Aaron on Ruby @RubyonAaron

now the depth stream can be added to the controller and switched with plain video

A robot policy that packed a toolbox 13% of the time now does it at 80%. The change came from a handful of human corrections, with the base model frozen. No retraining. FlowDAgger patches the policy by steering it in latent space instead of rewriting its weights. 🧵 👇 (1/5)

I just released a first video deep dive into my LeHome Challenge solution (🥇 sim round, 🥈 real-robot final at ICRA 2026). I already shared what I did and the video series is about why: the reasoning, the alternatives I rejected, and the intuition behind each choice. I split the whole video explanation into 3 parts: Part 1 is about RL for VLAs 👇

They say, Any Sufficiently advanced engineering is indistinguishable from art. This is what that looks like. A 5-axis Hermle machining centre isn't simply shaping metal. It is coordinating five axes of motion simultaneously while its spindle turns at up to 18,500-41,000 RPM, following tool-paths that can contain hundreds of thousands or even millions of lines of CAM-generated code. It cuts materials that would be impossible to drill by hand, titanium, Inconel, hardened tool steels, cobalt-chrome and aerospace-grade stainless alloys,while holding tolerances measured in microns, smaller than a fraction of a human hair. A fully equipped machine like this can cost well over €2.5 million. Because it can repeatedly manufacture aerospace-grade components with micron-level precision in a single 10-20 hour machining cycle, parts that would otherwise require multiple machines, repeated setups, days of manual work, and still struggle to achieve the same accuracy and tolerances needed for high performance parts.

The daily life of warehouse humanoid workers... Geek+ showcased its wheeled humanoid robot GlON sorting,picking bananas using specialized three-fingered gripper --- Humanoid robots are part of their unmanned warehouse automation system and will collaborate with other robots(AMRs,forklift robots…)to fulfill orders.

CyberRobo @CyberRobooo

5 months ago

Will warehouse workers be replaced by Humanoid worker's?🦾🤖👷‍♂️ Beijing Geek+ just officially launched Gino 1:World’s FIRST general-purpose humanoid robot made specifically for warehouses One single robot can do: picking + moving boxes + packing + inspection,basically replaces

9 60 215 16K 53

✨🇨🇳At the 2026 World Artificial Intelligence Conference in Shanghai, humanoid robots are already taking over the repetitive physical work of package sorting. With 3D vision identifying package positions, force sensors dynamically adjusting grip strength, a single-arm load capacity of 18kg, and a sorting speed of 1,200 packages per hour, these robots are now operating in over 10 logistics centers across China. The era of 24/7 automated logistics operations has officially arrived.

China just dropped an open source model that turns any phone into a 3D scanner you point it at a room and walk through it the whole space builds itself into a 3D point cloud in real time as the camera moves, no LiDAR and no depth sensor anywhere it held a full 13 minute walkthrough at 25,000 frames without the map drifting or collapsing > ~20 fps on a single GPU > works indoors and outdoors from plain footage > the scan normally needs a $50,000 rig and a technician > Apache 2.0 and runs fully offline on your machine repo in the reply

savip @savipww

6 days ago

SURVEYORS PAY $40,000 FOR A SCANNER THAT DOES THIS AND HE BUILT ONE WITH A RASPBERRY PI THEN WALKED INTO A CAVE TO TEST IT a spinning lidar an imu and a pi bolted into a 3d printed shell it builds a full 3d point cloud of everything around him as he walks no gps no signal

10 9 140 217K 142

A robot sees a glass and a towel. The difficult part is understanding every physical action between them. This Perceptron Egocentric demo captures a task that demands continuous coordination. ➝ One hand stabilizes the glass ➝ The second hand folds and reshapes the towel ➝ The grip changes as the towel enters the glass ➝ Contact continues while the object rotates ➝ Pressure shifts across the inner and outer surfaces ➝ Each action receives its own temporal boundary ➝ Both hands remain tracked through the full sequence The result is structured supervision built from a normal human task. This type of egocentric data can help manipulation models learn the exact movements behind household actions instead of relying on one broad video caption. Early access at perceptron.inc #EmbodiedAI #RobotLearning #PhysicalAI #Robotics

A mini model of a real dredging vessel, reshaping the seabed just like the real thing

China's robots just started building other robots on a factory floor. 44 seconds of footage, no theatrics this time. The G1 is assembling precision motor components on a real production line, running on a new embodied AI model called UnifoLM-X1-0. This isn't a stunt. It's the same playbook every frontier AI lab uses, just with hardware attached. China's robots are running on their own assembly line specifically to generate the motion data needed to train the next generation, a closed loop where the robot's work becomes its own training set. The timing tells you what this is really for. The company behind this is heading toward a mid-2026 IPO, and until now almost every public G1 appearance has been a gala dance routine or a boxing demo. A robot doing repetitive, unglamorous assembly work is a far stronger data point for investors than another choreographed performance. The CEO has been explicit about the bar he's measuring against, what he calls the 80/80 goal, a robot completing 80% of tasks across 80% of real-world environments. One in-house factory line doesn't clear that bar on its own, but it's the first public sign anyone is building toward it instead of performing for cameras.

