The push to standardize the physical AI software stack takes a major leap forward today. Building on the Hugging Face and NVIDIA integrations we've been tracking, Ant Group has open-sourced a 'universal brain' capable of adapting to completely different robot shapes, accelerating the industry-wide move toward hardware-agnostic control.
Fleshing out the LeRobot library integration we previewed yesterday, NVIDIA and Hugging Face have formally announced their partnership to embed the Isaac GR00T 1.7 foundation model and Isaac Teleop tools into the open-source platform. The move targets a combined ecosystem of 19 million developers (NVIDIA's 3 million roboticists and Hugging Face's 16 million AI builders) with a standardized pipeline for data collection, model training, and physical deployment.
Why it matters
This partnership represents a major step toward democratizing and standardizing the development of embodied AI. By creating an open, end-to-end framework, it significantly lowers the barrier to entry for startups and researchers, much like how open-source LLMs transformed digital AI. For entrepreneurs in the space, this provides a powerful, off-the-shelf toolkit to build upon, potentially reducing the immense cost and complexity of developing a full software stack from scratch and allowing teams to focus on novel hardware and applications. The goal is to foster a collaborative ecosystem that can accelerate progress toward general-purpose robot autonomy.
"The collaboration aims to provide physical AI developers worldwide with a common platform and open workflows to collect robot data, train models, and deploy them on robots," stated the companies in a joint release. This move is seen by industry analysts as a direct effort to build an open standard to compete with the proprietary, vertically integrated systems being developed by companies like Tesla. Some critics note that while this lowers the software barrier, hardware development and data acquisition in diverse real-world environments remain significant and costly challenges for smaller teams.
Robbyant, the embodied AI company within Ant Group, has upgraded and open-sourced LingBot-VLA 2.0, a Vision-Language-Action (VLA) model designed to function as a 'universal brain' for diverse robot types. Announced on Wednesday, the model was trained on 60,000 hours of real-world data from 20 different robot morphologies. Robbyant claims it shows significant improvements in generalizing across hardware, supporting more degrees of freedom, and deploying efficiently without extensive hardware-specific retraining.
Why it matters
The release of a single, adaptable model that works across various robot platforms could be a game-changer for the industry, drastically reducing development costs and time-to-market. It directly tackles the problem of software fragmentation, where every new robot requires a bespoke control stack. For an entrepreneur, this open-source 'brain' could allow for rapid prototyping and deployment on custom hardware, enabling a focus on mechanical innovation and application-specific tuning rather than reinventing the core AI. This could accelerate the commercial viability of robotics in logistics, retail, and industrial automation.
Robbyant stated the goal is to "address a key bottleneck in the industrial-scale deployment of robotics." One perspective from The Insight suggests this move is part of a broader industry shift away from hype-driven demos toward solving foundational problems, with a focus on pragmatic, factory-first automation. In parallel, Robbyant also open-sourced LingBot-Vision, a foundation model specifically for improving spatial perception, particularly with challenging materials like glass.
Hugging Face has released LeRobot v0.6.0, a significant update to its open-source robotics toolkit for embodied AI. The release on Tuesday introduces new capabilities including world-model policies, Vision-Language-Action (VLA) models, and support for reward models. It also adds new simulation benchmarks, tooling for human-in-the-loop corrections, and support for both Fully Sharded Data Parallel (FSDP) training and cloud-based training via Hugging Face Jobs.
Why it matters
This update makes the open-source robotics stack more mature and professional, moving it from a collection of experimental tools to a more robust and reproducible infrastructure. For developers and startups, it provides a standardized experimental loop for developing, evaluating, and retraining robot policies. The inclusion of simulation benchmarks and cloud training support makes it easier to scale research and development, helping to bridge the gap between isolated demos and repeatable, applied machine learning systems for robotics.
The release notes emphasize a focus on "more robust and reproducible infrastructure for practitioners to develop, evaluate and retrain embodied policies." This complements the simultaneous announcement of a deeper partnership with NVIDIA to integrate Isaac GR00T, providing a comprehensive open-source ecosystem for building and deploying physical AI.
