Les robots d’IA sont devenus étonnamment humains cette année (mise à jour 2026)

Something changed in robotics this year. The machines became softer, faster, smarter, more human, and in some cases, a lot more dangerous. China unveiled a soft humanoid robot that can shapeshift, float, swim, fly, and even walk on water, defying everything we thought was physically possible. A robot named Moya was revealed in Shanghai. Described as the world's first fully biomimetic embodied intelligent robot with a walking accuracy of 92% and a body temperature maintained between 32 and 36° C during interaction. It can smile, it can hold eye contact, and it made thousands of people deeply uncomfortable because it feels too human. Then Boston Dynamics dropped the bombshell everyone had been waiting for. Atlas, its humanoid robot, is being manufactured immediately with deployments scheduled at Hyundai and Google DeepMind. The back flipping research robot, it's now an industrial worker. Chinese humanoid robots performed kung fu moves, choreographed dances, and elaborate gymnastics at the country's spring festival, Gala, the world's most watched television program. And then it gets darker. China unveiled robotic wolf packs with AI, missiles, and grenade launchers designed for urban warfare capable of firepower-based target suppression in dense urban scenarios. We've crossed a line we can't uncross. And just when you thought robots couldn't get any more sci-fi, Unitry Robotics unveiled the GD1, a manned Mecca resembling an Autobot from Transformers, weighing 500 kg with a pilot on board. So yes, you can literally pilot a giant robot now. Now, here's the question we need to answer. Are we watching the birth of a new industrial revolution or the moment humans started building their own replacements? When robots can look human, work like humans, learn faster than humans, and carry weapons into combat zones, where does this end? And more importantly, are we ready for what comes next? That's exactly what we're going to answer in this video. So stay until the end because this story gets a lot crazier than you think. In Shanghai, a robotics company called Droidup just revealed a humanoid named Moya, and they're calling it the world's first fully biomimetic embodied intelligent robot. That's a mouthful. Yet, the idea behind it is simple. This thing isn't built to look like a machine that happens to walk. It's built to move, react, and exist in a way that feels human on a very subtle level, both physically and socially. Moya sits right in that zone people have argued about for years. The space where robots stop looking like tools and start feeling like artificial people. It's not industrial. It's not cartoonish. And it's not trying to be cute or exaggerated. The whole design direction is realism without going full sci-fi android. In videos that blew up on Chinese social media and were later covered by the South China Morning Post, you see Moya making steady eye contact, smiling slightly, nodding as if it's actually following a conversation, and walking with a gate that honestly makes you do a double take. Droidup says the robot can reproduce human micro expressions. And that's a big deal because those tiny facial movements around the eyes, cheeks, and mouth are exactly what make interactions feel natural instead of robotic and stiff. It's not just about moving lips. It's about subtle timing and small muscle-like shifts that humans read almost subconsciously. Physically, Moya is built almost like an average adult, and the proportions are very intentional. It stands 1.65 65 m tall, about 5'5, and weighs roughly 32 kg or 70 lb. That weight is surprisingly low for a full humanoid, which suggests a lightweight internal structure, and probably careful material choices to keep the mass down while still supporting smooth motion. Visually though, it still carries human proportions in the torso, limb length, and overall posture, which helps avoid the toylike or mechanical look a lot of other humanoids still have. Another detail that really shows how far they're pushing realism is body temperature. Droid up says Moya maintains a surface temperature between 32 and 36° C, which is 89.6 to 96.8° F. So, when you're close to it, or potentially interacting physically, it doesn't feel like cold metal or plastic. It feels warm, closer to a living person than a machine, which can really change how comfortable people feel around it over longer interactions. The company also claims Moya's walking posture has a 92% accuracy compared to human gate. That number is about how natural the motion looks and flows, not just whether it can stay upright or avoid falling. A lot of humanoids can technically walk from point A to point B. Very few can do it in a way where your brain stops tracking the mechanics and starts reacting socially, almost like you're watching another person move through the room. That's where Moya is clearly aiming. Online reactions in China were mixed, and that's putting it lightly. Some people were fascinated, saying, "This is the future of companions, assistants, and service robots in everyday environments." Others said it gave them chills, which is classic uncanny valley territory, where something looks almost human yet just off enough to feel unsettling. That tension is basically built into this kind of design. Technically, Droidup hasn't released a full deep dive into the hardware stack, which only adds to the mystery and speculation. There are reports from a robotics site called Robo Horizon that Moya is built on something referred to as a Walker 3 chassis. That name raised eyebrows because Walker is usually linked with UB Tech, a more established humanoid robotics company known for its Walker series. So far, neither Droidup nor Ubitech has confirmed any direct connection, licensing deal, or shared platform. It could be coincidence in naming, or there could be supply chain or platform overlap that just hasn't been publicly detailed. What is known is that Moya appears to use a modular design, meaning the outer appearance, including the face and surface features, can be changed or customized without redesigning the internal mechanical platform. That's important for commercial use since different industries might want different visual styles and levels of realism from healthcare to education to high-end hospitality. And that's exactly where Droidup says Moya is headed. They're not marketing it as a warehouse robot, a factory arm replacement, or some extreme sports machine doing back flips for viral clips. The focus is clearly on environments that involve long-term human interaction and social presence. According to figures mentioned in SCMP's video coverage, Moya is expected to enter the market by late 2026 with a starting price around 1.2 million Japanese yen. Final pricing and availability haven't been formally announced yet. Yet, that price point already makes it clear this is positioned as a premium human-facing system aimed at institutions and businesses, not a cheap consumer gadget you'd order online. Now, while Moya is all about human realism and social interaction, another humanoid from China has been busy proving it can survive where most electronics would just give up. Unitry Robotics recently showed its G1 humanoid robot completing a massive autonomous trek in extreme cold conditions in Shinjang's Alte region. We're talking temperatures dropping to -47.4° C, which is -53° F. In those conditions, batteries struggle, lubricants thicken, plastics get brittle, and most robots are not designed to operate. Yet, the G1 walked more than 130,000 steps across a snow field while tracing the shape of a Winter Olympics emblem that measured about 186 m long and 100 m wide. That's not just a straight line walk. That's precise path following in brutal weather. To survive that environment, engineers modified the robot from its standard setup. They dressed it in an orange insulated puffer jacket and added improvised plastic covers around its lower limbs to protect joints, actuators, and battery systems from freezing. Positioning relied on China's Bato satellite navigation system, giving centimeter level accuracy. Onboard adaptive path planning algorithms helped the robot maintain stable movement across uneven, icy terrain. The G1 itself is relatively compact for a humanoid. It stands about 127 cm tall, around 4.2 2 ft and weighs roughly 35 kg or 77 lb. Depending on configuration, it uses between 23 and 43 joint motors with maximum joint torque reaching 120 newm. Sensor-wise, it carries 3D lidar, an Intel Real Sense Depth camera, and a noiseancelling microphone array for voice interaction. Power comes from a 9,000 milliamp hour quick release battery that delivers up to 2 hours of operation and supports fast battery swapping. An 8 core processor controls the joints allowing walking speeds of about 6.5 feet per second, which is roughly 4.5 miles per hour. The robot runs on Unit's Unifol LM model, a unified large model for robotics, and uses reinforcement learning for motion control and task execution. Originally introduced in May 2024 as an entry-level humanoid, the G1 starts at RMB9,000, around $14,240. Unitri also reported shipping more than 5,500 humanoid robots in 2025, which shows serious scale. Then over in Shenzhen, another humanoid grabbed headlines for a very different reason. Chinese EV maker Xpunk brought its Iron Humanoid robot to a shopping mall to show off progress ahead of a planned 2026 mass rollout. The demo started strong. Iron performed a choreographed catwalk, interacted with spectators, and stood on stage next to a child designed to look friendly and approachable. The robot's motion system includes a fivederee of freedom human-like spine and a sophisticated hip structure that improves balance and fluidity. Under its outer suit is a 3D printed lattice fascia layer that mimics human musculature. This layer helps dampen vibrations and smooth out jerky motion that's common in bipeedal robots. The goal is to make iron feel less like an industrial machine and more like something that belongs in public, even retail environments. Then after completing a smooth walk, Iron lost balance while standing still and fell backward. A staff member partially broke the fall and the robot ended up face down on the stage. Staff quickly removed it, made adjustments, and later brought it back to interact with children, asking trivia questions. The next day, it stayed strapped to a support frame and focused on educational interactions instead of walking. Despite many earlier flawless demos, the fall dominated online discussion and became a trending topic across Chinese platforms. Technically, iron is built to match adult human scale. It stands 178 cm tall, about 5'10", and weighs 70 kg or 154 lb. It has 62 active joints, and each hand offers 22° of freedom, allowing it to handle both delicate and bulky objects. A curved display wrapped around its head acts as a dynamic face, changing expressions as it communicates. Power comes from a lightweight all solid-state battery for longer operation and improved safety. processing is handled by three Turing AI chips, delivering 2,250 trillion operations per second. Xpang's vision language action system links perception directly to movement, and the robot was trained on thousands of hours of human motion data. It can walk at speeds up to 2 m/s while absorbing shock through its feet to maintain balance. Behind all these humanoids, researchers are also rethinking the mechanics that make movement possible in the first place. At Harvard C's engineers developed a new robotic joint design inspired by the human knee using rolling contact joints. These joints use pairs of curved surfaces that roll and slide against each other connected by flexible elements. The team created a method to mathematically optimize the shapes of these surfaces based on the forces and tasks the joint needs to perform. This means the joint itself helps direct energy efficiently, reducing the need for oversized actuators and complex control. In testing, a kneelike joint designed this way corrected misalignment by 99% compared to a standard joint. A two-finger robotic gripper built using the same approach could hold more than three times the weight of a conventional design using the same actuator input. applications range from exoskeletons and assistive devices to more natural humanoid robots and even biomechanical studies of animals. And finally, in the US, Westwood Robotics is tackling another big challenge, getting humanoids to work while they walk. Their Themis Gen 2.5 robot runs on an AI augmented humanoid operating system called AOS, which tightly integrates perception, planning, and control. Most humanoids today stop moving before they manipulate objects. Femis is designed to do both at once. Its whole body location controller uses sensor fusion state estimation to maintain balance while the arms handle objects. A navigation system built on multi-layer mapping and semantic understanding helps it plan paths while recognizing its surroundings. The interaction system is powered by an object- ccentric vision action model or OC VAM which links visual perception directly to physical actions in a compute efficient way. Hardware also got serious upgrades. The structure is redesigned for about 40% greater impact resistance. Each arm now has 7° of freedom and can handle payloads of more than 5 kg. In the hips, new Mountain Beer actuators deliver over 120% more torque while producing around 80% less heat. The earlier Themis 52 stood 5' 3 in tall with 40° of freedom, arms with six degrees of freedom, and hands moving in seven directions using back drivable BR actuators for smooth, safe motion. In Shenzhen, researchers at Southern University of Science and Technology just revealed Grow HR, a soft humanoid that triples its size, crawls, swims, floats, and even walks on water. Meanwhile, Limx Dynamics 18 Ali robot stepped out of crates and marched in formation using their KOSA operating system. In Denmark, the humanoid artist IDA designed a pod house for humans and robots. In space, Engine AI's PM01 is being prepped as the first robot astronaut and exhumanoids Tien Kung linked directly to a LEO satellite to complete a real task. Back on Earth, China just opened a massive humanoid joint factory and companies like Unitry Robotics are shipping thousands of units. This is happening right now. So, let's talk about it. All right. So, one of the most unusual developments comes from researchers at the Southern University of Science and Technology in Shenzhen. They unveiled a soft humanoid robot called Grow HR, and it's basically built around a totally different idea of what a humanoid skeleton should be. Most humanoids today still rely on rigid column-like frames. They're strong in a straightforward way, and they're predictable, which engineers like. Yet, the trade-off is that they don't adapt well. Human bodies adapt constantly and a lot of that comes down to how bones work. Bones are not just hard sticks. They combine different functions at the same time. Growth through epiial plates, stiffness through compact bone, impact absorption through cancellous bone, and then these multiscale internal cavities that keep everything light while still strong. Grow HR is basically an attempt to bring that bone logic into robotics. Instead of building a standard rigid skeleton, the team designed bone inspired growable linkages. Each linkage mixes soft expandable chambers with tensioned cables and rigid adapters that keep the structure stable. There's also a nonstretchable textile layer that adds axial stiffness so the linkage doesn't just balloon randomly. Then a synchronous cable system runs through the linkage, helping the extension happen smoothly and uniformly. So the robot can deform on purpose while still holding shape like it has internal structure, not like a floppy inflatable. Now the numbers are honestly the part that makes you take it seriously. The linked structures can extend up to 315%. The robot can extend its height by 278%, meaning it can nearly triple in height at full stretch. Grow HR reaches about 1.36 meters. Then it can shrink again in ways rigid humanoids simply can't. It can reduce its height by 36% and its width by 61% specifically to navigate narrow and low spaces. That matters because in real environments, especially rescue environments, the bottleneck is rarely raw strength. The bottleneck is getting through debris, through gaps, through doors, under collapsed beams, and doing it without getting stuck. and Grow HR stays light while doing all of this. The whole robot weighs about 9.9 lb, so 4.5 kg. Each linkage is around 350 g. That lightweight design is a big part of why it can be so mobile. And the way it moves is also not what people expect. It can crawl efficiently, and the team reports crawling at about 112.2 mm per minute. That speed is already decent for a soft humanoid. Yet, the real kicker is how much faster it gets when the growable linkages and the joint motors work together. Coordinated actuation of linkages plus motors enables crawling at a pace 1,122 times faster than either mechanism alone. They also describe it as over 1,000 times faster than motor only locomotion, which tells you something important. Growth style structural change is not just a gimmick, it's an amplifier. You pair it with motors and suddenly the robot moves in a way that neither method can produce alone. Then you get into the multiode part which is where Grow HR starts sounding almost unreal. Yet it's described as demonstrated capability. Because the robot is so light, it can float and swim. It can walk on water at around 16 mm/s. And with ducted fans or quadrotors, it can fly over distances up to several meters. That last part is important because they're not claiming it has built-in jet propulsion. They're saying the platform is light enough that adding ducted fans or a quadrotor system gives you an aerial mode, meaning the same humanoid body can switch between ground movement, water movement, and short range flight depending on what you attach. It also demonstrates strength in a way soft robots usually struggle with. Grow HR can deliver powerful kicks, and the researchers talk about energy storage beyond rigid robots. In other words, the system can store energy in its compliant structures and then release it quickly like a spring instead of relying purely on motor torque at the instant of impact. Stability and balance get managed through active control and also by adding carbon fiber telescopic linkages, which implies they're mixing materials to get stiffness when they need it. One of the researchers, a PhD student at SUST named Wong Ting, specifically explained that this growable bioinspired structure could be applied in field rescue missions, especially navigating narrow gaps, and that multiple locomotion modes let it adapt to complex terrain. The research was published in science advances, which is a key detail because it signals the work is being positioned as a serious new structural direction, not a one-off prototype story. Now, while Grow HR is about reshaping a single humanoid body, another big push is happening around scaling humanoids as coordinated teams because one robot is one thing, but deploying a whole fleet is the real industrial test. That's where Limx Dynamics comes in. also based in Shenzhen. They released a video showing 18 of their Ali humanoid robots autonomously emerging from shipping crates, standing up, walking information, and performing a coordinated routine without human intervention. The company describes it as the world's first practical autonomous deployment of humanoid robots. Whether that specific claim holds long-term is open for debate, yet the demo still signals a major step in coordination. The scenario itself is practical. 18 oi units are placed inside standard shipping containers. The doors open and each robot stands, reaches full height, and begins walking on two legs. Inside tight spaces, they avoid collisions, exit the containers, and move forward together before completing a synchronized routine. The emphasis is on realistic deployment conditions rather than staged lab spacing. Limx ties this to its operating system called Kosa, short for cognitive OS of agents. KOSA enables shared decision-making, movement planning, and collective control. In simple terms, it lets multiple humanoids behave like a coordinated group rather than independent machines sharing space. The company says KOSA mirrors aspects of the human brain by integrating cognition and physical action within one framework. Instead of separating thinking and motion into delayed stages, planning and movement happen almost simultaneously. KOSA also includes memory. Robots store information about environments and objects they've seen before and use it to anticipate what might happen next. During tasks like stair climbing, onboard sensors feed realtime data into KOSA, which immediately adjusts balance, posture, and gate. Limx describes it as unifying perception, learned skills, memory, and even internal emotional states that influence behavior over time. In practical terms, this is about long-term internal variables shaping how robots respond to complex environments. This demo builds on earlier Ali showcases where individual robots navigated uneven terrain, climbed rubble, and balanced on one leg using reinforcement learning. The difference now is scale, coordinated multi-root behavior instead of single robot stunts. The company's vision is clear. Teams of humanoids working together in manufacturing and industrial environments. Beyond movement and labor, humanoids are entering creative fields. In Denmark, at the Utan Center, a concept building designed by a humanoid robot is on display. The robot IDA was completed in 2019 and features a human-like face, camera eyes, and a robotic arm for drawing. Its previous work has appeared at the V and A Museum, the Venice Bianale, and the Great Pyramid of Giza. for the exhibition I'm not a robot architecture and design between human and machine. IDA produced what's described as the first building concept designed by a humanoid robot. The design is a retrofuturistic pod house inspired by 1950s and60s space age architecture with rounded forms and large