Ever wanted to play Counter-Strike in a neural network? These videos show people playing (with keyboard & mouse) in 💎 DIAMOND's diffusion world model, trained to simulate the game Counter-Strike: Global Offensive. 💻 Download and play it yourself → github.com/eloialonso/diamon… 🧵

Live demo (4 players): generalintuition.com/mira Full code (training & inference): github.com/mira-wm/mira Dataset (multiplayer, aligned video/action/game state): huggingface.co/datasets/kyut… Paper (loads of nuggets): mira-wm.com/paper/

Introducing MIRA, a multiplayer world model, a dream of Rocket League (links in 🧵)

Today we’re introducing MIRA, a new multiplayer world model, built with @gen_intuition, in collaboration with Epic Games. We release an in-depth technical report, dataset, as well as an online demo that you can try right now (link below).

Really proud of the team. Genuinely a good read!

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.

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).

Very proud of the team, and excited for what's ahead!

Announcing our $320M Series A at a $2.3B valuation, led by @khoslaventures, with @generalcatalyst, @ericschmidt and @JeffBezos. General Intuition is the frontier lab for acting in space and time. We build large action foundation models trained on billions of ground truth action-labeled gameplay clips from 17M monthly active users on Medal, and push the frontier of world models to generate infinite training environments.

We’ve just raised $320M to do it at scale, and we're hiring. Thank you @khoslaventures @generalcatalyst @ericschmidt @JeffBezos for backing us.

And unlike any other video, gameplay comes with ground-truth actions. We know the exact input behind every frame. That's the whole difference: a model can learn what an action causes, not just how a scene looks.

@gen_intuition we build foundation models to act: in games, in simulation, eventually in the physical world. The embodiment is downstream.

The largest existing source of action is gameplay. The usual objection writes it off as too far from the real world. Fair point. We think it's backwards. The coming explosion of embodiments won't converge on the human body. It'll favor the simplest ones, with control clean enough to drive like a game controller. Games already live at that layer.

Everyone wants to scale action models. Almost no one can, because action doesn't sit on the internet the way text does. There's no corpus of what to do, only of what to say.

1/ What happens when a denoiser trained at one noise level is reused inside an iterative sampler? We trained SwinIR with Noise2Noise at σ=10, then dropped it unchanged into the constrained sampler of @ZKadkhodaie & @EeroSimoncelli (NeurIPS ’21) for 10% random inpainting. Baseline SwinIR: 6.08 dB on Set12. SwinIR-WNE: 23.87 dB. Same backbone. Same N2N pairs. Same sampler. With François Fleuret @francoisfleuret. ICML 2026. 🧵

We made the best voice cloning in the industry, capturing accents, prosody, and identity perfectly. You don’t have to trust us: run the blind test, tweak intensity with classifier-free guidance, and start building with our API today: gradium.ai/blog/voice-clonin…

Me defending my O(n^3) solution to the coding interviewer.

We’re excited to introduce Pocket TTS: a 100M-parameter text-to-speech model with high-quality voice cloning that runs on your laptop—no GPU required. Open-source, lightweight, and incredibly fast. 🧵👇

Yann LeCun (Chief AI Scientist, Meta, @ylecun), @PimDeWitte (CEO, General Intuition), and Aude Durand (Kyutai, @aude_drn), talk about world models, embodied agents, Yann's new company, and the limitations of LLMs 0:00 - Introduction to World Models 5:00 - Why World Models, Intuition & Introducing Yann's new company 10:00 - Architectures Merging Language & Interaction Data towards General Agents 20:00 - Open Source, Sovereign AI & @kyutai_labs Partnership Keynote for #aiPULSE2025 at Station F in Paris 🇫🇷

General Intuition, l'IA qui "prédit" le monde 🌐 "La prochaine frontière de l'IA, c'est de développer des modèles capables de percevoir et d'interagir avec le monde réel" 💬 @micheli_vincent, Cofondateur de @gen_intuition 🎙️ @Fsorel

Excited to do the keynote on World Models with @ylecun at AI Pulse on Thursday. still not quite sure why I got invited to share the time with this legend (if it’s a typo, you can still tell me!), but hopefully we can all go find out in Paris, I will be as curious as you! 🤣