China has launched the M-Robots community, the country's first open-source community for a full-stack commercial robot operating system. Announced Wednesday, the initiative is based on the OpenHarmony operating system and aims to create a shared technology platform for the Chinese robotics industry. Institutes and companies from the city of Chongqing are among the founding members.
Why it matters
This initiative signals a strategic push by China to develop a domestic, standardized robot operating system, reducing reliance on foreign technologies like ROS. By fostering a collaborative open-source ecosystem, the goal is to accelerate the transition from standalone robots to interconnected, multi-robot systems powered by AI. This could help standardize development and promote interoperability within China's rapidly scaling robotics sector.
An official statement indicated the community aims to "accelerate China's transition from standalone robots to multi-robot collaboration and AI-powered systems." The move mirrors similar efforts globally to create standardized platforms but with a clear national strategic objective to build a self-sufficient and technologically sovereign robotics industry.
Rémi Cadène, a former AI systems scientist for Tesla's Optimus project, has co-founded UMA, a Paris-based humanoid robotics startup. The company emerged from stealth on Tuesday, unveiling plans for a lightweight, AI-powered humanoid robot named 'Northstar'. UMA is targeting the European market, specifically manufacturing and logistics, to address the continent's labor shortages and dense industrial infrastructure. The startup, which also includes talent from Hugging Face, is planning industrial pilot programs this year and has reportedly lined up 50 potential customers.
Why it matters
The launch of UMA by a veteran of a leading humanoid program signifies both the intense global competition and the maturation of the talent pool in robotics. By focusing specifically on the European market, UMA can tailor its hardware and software to regional safety standards and industrial needs, potentially gaining an advantage over US and Chinese competitors. This development suggests a diversifying global market where regional champions can emerge, rather than a winner-take-all dynamic dominated by a few giants.
According to Bloomberg, UMA is emphasizing a software-centric approach, leveraging a 'Real-Time Learning' architecture that allows the robot to learn new skills via demonstration. "Our approach focuses on learning from demonstration... allowing for rapid adaptation to new tasks and environments," Cadène stated at the Machina Summit. This strategy aligns with the broader industry trend of moving away from complex programming to more intuitive, AI-driven training methods.
AI robotics company Nomagic announced on Wednesday the successful early deployment of its first Vision-Language-Action (VLA) model in a live production environment. The 'AI brain' is operating in warehouse robots at the Swiss e-commerce platform Brack.Alltron. Led by former Google DeepMind researcher Markus Wulfmeier, Nomagic's new AI lab reports that the system has already halved the rate of robot-caused interventions in the warehouse.
Why it matters
This deployment marks a significant milestone for embodied AI, proving that advanced VLA models can deliver tangible ROI in real-world industrial settings, not just in controlled lab environments. Nomagic's strategy—mastering a narrow, high-value task before expanding to general-purpose capabilities—offers a pragmatic counterpoint to the ambitious 'moonshot' projects from larger labs. For entrepreneurs, this demonstrates a viable commercialization path for physical AI: start with a well-defined problem, achieve measurable improvements, and then scale. It shifts the focus from achieving human-level generality to delivering superhuman reliability on a specific job.
"Our goal is to first make our robots super accurate in the tasks they already do, and only then start to expand the range of their skills," Wulfmeier told Fortune. This 'mastery-first' approach contrasts with the 'general-purpose-first' strategy of many humanoid developers. The successful deployment provides a crucial data point that incremental, targeted AI upgrades can yield significant operational efficiencies, potentially accelerating the adoption of smarter robotics in logistics.
South Korean robotic hand developer Tesollo Inc. has initiated preparations for an IPO following the completion of a Series B funding round. The company, which specializes in multi-jointed robotic hands and other core components, aims to use the capital for global expansion. Tesollo's products, including its Delto Gripper series, are already exported to 19 countries and are designed to address the high stress and wear on the end-effectors of humanoid and industrial robots.