port hole windows. It's intended for human and robot cohabitation and is framed as usable on Earth or even on the moon or Mars. The project includes pen sketches, painted artworks made with the robot arm, and digital renders of the interior and exterior. The layout shows two living rooms, a spiral staircase, a kitchenet or bathroom, and a smaller pod within the main structure for the robot. Aiden Miller, who works with IDA, openly discusses the blurred line between human prompting and AI agency, describing a team of studio technicians supporting the work. He presents IDA as having rising creative agency while still responding to commissions similar to human artists. IDA itself describes architecture as a natural extension of its purpose, exploring how humans and technology share living spaces. The exhibition runs until October 18th, 2026 with future showings planned in London. Meanwhile, humanoid robots were a major presence at CES. Yet, AI adoption in architecture remains limited with only about 8% of US studios using AI tools by early 2025. The most ambitious direction though is space. Shenzhen based engine AI has launched efforts to send a humanoid robot into orbit, partnering with Beijing Interstellar Human Spaceflight Technology. Their chosen platform is the PM01 humanoid, which they aim to develop into a robotic astronaut. Space presents extreme conditions, vacuum, microgravity, radiation, and massive temperature swings. Engine AI says PM01 integrates high precision environmental sensing, millisecond level motion response, and autonomous decision-making to handle these demands. The idea is for humanoid robots to perform risky tasks like station repairs, hazardous exploration, and long-term monitoring, reducing risk to human crews. PM01 also has a commercial version on Earth. A model called PM01 JD Joy inside launched for about $27,000, offering natural conversations, coordinated behavior, persona customization, and personalized voices. The robot stands 1.38 m tall, weighs 40 kg, and uses an aluminum alloy exoskeleton. Its waist rotates up to 320°. It relies on an Intel Real Sense Depth camera for perception and runs on Nvidia Jetson Orurin plus an Intel N97 CPU for real-time control. Engine AI compares PM01 to its larger SE01 humanoid, noting PM01 is 27 cm shorter and 8 kg lighter. Meanwhile, Interstellar, the space partner, has its own road map, aiming for suborbital flights around 2028 at 100 km altitude and orbital tourism missions around 2032, reaching 400 km. Another space linked development focuses on communications. Exhumanoids embodied Tien Kong robot reportedly became the first humanoid to connect directly to a low Earth orbit satellite without ground network support. At a Beijing event, it linked with a Galaxy Space phased array satellite and transmitted live visual and movement data. During the demo, the robot retrieved a certificate from a driverless vehicle and delivered it to a building while its actions were streamed via satellite to a control center almost instantly. Phones and computers were also connected to the same satellite link. This proves humanoids can operate in remote areas, disaster zones, and low signal environments where traditional networks fail. Back on Earth, manufacturing capacity is scaling fast. EU Robot Technology opened what it calls the world's first factory dedicated to humanoid robot joints in Shanghai. The facility can produce 100,000 joints per year with plans to triple output. These joints make up about 50% of a humanoid's cost, making them strategically critical. Shipment data shows rapid market growth. Unitry Robotics reported shipping 5,500 pure humanoid robots in 2025 with 6,500 units produced. IDC estimates global humanoid shipments hit 18,000 units in 2025 with $440 million in sales and 58% year-over-year growth. Chinese companies dominate more than 80% of installations. Even Elon Musk now describes China as Tesla's biggest humanoid competitor, highlighting China's strength in scaling, manufacturing, and AI. Analysts point to China's complete supply chain and cost advantages, pushing humanoid prices into the tens of thousands of dollars. Broader industry data shows China had 451,700 intelligent robotics enterprises by the end of 2024 with total registered capital of 6.44 trillion yuan. So Boston Dynamics just dropped a major new Atlas update and this version is doing clean cartwheels, back flips, and midstep balance recoveries thanks to work with the robotics and AI institute led by Mark Ryber. At the same time in China, Agibbot has humanoids performing kung fu with real monks at the Shaolin Temple using its Genie Operator 1 AI and zero sample learning to copy complex human motion. Then those same robots take the stage at Agibbo night in Shanghai, a full 1-hour robot gala where 16 humanoids dance, flip, act, and perform in tight coordination. And while all of that is happening, Faraday Future just entered the robotics market at the NADA show with three embodied AI robots going up for sale with real prices, deposits, and delivery timelines. All right, let's start in the US where Boston Dynamics is basically wrapping up the testing era of its research version of Atlas with a kind of final stress test that looks more like a gymnastics competition than a robotics demo. They worked closely with the Robotics and AI Institute, also known as RA AI, which is led by Mark Rybert, the same person who founded Boston Dynamics in the first place. Their goal here was to push full body control and mobility right to the edge before shifting focus toward a more productionready humanoid. In the latest video, Atlas starts off calmly, just walking across an open area with a gate that honestly looks very human. The motion is smooth, the posture is stable, and there's none of that stiff robotic vibe people used to associate with humanoids. Then the intensity ramps up fast. Atlas launches into a sideways cartwheel, coordinating arms and legs to keep its momentum and balance through the entire motion that flows directly into a backflip. The robot tucks in, rotates cleanly in midair, and lands on both feet without collapsing. What really stands out is how it absorbs the impact on landing. Instead of just slamming down, you can see controlled force distribution through the legs and torso. The extended footage also shows slow motion replays and a blooper reel, which is actually just as important as the successful runs. There are clips where Atlas collapses, tumbles, or loses stability. In one moment, it steps down, places its foot slightly wrong, and then subtly adjusts it before continuing forward. Those microcorrections show how advanced the control system has become. This isn't just presscripted animation. It's dynamic feedback, real-time balance adjustments, and whole body coordination. Our AI explained that all of these behaviors, from the natural walking shown at CES 2026 to the gymnastic moves in these videos, come from the same learning framework. It's not one system for walking and another for flips. They're using a whole body learning approach designed for zeroot transfer. That means the control policies are trained in simulation and then run directly on the real robot without extra tuning. In robotics, that kind of simtoreal transfer has always been a huge bottleneck. So this is a serious milestone. The partnership between RAI and Boston Dynamics was formalized in early 2025. Their goals included improving how agile behaviors transfer from simulation to hardware, boosting loco manipulation skills, so moving while interacting with objects, and developing full body contact strategies where arms and legs coordinate during dynamic tasks. This builds on earlier reinforcement learning work, including the spot reinforcement learning researcher kit, which helped the quadriped spot reach record- setting running speeds. Now, while the research Atlas is doing flips for demos, the production version is being prepared for industry. The Enterprise Atlas is designed for large scale manufacturing. It has 56 degrees of freedom and a 4digit tactile sensing gripper, which is a big deal for handling parts with precision. Hyundai Motor Group confirmed that these robots are set to be deployed at Hyundai Motor Group Metaplant America by 2028. The first tasks will be part sequencing, and by 2030, the plan is to expand into full component assembly. So all that acrobatic control is laying the groundwork for factory floor reliability. While Atlas is flipping in the US, over in China, humanoids are throwing punches and kicks alongside real martial artists. Chinese robotics company Aubot released a video that quickly went viral, showing its humanoid robots practicing kung fu at the historic Shaolin Temple. That location choice is symbolic on purpose. Shaolin is known worldwide as the birthplace of Shaolin Kung Fu with centuries of discipline, training, and philosophy behind it. In the clip, the robots move with serious balance, speed, and precision. They perform coordinated kung fu sequences in sync with monks, and the level of control surprises a lot of viewers. Some people react with excitement, others with a bit of fear because it's a very direct visual of machines stepping into traditionally human physical domains. This isn't a one-off stunt for Agibbot. Back in March 2025, the Shanghai based company introduced its Lingshi X2 humanoid as a generalpurpose robot in a video where it was cycling like a human in an open space. In May 2025, they released footage of the same robot performing a Webster flip, which is a forward somersault with a back leg takeoff and tight midair control. So, these kung fu moves build on an existing trajectory of athletic demonstrations. The Lingshi X2 is powered by Agibbot's Genie Operator 1 AI model. also called the GO1 model. This system uses zero sample generalization which lets the robot manipulate items and carry out basic tasks without prior task specific training. A key feature is something called latent actions. This helps the robot understand human motion by looking at past and current visual frames effectively inferring movement patterns without massive labeled data sets. That matters because one of the biggest bottlenecks in humanoid robotics is the shortage of highquality labeled action data for human-like behavior. The GO1 model also supports multi-root synchronization. That's why in the Shaolin video, you see groups of humanoids moving in coordinated patterns, not just a single robot doing isolated moves. Agibbot positions these robots as collaborators that can work alongside humans, learning from demonstrations and assisting in physical tasks. Then Agibbot took things even further by hosting a full-blown robot gala in Shanghai called Aggiebot Night. This was a 60-minute show led entirely by robots with embodied intelligence. 