It's hard to believe it's been almost two years since we started XDOF.
What began as a conviction that robotics needed a new infrastructure layer has grown into a company supporting some of the most ambitious teams building robot foundation models.
This milestone belongs to our incredible team. Thank you for the relentless hard work, persistence, and belief through every challenge along the way.
Thank you to our customers, partners, and investors for your trust and support. We're grateful to be building alongside you.
We're still at the beginning. Excited for what's ahead.
Announcing @xdofai: We’ve raised $70 million to build the core robotic infrastructure ecosystem for robot foundation models.
My cofounders Fred (@YideShentu), Nemo (@itsnemojin) and I have been pursuing the dream of general purpose robots for our entire lives. After work at Covariant, Meta and Tesla, it became clear to us that general purpose robots are coming, and we are building XDOF to help make them a reality.
For the last two years, we’ve been working behind the scenes to support major labs and companies deploying robots. In us, they have a partner with full-stack expertise, from hardware to operations to policy training.
As our first public contribution to the space, we are open-sourcing ABC-130K, the largest open source teleoperation dataset, in collaboration with our partners from UC Berkeley, Carnegie Mellon, MIT and Amazon FAR.
Thank you to our customers, partners, collaborators and investors for your trust and conviction in us. Together, we can accelerate the future of robotics!
To truly capture human intelligence, robot data collection must align with human nature along two fundamental dimensions: manipulation and perception.
We built ActiveGlasses around this principle, enabling robot active vision and manipulation as a naturally coordinated behavior.
[Accepted to ICRA 2026!] 🚀 Introducing EgoMI: An egocentric manipulation interface that captures synchronized 6-DoF head and hand trajectories from egocentric human demonstrations!
Transfers to IL policies zero-shot w/o visual augmentation or on-embodiment data.
1/n
@chris_j_paxton The box folding is probably faked (teleoperated)
Such a long traj uncertain task even if you can reset the env to the same starting state I don't think replay will work.
@benjamin_bolte if all these are from the same model that's already very impressive. Can you robot just f*** walk like a normal humanoid? (not even like a human)
If no please stop saying this. Talking is cheap.
@kscalelabs Spotted on Chinese social media: another “innovation” outsourcing everything to China. But I guess that's still better than the garage-built 3D printed humanoids that can never walk?
Very excited to to have ROS integration for GELLO, built by Franka Robotics! Robot teleoperation data is becoming more and more important for large scale Robot foundation model training. Great to see larger adaption for low cost, scalable teleoperation!
@trossenrobotics Next generation QDD = Damiao Motor from China?
https://t.co/kmBaD4DU95
this robot looks like off the shelf motor + simple cnc parts connecting them together, unsure why you call it breakthrough AI platform.
I first learned the term "holonomic" during a conversation with @jimmyyhwu while experimenting with the AgileX Ranger Mini. Despite being omnidirectional, the Ranger Mini isn't holonomic. Initially, I thought this distinction didn't matter—until I tried to make the base follow human pelvis motion.
A non-holonomic base requires steering before turning, introducing additional latency and jerkiness. This made the teleoperation experience quite frustrating.
Congrats on the true full open-source release! 🎉 I hope more people get to enjoy this incredible base concept! 🚀
When will robots help us with our household chores?
TidyBot++ brings us closer to that future. Our new open-source mobile manipulator makes it more accessible and practical to do robot learning research outside the lab, in real homes!
(No functional robots were harmed in this project) 🤖🔧
Since finding a computer with FireWire input is nearly impossible these days, I cracked open the OG Torch X to peek at how these mini robots are built.
Turns out, it’s a cable-driven transmission for ultra-low friction while maintaining a solid gear ratio, paired with a coreless motor for zero cogging torque.
This one can only provide force feedback on the xyz translation, rotation part it only has encoders to read the user input.
Huge thanks to @3dsystems for the loaner haptic device! I've always seen a really old one sitting at our lab even since I was a undergraduate but never actually tried it. It’s pure magic—being able to feel the shape, texture, weight, and physics of objects right in mid-air. It’s like a precisely-engineered mini robot arm with super smooth, low-friction transmission and ultra-low inertia. A whole different level compared to zero-gravity drag modes on standard collaborative robots. Can not wait to see how this force feedback mechanism could combine with GELLO https://t.co/w1sqEpbMMq to enhance data collection quality!