Here’s Claude absolutely RIPPING building a modern mansion draft with primitives—using UnrealMCP to control Unreal Engine by himself. You prompt what you want, he does what you ask. He can be creative too. Tiny little mini tutorial thread 👇 🌸

Here is an example of why I say "it depends" when someone asks what is the best monocular mocap solution. Most would say that PromptHMR is the best one today. Is it? I was just giving a try on a video and turns out, that, in my opinion 1º- 4d humans 2º- GVHMR 3º-

x.com/i/article/2010…

We believe robots need instinct, not only reasoning. Introducing Project-Instinct — a full-stack, instinct-level whole-body control toolkit for legged & humanoid robots. 🔗 project-instinct.github.io (1/3)

Check out this repo if you want to play with this real-time character controller (need RTX 4090): github.com/mshoe/PARC The dataset and models are here: huggingface.co/datasets/mxucg… I also wrote a blog post about the autophagous data augmentation method: michaelx.io/blog/firstpape…

Introduce HumanX, a full-stack framework that compiles human video into generalizable, real-world interaction skills 🏀⚽️🥊📦 for humanoids, without task-specific rewards. Paper: arxiv.org/abs/2602.02473 Page: wyhuai.github.io/human-x/ #humanoid #ai #hkust #robotics #sports

🚀 Introducing CHIP: Adaptive Compliance for Humanoid Control through Hindsight Perturbation! Current humanoids face a trade-off: they are either Agile & Stiff OR Slow & Soft. CHIP breaks this barrier. We enable on-the-fly switching between Compliant (wiping 🧼,

We have seen rapid progress in humanoid control — specialist robots can reliably generate agile, acrobatic, but preset motions. Our singular focus this year: putting generalist humanoids to do real work. To progress toward this goal, we developed SONIC (nvlabs.github.io/GEAR-SONIC/),

🚀Excited to share that our paper EgoX👀 has been accepted to #CVPR2026 ! Huge thanks to my co-first authors (taewoongkang, dohyeon), co-authors (Minho Park,junhahyung) and Prof. Jaegul Choo. See you in Denver!🏔️ #VideoGeneration #WorldModeling #Robotics

Whole-body intilligence doesn’t need to be huge models; it needs to scale on the right task and compute Proud to be part of the team for bringing GEAR-SONIC to life. Open sourced!

InterPrior is accepted to #CVPR2026. Meanwhile, we’d love to share our ongoing open-sourcing efforts: InterAct: github.com/wzyabcas/Inter… InterMimic: github.com/Sirui-Xu/Inter… 1. InterMimic now supports multi-GPU training, as well as IsaacLab replay and inference. 2. InterAct—our

Animation 🤝 Robotics

💡Introducing 𝑼𝑴𝑶 -- one unified model that unlocks motion foundation model (HY-Motion Tencent HY) priors for 𝟏𝟎+ 𝐭𝐚𝐬𝐤𝐬: 𝐞𝐝𝐢𝐭𝐢𝐧𝐠, 𝐫𝐞𝐚𝐜𝐭𝐢𝐨𝐧 𝐠𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐨𝐧, 𝐬𝐭𝐲𝐥𝐢𝐳𝐚𝐭𝐢𝐨𝐧, 𝐭𝐫𝐚𝐣𝐞𝐜𝐭𝐨𝐫𝐲 𝐜𝐨𝐧𝐭𝐫𝐨𝐥, 𝐨𝐛𝐬𝐭𝐚𝐜𝐥𝐞

What’s the right representation for a world model? 3D, pixels, or something else? Excited to release our new paper “Forecasting Motion in the Wild” where we propose point tracks as tokens for generating complex non-rigid motion and behavior From @GoogleDeepmind @Berkeley_AI

We scaled off-policy RL to sim-to-real. To our knowledge, FlashSAC is the fastest and most performant RL algorithm across IsaacLab, MuJoCo Playground, and many more, all with a single set of hyperparameters. Project page: holiday-robot.github.io/FlashSAC Paper: arxiv.org/pdf/2604.04539