1/ Better robot models come from better data.
That takes two things: hardware that captures it fully, and tooling that turns it into training signal.
We built the first. Today, @GroundedSI launches the second: Grounded API. Access both hardware + API ↓
We’re heading to CoRL with the best of BitRobot’s ecosystem.
Visit our booth for research lightning talks, hands-on egocentric demos from our partners, and a look at new product prototypes.
@micoolcho will also be taking the stage for a keynote. See you in Austin, Nov 9–11!
RT @ZenO4AI: We’re excited to announce our collaboration with @BitRobotNetwork, an open robotics network bringing together contributors, data, and infrastructure for Physical AI.
ZenO is launching SN/05 on BitRobot, connecting real-world Physical AI data collected through the ZenO iPhone app with BitRobot’s broader robotics ecosystem.
Through this collaboration, approved data contributions from ZenO can become part of a broader robotics data network, while opening up additional rewards for contributors.
More details below ↓
1/ SN/05 is live on BitRobot, built with @ZenO4AI.
We’re scaling egocentric data with ZenO: first-person iPhone demos of real-world tasks, captured with ARKit spatial tracking and structured for robot learning.
Want to help grow our open robotics network? ↓
5/ Each recording pairs first-person video with iPhone's ARKit spatial tracking, capturing the camera’s position and orientation in 3D, frame by frame.
The video shows hand motions and objects, and the depth information anchors them in real space. Both are key signals for world models and manipulation policies.
6/ Every task you record helps build the open robotics lab. Earn Bolts, level up your Lab Access ID, and advance robot learning.
Record:
iOS App Store Download:
Track Bolts:
Congrats to the Princeton team on their #CoRL2026 acceptance, tested zero-shot on the @frodobots Earth Rover!
Action Self-Distillation (ASD) lets a VLM teach itself language-conditioned navigation from robot trajectories alone and no human labels.
It hit 84% success indoors and 95% outdoors across 13 real-world tasks.
Share
Copied
Bowen Feng@BowenFengbf
Can robots learn to reason and act with NO human-annotated language or reasoning labels?
Introducing Action Self-Distillation (ASD), one model that teaches itself language goal-conditioned reasoning and navigation directly from robot demonstrations. #CoRL 2026🤖
Can robots learn to reason and act with NO human-annotated language or reasoning labels?
Introducing Action Self-Distillation (ASD), one model that teaches itself language goal-conditioned reasoning and navigation directly from robot demonstrations. #CoRL 2026🤖
RT @micoolcho: Sooo cool to still see folks pushing navigation research! Equally awesome to see our Earth Rovers being used in this research! Congrats @BowenFengbf@shahdhruv_ & team!
Share
Copied
Dhruv Shah@shahdhruv_
We find that a VLM can teach itself to understand language goals, reason about them, and command robot actions -- we call this Action Self-Distillation!
ASD can perform language navigation by outperforming any VLA trained on hand-annotated data.
We find that a VLM can teach itself to understand language goals, reason about them, and command robot actions -- we call this Action Self-Distillation!
ASD can perform language navigation by outperforming any VLA trained on hand-annotated data.
Share
Copied
Bowen Feng@BowenFengbf
Can robots learn to reason and act with NO human-annotated language or reasoning labels?
Introducing Action Self-Distillation (ASD), one model that teaches itself language goal-conditioned reasoning and navigation directly from robot demonstrations. #CoRL 2026🤖
Another subnet is coming online to the BitRobot Network tomorrow.
Have you set up your Lab Access ID to earn Bolts?
Here's a sneak peak of the gear you will need to position yourself for dual rewards again.
RT @corl_conf: 💎 Thanks to @BitRobotNetwork for being a Diamond Sponsor of #CoRL2026! 🤖
📊 Scaling robot learning starts with better real-world data
🧢 Their RoboCap wearable enables scalable egocentric data collection, with synced head + hand tracking
Share
Copied
BitRobot 🦾@BitRobotNetwork
1/ Better robot models come from better data.
That takes two things: hardware that captures it fully, and tooling that turns it into training signal.
We built the first. Today, @GroundedSI launches the second: Grounded API. Access both hardware + API ↓
@robocurve Anywhere you have an internet connection, you can prompt Dream-1 to control your robot.
Here are rollouts from beta testers from 8 countries, and from our office in SF.
@robocurve Anywhere you have an internet connection, you can prompt Dream-1 to control your robot.
Here are rollouts from beta testers from 8 countries, and from our office in SF.
Congrats to @axisrobotics on their TGE.
SN/04 contributors who completed the Axis Alliance tasks were rewarded twice, in Axis points and BitRobot Bolts.
We're building the world's open robotics lab. Start here and earn:
Thanks @KhalaResearch for this deep dive into robot learning and how industry leaders are tackling the data bottleneck.
Great to see our subnet partner @axisrobotics featured alongside BitRobot’s ecosystem crowdsourcing data across teleoperation, simulation, and ego video!
Robotics is still waiting for its "ChatGPT" moment
Language models had the internet to learn from, while robots need diverse physical demonstrations that are expensive to collect
Bessemer estimated just 300,000 hours of robot manipulation data globally in April, while others believe robotics could require 100M+ training hours
Bessemer projects $3B+ in robot-data spending over two years
Open networks are using crypto incentives to coordinate contributors across simulation, teleoperation and human video collection
Axis reports 7M+ trajectories from 220K+ contributors, with its data lifting task success from 83.9% to 88.8% on the LIBERO-Plus simulation benchmark
Our latest report explores the data opportunity through Axis, BitRobot, Eastworlds & THOR
Full report in the next post below
Robotics is still waiting for its "ChatGPT" moment
Language models had the internet to learn from, while robots need diverse physical demonstrations that are expensive to collect
Bessemer estimated just 300,000 hours of robot manipulation data globally in April, while others believe robotics could require 100M+ training hours
Bessemer projects $3B+ in robot-data spending over two years
Open networks are using crypto incentives to coordinate contributors across simulation, teleoperation and human video collection
Axis reports 7M+ trajectories from 220K+ contributors, with its data lifting task success from 83.9% to 88.8% on the LIBERO-Plus simulation benchmark
Our latest report explores the data opportunity through Axis, BitRobot, Eastworlds & THOR
Full report in the next post below
We're joining @TheHumanoidHub for their inaugural conference in SF today and tomorrow alongside @GroundedSI
Meet the team, get hands-on with our RoboCap + RoboWrist bundle, and see SOTA data enrichment in action.
Register below ↓
Share
Copied
The Humanoid Hub@TheHumanoidHub
Speakers from 1X, Dyna Robotics, Skild AI, Google DeepMind, Physical Intelligence and more.
Humanoid robots, dexterous hands and more cool robotics demos on the floor.
The robotics community in one floor, and four NVIDIA Jetsons to give away.
Agenda:
Speakers from 1X, Dyna Robotics, Skild AI, Google DeepMind, Physical Intelligence and more.
Humanoid robots, dexterous hands and more cool robotics demos on the floor.
The robotics community in one floor, and four NVIDIA Jetsons to give away.
Agenda: