World’s largest AI × robotics channel. Direct access to 200+ robotics companies. Original, tech-grounded reporting. 500K+ followers. 📬 [email protected]
Joined July 2012
- Tweets2.5K
- Following1.1K
- Followers22.6K
- Likes4.4K
Pinned Tweet
So glad we got to visit so many amazing robotics companies together.
Hope you come back every year. This industry moves fast and never disappoints.
There are so many great robotics companies here, and everyone is welcome to come see them.
That's a wrap on an amazing 17-day journey across China.
I didn't do this alone. My friend @dolylupec came along for the whole ride. She's got a background in AI and robotics investing, and I'm grateful she chronicled the tours on her account and pitched in on my channel.
@XRoboHub is one of the most resourceful people I know. He understands the humanoid space well and helped us schedule most of the important visits, then came along too.
And a shout-out to @Robo_Tuo, who encouraged me to do this trip and supported us all along.
It was my first time in China, and the trip was packed so tight with tours, we didn’t get to do much touristy stuff. Can't complain. It was a hell of a ride.
4 cities on the mainland. 18 humanoid makers. Warm, welcoming people everywhere. We had the privilege of chatting with founders, senior engineers, researchers, and marketing and strategy leads, a lot of whom said yes on very short notice. Grateful for that.
The energy is extraordinary. It's the trifecta: engineering talent, a deep supply chain, and breakneck adoption of new tech. There's intense competition, but underneath it, a real spirit of sharing and collaboration.
I've got a series of tour videos lined up, dropping over the coming days.
And something tells me this won't be my last trip. Robotics and embodied AI are just getting started. The next ten years are going to reshape how we think about physical work.
We ended on the Great Wall, a 2,000 year old marvel of engineering, after two weeks staring at the next one.
The robots are coming. I just went to meet them first.
This is wild... 🤯 Unitree just unveiled H2-X, and it’s an absolute beast.
Spinning kicks, aerial flips, powerful knee and elbow strikes. It jumps from a 4-meter platform onto mats, rolls with the impact, and gets back up. That whole sequence is a serious test of joint strength and whole-body control.
The launch video even casts it as a personal bodyguard. It faces off with an intruder and, in a staged rescue, lifts the front of a car while someone pulls a dummy out from underneath.
Unitree says the footage is real and not sped up. It also reports roughly eight hours of battery life while standing.
Honestly, Figure should bring F.03 to the next World Humanoid Robot Games.
Try the 100m and the manipulation events.
Beijing’s robot half-marathon would make a good outdoor endurance test, too. Haha.
RoboHub🤖 retweeted
Meet X2Real—a sim-to-real benchmark for generalist robot policies, built on @NVIDIARobotics Isaac Lab-Arena - an open-source framework for scalable policy evaluation in simulation.
X2Real evaluates 10 capability dimensions across 44 hierarchical, long-horizon tasks—from visual and language reasoning to fine manipulation, bimanual collaboration and tool use.
Policies are tested across 3 OOD axes: environment appearance, task settings and robot embodiment.
On policies trained only with real-world data, X2Real achieves a 0.84 linear correlation between simulation and real-world evaluation scores.
Powered by our Physical DSL, the platform creates an agent-driven loop from task composition and environment compilation to policy rollout and trajectory evaluation.
Explore X2Real:
developer.x2robot.com/develo…
#PhysicalAI #Robotics #EmbodiedAI #Sim2Real
10,000 sewing robots a year. That’s the planned capacity of Aitu’s newly opened factory.
Its latest footage shows them at work at Bosideng, the Chinese down-jacket maker.
Automated sewing machines can follow a stitch pattern, but loading them often still takes skilled hands. Soft fabric can fold, slip or bring another layer along when you try to pick just one.
Aitu says its robots use a vision-language-action model built for sewing, machine vision and adaptive grippers to separate and line up the fabric panels.
The dual-arm robots work with automated fixtures and sewing machines at the same workstation.
The sewing equipment comes from Aitu’s parent company, Jack Technology. It sells more industrial sewing machines than any other manufacturer.
Japan’s JR West is adding two new railway maintenance robots, including a giant semi-humanoid with a large gripper and a chainsaw.
The dual-arm ZX-3 is mounted on a maintenance vehicle’s boom and can cut branches up to 15 meters (49 ft) above the ground.
An operator on the ground controls it while watching the feed from cameras in its head. One arm holds the branch while the other cuts it.
The ZX-3 is smaller and lighter than the ZX-1, which JR West has been using since July 2024. It has a longer boom and is scheduled to start work in November.
The single-arm ZX-2 is scheduled to start work this month. It can hold loads of up to 265 lb (120 kg) to help crews replace heavy overhead-line components.
Both belong to the ZIZAI railway robot series, developed by JR West with Nippon Signal, robotics firm Jinki-Ittai and rail vehicle maker Toyo Sharyo.
JR West aims to reduce the number of workers needed for these maintenance jobs by about 30%.
RoboHub🤖 retweeted
RoboParty just open-sourced RP1, a high-performance humanoid platform for embodied AI research.
RP1 uses RoboParty’s own ROMOMO actuators. The release covers mechanical and electrical designs, low-level control, simulation, training and deployment.
MimicLite includes pretrained whole-body tracking policies for RP1, along with matching motion data and training configs. RoboParty Lab also provides hhtools for motion retargeting and UFO for unsupervised reinforcement learning.
That’s the draw for embodied AI teams: a trained motion controller to build on, and the freedom to develop the body and its policies around their own experiments.
🚀RP1 is officially here.
🌍Putting control of robotics back in developers’ hands.
Introducing RP1 (ROBOTO 01) by RoboParty—a high-performance, fully open-source humanoid robotics platform for frontier robotics research and embodied AI model development.
🤖 Open across the full hardware and software stack:
• Robot platform: mechanical CAD, bill of materials, and assembly documentation
• Actuators: in-house ROMOMO joint designs and related technical resources
• Control: electrical systems, firmware, low-level drivers, and motion control implementations
• Development: simulation models, SDKs, data processing tools, and training tools
• Engineering foundations: technical documentation and research materials
Open the black box.
Understand the system.
Make it your own.
A deviation in motion can originate in the model, control policy, actuator response, or mechanical structure.
RP1 opens up the design and implementation of these layers so developers can trace problems, modify the system, and validate changes on real hardware.
Open-source algorithms for advanced motion control
RP1 has demonstrated continuous walking, recovery from disturbances, getting back up after a fall, and human motion tracking.
UFO and MimicLite connect motion learning, control policies, and execution on real hardware.
The open toolchain connects data collection, simulation training, hardware deployment, and experimental validation—supporting joint iteration across models, control, and hardware.
From RPO to RP1, we’re turning one team’s engineering experience into a starting point for many more.
Opening the future of embodied AI to everyone ready to explore it.
🔗RP1 on GitHub:
github.com/Roboparty/RP1
#RoboParty #RP1 #OpenSourceRobotics #EmbodiedAI
RoboParty just open-sourced RP1, a high-performance humanoid platform for embodied AI research.
RP1 uses RoboParty’s own ROMOMO actuators. The release covers mechanical and electrical designs, low-level control, simulation, training and deployment.
MimicLite includes pretrained whole-body tracking policies for RP1, along with matching motion data and training configs. RoboParty Lab also provides hhtools for motion retargeting and UFO for unsupervised reinforcement learning.
That’s the draw for embodied AI teams: a trained motion controller to build on, and the freedom to develop the body and its policies around their own experiments.
🚀RP1 is officially here.
🌍Putting control of robotics back in developers’ hands.
Introducing RP1 (ROBOTO 01) by RoboParty—a high-performance, fully open-source humanoid robotics platform for frontier robotics research and embodied AI model development.
🤖 Open across the full hardware and software stack:
• Robot platform: mechanical CAD, bill of materials, and assembly documentation
• Actuators: in-house ROMOMO joint designs and related technical resources
• Control: electrical systems, firmware, low-level drivers, and motion control implementations
• Development: simulation models, SDKs, data processing tools, and training tools
• Engineering foundations: technical documentation and research materials
Open the black box.
Understand the system.
Make it your own.
A deviation in motion can originate in the model, control policy, actuator response, or mechanical structure.
RP1 opens up the design and implementation of these layers so developers can trace problems, modify the system, and validate changes on real hardware.
Open-source algorithms for advanced motion control
RP1 has demonstrated continuous walking, recovery from disturbances, getting back up after a fall, and human motion tracking.
UFO and MimicLite connect motion learning, control policies, and execution on real hardware.
The open toolchain connects data collection, simulation training, hardware deployment, and experimental validation—supporting joint iteration across models, control, and hardware.
From RPO to RP1, we’re turning one team’s engineering experience into a starting point for many more.
Opening the future of embodied AI to everyone ready to explore it.
🔗RP1 on GitHub:
github.com/Roboparty/RP1
#RoboParty #RP1 #OpenSourceRobotics #EmbodiedAI
Tesla’s flying Roadster is wild, but I’ve got to give it to the Gen Z maker in China who built his own flying skateboard.
Ayu’s inspiration was the New Goblin’s board in Spider-Man 3. His first prototype used five jet engines and caught fire during ground testing, so he switched to 10 electric rotors, which he says produce over 660 lb of thrust combined.
Small shifts in his weight could tilt the board. He practiced in a safety harness and used a phone-based assistant that tracked the board’s tilt and gave him spoken cues to adjust his stance.
He says the project took about two months, with more than 10 days spent learning to fly it. Now he’s riding it over rivers and through a canyon near his hometown in Hubei.
@elonmusk, got room for this guy on your team?
Source and full video:
【我造了一块会飞的滑板...】 bilibili.com/video/BV1KiJk67…
I remember Elon saying he wasn’t sold on flying cars. Now Tesla has unveiled a Roadster that can fly.