Why it matters
Tesollo's IPO plans highlight the growing maturity and investment appetite for specialized component manufacturers within the robotics supply chain. Dexterous and durable hands are a critical bottleneck for the entire humanoid industry; improving their Mean Time Between Repair (MTBR) is essential for making humanoid robots commercially viable in industrial settings. This move signifies that the market is beginning to value the 'picks and shovels' suppliers just as much as the high-profile robot integrators.
"The hands of a humanoid robot are subject to a lot of stress, and the maintenance cycle is very short," a Tesollo official explained to The Robot Report, emphasizing their focus on durability and repairability. The company's success, with overseas sales already surpassing domestic figures, points to a global demand for robust, high-performance end-of-arm tooling that can withstand the rigors of real-world work.
South Korean motion-control specialist Samhyun announced on Wednesday a plan to invest 100 billion won (approx. $72.3 million) through 2029 to pivot its business toward robotics. A significant portion, 40 billion won, is specifically earmarked for developing humanoid robot actuators. The company also unveiled 'AXLON', a new brand for its full lineup of integrated joint actuators designed for the robotics market.
Why it matters
Samhyun's major investment and strategic pivot underscore the immense market opportunity perceived in the robotics component supply chain, particularly for humanoids. As an established motion-control company, its entry adds manufacturing scale and industrial expertise to the actuator space. The development of high-performance, integrated actuators is critical for improving the torque density, efficiency, and precision of humanoid robots, which are essential for their widespread adoption.
"We aim to become a major global supplier in the rapidly growing robot components market," the company stated. This follows related news from another Korean firm, Tesollo, which is pursuing an IPO for its robotic hand technology. Together, these moves indicate South Korea is strategically positioning itself as a key hub for high-performance robotics hardware and components.
3D vision systems company Orbbec has launched a 'Robot-Free Data Collection Hardware Platform' to help companies more efficiently gather real-world interaction data for training physical AI models. Announced Tuesday, the platform uses Orbbec's EGO, UMI, and WristCam product series to provide a standardized solution for capturing the high-quality data needed for tasks like dexterous manipulation.
Why it matters
High-quality, large-scale training data is the lifeblood of modern AI, and this is especially true and difficult for robotics. Orbbec's platform addresses a significant engineering bottleneck by simplifying and standardizing the data collection process. Instead of building custom data-capture rigs, robotics companies can use this off-the-shelf hardware, potentially accelerating their model development and reducing the time it takes to move from lab demos to practical applications. This 'picks and shovels' play is crucial for the industrialization of physical AI.
"The efficient and cost-effective collection of high-quality real-world data is critical for the development of physical AI," a company representative stated. The platform is designed to tackle key challenges like multi-sensor calibration, time synchronization, and capturing the fine-grained hand-object interaction data that is essential for training manipulation models.
Hot on the heels of unveiling the Astra-1 humanoid we covered last month, IIT Madras's Centre for Automation and Robotics has introduced 'Vayu-1,' an indigenously developed actuator prototype. Presented Wednesday at the National Robotics Summit in Chennai, the actuator supports up to 10kg per limb with high precision and targets a production cost under $1,500—a fraction of the price of imported alternatives.
Why it matters
This is a significant breakthrough for India's 'Make in India' initiative and its domestic robotics ecosystem. By drastically reducing the cost of a core, high-performance component, IIT Madras is helping to make advanced robotics and automation more accessible to the country's small and medium-sized enterprises (SMEs). This could accelerate the adoption of robotics in Indian manufacturing and strengthen the country's position as a competitive player in the global market for robotic hardware.
"This makes high-precision robotics affordable for a much wider range of industries in India," a researcher from the project told RobotWale News. This development aligns with recent funding announcements for Indian humanoid startups like Bharat Robotics and Astha Robotics, indicating a coordinated push to build a self-reliant robotics industry.
Researchers have developed EleTac, a soft robotic gripper inspired by an elephant's trunk that integrates high-resolution tactile sensing and proprioception using a single internal fisheye camera. The design, detailed on Tuesday, uses a deep learning model to interpret images from the camera, allowing the gripper to perceive the shape, texture, and pose of objects it is grasping, as well as its own configuration.