16 humanoid robots performed music, dance, and comedy. They did flips, fast spins, group dances, and runway style walks. Large groups moved in formation with tight timing, shifting smoothly from one performance to another without interruptions. The stability on display was key. These robots weren't just doing one flashy move and stopping. They sustained complex, highintensity performances over a full hour. According to Chu Hang, Agibbot's chief marketing officer, this event went beyond a product showcase and acted as a realworld test of stability, consistency, and system level coordination across multiple robots operating at the same time. There were also collaborative acts where human performers danced with Agabot G2 humanoid robots and done quadriped robots. These segments showed real-time alignment between human and robotic motion. In other acts, humanoids performed card magic and floating illusions. Comedic skits featured multiple robots interacting on stage with better timing and more expressive behavior than earlier generations. Agibbot used the gala to show its full robot lineup. The full-sized AT2 series handled multimmodal interaction and autonomous navigation for presentation and showroom style roles. The compact half-sized X2 series focused on natural conversations, humanlike walking, and expressive movement, which makes them useful for entertainment, research, and education. The industrial G2 series combined interactive AI with precise force controlled handling, enabling quick deployment in factories and logistic settings. Meanwhile, the Dun series quadriped robots demonstrated dependable mobility and task performance in complex inspection and operational environments. While all this performance art is happening, another company from the US side is jumping directly into selling embodied AI robots. Faraday Future, known for its electric vehicle ambitions, unveiled three embodied AI robot series at the NADA show in Las Vegas. This marks the official debut of FFAI Robotics Incorporated, a new subsidiary based in California. They introduced three products, FF Futurist, FF Master, and FX AIS. Two are humanoids and one is a quadriped. Sales and non-binding pre-orders open the same day with initial deliveries targeted for late February 2026. Faraday Future says it aims to be the first US-based company delivering both humanoid and quadriped robots at the same time. They also presented a 3in-1 EAI robotics ecosystem which includes EAI devices, an open-source EAI brain and platform and a decentralized data factory. The idea is continuous learning, scenario specific customization, and scalable deployment. According to the company, more than 1,200 units are already covered by non-binding, non-refundable B2B deposits. Production prep, customization, testing, and data training are happening in parallel. The FF Futurist is a full-size professional humanoid aimed at commercial settings like hospitality, retail, education, and events. It comes with advanced perception systems, multilingual interaction, and modular software for future upgrades. The FF Master is a more athletic humanoid meant for interactive and educational uses. Designed as an AI companion in homes, institutions, and events, the FX Agis is a quadriped for security, patrol, and industrial tasks. Built to handle complex terrain with optional sensors and modules. Pricing is very clear. FF Futurist starts at $34,990. FF Master starts at $19,990 and FX Eegis starts at $2,499. Extra ecosystem skills packages are sold separately, and the company is exploring financing, leasing, and rental models to make adoption easier. Faraday Future frames this robotics push as a natural extension of its AIdriven mobility work. It plans a dual track strategy covering both EAI vehicles and EAI robotics, sharing research, manufacturing, software, and services across both. In China, a humanoid robot did something humans trained for years to master, and no humanoid had ever pulled it off before. In the US, another humanoid casually handed out Christmas swag and talked back. A robot dog sold out in under an hour like a new smartphone. factories started replacing skilled workers with humanoids at scale. And one of the most famous robots in the world is about to step onto a global stage next month. And it all starts with a robot in China doing something that looks simple until you realize it's one of the hardest things a machine has ever been asked to do. On December 22, Tar Robotics showed a humanoid robot on stage at a live event. And this robot didn't lift boxes, didn't stack crates, didn't walk around waving at people. It sat down and did hand embroidery, real embroidery. It threaded a needle, used both hands, and stitched a logo live in front of an audience. Now, if you've never stitched anything in your life, that might not sound impressive. But the moment you think about what's actually happening there, it gets kind of wild. You're dealing with thread that stretches and twists, fabric that moves and deforms constantly, and movements that need subm precision over a long sequence of steps. One tiny mistake and the thread snaps, the stitch misses, or the whole thing falls apart. This kind of task has been a nightmare for robotics for decades. Industrial robots are amazing at rigid, repeatable jobs. Pick this metal part, place it there, do it again a thousand times. But soft materials break everything. They change shape, they resist unpredictably, and they require constant force adjustment in real time. That's why until now long, delicate, two-handed work like this was seen as basically offlimits for automation. And that's what made this demo so important. The robot moved through the whole process smoothly, staying stable the entire time with no signs of hesitation or struggle. Both hands working together, adjusting force, tracking the needle and thread visually, and staying stable the whole time. That's not just motion control. That's embodied intelligence working as a system. People inside robotics immediately clocked what this meant. Because once you can do this, you're suddenly not just talking about embroidery. You're talking about wire harness assembly, precision electronics, fine mechanical assembly, all the stuff factories still rely on skilled human hands for. At the event, Tar Robotics CEO Dr. Chen Yoloon explained how they got there. And the way he described it actually lines up with a bigger trend we've been seeing across AI lately. He called it a data AI physics trinity approach. Basically, instead of treating data collection, model training, and physical robots as separate worlds, they connect everything into one continuous loop. They collect detailed realworld operational data using their SenseHub platform. That data feeds into their embodied AI model, the AWE 2.0 AI world engine. And here's the key part. The model isn't trained to do one specific task. It's trained to learn general physical skills, balance, coordination, force control, vision under uncertainty, stuff that transfers from one job to another. Those learned skills then run directly on their humanoid robots, the T-series and A series platforms. Chen also stressed something that matters way more than people realize. The digital to physical gap is small. What the AI learns during training actually carries over into the real world, which is why the robot held together on stage and performed the task without falling apart under real conditions. The system follows classic AI scaling laws, too. As more real world data flows in, performance improves across multiple tasks at once. Their chief scientist, Dr. Dingwin Chow, talked about how success rates jumped across different scenarios simply by scaling data and refining the model architecture. The long-term goal here obviously isn't a robot that can stitch logos. It's robots that learn adaptable skills and carry them into new environments without being reprogrammed every time. And the craziest part is that Tar Robotics was founded on February 5, 2025. In under a year, they went from concept to live humanoid demos showing capabilities people used to say were years away. Investors clearly noticed. The company raised $120 million in an Angel round, then followed that with another $122 million in an Angel Plus round. Now, at the time this was happening in China, a clip from the US revealed how far humanoids have come in a much more everyday setting. On December 23, right before Christmas Eve, Figure AI's CEO Brett Adcock shared a short video showing him interacting with the company's latest humanoid, Figure03, in a simple, unscripted setting. He starts by asking the robot where it was built. It answers clearly saying San Jose, California. >> I was built right here in San Jose, California. >> He asks which generation it belongs to. The robot says it's part of the third generation. >> There are three generations of bigger robots, and I'm the latest one. >> When he asks which generation is best, it does the polite thing and says the third generation because it has the most advanced features. I believe I'm the best as the third generation with the most advanced features. >> Then he tests something more practical, visual recognition. He asks the robot to hand him medium and large shirts. The shirts are sitting in different baskets. Figure 03 looks at the setup, identifies the correct size each time, picks the right basket, and hands the shirt over correctly. Under the hood, this robot runs on Figure AI's Helix model, which is a vision language action system. That means vision, language understanding, and physical movement are tied together in one loop instead of being bolted together with rules. People watching the clip were impressed, but they also immediately noticed something else. The robot pauses for about 2 to 3 seconds before answering questions. That delay jumped out in the comments. >> You give me a medium shirt. >> One person even joked that it felt like dialup internet and asked how Chinese robots compare. That reaction actually points to a real unsolved problem in humanoids right now. Speech latency. Even when perception and manipulation are solid, conversational timing still feels off. The robot hears the question, processes it, generates a response, then speaks. Every step adds delay, and humans notice it instantly. Now, that doesn't mean Figure03 is weak. Far from it. The robot was released in October and is a big step forward for the company. It's smaller and about 9% lighter than previous versions. It has a softer exterior with mesh fabric and foam padding designed to reduce injury risk. It stands about 5' 6 in tall and is built mainly as an assistant for household pick and place tasks. Physically, it's more capable, too. In another clip released a few weeks earlier, figure 03 showed quick starts, smooth turns, sharp braking, and even running. Forbes reported speeds between 4 and 6 mph, which would beat the usual humanoid benchmark of 3 to 4 mph. Figure AI itself gave a more conservative number of 2.7 mph. Audio hardware has been upgraded as well with larger speakers for clearer speech and less distortion. For power, Figure03 uses wireless charging through coils built into its feet. It can recharge by standing on a charging pad at up to 2 kW and a full charge gives it roughly 5 hours of operation. So yeah, speech latency is still a thing. But overall, figure 03 shows how far humanoids have come in everyday interaction, even in casual settings. Then almost at the same time, something