Building @Pranalabs_ Prev: @Anuttacon_ @miHoYo @PKU1898 @erasmusuni
Palo Alto, CA
Joined April 2018
- Tweets203
- Following994
- Followers202
- Likes770
Taier cooked
ps. we don't really try to figure out what a world model is. we just focus on making models spatially smart :)
After 5 yrs internet frens, finally met @VoidAsuka 😭😭 and met new fren @Jsavetheworld working on world model and 3d
The real one-shot RSI opportunity for startups isn't in LLMs — it's in real spatial intelligence.
We are excited to announce that World Labs is joining @AMD.
The research and technical breakthroughs we have achieved since our founding in 2024 have given us a clear vision for AI’s potential to solve problems in the spatial and physical world.
Accelerating the future of spatial and physical intelligence requires scaling our efforts, scaling our reach, and getting closer to the hardware.
Pho retweeted
Great to work with @tripoai at SIGGRAPH 2026!
Their industry-leading 3D generation, our real-time contact physics. Together we showed a generated world can hold up under grasping and contact, not just look good.
Intelligence needs environments that push back. Excited for more.
At SIGGRAPH 2026, we partnered with @PranaLabs_ around a simple question: can a generated world do more than look real?
Tripo provided functional, simulation-ready 3D assets. Prana’s FLASH engine handled contact, grasping, and object dynamics, while a generative renderer developed through our Project Eden research turned the underlying simulation state into high-fidelity visuals.
The result was a world that looked convincing and responded realistically to robotic interaction. The demo showed how generated worlds are evolving into spaces that machines can perceive, test, and act within.
Excited to announce our SIGGRAPH Stage Session next Wed, July 22, co-hosted by Tripo × @PranaLabs_
Topic: "Seeing is Believing, Grasping is Physics" — a deep dive into how perception meets physical interaction.
Come witness it live. Be there. 🚀
We're excited to be part of SIGGRAPH 2026, a world-leading conference for computer graphics and interactive technology.
This year, Tripo is heading to Los Angeles with five accepted papers, a featured keynote, and technical sessions. During SIGGRAPH week, we will also take part in the Worlds in Action Hackathon and host various side events.
Meet the Tripo team and explore generative 3D, world models, and the future of interactive media.
See our full SIGGRAPH 2026 schedule: tripo3d.ai/tripo-siggraph-20…
Pho retweeted
Contact does not always deform surfaces, especially when interacting with liquids and soft objects. However, most existing tactile sensors rely on surface deformation to infer contact.
Excited to share LightTact, a visual-tactile fingertip sensor that makes contact directly visible. It provides:
- Deformation-Independent Sensing: contact is detected optically, not inferred from surface deformation.
- High-Contrast Raw Images: non-contact pixels stay near-black.
- Pixel-Level Contact Segmentation: robust across materials, forces, appearances, and external lighting conditions.
This project was co-led with Boda (bodahuo.github.io/). Grateful to all co-authors for their contributions, and many thanks to @yxyang1995 for the insightful discussions.
We will present LightTact at RSS 2026 in Sydney next week!
Paper: arxiv.org/abs/2512.20591
Open-source: github.com/linchangyi1/Light…
Indeed. Gaming world is a natrual bridge between software and the real physical world. So much unexplored.
The term “world model” is tossed around a lot but this is the real deal. From video and actions, the model learns a consistent representation of the physics, 3D environment, and game state of Rocket League.
I'm excited to have been part of this project because I see this as a step towards learning world models that can generalized to the real world. I’m a believer in learning from synthetic data (Sintel, SURREAL, AGORA, BEDLAM, BEDLAM2) and I’ve seen how models trained on synthetic data can generalize.
MIRA is an important step because it’s learned from video gameplay generated completely by bots. This provides a pathway to scaling and scaling is key to real-world generality. Traditional graphics and games provide a path to learning rich models of the world.
Check out the detailed tech report: mira-wm.com/paper/
Plenty of people chase the "world model" label for the fundraising narrative.
Very few simulate the world it's supposed to describe.
Emotional Intelligence is one of the hardest problems worth solving. Probably no real breakthrough until robotics catch up.
Pho retweeted
We've spoken with hundreds of ad creatives, marketing designers, filmmakers, and animation teams — and heard the same thing: the outputs look great… until they don't 😅. When they fail, it's incredibly hard to tell why. Is it the prompt, the model, or the world itself quietly breaking? That ambiguity is the real bottleneck.
Physion-Atlas 1.0 introduces a more objective, diagnostic way to evaluate video world models — moving beyond high-level comparisons to surface what actually matters. It disentangles prompt misalignment from physical and visual inconsistencies, grounding every judgment in explicit spatiotemporal evidence. Not just which output is better, but what breaks, when, where, and why. From abstract comparisons → diagnosable reality 🔍
📄 Blog: physionlabs.ai/blog/physion-…
📝 Evaluate your model: docs.google.com/forms/d/e/1F…
Pho retweeted
I am once again reaching out to you all to apply @ nous for the post training team to work with me and our team on advancing several core capabilities of our models!
If you want to work on:
- Creativity, Roleplaying, and Simulation
- STEM
- Math
- Code and Code Agents
- Innovative New Agents
- Instruction Following and RLHF
- Safety and Uncensoring
- Multimodality
- Training and Data Synthesis Infrastructure
*and* have experience training models, doing RL or ML, data engineering, or working with synthetic data.. and are a highly self-motivated individual, then you should apply.
This is a full time job, is fully remote, has competitive pay, and will be hard work!
You can DM me directly or through nousresearch.com/careers
We want those who will dedicate their time toward advancing our models and want to do so with a passion.
Pho retweeted
For our AI game, we didn't use ChatGPT. We trained our own LLM for more immersive roleplay and more natural conversations
#WhispersfromtheStar #AI #Gaming #Anuttacon #ArtificialIntelligence
Pho retweeted
Thrilled to announce the @Kaggle Game Arena, a new leaderboard testing how modern LLMs perform on games (spoiler: not very well atm!). AI systems play each other, making it an objective & evergreen benchmark that will scale in difficulty as they improve.
kaggle.com/game-arena