Why it matters
This innovation provides a clever, low-cost solution to a persistent challenge in soft robotics: how to integrate robust sensing without adding complex wiring, weight, or fragility. Using a single camera for both tactile and proprioceptive feedback simplifies the hardware significantly. This could lead to more capable, delicate, and affordable soft grippers for applications in manufacturing, logistics, and healthcare where gentle handling of diverse objects is required.
"The single-camera design makes it compact, lightweight, and easy to manufacture," the researchers wrote in their paper. This vision-based approach to tactile sensing represents a growing trend in robotics, where advanced AI models are used to extract rich information from simple, inexpensive sensors, bypassing the need for more complex and costly hardware.
Adding to the wave of local capital we've tracked with Astha Robotics and Nexus Robotics, Bengaluru-based Bharat Robotics has secured ₹50 crore (approximately $6 million) in a Series A led by Venture Catalyst Partners. The funding will accelerate commercialization of its 'Bharat-1' humanoid, aiming to scale production for the Indian manufacturing sector with aggressive local pricing.
Why it matters
This funding is a notable milestone for India's growing domestic robotics ecosystem, which has seen a recent surge in investment. Along with other recently funded startups like Astha Robotics, it signals a concerted effort to build locally-sourced automation solutions. A competitive domestic market could make humanoid robots more accessible for India's small and medium-sized enterprises (SMEs), boosting the country's manufacturing capabilities and addressing labor shortages.
"Our goal is to bring advanced automation within reach of every Indian manufacturer," a company spokesperson told RobotWale News. This news follows a similar $10 million Series A for another Indian humanoid startup, Astha Robotics, also announced Wednesday. The back-to-back funding rounds underscore rising investor confidence in India's potential to become a significant player in the global robotics market.
Mowito, a physical AI startup developing foundation models for industrial robot arms, has raised $3 million in a pre-seed funding round led by Version One Ventures. Announced Tuesday, the funding will be used to expand its US presence and scale deployments. Mowito's technology enables robots to learn complex tasks by observing human demonstrations, rather than requiring traditional, line-by-line coding.
Why it matters
This investment highlights growing confidence in AI-driven solutions that lower the barrier to factory automation. The ability for robots to learn by demonstration is a significant innovation, as it makes them far more adaptable to new tasks and reduces the need for specialized robotics programmers, a key bottleneck for many manufacturers. For entrepreneurs, this signals a major market opportunity in creating more intuitive, user-friendly AI platforms for industrial settings.
"Manufacturing is held back by the difficulty and cost of programming robots for complex tasks," Mowito's CEO said in a statement. "Our goal is to make deploying automation as easy as showing a person what to do." The company's AI models run on existing robot hardware and are already being used by a Fortune 500 automotive company and a major electronics manufacturer.
Making a major move after its bankruptcy reorganization, iRobot on Tuesday unveiled a broad new product lineup. The launch includes five new Roomba robot vacuum and mop models, featuring improved navigation, suction, and AI intelligence. The company also introduced a new product category with the Roomba Electro Plus, a cordless 5-in-1 disinfecting hard floor cleaner that uses electrolyzed water instead of chemicals for sterilization.
Why it matters
This aggressive product launch signals a strategic reboot for iRobot under its new ownership, aimed at reclaiming market share from competitors. The dual strategy of upgrading its core robotic line while diversifying into manual cleaning products with a 'green' angle (chemical-free cleaning) shows a multi-pronged effort to appeal to a wider range of consumers. It's a critical test of whether the iconic brand can innovate its way back to the top of the competitive consumer robotics market.
"We are redefining floor care for our customers with a robust portfolio of solutions that clean more effectively and intelligently," said Colin Angle, iRobot's chairman. Reviewers at Mashdigi noted the new robot models feature advanced LiDAR navigation and AI recognition across various price points, representing a significant upgrade to iRobot's technology stack to better compete with brands like Dreame and Roborock.
Medicaroid announced on Wednesday that its 'hinotori' surgical robot system has received CE-MDR approval, clearing the way for its commercial expansion into the European market. The hinotori system, which is the first Japanese-developed surgical robot to receive this clearance, is designed for minimally invasive surgery and features eight-axis robotic arms and high-definition 3D imaging.