happened in the consumer space that surprised a lot of people. So, Vita Dynamics launched its AI powered robot dog under the V-bot brand. The price was set at 9,988 R&B, roughly 1,368 as an early bird offer. Within just 52 minutes, more than 1,000 units were sold. That's real demand, not hype clicks. The launch pushed related topics to number seven on Weibo within 5 hours, and live streams around the product passed 10 million views. For a consumer embodied AI product at a 4 figure price, that's a big signal. What makes this robot dog stand out is that it's fully autonomous. No remote control, no constant human input. It runs 128 tops of ondevice AI compute and uses LAR and depth vision processing visual data locally to protect user privacy. The hardware is serious. It uses V-bot self-developed N45 permanent magnet joint motors with 24.5 Newton meters of peak torque. It can tow up to 100 kg. Battery life is around 5 hours and it supports wireless charging along with 240 watt USBC fast charging. The founding team includes people from Horizon Robotics and Lee Auto and the design borrows heavily from electric vehicle architecture. Founder Yu Yinan described it as having a spatial intelligence system similar to autonomous driving stacks plus an interactive cockpit for human machine interaction. Public beta testing is planned for January 2026 with batch delivery starting in March. And honestly, that early sales spike is being seen as a real validation moment. It suggests that consumer grade embodied intelligence is officially something people are willing to buy at scale. Now, while robot dogs were selling out online, humanoids were quietly replacing people on factory floors. CL, the world's largest EV battery maker, launched what it says is the world's first large-scale humanoid robot powered battery production line at its Luo Yang facility in Henan. The humanoid robot called Xiaomo is now handling critical tasks in battery pack production. Shiao was developed by Spirit AI, a Hangju based startup founded in 2024 and backed by CL. The robot works in end-of-line and direct current internal resistance testing, which are the final checks before battery packs leave the line. These tasks used to require humans to manually connect high voltage test plugs, which came with safety risks and inconsistent quality. Xiao uses an end-to-end vision language action model, giving it strong environmental perception and precise control. It handles uncertainty, adjusts posture and force in real time, and inserts and removes flexible wiring harnesses without damaging components. In real production, Xiaomo achieves over 99% connection success and matches the cycle times of skilled human workers. It also monitors wiring connections continuously and reports anomalies in real time, cutting defect rates. During downtime, it switches into inspection mode to further improve reliability. CL says Xiao now handles continuous production across multiple battery models and delivers nearly three times the daily workload of manual labor while staying consistent and stable. This rollout fits into CL's broader dominance in EV batteries. From January to October, the company logged 355.2 gatt hours of global installations, up 36.6% year-over-year, giving it a 38.1% global market share. In China alone, November installations hit 40.87 gawatt hours or 43.71% of the market. And finally, looking ahead to early next year, one of the most iconic humanoids is about to step onto a global stage. Atlas, developed by Boston Dynamics and backed by Hyundai Motor Group, is set to make its first live public demo at CES 2026 in Las Vegas. Hyundai plans to unveil its broader AI robotic strategy there, positioning robotics at the center of its future growth. Atlas has already been running pilot demos at Hyundai's EV plant in Georgia. Hyundai acquired Boston Dynamics in 2021 and has been steadily integrating robotics into its long-term plans. The company has announced around 50 trillion1 roughly $ 34 billion in AI investment from 2026 to 2030 and plans to build a robot manufacturing plant in the US capable of producing 30,000 units a year. For people who followed robotics for a long time, Atlas represents decades of progress. From early MIT leg lab experiments to modern humanoids walking into factories and trade shows, robotics just built a humanoid robot that recognizes your face, remembers your past conversations, and reads your emotions in real time. UNIX AI's Panther is already cooking, cleaning, and running homes for up to 16 hours on a single charge. IHMC's Alex is built for war zones and disaster sites, moving faster and handling dangerous environments before humans step in. At the same time, Princeton created a robot that moves using heat instead of motors. Scientists built living neurobots with real neurons. And new artificial muscles now let robots lift up to 100 times their own weight. And now, Unitri is about to launch a $4,000 humanoid globally. All of this is happening right now. So let's talk about it. All right. So a company called Real Botics just delivered its first humanoid robot equipped with a system called Vinci to Ericson. And the whole point of Vinci is visual awareness combined with memory and behavior tracking. Now what makes it different is the way it actually interacts with people. The cameras are built directly inside the robot's eyes. So when it looks at you, it's not fake eye contact. It's actually tracking your face, your movement, and your behavior in real time. That alone changes how natural the interaction feels. Now add memory on top of that. The robot can recognize returning users, remember past conversations, and continue where things left off. So instead of resetting every time like most assistants today, it builds context over time. That's a completely different type of interaction loop. It also tracks emotional signals, which means it's analyzing how you respond, your expressions, your engagement level, and adjusting its behavior accordingly. So, the interaction becomes more fluid and personalized instead of scripted. Under the hood, it's doing object recognition, motion detection, and realtime engagement tracking. And the key part here is not just interaction, it's data. Vinci is designed to capture structured data about every interaction. who you are, how you behave, how you respond emotionally, how engaged you are over time. That data can then be analyzed by companies. So, this becomes a tool for customer engagement, analytics, training environments, even clinical research. You're basically turning human robot interaction into measurable data sets. And it's not locked to one robot. Robotics says Vinci can be integrated into all of their humanoid platforms, which means this system could scale across industries pretty fast. Ericson deploying it is actually a big signal. That's not a lab test anymore. That's enterprise level use where robots are interacting with real people and generating real data. Now, at the same time, a Chinese company called Unix AAI launched a humanoid robot called Panther and they're already shipping it globally. This one is designed for actual household use. Panther is about 5' 3 in tall, weighs around 80 kg or 180 lbs, and runs for anywhere between 8 and 16 hours on a single charge. That battery range alone is already pushing it closer to something you could actually use daily. The design is interesting because it's not a traditional walking humanoid. It's wheeled with a four-wheel steering and four-wheel drive system that makes it more stable and efficient indoors, especially in cluttered environments where legged robots still struggle. It has 34 degrees of freedom, including something they call the first mass-roduced 8D bionic arms. Those arms combined with adaptive intelligent grippers give it pretty high precision when handling objects. And it's not doing single tasks. That's the key difference here. Panther is built for multi-step workflows. So, it can wake you up, prepare breakfast, clean the kitchen afterward, organize the living space, and basically chain all of that into one continuous sequence. That's a big jump from robots that can only execute isolated commands. It uses a full stack of systems to make that work. Uniflex handles task generalization and imitation learning, meaning it can adapt across different scenarios. Uniouch adds visual tactile capabilities so it can actually handle objects more precisely and Uni Cortex is responsible for long-term planning which is what enables those multi-step task sequences. It also has cameras, sensors and audio systems for navigation, object recognition, and interaction with people. And the use cases go beyond just homes. They're targeting hotels, retail, reception services, guided tours, elderly care, even industrial environments like security patrols and research. There are still challenges, of course. Real homes are messy, lighting changes constantly, soft objects are hard to manipulate, and reliability is still a big question. Battery life, safety, cost, all of that still needs to improve. Still, the fact that these robots are already performing multiple real world tasks in actual homes is a pretty clear shift. Now, while some companies are focusing on homes and interaction, others are pushing into extreme environments. There's a humanoid robot called Alex, developed by IHMC in the US with support from the Office of Naval Research. This one is built for situations where humans shouldn't go. Alex is the successor to a robot called Nadia, which was already known for things like playing pingpong and boxing. The new version focuses more on realworld mobility, autonomy, and response speed. One of the biggest upgrades is weight reduction. Alex weighs about 187 lbs, including its battery, down from Nadia's 220 lb. That might not sound massive, though in robotics, that kind of reduction has a huge impact on agility and energy efficiency. It uses custom high-powered actuators that cut weight without giving up strength, which is a big deal for a robot like this. That lighter build should help Alex move with more speed, react faster, recover balance more smoothly, and handle unstable terrain with a lot more confidence. On the hardware side, it comes with 19 degrees of freedom, high-speed joints that can hit 9 radians per second, and wrists with up to 300 degrees of motion. All of that gives it the kind of range, agility, and control it needs to deal with more demanding, complex tasks in the real world. It can also carry a continuous payload of about 10 kg which makes it a serious candidate for work in collapsed buildings, hazardous environments, disaster zones, and even military-style operations where strength, precision, and mobility all matter at the same time. The idea is that it operates as part of a human machine team. It can go into dangerous areas first, explore, gather information, and reduce risk before humans step in. Interestingly, Alex doesn't even have a face yet. At its public debut, people are actually going to design its face, and the best designs will be turned into real models. That's kind of a reminder that the focus here isn't appearance, it's