Yunqi Capital published a talk on July 17, when WAIC 2026 opened, titled "The Year of Action," using five numbers to observe AI moving from "answering questions" to "completing tasks." As of early July, OpenClaw had received 380,000+ GitHub stars; after the same model was switched to a more complete Harness architecture, the task success rate on a coding Agent benchmark rose from 71.75% to 81.36%.
Yin Qi delivered a keynote titled "When Agents Enter the Physical World" at the WAIC 2026 opening ceremony main forum, proposing the formula "agent capability = model capability × agent operating system capability." The talk centered on three structural shifts in the agent era—new systems, new devices, and new networks—and noted that an Agentic OS and the A2A network will become key infrastructure. StepFun is conducting frontier research with the Shanghai Qi Zhi Institute, led by Turing Award winner Andrew Yao.
Ars Technica interviews MIT's Vincent Sitzmann, Runway CTO Anastasis Germanidis, and World Labs co-founder Ben Mildenhall to explain the definition of world models, the diverging technical approaches, and the questions that remain unsettled.
Every grasp is data. Every failure is feedback. Every iteration makes Daksha smarter. Humanoid intelligence is being built through action. #Daksha #HumanoidRobot #EmbodiedAI
Yunqi Podcast Attent!on speaks with Wang Ximing, co-founder of the brain-computer interface project Kongshanci, about its approach of using brain-computer interfaces and neuromodulation to address psychiatric and cognitive disorders: the company was founded in October 2024 and has completed angel, Pre-A, and A rounds of financing (with Yunqi continuing to invest); it currently has three products, one of which has entered clinical use at hospitals, one is a home-oriented sleep product combining EEG with electrical stimulation (internal testing completed), and another ultrasound brain-computer interface technology is in development.
Union Square Ventures believes that as the cost of models handling messy unstructured input falls and sensors and hardware become widespread, data loops are expanding beyond software into areas that were previously out of reach. Ambient speech, once transcribed and structured, becomes a usable dataset, giving rise to vertical applications and horizontal tools such as Abridge and Granola; in the physical world, robots are becoming more capable and cheaper, and related scenarios are moving from experiments toward commercialization.
Ars Technica interviewed several robotics researchers and founders, outlining how AI is driving the development of general-purpose autonomous robots and the data, safety, and reliability obstacles they face.
KEENON Robotics announced it will attend Interclean Amsterdam 2026, presenting cleaning service robot solutions related to sustainable intelligence. The body text on which the summary is based is actually an industry analysis: China is already the world's largest and fastest-growing robotics market, with the service robot market exceeding 28 billion RMB and growing nearly 40% year on year; funding jumped 344% year on year in 2020, and in the first half of 2021 most leading manufacturers were at B/C rounds, with funding mostly between 300 million and 1 billion RMB.
We have been approaching reward supervision for robots the wrong way. I think free-form preferences are part of the answer. Short thread 🧵
The Diffusion PR team spent four days at the Automate 2026 trade show in Chicago, where they saw robot baristas making coffee on site. Physical AI was the buzzword of this year's show, and Alexandre Boffi of Kawasaki Robotics explained it as the mechanical embodiment of artificial intelligence in the physical world.
Dexmate argues in a post that the deployment ceiling for industrial Physical AI is set by hardware rather than models, using its own Vega 1 Pro as an example.
Enchanted Tools presents the application positioning of its humanoid robot Mirokaï for real commercial environments, with target scenarios covering healthcare, senior living, retail, museums, hospitality, and public spaces. Mirokaï is designed to take on repetitive reception tasks such as greeting visitors, answering questions, and guiding people through buildings, so that staff can devote their time to interactions that require a human presence. This is promotional product content; the original text discloses no technical specifications or verified deployment results.
Nobody buys a robot. People buy back time. Caregivers' time. Employees' time. People's time. → This is the true original design intent behind Mirokaï. ⊕ More details: https://enchanted.tools/robot ⊕ Book a meeting: https://enchanted.tools/contact
StridingAI introduces its Physical AI vision in animated form, in which multimodal intelligence drives real-world robotic action. The company aims to connect processes from inventory management to customer service, improving efficiency in retail, hospitality, industrial, and everyday environments.
The chapter "Trends in Open-Source Commercialization" from the China Open-Source Development Deep Report (2025), written by the Yunqi team, has been officially released. It notes that global open-source commercialization is moving from a service-oriented model into an intelligence- and ecosystem-oriented one. In 2025, global commercial open-source software companies raised about 6 billion USD, of which open-source AI accounted for about 3.8 billion USD and China about 1.1 billion USD. The report adds dedicated observations on the commercialization of open-source AI hardware and embodied intelligence, noting that open-source robotics commercialization unfolds along three main lines: hardware sales, operations and maintenance services, and scenario-based solutions.