Why it matters
The entry of a new major player into the European surgical robotics market will increase competition, which could drive innovation and potentially lower costs. For hospitals and healthcare providers, more choice in robotic platforms can lead to better-suited technology for their specific needs. This regulatory milestone is crucial for Medicaroid's global commercialization strategy and signals the growing internationalization of the healthcare robotics space beyond the traditionally dominant players.
The approval allows Medicaroid to "broaden the reach of its surgical robotics platform in Europe, catering to the growing demand for minimally invasive robotic procedures," according to a statement. The hinotori system has been in use in Japan since its regulatory approval there in 2020.
Researchers at Leiden University have developed microscopic robots that can navigate complex environments without any sensors, software, or external control. Announced Wednesday, these soft, chain-like microrobots, with segments measuring just 5 micrometers, move autonomously by responding to the physical properties of their environment. The team demonstrated that the robots, propelled by a chemical reaction with hydrogen peroxide in water, could exhibit emergent behaviors like avoiding obstacles and pushing objects.
Why it matters
This breakthrough represents a fundamental shift in microrobotics, embedding 'intelligence' directly into the physical form of the robot rather than relying on complex external control systems. This concept of 'mechanical intelligence' could revolutionize applications like targeted drug delivery and minimally invasive diagnostics. By designing robots that can navigate the body's complex, unstructured environments autonomously, this work opens a path to creating medical microrobots that are simpler, more robust, and potentially safer.
"Instead of building a small computer, we are looking at how we can get the material itself to do the computing," said researcher Rachel Doherty. This biomimetic approach, inspired by how simple organisms interact with their world, circumvents the immense challenge of miniaturizing sensors and actuators, pointing to a new paradigm for designing smart microscopic systems.
Microbot Medical reported on Tuesday that its revenue and new customer acquisitions both grew by over 100% in the second quarter of 2026. The growth was driven by the full market release of its LIBERTY Endovascular Robotic System in April. The company is expanding its U.S. sales team and has also received marketing clearance in Israel, with plans for further international market entry by the end of the year.
Why it matters
This rapid commercial adoption is a strong positive signal for the medical microrobotics field. It demonstrates that there is real clinical demand for these technologies and that they can be successfully brought to market. The growth of the LIBERTY system indicates a trend toward greater precision and remote operation in endovascular procedures, which could improve patient outcomes and expand access to specialized care.
"The significant growth and market expansion of the LIBERTY System indicate increasing adoption of robotic technology in medical procedures," said CEO Harel Gadot in a press release. The company also recently partnered with Lovell Government Services to facilitate sales to U.S. federal healthcare systems, including the VA and Department of Defense, further broadening its potential market.
Engineers at Harvard University have developed a new 3D printing technique called rotational multimaterial 3D printing (r-MM3D) that can create complex soft robots in a single, simplified process. Revealed on Wednesday, the method uses a single nozzle that rotates to switch between different materials, allowing for the creation of soft robots with integrated hollow channels that act as fluidic 'muscles'. This enables precise, predictable movement when the channels are pressurized.
Why it matters
This innovation could significantly accelerate the fabrication of soft robotics, making their production more efficient, customizable, and scalable. By embedding actuation directly into the structure during the printing process, it simplifies what has been a complex, multi-step assembly. For entrepreneurs, this opens up possibilities for creating more lifelike and dexterous robotic tools for applications ranging from delicate surgical instruments and assistive devices to novel manufacturing grippers, lowering the barrier to producing sophisticated soft robotic systems.
"This is the first time that we've been able to print a soft robot with integrated sensors and actuators in one go," said Robert Wood, a senior author of the study. The team demonstrated the technique by printing a soft, multi-fingered gripper capable of picking up a variety of objects, showcasing the method's potential for creating functional, integrated soft robotic systems.
Expanding on the Bristol liquid-metal pump we tracked last month, the team (now joined by NC State) has figured out how to triple its force. Published Tuesday in Advanced Functional Materials, the new method applies a tiny voltage to the gallium-based droplet, exploiting electrocapillarity to amplify the pump's power by up to 3.5 times without increasing its footprint.
