We hosted Prof. Alyosha Efros (UC Berkeley) at @SkildAI! He didn't believe that robots could actually cook eggs reliably. :) Tested back-to-back 5times without fail! One batch of scrambled eggs every ~2.5mins nonstop. The same model assembles a GPU on a server rack too.

Congratulations to @berkeley_ai alumnus @pathak2206 who has been awarded the PAMI Young Researcher in Computer Vision Award! This top award for young researchers in computer vision is given to two recipients yearly. thecvf.com/?page_id=413#YRA

Force is arguably the most overlooked ingredient in modern robot learning. Introducing FACTR 2: it turns *any* commodity robot into a force-aware system with no force sensors required. Train a tiny force network in <1min with <10mins of data and drop it into any existing teleop pipelines: ✅ Free force sensing for both the robot and the operator arm ✅ Makes demos higher-quality → fewer of them needed. ✅ A new force-aware learning algorithm (FIRST) uses those recovered forces to figure out which parts of a demo actually matter, making learning data-efficient. ✅ Strong performance on complex tasks with fewer demos and even no pretraining! More details below.

I’m deeply grateful to receive the PAMI Young Researcher Award. This recognition is a reflection of the exceptional students, colleagues, collaborators, and mentors I’ve had the privilege to learn from and work alongside. 🙏😇 Thanks, TC PAMI and @CVPR, for this honor.

Update: This work has been accepted at ICML 🎉 We’ve also released our code, model weights, dataset and all resources here: github.com/Sim2Reason/Sim2Re…

🚀 Excited to share ViPRA: Video Prediction for Robot Actions 📍 Accepted to #ICLR2026 @iclr_conf 🏆 Best Paper — #NeurIPS2025 Embodied World Models Workshop Robot learning today still needs millions of action labeled videos. Yet videos are abundant — from humans and the web — but lack action labels. Meanwhile, pretrained video models already learn rich dynamics. ViPRA is a recipe for turning pretrained video models into robot policies while enabling robot learning to scale with actionless videos. 🧵 Thread ↓

What if one AI brain could run every robot on the planet—from a humanoid to a warehouse arm—all at once? 🧠 @pathak2206, CEO and Co-Founder, and Abhinav Gupta, President and Co-Founder of @SkildAI, explain how they are building "OmniBrain," a universal foundation model designed to generalize intelligence across any robot form factor and task. 📺 Watch the episode: nvda.ws/4mKYvVu

Given how everyone are refining their research ideas with LLMs, I definitely think training LLMs with simulated and real world videos and data would go a long way in research iterations. Probably OAI and Anthropic should use such pipelines with the growing robotic data!!

Let models learn directly from simulated worlds, observing how objects move, collide, and interact, much like scientists build intuition through experiment. 모델이 물리적 상호작용에 대한 예측을 할 수 있다면, 곧 내부적으로 월드모델을 가지게 될것이다. 마치 사람처럼.

Excited to share Sim2Reason -- training LLMs in simulation to learn Olympiad-level physics (mechanics)! Today, LLMs learn science by reading what humans have already written, absorbing distilled knowledge from textbooks and the internet. But human-annotated physics data is fundamentally scarce, and that bottleneck isn't going away. Analogy to robotics: Sim2Real transformed robotics, where we train in simulation and deploy zero-shot in the real world. We do not try to teach robots by describing physics to them, but they have to experience it. Approach: Our Sim2Reason makes the same bet we made in robotics -- skip the descriptions, go straight to the source. Let models learn directly from simulated worlds, observing how objects move, collide, and interact, much like scientists build intuition through experiment. Result: Models trained purely on simulated experience develop transferable physical reasoning skills, improving even on problems that were never simulated. Zero-shot gains on IPhO, IIT JEE Advanced, OlympiadBench — problems the model never saw during training.

🚨 BREAKING: Skild AI acquires Zebra Technologies robotics division! 🤯 @SkildAI acquired the robotics division of @ZebraTechnology (formerly Fetch Robotics) to deploy their omni-bodied brain across warehouses, unlocking massive productivity gains and accelerating their data flywheel. Most warehouse robotics solutions use classical approaches for navigation and routing, but many parts remain human-bottlenecked like moving objects between receptacles. Zebra Technologies brings one of the most battle-tested warehouse robotics platforms in the industry. Their Symmetry Fulfillment orchestration platform already coordinates tasks between robots and frontline workers using real-time data from Zebra wearable devices, proven in logistics environments where reliability is mandatory. The Skild Brain is an omni-bodied foundation model that generalizes across embodiments without retraining from zero. Quadrupeds, humanoids, tabletop arms, mobile manipulators—the same underlying model operates all of them. Adding the Skild Brain to Zebra's Symmetry platform means robots don't just follow instructions, actually they make informed decisions. The Symmetry platform will expand beyond its current footprint into new verticals, new use cases, and a wider range of robot form factors. This accelerates Skild's data flywheel, bringing in more diverse data to train the omni-bodied brain. @pathak2206 🔥 ~~ ♻️ Join the weekly robotics newsletter, and never miss any news → ziegler.substack.com

Skild Brain preparing an omelet with everyday human tools. The robot drops an eggshell into the bowl at one point but recovers and continues the task. The ability to self-correct during edge cases is what will make robots dependable for complex, long-horizon missions.

AFAIK this 2021 paper was sort of first “proof” (analysis by synthesis) of Hoyt and Taylor’s (Nature 1981) hypothesis from Biomechanics: x.com/pathak2206/status/1518…

Can't remember things from 5 years ago apparently