The Rocket version uses cold gas thrusters adapted from SpaceX technology. They release compressed gas from tanks wrapped in carbon fiber to lift the car and control its movement in the air.
Roadster reservations are already open in the U.S. and China. For Performance, the deposit is $50,000 in the U.S. or RMB 332,000 in China. Thrust adds rear thrusters for acceleration, with a $250,000 U.S. deposit.
Rocket is available by invitation only and isn’t approved for public roads. Tesla hasn’t announced final prices, and Musk is aiming to start Roadster production in 2028.
RoboHub🤖 retweeted
Something new is coming.
October 29. Shanghai World Expo Center.
At AGIBOT Innovation Day 2026, we’re unveiling what’s next in embodied AI.
New robots. New models. And a bold new vision for how robots create value.
The next chapter of embodied AI is about to begin.
Stay tuned.
#AGIBOT #AGIBOTInnovationDay #EmbodiedAI
Leju has 15 KUAVO humanoids loading auto parts at Zhaofeng’s factory in Hangzhou, China.
These wheeled robots handle wheel-hub bearing components across 15 stations. Zhaofeng reports a 98% loading success rate across more than 10 part types and says the robots have helped produce over 60,000 components.
There’s still a speed gap: reported loading cycles take under 40 seconds, versus about 15 for a worker at the same station.
Leju says the same loading skills can be reused across robots, and data from the line goes back into training. That could make adding the next robot less of a custom engineering job.
This isn’t Figure’s June promo: one robot working, three posing for the camera. 😂
Every 30 minutes, a brand-new humanoid robot rolls off the line here. 🤖🇨🇳
China’s first fully automated assembly line for humanoids just went operational in Foshan, scaling production to 10,000 units annually.
The line features 24 digitalized, precision assembly processes, boosting manufacturing efficiency by over 50% compared with traditional manufacturing methods.
A total of 77 procedures are in place to test the safety performance of the components and finished units, ensuring industrial-grade stability.
Based on the footage, the robot currently in production is the all-new Leju KUAVO 5 humanoid.
The line also features high flexibility, capable of mixed-model and multi-unit assembly to adapt to different robot designs on the fly.
We are officially moving past the lab stage and into a new era where mass production drives down the cost of embodied AI for global markets.
Video By：CCTV+
Boston Dynamics says it takes about a day to build an Atlas. Its target for 2027: about an hour.
Inside its prototype shop, precision machining, assembly, testing and repair all happen under one roof. Some newly machined parts go into robots the same day or the next.
That shortens the wait between finding a hardware problem and testing a fix. The team can improve Atlas and the way it’s built at the same time.
A strong speaker lineup, live robot demos, and lots of robotics folks coming together... this should be fun.
RoboHub🤖 retweeted
What's stopping teams from getting value out of the data they already have?
Aurora Feng (Founder, Neural Motion) and Joe Harris (Founder & CEO, Alloy Robotics).
Tomorrow
@aurorafeng_01 @_joe_harris_
I’ll be speaking at the SF Humanoid Hub Conference tomorrow 10/15! Thanks @TheHumanoidHub for inviting me😆
Mastering gongfu tea isn’t easy for humans or robots.
Natura-1 makes the whole brewing routine look easy. That takes both a capable model and dexterous hands.
The final pour is a little off, but it’s still pretty impressive, haha.
This robot is making wontons. One hand cups the wrapper while the other adds the filling, then both hands fold it closed.
It’s running Natura-1, @naturewill’s 50B-parameter mixture-of-experts model. The team says it was pretrained on roughly a million hours of first-person footage, then adapted with robot data.
The team reconstructs 3D hand movements from that footage and pairs them with task descriptions. During training, Natura-1 learns to predict future video and actions together, using past video as context.
On the robot, it takes in language and vision and controls 20-DoF hands, with a reported 100 ms observation-to-action loop. The hard part here is supporting the thin wrapper while leaving its edges free to fold, then adjusting the grip as its shape changes.
The same foundation model also powers demos of sealing shipping boxes, applying phone screen protectors and writing with a brush. These longer workflows, including making wontons, use additional task-specific training.
Nature Will separately reports zero-shot tests with instructions and setups absent from its robot post-training data. In the longer interaction video, the presenter moves plates midway through fruit sorting, and the robot adjusts where it puts the fruit.
That video also leaves in missed grasps, retries and spoken corrections. You can see where it adapts and where the presenter prompts it to try again.
PI’s frontier models are powering everything from harvesting and household robots to assembly systems and wheel loaders.
Formic’s demo looks really good. The robot is handling the full box-packing cycle with PI’s models. Looks like they’ve put an Astribot S1 to work on the packing line.
The Pi API beta gives more teams a way to turn PI’s latest models into working robot applications. They can adapt the models with demonstrations from their own robots, with PI handling training compute and cloud inference.