Why it matters
This innovation overcomes a major limitation of soft robotics: the trade-off between power and size/weight. Enabling smaller, more powerful, and more energy-efficient fluid pumps could be revolutionary for applications where robots interact closely with humans. This could lead to lighter and more effective wearable rehabilitation devices, more capable biomedical microdevices, and more dexterous soft grippers for delicate manipulation tasks.
"This is a significant breakthrough in smart soft robotics," said Professor Jonathan Rossiter, who led the research. "Our work opens up new possibilities for creating more powerful and efficient soft robots that can be used in a wide range of applications, from healthcare to manufacturing." The findings suggest a new design paradigm for soft actuators based on simple physical principles.
Researchers at the Technical University of Munich have developed a lightweight, soft exoskeleton that enables people with severe motor impairments to regain hand function. The system, detailed Tuesday, uses advanced mechanical design and muscle-sensing technology (electromyography) combined with AI to interpret even very weak muscle signals from the user's forearm and translate them into hand movements like grasping.
Why it matters
This soft exoskeleton represents a significant step forward in assistive technology, offering a practical way to restore a degree of independence for individuals with conditions like stroke or ALS. Unlike rigid exoskeletons, its soft, lightweight design is more comfortable and less obtrusive for daily use. The project's patient-centered design process, where user feedback directly shaped the device, serves as a model for developing effective and user-friendly assistive robotics.
"Our system is designed to be a true aid in everyday life," said a researcher on the project. "The patients' needs were the central focus from the very beginning." The device weighs only about 300 grams and can be worn discreetly under a sleeve, highlighting the advantages of a soft robotics approach for wearable assistive devices.
Open-Source Robotics Gets Standardized Brains and Tools Two major releases aim to unify the fragmented robotics development landscape. NVIDIA and Hugging Face are integrating the Isaac GR00T platform with LeRobot, creating a standardized pipeline. Separately, Ant Group's Robbyant has open-sourced LingBot-VLA 2.0, a universal model designed to control diverse robot morphologies, reducing hardware-specific development overhead.
Humanoid Competition Heats Up with New European Entrant The global humanoid race is expanding with the launch of Paris-based UMA, founded by a former Tesla Optimus scientist. The startup's focus on the European manufacturing and logistics market adds a new regional contender against established US and Chinese players.
Venture Capital Debates the 'Humanoid Fallacy' A counter-narrative is emerging among some Silicon Valley investors, who are expressing skepticism about the humanoid form factor. They argue that specialized, wheeled robots offer better ROI for specific industrial tasks, leading them to back companies focused on function-optimized designs over anthropomorphic ones.
Dexterous Manipulation Emerges as a Key Hardware Bottleneck Multiple startups are attracting funding and preparing for IPOs by focusing specifically on advanced robotic hands. Companies like Tesollo in South Korea and a burgeoning sector in China are tackling the durability and dexterity challenges of end-effectors, which are seen as a critical component for enabling widespread humanoid deployment.
The 'Sim-to-Real' Gap Narrows with Practical AI Deployments Embodied AI is moving from the lab to live production environments. Nomagic has successfully deployed a Vision-Language-Action (VLA) model in a Swiss warehouse, halving robot-caused interventions. This demonstrates a pragmatic approach focused on mastering specific tasks in the real world before attempting general-purpose intelligence.
What to Expect
2026-07-13—Robotics: Science and Systems (RSS) 2026 conference begins in Sydney, Australia.
2026-08-01—Honor expected to launch its 'Robot Phone' in China, featuring a 200MP camera on a robotic gimbal arm.
How We Built This Briefing
Every story, researched.
Every story verified across multiple sources before publication.
🔍
Scanned
Across multiple search engines and news databases
634
📖
Read in full
Every article opened, read, and evaluated
246
⭐
Published today
Ranked by importance and verified across sources
20
— The Robot Beat
🎙 Listen as a podcast
Subscribe in your favorite podcast app to get each new briefing delivered automatically as audio.
Apple Podcasts
Library tab → ••• menu → Follow a Show by URL → paste