation for this but the video game unlock seems absolutely legit esp if you have the full end to end loop of consequence mapping for ai to learn from that’s insanely rich (e.g. controller input → camera movement → game state transition → physics/collision response → npc reaction → player choice → reward/failure/social outcome). i personally grew up playing a ton of videogames as a kid including one of the first games to ever have open ended world simulation (it was called shenmue, absolutely pioneering, look it up). these guys seem to have a good understanding that these feedback loops in actual game play is the best data set for ai to behave in human contexts because that’s precisely how we learn & explore when we are young (like watch a kid play around).

.@gen_intuition CEO @PimDeWitte discusses using data from gamers to train AI & broader implications for the labor market: cnb.cx/3QHDIXj

General Intuition CEO @pimdewitte, who's building foundation models trained on video game controller input data ("action-labeled gameplay clips"), says general intelligence won't "taste like an LLM": "We have a scale of data that's going to allow us to jump to the frontier in one capability — which is any system that can be controlled with a game controller (which is most robots) — and then, you can use that to create a sufficiently general intelligence." "As humans, the decision to talk or type is a very, very small subset of the actions that we can actually take." "So in order to create a sufficiently general intelligence to play 10,000 video games, the model has to be able to predict across the entire action space of human cognition when they're interacting with these environments. Which are 2D and 3D environments, interfaces, long-horizon tasks, short-horizon tasks, [etc.]." "It has to be a sufficiently general intelligence in order to predict actions. Therefore, the type of model you get out is not going to taste like an LLM. This model is going to be incredibly good at navigating unforeseen environments. It's going to be incredibly good at zero-shotting any task that can be done with a game controller."

Time to push for the frontier with the best team to ever do it. Come join us.

General Intuition raises $320 million to develop AI from gaming axios.com/2026/06/26/general…

ICYMI: This is a pretty monumental story for games and AI. Pim de Witte there is a path forward for embracing innovation and respecting art. gamesbeat.com/general-intuit…

Nice writeup from @alexrkonrad which includes potentially the best quote ever recorded from a GI team member 🤣

"We are as focused on spatial-temporal dynamics as Anthropic was on code." @gen_intuition's @PimDeWitte explains why the next frontier in AI is teaching models how the physical world unfolds in time. These world models unlock robotics, simulation, and scientific discovery. Congratulations to Pim and the team on their Series A. We’re proud to keep backing them. Revisit @max_rimpel's full conversation with Pim below.

Two Point Three Billion reasons to understand what General Intuition is up to: notboring.co/p/world-models

While visiting @gen_intuition's NYC R&D lab, the team demonstrated that the same underlying model powering an agent playing 100 hours of video games was also powering a large quadrupedal robot. 1/ techcrunch.com/2026/06/25/ge… via @techcrunch

SITUATION EXPLAINED: What is the difference between a world model and a video model? We asked @PimDeWitte, CEO of @gen_intuition: "A video model is going to predict the next likely frame or next entertaining frame." "But it shouldn't be the next likely or entertaining. It should be the most precise reaction to what the action was taken." "A world model needs to be able to predict the full space of possible outcomes for any given action input. That's why these models are typically so hard to train. You need a lot of data in order to get that entire action and state space essentially mapped." "What it enables: you can simulate what would be really expensive either in the real world or in classical simulation. Car accidents, for example. These are very expensive to do in the real world, and quite difficult to do in simulation to an accurate degree." "You want to train on lots of data of people crashing cars, simulate that, in order for your self-driving car model to avoid such incidents. And that capability would be useful for every self-driving car company."