Vishal

Construction_and_Engineering

Robbyant just released LingBot-VA, a robot control system that predicts what the world will look like after taking those actions, then checks if reality matches its prediction.

The approach works in three steps: predict future video frames based on current observations and language instructions, decode actions from the predicted video, then replace predictions with real observations after execution.

This creates a closed-loop where the robot continuously grounds itself in reality.

LingBot-VA excels across diverse tasks: long-horizon manipulation (make breakfast, unpack delivery), precision control (insert tubes, pick screws), and deformable objects (fold clothes, open drawer). Results on RoboTwin 2.0 and LIBERO benchmarks show consistent improvements over state-of-the-art methods.

What makes this powerful is long-term memory. Traditional models only see the current observation and get confused by repeated states. 

For example, if a robot needs to open the right box, close it, then open the left box, a memoryless model sees the closed right box twice and gets stuck in loops. LingBot-VA remembers the full sequence and completes the task correctly.

![](https://m.stacker.news/128430)

Video world models for robot control! 🎬

SpaceX is requesting authorization to launch and operate a constellation of 1 million satellites with unprecedented computing capacity (orbital data centers) to power advanced AI systems, according to a new document filed with the FCC. 🚀🚀🚀

SpaceX: “Launching one million satellites that function as orbital data centers is the first step toward becoming a Kardashev II-level civilization—capable of harnessing all the energy of the Sun—while supporting AI applications used by billions of people today and ensuring humanity’s multiplanetary future among the stars.”

In the SpaceX document:
• SpaceX intends to deploy a system with up to one million satellites, operating in narrow orbital shells of up to 50 km each (leaving enough space to avoid conflicts with other systems with similar ambitions).

•The system will operate between 500 km and 2,000 km altitude, with orbital inclinations of 30 degrees and a sun-synchronous orbit.
•SpaceX plans to design and operate different versions of the satellite hardware, optimized for each orbital shell.

•The system will rely almost exclusively on high-bandwidth optical links for communication. These optical links will route it within the network and with satellites in the Starlink constellation, using its high-capacity (petabit) and high-reliability laser mesh, which in turn will transmit the traffic to authorized ground stations.


SpaceX added:
“With Starship’s ability to deliver unprecedented tonnage into orbit for AI computing, intelligence processing in space could surpass the electricity consumption of the entire U.S. economy — without the immense cost and enormous disruption of rebuilding the already overburdened terrestrial power grid to meet the explosion in demand for data centers. Thus, satellites functioning as solar-powered orbital data centers are the most economical, efficient, and environmentally responsible way to build the infrastructure needed to meet the growing demand for AI-enabled goods and services.”

“With the inherent efficiency of solar data centers in orbit and the rapidly falling launch costs thanks to the development of Starship, SpaceX will be able to scale its constellation in an economically viable way as demand increases and computing capacity evolves. For example, launching 1 million tons per year of satellites generating 100 kW of computing power per ton would add 100 gigawatts of AI capacity annually, with minimal operation or maintenance needs.”

![](https://m.stacker.news/128297)

SpaceX proposes orbital data centers to scale AI beyond Earth.

Stacker_Stocks

Another European robotics company steals the show.

RobCo, a Munich-based robotics company, just raised $100 million in Series C funding co-led by Lightspeed and Lingotto Innovation (an Exor investment company), with participation from Sequoia Capital, Greenfield Partners, Kindred Capital VC, Leitmotif, and The Friedkin Group.

The funding will advance RobCo's physical AI roadmap, expand enterprise deployments, and deepen the company's presence in the U.S. market. RobCo expanded into the United States in 2025 and now operates in San Francisco and Austin. 🇺🇸

Founded in 2020, RobCo enables increasingly autonomous robot operations inside real production environments through Physical AI. By combining perception, motion planning, and self-learning methods, the platform reduces friction between today's processes and end-to-end automation. Teams can focus less on setting up and maintaining systems and more on their core business.

What makes RobCo different is the full-stack approach. The company has been vertically integrated from day one, developing hardware and software as a single platform. 

Their robots can acquire task-specific skills through demonstration and self-learning rather than manual programming. This enables faster deployment, rapid iteration, and easier adaptation to complex or variable processes.