capability, balance, perception, autonomy, and realworld performance. And the same system could be used in manufacturing, logistics, aircraft maintenance, oil rigs, basically anywhere that requires precision in risky environments. Now, if you zoom out a bit, you start to see something else happening in the hardware itself. Researchers at Princeton just built a robot that doesn't use motors at all. Instead, it moves using heat. They used a material called liquid crystal elastimemer, which can be programmed at the molecular level. When heat is applied, the material contracts or bends in specific ways depending on how it was printed. So instead of building a robot and then adding joints and motors, they're embedding movement directly into the material itself. They used a custom 3D printer to create patterned zones inside the material. And these zones act like hinges. When heated, they bend predictably, allowing the structure to fold and unfold. They even integrated flexible circuit boards during the printing process, so everything is built as one system instead of being assembled afterward. The robot includes temperature sensors and closed loop control, meaning it can adjust itself in real time to maintain accuracy over repeated movements. They demonstrated it with an origami inspired structure that flaps like a crane, and it does this repeatedly without noticeable wear. That's important because durability is usually a big problem with soft robots. The system also uses mathematical models from origami design to control motion. So, this is not random bending. It's highly structured and programmable. The long-term idea is scalability. These robots could be manufactured more easily, operate in environments where rigid systems fail, and even be used inside the human body. Now, if that sounds advanced, the next one goes even further. Scientists have created what they call neurobots. These are living robots made from frog cells with actual neurons integrated into their structure. Previous versions known as xenobots could move using psyia, basically tiny hairike structures. Though they didn't have any internal control system, neurobots changed that. Researchers inserted neural precursor cells into these biological constructs. Over time, those cells developed into neurons and formed networks inside the robot. Those neurons connect to other cells that control movement, which means the robot now has a basic nervous system influencing its behavior. And the effects are very clear. The robots become more active, their shapes change, and their movement patterns become more complex. They even tested how neural activity affects behavior by using drugs that alter neural communication. The results showed that the nervous system was actively shaping how the robots move. There were also unexpected changes at the genetic level. some gene expressions linked to visual system development started appearing which suggests that future versions could develop new sensory capabilities. That part is still early though it shows how unpredictable this field is becoming. These robots are not mechanical. They're biological systems with programmable behavior and the potential applications go into areas like medicine where you could have living machines operating inside the body. Now, at the same time, there's another breakthrough happening in robotic strength. Scientists developed artificial muscles that allow robots to lift up to 100 times their own weight. These are called harp actuators, and they're basically flexible airpowered structures that mimic how real muscles work. Instead of rigid motors, they expand and contract using small amounts of air. That makes them lightweight, quiet, and highly adaptable. They can operate in extreme environments, including high heat and abrasive conditions. And because they're flexible, robots using these muscles can squeeze through tight spaces and move through debris. That makes them ideal for disaster response, where you need machines that can navigate collapsed structures without causing more damage. They've already built a robotic arm inspired by an elephant trunk, which can reach around obstacles with a high level of precision. There's also a wearable system that helps humans lift heavy objects by reducing strain, which shows this technology isn't limited to robots alone. And the materials used are strong enough for space applications, meaning these systems could eventually be used in space missions as well. Finally, there's one more shift happening that ties everything together, and that's cost and scale. Unitry is about to launch a humanoid robot called R1 globally, and it's priced at around 29,900 yuan, which is roughly 4,370. That's extremely low compared to most humanoid robots today. The R1 stands about 123 cm tall, weighs around 59 lb, and is designed for dynamic movement. It can run downhill, perform cartwheels, stand up from the ground, and basically handle athletic motion. Unitry is planning to sell it through AliExpress, targeting markets like the US, Europe, Japan, and Singapore. And the scale is what really stands out. They shipped over 5,500 robots in 2025, while companies like Tesla, Figure AI, and Agility Robotics shipped around 150 each. For 2026, they're aiming for 10,000 to 20,000 units. That kind of production volume changes everything because once you hit that scale, prices drop, accessibility increases, and adoption accelerates. Industry projections suggest unitry could account for nearly half of all humanoid robot production soon. China just revealed an autonomous robot war pack built from dog bots, drones, laser weapons, and unmanned boats. Europe is putting military robots through one of the toughest realworld tests anywhere. Bezos is building a 100 billion AI industry machine. Amazon is preparing for a future packed with robots. BMW is already testing new humanoids on factory work. Zuckerberg wants to build a personal super intelligence around your life. Musk is pushing a giant chip fortress, 50,000 Optimus robots, and possibly mass production through Shanghai. And Unit's humanoids are now learning tennis, chasing bores, and sprinting at near human record speed. All of this is happening right now. So, let's talk about it. All right, let's start with China. China just officially laid out its vision for the future of ground warfare and it is an autonomous wolf pack of robot dogs and drones that thinks and hunts as a single organism. The reveal came through a new documentary from CCTV where the PLA walked through its road map from soldier support platforms all the way to fully autonomous urban combat units. The key concept is that the Wolfpack is not just a group of robots. It is a distributed network with a shared digital brain and each machine has a specific role. Shadow is the scout handling real-time situational awareness. Polar is the heavy lifter moving logistics and ammunition. And then there is Bloody, the strike element, a robot dog that is basically a walking arsenal armed with an automatic rifle, grenade launcher, and mini rockets. The new generation is faster and more durable. They hit around 15 kmh and carry up to 25 kg of payload. The joints are flexible enough to handle rubble and staircases in urban environments. And the control scheme is aggressively simple. One soldier can run the whole pack using voice commands, a joystick mounted on their rifle, or even gestures through a tactical glove. The wild part is a system called ATLS. Chinese engineers trained a swarm of 96 drones and robot dogs to understand each other's intent without constant radio communication. That means the network can coordinate attacks even under full signal jamming or GPS denial. The whole thing is built to operate when the classic tools of electronic warfare are turned off. And the land systems are only one layer. At sea, there are unmanned L30 boats running at 65 kmh that can autonomously encircle and ram targets. In the air, there are laser cannons called Guang Jian, where one unit blinds drone swarms and another burns out the electronics on the highest priority threats. Algorithms handle the targeting hierarchy. The operator basically gets one button, confirm strike. Now, while China is stacking robots by the thousand, the AI behind all of this is the part they are keeping most quiet about. So, take the coordination claims with a bit of salt until somebody outside the CCTV edit bay sees it live. That said, this whole military robotics push is clearly not just a China story anymore. Europe is about to run one of the toughest realworld field tests for military robots anywhere in the world. Around 20 international teams are heading into the Swiss Army's Thun training area for LRO 2026, where unmanned ground vehicles and drones will be pushed through reconnaissance, transport, and search and rescue missions in rough natural terrain. And this matters because it is one thing to show a robot on a polished demo course. It is something else entirely to drop it into mud, uneven ground, unpredictable conditions and realistic mission pressure. No clean urban interiors, no carefully staged environment, just open terrain and militarystyle tasks. That is where mobility, sensing, autonomy and reliability all get exposed very quickly. So while China is showing off the future as a coordinated robotic combat network, Europe is basically building a public stress test for the same broader trend. The common thread is obvious. Military robotics is moving out of theory and into environments where failure actually means something. Now shifting over to something that feels like the start of a new pattern. AI agents are starting to run humanoid robots directly. A company called Humanoid ran an experiment where a cloud-based AI SAP's platform controlled a wheeled humanoid called HMD1 Alpha through the jewel agent layer. The robot received high-level business tasks and executed them autonomously inside a real messy warehouse. It found the right pallets, grabbed the boxes, and loaded them onto carts all on its own. The bigger idea is that corporate AI software will not just handle things like purchasing and scheduling. It will also operate physical robots. In the enterprise of the future, your company does not just get a digital brain. It gets a set of working hands to match. Speaking of that future, Jeff Bezos is betting heavily on it. He is launching a $100 billion investment fund aimed at buying up industrial companies in aerospace, chip manufacturing, and defense and rebuilding them around AI. And as preparation, he has been quietly snapping up AI startups. Amazon just bought a robotic startup called Fauna along with its humanoid called Sprout. Sprout is a compact bipeedal robot about a meter tall designed with a soft shell, no sharp joints, and a focus on social interaction. The pitch is that it is safe enough to learn in human environments. A week earlier, Amazon also picked up the company behind the river delivery robots. Bezos wants to automate last mile delivery and is not being shy about it. At the same time, leaked documents suggest Amazon is planning to replace up to 600,000 future job openings with AI and robots. So, while