One human demonstration. Any multi-fingered hand. Zero-shot sim-to-real visuomotor policy. morphometricimitation.github… Collaborators: @he_siming @ckwolfeofficial @HaozhiQ @LeaMue27 Shankar Sastry, Claire Tomlin, @JitendraMalikCV

Egocentric human data is the key to a major breakthrough in robotics foundation models. Huge congratulations to the GSI team, especially @AdemiAdeniji, whose work is making this vision a reality

Grounded API is live. - SOTA on hand-tracking benchmarks (< 1 cm) - SOTA on SLAM benchmarks - In-the-wild ego data -> enriched data in minutes - Integration with @huggingface @LeRobotHF & @rerundotio - Built for @BitRobotNetwork RoboCap suite Technical report & more↓

I'm excited to release mjbatch, a simple tool that lets you step thousands of MuJoCo simulations in parallel on your CPU, all from the comfort of Python! github.com/kevinzakka/mjbatc… To show how far this primitive can take you, I built some fun, self-contained examples with it 👇

I came from a world where I measured robotics research in days: collect hundreds of demonstrations, train, then watch the policy break as soon as the setup changed. Since I joined Sunday, I’ve watched that loop compress dramatically. ACT-2 can learn a new folding strategy from one demonstration, with just five minutes of fine-tuning, then transfer it to unseen garments and beds. What once took a week can now happen in an hour. That’s the deeper impact of scaling pretraining: it doesn’t just produce a better policy. It produces a better learner.

Introducing ACT-2 Preview The first robotics model to unify broad generalization with high reliability. A single fine-tuning example can teach Memo a new behavior that generalizes. Zero shot, real unseen homes, 99% success rate.

FPO++ code is out! We’re releasing the full training and evaluation code for locomotion, motion tracking, and manipulation. github.com/amazon-far/fpo-co… Co-led with @brenthyi

we started a company!! so, we’re tackling continual learning: what’s the learning algorithm to take arbitrary data — documents, conversations, the models’ own experience — and make better models? how do we scale compute in the same way we’ve already seen with pre-training and inference time, but scaling on the same data we see as humans, day after day with no labels, no rewards? A lot of the ingredients are out there already (rl, distillation, long-context, sparse / param-efficient architectures, etc.). our team is at the frontier of these topics, and we’re singularly focused on this. we want to understand this problem better than anyone else in the world. nobody’s solved this problem yet, but even today it’s extremely greenfield opportunity to co-develop research & useful products. in our space, how people interact with the models defines what the data distribution is - and working on this problem end-to-end, from core science to end user, gives us incredible freedom to define the problem and imagine new kinds of experiences. i expect we’ll use models that continually learn much differently than we’re using them today. it’ll feel different when the models _just know_, and build on our thinking and direction in ways we can’t even imagine. we don’t even know the queries we’re not asking, the things we would do but aren’t able to today. i’m so excited to share what we’re doing with the world in the coming months!! and the team is extremely cracked :) tackling this grand challenge and working alongside @jxmnop @EyubogluSabri @dan_biderman @MayeeChen @__howardchen @shizhehe and many others has made every day so fun. come work with us!

Introducing ABC: open data, training, and infrastructure for robotics. We release the largest teleop dataset to date, and extensively investigate design decisions, pretraining, and post-training techniques. @arthurallshire @Cinnabar233 @adamrasb @redstone_hong @davidrmcall

Philipp and team are some of the most hardworking and talented scientists I’ve met. So bullish on this team!!! Very much worth keeping an eye out for them 👀

Our new work, STITCH 2.0, can perform consecutive running sutures to close a sample wound with the daVinci robot.

World models are heavy. They don't need to be. Each frame is encoded as 1024 spatial tokens. What if it were just 1? In our #CVPR2026 Highlight from Amazon FAR, we compress frames into "delta" tokens for efficient generative world modeling. Paper, code & models below ↓ (1/7)

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 @TTIC_Connect

Robotics: coding agents’ next frontier. So how good are they? We introduce CaP-X: an open-source framework and benchmark for coding agents, where they write code for robot perception and control, execute it on sim and real robots, observe the outcomes, and iteratively improve code reliability. From @NVIDIA @Berkeley_AI @CMU_Robotics @StanfordAILab capgym.github.io 🧵

Excited to share our latest work on motion generation! We tackled multi-agent generation across diverse tasks using Diffusion Forcing. Check out the project page for more! 🚀

When people share a space, their movements become intertwined. Embodied agents need to understand these social dynamics to interact effectively. Introducing MAGNet 🧲, a unified autoregressive diffusion forcing model for multi-agent motion generation that captures these interactions. MAGNet is flexible: predict the future, fill in missing motion, or have people react to each other, all while naturally scaling to N>2 people and generating ultra-long motion sequences.

So basically the most valuable thing to build right now is friendship

The viser viewer in mjlab just got a huge QOL upgrade! - Real-time factor control: go slower or faster than real-time and viewer paces physics to match - Single step mode: advance one physics step at a time (super useful for debugging!) - Overall faster and smoother

New in mjlab from the amazing @ki_ki_ki1: 8 new terrains and a viser-based terrain visualizer 😎

We trained diffusion models on a billion LLM activations, and we want you to use them! New preprint: Learning a Generative Meta-Model of LLM Activations Joint work with @feng_jiahai, @trevordarrell, @AlecRad, @JacobSteinhardt. More in thread 🧵