I can confirm that because I had a chance to visit their HQ couple of times, and it was always an awesome experience in the heart of Munich! 

RobCo's robots are deployed across a range of industrial environments. The U.S. has become both a strategic priority and a major growth market.

![](https://m.stacker.news/128001)

Read more here: https://finance.yahoo.com/news/robco-raises-100-million-scale-130000452.html

RobCo raises $100M Series C to build scalable robotics!

Brain-computer interfaces (BCI) & Robotics lack large-scale datasets. 

Fourier Robots is developing BCI-driven robotic rehabilitation training. 

Imagine a stroke patient thinks of moving a hand, and the robot detects the intent to guide the motion.

![](https://m.stacker.news/127992)

BCI + Robotics: Training movements directly from the brain.

Micro-dispensing at the picoliter scale is incredible when you compare it to everyday numbers: a single raindrop is about 50 microliters, roughly 50 million times larger than a picoliter.  

Being able to place droplets this small without touching the surface is key for biotech, diagnostics, and micro-electronics, where tiny volumes matter.  

And at this scale, two basics become critical: 
→ Accuracy = how close you are to the real value. 
→ Precision = how repeatable each droplet is.  

I could watch it all day long!

![](https://m.stacker.news/127890)

The absurd engineering of picoliters.

A robot that looks like a hamster ball and can fly over obstacles on its own.

Revolute Robotics created a machine called the Hybrid Mobility Robot. I had a chance to meet their team in Boston earlier this year.

Two special rings allow this robot to roll in any direction and whirl like a helicopter.

A lightweight exoskeleton absorbs shock during landing and prevents vibrations from damaging electronics. Typically, the HMR rolls on the ground, but can also fly to overcome obstacles or climb steep inclines.

This robot can go places that humans might not be able to reach, including confined spaces, performing security tasks, and even performing search and rescue missions.

![](https://m.stacker.news/127751)

Hamster-powered robot!

This is a fully autonomous tractor from Dutch company AgXeed, designed to work on fields without any human supervision. 

It's powered by a diesel engine that drives a generator, which then powers electric motors driving the tracks. The fuel tank allows for 24 hours of continuous operation. (nice!)

The rear hitch has an 8-ton lift capacity, the front has 3 tons. It's equipped with electric PTO, hydraulic pump, and all the connections you'd find on a traditional tractor, meaning you can attach standard farm implements.

Safety and navigation are handled by a LiDAR sensor on top, ultrasonic and radar sensors, and contact-sensitive bumpers. RTK GNSS receiver provides positioning accuracy down to 2.5 cm, with a module for bidirectional data transmission and RTK correction.

Love seeing that so many companies are trying to make farmer's life easier! 🚜

![](https://m.stacker.news/127724)

Autonomous tractor from Netherlands!

Unbox Robotics just closed a $28 million Series B led by ICICI Venture and Redstart Labs (Infoedge), with participation from F-Prime, 3one4 Capital, Navam Capital, Force Ventures LLP, and other existing investors.

The Pune-based company builds modular robotic systems for warehouse and logistics operations. Their platform combines proprietary swarm-intelligence software with modular 3D robotic sortation hardware, allowing large fleets of robots to coordinate dynamically and scale throughput with minimal fixed infrastructure.

The funding will be used to strengthen leadership and engineering teams, accelerate new product development, and expand market presence across India and select international markets. 

The company also created meaningful liquidity for employees through its ESOP program as part of the transaction.

Unbox Robotics has built a growing customer base across Europe, the United States, and India, supporting global e-commerce, retail, and third-party logistics operators. 

The supply chain robotics market is heating up. With labor shortages, rising fulfillment costs, and demand for faster delivery, companies like Unbox that can deploy modular, software-driven robotic systems are in the right place at the right time.

P.S. India producing robotics companies that compete globally in warehouse automation is a sign of how the robotics ecosystem is maturing beyond just the U.S., Europe, and China. 🇮🇳

![](https://m.stacker.news/127610)

Indian robotics company Unbox Robotics raises $28M Series B!

The owner of the biggest YouTube channel visited the Starship factory in Starbase, Texas.

The YouTuber was guided by Bill Riley, vice president of engineering for the project, and was able to install a piece of the heat shield on one of the flaps that will go into space on a future Starship launch.