The article discusses how teleoperation technology could move work once considered necessarily local, such as excavator and loader operation, into the global labor market.
Our message at @VivaTech 2026: we don't build robots for show; NAO delivers real-world solutions 🦾✨ With unwavering conviction, we came to show what we build in France: a social robotics technology that is useful, easy to use, certified, highly human-centered, and brings people together.
Invasive brain-computer interfaces are moving from technology demonstrations toward early validation of commercial pathways, with a triple inflection point in regulation, clinical use, and AI decoding arriving together in 2026. China's first invasive brain-computer interface product has been approved and granted a national medical insurance billing code; US companies including Neuralink, Synchron, Paradromics, and Precision are entering multi-path clinical trial stages along different routes such as microneedles, intravascular approaches, electrode arrays, and intraoperative methods. The combination of large language models and neural decoding algorithms is pushing the valuation ceiling of brain-computer interfaces from motor control toward language and more complex human-computer interaction.
A new method for off-policy RL with diffusion: given off-policy data, we need to determine what its diffusion latent steps would be under the *current* policy (rather than the policy that collected it), which requires inverting the diffusion process on the off-policy data.
This article is an interview with Dr. Jianyuan Wang, first author of VGGT, which won the CVPR 2025 Best Paper Award, and of VGGT-Ω, a CVPR 2026 Best Paper finalist. He is currently pursuing a PhD through a joint program between the Oxford Visual Geometry Group and Meta AI.
Physical Intelligence is developing a general-purpose control brain that can work across many kinds of robots, using a vision-language-action (VLA) model to teach robots tasks such as making coffee, folding clothes, and peeling vegetables, rather than focusing on a single machine as Tesla, Boston Dynamics, or Amazon do.
When watching viral humanoid robot demonstration videos, first confirm whether the robot is being controlled by a human through teleoperation; unless a paper or company explicitly states it is fully autonomous, be skeptical. Next, judge whether the demonstration was completed in a new test environment encountered for the first time, and whether the video is played at 2x or 4x speed, because for safety and other reasons robots usually move slowly in reality. The informational value of demonstration videos varies widely; some are just entertainment or promotional clips, and genuine capability progress is not easily conveyed through packaging.
summary_z
Chen Yu, managing partner of Yunqi Partners, explains his views on judging long-term opportunities amid technological paradigm shifts in a video interview with PEdaily, with Yunqi Insights excerpting the core points. Yunqi began focusing on robotics and autonomous driving as early as 2016 and invested in MiniMax's angel round in 2021, favoring the approach of a foundation model solving multiple types of tasks. He believes large models enable a team of about 10 people to build products that previously required 100 people, and that the moat for AI applications lies in scenario understanding, data barriers, and multi-model collaboration.
If you're at CVPR 2026, I'll be giving another talk tomorrow (June 4) at the ScaleBot workshop, room 610/612, at 1:30 pm. Topic: making scaling robot data easier 😃 Scaling data matters, and in robotics there's a strange trick...
In an interview, Boston Dynamics mechanical design engineer Chastity Kelly explains the foot design of the Atlas humanoid robot. The new-generation foot has a symmetric structure, with the ankle joint located at the middle of the foot; combined with knees that can bend backward, this allows Atlas to walk forward and backward without turning around. The sole has multi-directional treads to provide two-way grip and prevent slipping on factory floors. She also shares her own journey: starting at a community college, she spent 15 years earning a PhD in mechanical engineering and entering the robotics industry.
The Yingzhi Technology XBOT coffee robot system can operate continuously 24 hours a day without on-site staff, producing hundreds of cups per day while maintaining consistent output standards. AI custom latte art supports printing selfies, logos, and more onto drinks, creating social exposure through customers' spontaneous sharing. Operators can remotely monitor revenue, order volume, and equipment performance through a centralized dashboard, turning the store model into scalable retail infrastructure.
Can robot soldiers really make us safer? Our CEO @sankaet's answer may surprise you. https://youtu.be/dKzbRMuu77w
Sanctuary AI CEO James Wells took part in a panel discussion and gave an interview at Web Summit Vancouver, arguing that Canada must adopt and invest in Physical AI and robotics early or its economic sovereignty could be held by others. He introduced the concept of "labor sovereignty," warning that if foreign robots take on work in Canada's key industries, control over labor productivity will flow abroad, and he called on the federal government to support domestic innovation, commercialization, and large-scale deployment on top of the six pillars of the national AI strategy.
RoboScience founder and CEO Tian Ye judges that the "GPT moment" for embodied intelligence will arrive within 5 years, marked by robots reaching the hands-on ability of a five- or six-year-old child and ordinary users being able to get started within 5 minutes.