ESP32-S3-DevKitC-1 Development Board BT 2.4G Wifi Module for Arduino 8MB PSRAM 16MB FLASH N16R8 42Pin CH340C Type-C ESP32 S3 s.click.aliexpress.com/e/_Ey6SXrY #ESP32S3 ⚡ | #ESP32 📡 | #DevKitC 🛠️ | #IoT 🌐 | #Arduino 💻

✨🇨🇳 Chinese tech firm UniAI unveils the world’s first 720° continuous backflip quadruped robot at WAIC. It shows advanced power and control capabilities for real-world applications.

夫と、どハマりしてるんだけど くだもの凍らして、生クリームいれて アイスにして食べるの ほんとに、めっちゃくちゃ美味しい。 ちょっと贅沢なんだけど、 これが楽しみで夏も頑張れてる！！！ 次はマンゴーもいいなぁ🤤🥭🩷

Someone open-source harness that lets Claude Code and Codex generate real 3D CAD models from plain English. It’s called text-to-cad, an open source harness that turns your coding agent into a CAD engineer.. Building a robot used to mean stitching 6+ tools, CAD software, URDF exporters, kinematics solvers, viewers. Days of work. Broken pipelines. Files no agent can read. text-to-cad kills all of it. → Exports STEP, STL, DXF, GLB, 3MF → Generates URDF robot descriptions → Slices straight to G-code → Sends jobs to your Bambu Lab printer → Runs fully offline, no backend One demo generates a full 7-DoF robot arm, working kinematics, custom GUI, the whole thing, almost entirely through prompts. 100% open source.

#3DPrinted #Robotic Hand Mirrors Human Gestures in Real Time by @YKwolfpec #EmergingTech #Innovation #TechForGood #Tech

If you don't believe AI will replace entire physical job market as well very soon, you are very naive

Needed a break, made a thing. Proof of concept, 2v2 arenas PvE/PvP, armor/modules simulation. Wanna play? Well.. test.. might be broken.. link to discord in comments. #indiegame #gamedev

Another awesome robotics company raised funds in Europe. 🇪🇺 This time a company that is shipping robots that build! @BuildMonumental has raised $32M Series B led by Khosla Ventures, and its fleet of autonomous bricklaying robots has already built more than 100 real structures across Europe. Houses. A school. A community centre. A hotel. Canal walls. Founded by @salar al Khafaji and Sebastiaan Visser, the co-founders of data visualisation firm Silk, acquired by Palantir in 2016, this is a team that knows how to build and sell a company. The technology is serious: → Fleet of 100+ electric autonomous robots operating on live job sites today → Advanced sensors, computer vision and small cranes laying brick to millimetre precision → All driven by Atrium, their proprietary AI software platform → Nearly half of all homes built in the last three months alone, pace is compounding fast The business model is also worth taking a look at. Contractors don't buy the robots. They hire Monumental as an autonomous subcontractor and pay for finished wall. So in the end you only pay for the output. This is the same forward-deployed engineering model Palantir pioneered in software. Monumental brought it to physical robotics years before anyone else in the industry caught on. The backdrop makes this urgent. US construction is short 200,000-400,000 workers every month. Since 1945, US manufacturing productivity rose eightfold. Construction gained just 10%, and has actually declined since the 1960s. LET'S MAKE CONSTRUCTION BUSINESSES SEXY AGAIN! 🧱 ~~ ♻️ Join the weekly robotics newsletter, and never miss any news → ziegler.substack.com

robotics is the future. not content. not another SaaS wrapper around an LLM. robots. • machines that sense, think, and act in the physical world. • machines that move energy, matter, goods, people. • machines that collapse labor, expand productivity, and rebuild entire industries. the future isn’t digital-first anymore. it’s physical-first with intelligence. manufacturing → robotics logistics → robotics agriculture → robotics construction → robotics defense → robotics healthcare → robotics homes → robotics every sector that still depends on human hands becomes a robotics problem. and the demand is bigger than the talent pool by orders of magnitude. learn perception. learn control. learn actuation. learn embedded. learn kinematics. learn systems engineering. learn how to make machines move. the next trillion-dollar companies will be built out of metal, motors, sensors, and code. don’t stay stuck in 2D screens. the future is 3D, autonomous, and mechanical.

1/ Collect once. Retarget to any bimanual arm with parallel grippers. I firmly believe bimanual robotic arms are the right embodiment to start adding value in the real world, and HandUMI is here to help startups accelerate deployment and researchers run more experiments. Record data with HandUMI, no robot in the loop, replayed straight onto Agilex PiPER, OpenArm, TRLC-DK1, and I2RT YAM. - Robot-free capture. - Modular: swap the gripper on the arm you use and start collecting data - LeRobot v3-compatible dataset format - Built-in calibration + QA before conversion - Sim replay and real teleoperation - Apache-2.0, fully open source ft. @alvax64 @leoperzz @raulb4s @mbrq_13 @robonet_ @0xnonhuman 🧵👇