![](https://m.stacker.news/127496)

Owner of the biggest YouTube channel visit Starbase and installs a Starship part

ProtoMotions just released v3.1, a HUGE upgrade focused on making it easier to train robot motion in simulation and deploy it on real hardware.

It already has +1k ⭐️ - let's see in couple of weeks after this upgrade is released.

The new modular setup is a big improvement. Instead of one giant block of code, they split everything into separate pieces for Control, Observations, Rewards, and Terminations. 

This means you can swap parts in and out without rewriting everything.

Training got way faster, meaning that you can now train a motion tracker to learn the entire AMASS motion dataset perfectly in under 24 hours. Before this, it took much longer.

Sim-to-real transfer is now built in. They added realistic sensor noise, random pushes to test balance, and automatic friction settings that work across IsaacGym, IsaacLab, and Newton. This means what works in simulation actually works on real robots without spending weeks tweaking parameters.

Last but not least, the deployment pipeline is clean. ONNX export now includes the full control loop: sensor data goes in, motor commands come out. No manual conversion needed.

Training is faster, the code is cleaner, and deployment is straightforward.

If you're working on humanoid walking, robot manipulation, or any character animation for robots, this is worth checking out.

![](https://m.stacker.news/127495)

Sim-to-real motion tracking repo!

Verobotics has built autonomous robots that clean and scan building exteriors, allowing building owners to proactively maintain and upkeep their buildings faster, cheaper, and smarter.

The problem is decades old. The only way to clean or inspect a building's exterior has been to lower a person from a BMU (Building Maintenance Unit) at the top of the building. 

BMUs are heavy, can be out of service, complicated to operate, and limit cleaning to a couple of times per year.

Verobotics breaks this paradigm completely. Their robots are lightweight, portable, and need no equipment on the roof. 

One person can deploy multiple robots to the building's facades and let the robots do the rest. 

Once they're done, the building owner not only has a clean building but also a digital twin of the building exterior for documenting health and identifying issues before they become risks.

Robot spots a crack? Noted! 📄

They're equipped with onboard cameras that scan the surface, detect window frames to crawl over, and ensure not a spot is missed. That means a consistent result and thorough inspection every time.

Onboard sensors continuously detect motion, location, and the surface beneath and around it. 

The data feeds to a hub that analyzes the surface in real time, building datasets that become the backbone of a digital twin for predictive maintenance.

Love seeing the specialized robots bringing value from day 0! :)

![](https://m.stacker.news/127463)

Robots that climb skyscrapers to clean and inspect them!

I can't even imagine the day when robots start assembling other robots on their own. That would be considered Supervised Artificial Intelligence.

![](https://m.stacker.news/127405)

When robots start creating robots, the AI ​​timeline changes forever.

KOU-III, designed by researchers at Shandong University in China, blends drone flight with two-legged walking capability.

![](https://m.stacker.news/127403)

KOU-III, designed by researchers at Shandong University in China

Stanford researchers just released UMI-FT, a handheld data collection platform that puts compact six-axis force/torque sensors on each finger, enabling finger-level wrench measurements alongside RGB, depth, and pose data.

Many manipulation tasks require careful force modulation: too little force and the task fails, too much and you cause damage. But commercial force/torque sensors are expensive, bulky, and fragile, which has limited large-scale force-aware policy learning.

UMI-FT changes the economics. The platform uses an iPhone for RGB vision, ultrawide RGB, depth, and pose via ARKit, with each finger sensorized using a CoinFT sensor to capture per-finger wrench information during manipulation. 

This multimodal data trains an adaptive compliance policy that predicts position targets, grasp force, and stiffness for execution on standard compliance controllers.

The learned policy runs slowest and generates reference targets, while model-based compliance and force controllers provide delicate 6D compliance control and real-time force modulation.

They tested on three contact-rich, force-sensitive tasks: whiteboard wiping (locate eraser, grasp, wipe until clean), skewering zucchini (grasp slice firmly, push onto stick until punctured), and lightbulb insertion (grasp bulb, align bayonet pin with socket slit, insert while overcoming spring force, rotate to light up).

The results are clear. Policies without compliance struggle to modulate contact force and trigger safety faults from excessive force. Policies without force sensing fail to grasp unseen objects or resist reaction forces, causing slippage. 