Based on a review of nearly 10 months of real customer deployments, Dyna Robotics believes the timeline for converting robotics capabilities into commercial value is longer than current funding assumptions suggest, and that much of the market is a bubble. The article presents three judgments: hardware is not the same as a channel, pretraining data scale is not the same as downstream capability, and building deployment systems is underestimated. It says Dyna-1, released in April 2025, ran autonomously for more than 24 hours on tasks such as napkin folding with a success rate above 99%, but deployments did not self-accelerate, and no one in the industry has yet validated at scale that a second deployment is faster than the first.
Why it matters: Based on the author's review of nearly ten months of real deployments, the piece points out differences between the pace of robotics commercialization and the LLM curve, offering a useful reference for judging an industry bubble.
More U.S. self-service retail kiosk operators are turning to automation as a solution to rising labor costs and the challenge of maintaining consistency across multiple locations, with their evaluation focus shifting from technological novelty to long-term profitability. Yingzhi Technology's XBOT positions its automated coffee equipment as part of a scalable retail operating model rather than as a standalone robotic device, aiming to reduce labor dependence, stabilize beverage quality, and extend operating hours within a smaller footprint. When assessing such systems, operators focus on variables such as labor-reduction potential, throughput capacity, and payback period.
In an article, XBOT (Yingzhi Technology) argues that rising labor costs, inconsistent execution, and the difficulty of expanding across multiple stores are leading operators to evaluate automated coffee systems as retail infrastructure rather than coffee equipment. Their value lies in more repeatable service processes during peak hours, a smaller operational footprint, longer opening hours, and lower reliance on labor, making them suitable for high-traffic settings such as airports, office buildings, transit hubs, hospitals, and universities. Such systems are also connecting to data on ordering behavior, inventory management, and throughput analysis, becoming part of a more integrated retail operating model.
Shima Notes talks with Wen Bowen, an NVIDIA researcher and first author of FoundationStereo and Fast FoundationStereo, about how stereo depth foundation models are changing robotic 3D perception.
"Confirmation of the Subject" is far more than a simple interdisciplinary collaboration between PNDbotics and the artist; it is a profound affirmation of a frontier "subject" that has not yet emerged.
Linear Capital's "Between the Lines" Season 2 features a conversation with Tashi Zhiheng founder and CEO Chen Yilun, presenting for the first time a full picture of his entrepreneurial journey and career choices. Tashi Zhiheng set the record for China's largest angel round in embodied intelligence at $242 million, with Linear Capital backing it for three consecutive rounds; the company is going straight after industrial wiring harnesses, a field where 1 million industrial workers are currently employed in China. Chen Yilun recounts his path from physics competitions to Eaton, DJI, and Huawei's intelligent driving.
Linear Capital shared an interview with Bot Auto founder Hou Xiaodi on Zhijian autoweek, in which he said that on April 30, the day before Musk announced the production version of the SemiTruck had rolled off the line, Bot Auto completed the first fully driverless autonomous truck commercial order in the United States, with no safety driver and no real-time remote human backup.
Why it matters: The interview systematically lays out a commercialization judgment framework centered on CPM, which can be used to understand the differences between L4 companies and L2+ companies on metrics and issues.
Yuan Minglun, Executive Deputy General Manager of Kunwei Technology, said force sensing (six-axis force sensors) is becoming a must-have for humanoid robots to complete force-control tasks such as precision assembly and sorting; vision tells the robot "what it is," while force sensing determines "how much force to use." Kunwei's KWR series six-axis force sensors achieve maximum accuracy better than 0.3%FS (including crosstalk), and he emphasized temperature compensation and the challenges of open humanoid robot scenarios. The company will complete production capacity reserves of 60,000 units per year by the end of June this year and is advancing inspection automation, benchmarking against Switzerland's Kistler, to prepare for sensor demand.
Wang Letian, VP of Product at Robot Era, explains the "software-defined hardware" design logic of its XHand dexterous hand: the fully direct-drive architecture serves reinforcement learning, imitation learning, and mass-production consistency; backdrivability supports teleoperation data collection and Sim2Real transfer; and it has been used as a foundational platform in more than 20 top-conference papers worldwide. The hand has cumulative sales of over 1,000 units and weighs 880 grams; last year it reached annual production on the scale of a thousand units, a single logistics order reached 50 million, and it secured its latest funding round led by SF Express.
Elephant Robotics addresses the multi-vendor integration challenges faced by robotics labs at Latin American universities, offering a vertically integrated solution stack spanning robotic arms, vision systems, and mobile bases. With a robotics R&D background and HAX accelerator experience, the company shortens what once required weeks of cross-vendor debugging into a short cycle from unboxing to teaching. The solution stack is positioned as a complete solution for the Americas market.