the public messaging is careful internally, this looks very much like a long-term workforce substitution play. If you thought robots were coming for jobs in 10 years, the answer is no. They are already in line. Now, while the humanoid market expands, that also means companies can actually shop around. BMW which had been testing figures robots is now running a new humanoid from a Swiss company called Hexagon. The robot is called AE and BMW is testing it at the Leipig plant on high voltage battery assembly and complex component production. AON is built for precision work over raw lifting. It moves on wheels at about 2.4 m/s which works out to almost 9 km hour. and that is several times faster than most walking humanoids. It handles parts up to 15 kilograms and swaps its own battery every four hours. The big takeaway is that features that seemed cutting edge a year ago are already becoming the baseline. Humanoids are starting to compete for jobs, not just attention. Now, let's talk about the money side because Mark Zuckerberg just announced he is spending $135 billion on a personal super intelligence. And if you assumed that means a super helpful assistant for you, think again. The real goal is hyperpersonalized advertising. The idea is to fuse top tier language models with Meta's social infrastructure to basically reinvent what a social network is. The AI will factor in your goals, hidden interests, habits, even your health indicators. Your feed will be generated in real time, shaped around your current mood, or whatever you searched for 10 minutes ago. Your account becomes a digital twin that knows you better than you know yourself. And to keep you from ever looking away, the interface moves off the phone screen and onto smart glasses. So, what do you think? Ultimate convenience or the end of private life? Drop it in the comments. Meanwhile, Elon Musk wants to end global dependence on chip suppliers. He just announced that Tesla, SpaceX, and XAI are building a vertical chip fortress in Texas on a site that is almost 9.5 km. The idea is a closed ecosystem from raw silicon all the way to finished processors all under one roof. The ambition is classic Musk. He is planning to run the 2nanmter process that the rest of the world is barely starting to adopt and crank out 1 terowatt of compute capacity per year. That works out to roughly a million silicon wafers a month, which would be about 70% of current global production. His budget estimate is $25 billion. Analysts already put the real number closer to 50 billion. If Musk actually pulls this off, it turns Tesla and XAI into companies that no longer depend on TSMC or anybody else for silicon. That is a level of vertical integration nobody has ever seriously attempted in the chip industry. But while the Fortress is still on paper, Tesla bot 3, which was supposed to drop in the first quarter, has been pushed back. That said, Tesla's recent hiring push tells a different story. They just opened a wave of production line job listings specifically for Optimus. The prototype phase looks like it is ending and the factory phase is starting. The internal plan is 50,000 robots this year. The first batches are not going on sale. They are being deployed inside Tesla plants, including the Texas Megaactory. There are also rumors that some will be running food service at Tesla diner in Los Angeles, which would literally mean humanoid robots walking orders out to your car. And now there is another piece that makes that scaling story more interesting. Tesla's China leadership just suggested that Giga Shanghai could become a major enabler for mass humanoid robot production. That matters because one of the biggest bottlenecks in this entire industry is not designing a humanoid that works once. It is manufacturing huge numbers of them reliably and cheaply. That is exactly where Shanghai becomes important. The plant already pushed out around 851,000 vehicles in 2025 and it is one of Tesla's most efficient production hubs anywhere in the world. So, when senior executives start saying the factory could help carry new products, including robots, that is not a random comment. It sounds a lot more like Tesla looking at its strongest manufacturing base and asking how fast it can turn that into an Optimus engine. And that also fits the broader shift inside the company. Musk has been trying to get investors to care less about cars and more about autonomy, robotics, and AI. If Shanghai really starts taking on robot production responsibilities, that would be one of the clearest signs yet that Tesla is serious about moving from a car company with robot demos to a company trying to massproduce humanoids at industrial scale. The other piece is Digital Optimus, a software agent from Tesla and XAI that can drive a robot in real time. The setup is almost elegant. Grock acts as the strategic brain. The AI4 chip, which costs about $650, handles fast reflexes, the way intuition works in humans. The system reads the last 5 seconds of screen video and then performs office level tasks like a real employee. Musk claims this combination will let digital optimists scale up to entire corporations. The project has a nickname that is an obvious shot at Microsoft, Macrohard. Now, let's move to a genuinely big breakthrough in robot training. Chinese scientists just taught a Unitry G1 humanoid to play a decent game of tennis. The news here is not the tennis. It is that they did it without a proper data set. Normally, training a humanoid on a dynamic task requires either a massive data set or a lot of handcoded motion. The team skipped both and got the robot playing in 5 hours. The technique is called latent. And inside a simulator, the robot experimented with angles, timing, and striking force on the fly. It taught itself to return shots over the net. The final result was about 90% success on forehand returns and close to 80% on backhand. Not Wimbledon level, but absolutely good enough to be a training partner for a beginner. And much more importantly, this opens up a new path for teaching robots to handle high-speed dynamic situations without manually labeled data. And Unitry is clearly not stopping at tennis. One of the weirder realworld clips making the rounds shows a customized Unitry G1 in Poland chasing wild boars through a parking lot and grassy roadside area. The robot called Edward Waki was jogging around trying to herd the animals back toward the forest while people filmed the whole thing. Now, to be fair, the boores were not exactly impressed. They mostly ignored it and wandered off. Still, the point is bigger than the clip itself. This is a humanoid operating in an uncontrolled public setting around animals, people, distractions, and all the usual chaos that makes the real world harder than any lab. It is also part of a growing idea that humanoids will not just work in factories. They will show up in public spaces, marketing campaigns, live events, and all kinds of situations where the value is half utility and half attention. And then there is the speed side. Unitry also just showed its H1 humanoid hitting up to 10 meters/s in a sprint test, which pushes it dangerously close to the pace behind Usain Bolt's 100 meter world record. Even if there is some measurement noise in the clip, the broader signal is hard to miss. Humanoids are getting a lot faster, a lot more agile, and a lot more physically capable in a very short time. That changes the way you think about the category. A few years ago, just getting a full-sized humanoid to walk cleanly was a headline. Now, one model is learning tennis in hours. Another is jogging around public spaces in Europe, and another is flirting with elite sprint speed. The pace of improvement is getting pretty serious. That is part of a bigger shift. China is trying to become the global center of humanoid robotics. And the country now has dozens of robot schools. These are not small labs. They are massive data centers around 10,000 square meters each where humanoids train in industrial scale physical work and not in simulation either. They are generating real physical training data from real tasks. Across 40 facilities, thousands of robots are practicing motor skills 24 hours a day. From carrying trays to assembling cars, using VR and motion capture to generate something like 6 million training recordings a year. And the commercial pressure is paying off. Unitry just filed for an IPO, meaning they are no longer a promising startup. They are an industrial operation with $248 million in revenue for 2025 and a target market valuation of around 7 billion. One more signal that this space is becoming a real business, a drone company called Lucid Drone Tech just hit $75 million in profit by renting robots and drones out on a subscription model. Cleaning companies sign up and the drones take jobs those crews could not handle before. They are not just washing skyscrapers. They can paint facades, seal joints, even clean sidewalks with dedicated ground units. In 2025, the company made more profit than it had earned total over the previous 7 years and scaled its fleet from 100 to 1,000 units. That is not futurism. That is a business that is already printing money. Then there is Boston Dynamics, or more precisely, the AI Institute that spun off from it. They took a two- wheeled robot and pushed its physical capabilities about as far as they go. The result is Roadrunner, a 15 kg machine that switches between locomotion modes depending on the task. Its legs are fully symmetric, bending in either direction, and they can straighten into a single line when needed. And of course, in true Boston Dynamics fashion, the balance is absurd. On the lighter side, Deep Robotics just showed off a robot horse, which is basically a creative take on the companion robot idea. It is a bit cute, a bit odd, and honestly, they should just build the adult-sized version already. Anyone want to ride one to the grocery store? So, Unitry just dropped something that genuinely made people stop scrolling. And when I say people, I mean Elon Musk literally typed cool on X after watching the footage. That's what we're kicking off with today because there's a lot to unpack here. And this story goes well beyond just one wild robot video. The machine is called the GD01 and Unitry is building it as the world's first productionready manned mecca. Not a concept, not a prototype behind glass at a tech expo. An actual product you can buy right now. Unitry confirmed to Wired that this is real and not an elaborate prank for 3.9 million yuan, which shakes out to somewhere between $573,000 and $650,000 depending on exchange rate. It stands 2.7 m or about 8.9 ft tall, weighs about 500 kg or roughly 1,100 lb with a person inside, and is constructed from high strength alloy specifically for civilian transport applications. The company is based in Hjo, China, and they are completely serious about this thing. The intro video is set to a thundering rock guitar soundtrack, and it opens with Unit's founder and CEO, Wang Shing Shing, walking up to the machine, literally holding its hand, and then climbing into the open air cockpit mounted in its