Here's the project page: https://umi-ft.github.io 


![](https://m.stacker.news/127342)

Force-sensing fingers!

Zipline has reached 2 million commercial drone deliveries and raised over $600 million, with a valuation of $7.6 billion. It has been growing 15% per week in US deliveries and has already delivered over 20 million items with an average flight time of 3 minutes.

![](https://m.stacker.news/127290)

Zipline scales drone deliveries and reaches a valuation of $7.6 billion.

This work answers that with a rare yes.

The core problem has always been contact.
Soft contact models give smooth gradients but break realism.

Hard contact models are realistic but kill optimization due to discontinuities.

This paper proposes an analytically smooth contact formulation that keeps both.

• A quadruped locomotion policy trained entirely in a differentiable simulator
• Zero-shot transfer to real hardware
• More than 10× better sample efficiency compared to PPO

Differentiable simulation has promised better robot learning for years, but locomotion was the hard wall. 

This is, to my knowledge, the first real sim-to-real legged locomotion result trained purely with analytic gradients.

If this scales, it could change how we train robots that rely on contact, not just legs, but manipulation too.

Learning Deployable Locomotion Control via Differentiable Simulation:

![](https://m.stacker.news/127233)

Can differentiable simulation actually work for real robot locomotion?

Microsoft just announced Rho-alpha (ρα), their first robotics model derived from the Phi series of vision-language models.

Rho-alpha translates natural language commands into control signals for robotic systems performing bimanual manipulation tasks. Commands like "push the green button with the right gripper," "pull out the red wire," "flip the top switch on," or "turn the knob to position 5" get executed directly by dual-arm robots.

What makes this different from standard vision-language-action (VLA) models is the additional modalities. Rho-alpha is a VLA+ model that adds tactile sensing to the perceptual mix, with plans to incorporate force feedback. 

On the learning side, the model is designed to continually improve during deployment by learning from human feedback.

The training approach combines trajectories from physical demonstrations and simulated tasks with web-scale visual question answering data. 

Since teleoperation data is scarce and expensive, Microsoft is using NVIDIA Isaac Sim on Azure to generate physically accurate synthetic datasets via reinforcement learning. These simulated trajectories get combined with commercial and open physical demonstration datasets.

The model is currently under evaluation on dual-arm setups and humanoid robots. Microsoft is opening an Early Access Program for organizations interested in evaluating Rho-alpha.

Robots that can adapt to dynamic situations and human preferences are more useful in real environments and more trusted by the people operating them.

![](https://m.stacker.news/127054)

Microsoft's first robotics foundation model!

According to Dario Amodei, CEO of Anthropic, we are less than a year away from models that do everything engineers do from start to finish. In other words, AI itself generates better AI.

The only obstacle is access to hardware and time to train the chips.

![](https://m.stacker.news/126896)

AI poised to self-improve, says Anthropic CEO.

This is what optimized robotic press brake operation looks like.

 A well-tuned program, an additional seventh axis for extended reach, smart gripper design, and a workpiece reversal mechanism to handle parts without human intervention.

This combination is invaluable for automated press brake operation. Cycle time is on fire! 🔥

KUKA robots are a solid choice for metal forming. 

They're built for high-payload, high-precision tasks and have a strong track record in automotive and heavy industry applications where reliability matters the most! :)

![](https://m.stacker.news/126893)

A robot that will bend metal for you!

Tesla Cybercab production starts in less than 100 days from today, which will use Tesla's revolutionary unboxed manufacturing process for the first time.

One Cybercab will be able to roll off the production line in under 10 seconds versus 34 seconds for Model Y, with a long term cycle time goal of ~5 seconds.

![](https://m.stacker.news/126818)

Tesla's Cybercab promises an industrial revolution: one car every 10 seconds.

Surely one of the coolest and craziest robot project I've seen this year.

Buoyant Choreographies by RoMeLa is a fun art project shown at ICRA in Atlanta last year.

It uses special robots made from helium balloons and moving legs that float and walk around. People can play with the robots using touch, game controllers, or even by blowing air.

The robots react by changing colors, moving differently, and showing emotions.

![](https://m.stacker.news/126702)