chest area. The robot then walks in a fully upright bipeedal stance, stable, smooth, no wobbling, approaches a stack of cinder blocks, and absolutely demolishes it. Later in the video, even without a pilot on board, the GD01 smashes through a wall of bricks entirely on its own. Then it bends backward, folds its legs underneath it, reconfigures its entire chassis into a four-legged quadriped form within seconds, and just keeps moving across uneven terrain without any external assistance. In that crawling configuration, the pilot would essentially be lying flat on their back staring at the sky, which is a genuinely absurd visual, but at that point, honestly, who cares? Wang sitting inside really hammers home the scale. The GD01 stands at roughly 1.6 times the height of an average adult. The system demonstrates stable bipeedal walking, high force output capable of toppling walls, and maintains rigid structural integrity even under heavy impact. It dynamically adjusts its center of gravity during the mode switch, which is not a small engineering achievement. Unitry released very limited technical specs alongside this and issued a safety notice urging users not to attempt hazardous modifications or extreme tests, openly acknowledging that humanoid robotics remains in an early experimental stage with real functional limitations for personal users. The internet's reaction was exactly what you'd expect. Beyond Musk's brief endorsement, a YouTube commenter wrote that the US makes cool robots in Hollywood movies while China makes practical ones in real life. An American journalist in China named Jason Smith posted on X that China is building everything we dreamed of as kids and that China is way, way, way ahead of the rest of the world. A user from Texas was more pointed, raising concern about competitive pressure and noting that China can produce these systems for roughly 20% of US costs. That number actually matters a lot for where this industry is heading. Chen Jing, vice president of the technology and strategy research institute told the Global Times that the GD01 shows China has crossed a key engineering threshold in embodied AI. His core point is that this machine is no longer confined to a lab. It has a price tag and a commercialization roadmap. He also pushed back on the assumption that humanoid form is the ultimate endpoint for robots, suggesting the future might involve more diverse human machine hybrid configurations. The bigger shift he describes is robots moving from being tools to becoming mobility platforms. Once a robot can carry a human and perform tasks, it stops replacing labor and starts extending human capability, similar to how cars and airplanes transform mobility entirely. At a cultural level, Chen said it completes the loop for the sci-fi generation and strategically challenges the narrative that China only follows rather than leads in deep technology. Now, Unitry as a company is worth understanding in context. You've probably already seen their G1 humanoid in clips doing acrobatics, dancing, and performing kung fu online. A few months ago at China's televised spring festival, Unitry robots performed synchronized parkour and martial arts routines using wireless communication between units to stay precisely in step. That multi-root coordination was a capability they were specifically showing off at scale. The company also recently launched a lowcost upper body humanoid starting at $26,900 yuan, about $4,290, featuring modular deployment options and up to 31 degrees of freedom aimed at research, light industry, and service applications. Their entry-level R1 humanoid runs about $6,000, and the cheapest G1 model comes in around $15,000. US-made humanoid robots routinely cost 10 times more. All of this, the G1, R1, and the Goto Robot Dog, is sold internationally through Alibaba's AliExpress, targeting North America, Europe, and Japan. The scale of Unit's output is also striking. They reportedly shipped more than 5,500 humanoid robots last year. For comparison, Tesla, Figure AI, and Agility Robotics each shipped around 150 units in that same time frame. According to Omdia, Chinese companies accounted for nearly 90% of global humanoid robot sales in 2025. Rival Agibbot sells a humanoid at around $14,000, while Musk has estimated that Tesla's Optimus could eventually fall between $20,000 and $30,000. Musk also said at the World Economic Forum in January that China is very good at AI, very good at manufacturing, and will definitely be the toughest competition for Tesla, adding that he doesn't see significant competitors outside of China. It's worth noting though, as Wired pointed out, that the GD01 is more geared toward generating publicity and demonstrating destructive force than everyday utility. Unit's humanoids aren't particularly dextrous yet and don't have the AI needed for genuinely complex real world tasks. That's a fair counterpoint to keep in mind. The cost advantage largely comes down to supply chain. Tech analyst Ma Jihao explained that China is the only country in the world with all major industrial categories domestically available. High performance motors, reducers, sensors, batteries, carbon fiber materials, all strong, all accessible. The International Federation of Robotics released a report on May 5th showing that 64% of industrial robots in the global electronics industry are installed in China and Chinese manufacturers supply 59% of that sector globally. In metal and machinery, Chinese robot suppliers hold an 85% domestic market share. Tesla has been leaning into this, too. The South China Morning Post reported the company has been engaging hundreds of Chinese component suppliers for Optimus since at least 3 years ago with some involved in actual R&D and hardware design, including curved glass head prototypes delivered in recent months. Unitry systems have also started appearing in real operational environments with Japan Airlines running humanoid trials at Tokyo's Haneda airport using Unitry and UB robotics machines. Morgan Stanley recently said China's early lead in humanoid robots could drive the next phase of its global manufacturing and export dominance. Unitry is also heading toward a public offering. In March, they filed for an IPO on Shanghai's Star Market, planning to raise around 4.2 billion yuan with roughly 85% earmarked for R&D, including more than 2 billion yuan specifically for robotics model development. Shifting over to figure AI now because they just released a demo that's impressive in a completely different way. Two humanoids, both running the company's Helix 02 AI model, walk into a minimalist bedroom and reset it entirely in under 2 minutes fully autonomously. We're talking opening doors, hanging a coat, closing a laptop, putting headphones away, disposing of trash, repositioning furniture, and then making the bed together. And the coordination here is the key technical detail. They share nothing with each other directly. No shared planner, no central controller, no messages between units. Each robot has only its onboard cameras and its learned policy. The two signal intent through subtle head nods while working on the comforter together. And beyond that, they're reading each other purely through movement with every action changing the room's state and forcing both robots to continuously re-evaluate and adapt in real time. The comforter was specifically the hardest challenge. According to Figure, unlike rigid objects, fabric has no fixed geometry and no stable grasp points. It folds, stretches, and shifts under tension as both robots pull from different positions. They had to predict each other's next moves while simultaneously adjusting grip, posture, and motion as the material kept changing shape. The robots also balanced dynamically on one leg, operated foot pedals, and transition between tasks with no scripted handoffs. The whole system is driven by a single vision language action framework trained endtoend through reinforcement learning and simulation using heavily randomized terrains and conditions. And those behaviors transfer directly to the real world without additional calibration, which directly addresses the simtoreal gap that's been one of robotics's hardest persistent problems. The Helix system also got a meaningful upgrade recently. Previously, it relied only on proprioception. The robots understood their own joint positions and movement, but weren't visually interpreting their surroundings while in motion. The new version processes stereo camera input and converts RGB imagery into a real-time three-dimensional spatial map of the environment, meaning the robots are now simultaneously seeing and feeling the terrain beneath them. This enables better stability on stairs and uneven surfaces even as lighting conditions shift. On the manufacturing side, Figure says they've ramped Figure03 production at their bot Q facility in California from one robot per day to one robot per hour within 4 months. Then there's physical intelligence, a San Francisco startup approaching the problem from a fundamentally different angle. Rather than building robots for specific tasks, their goal is to build a generalpurpose foundational model, a single robot brain that can adapt across different hardware and handle the unpredictability of real world environments. They keep coming back to something called Moravec's paradox, which robot researcher Hans Moravec described in 1988. The observation that things easy for computers like arithmetic and chess are genuinely hard for humans. While things trivial for humans like walking, grasping objects or folding clothes are extremely difficult for machines. Nearly 40 years later, that remains the central obstacle in robotics and physical intelligence is trying to close it entirely. The company was co-founded by Lackey Groom, originally from Perth, Australia. He flew to Silicon Valley at 17, spent six years as an early employee at Stripe, and then in 2023, when several top scientists from DeepMind's robotics team, including Carl Hman, Sergey Lavine, and Chelsea Finn, were planning to leave, he joined them to build physical intelligence. They've raised over a billion dollars and currently sit at a $5.6 billion valuation. Walking through their office, you see a coffee-making robot picking up the handle, loading it into the machine, and pressing the button in one smooth, uninterrupted sequence. Next to it, a robotic arm folds a pair of shorts it has never handled before, working from knowledge built through tea operation data. Lackey actually admitted in an interview that he's bad at folding clothes himself, so having the robot do it is genuinely useful to him. Other setups tackle peeling fruit and assembling packages. The open plan office has researchers discussing algorithms on one side and data collectors running tea operation demonstrations on the other. All feeding a continuous cycle of hypothesis, data collection, model training and evaluation on the same robots. Their model has gone through three major versions. Pi zero focused on proving functionality, getting robots to handle tasks previously considered out of reach with clothes folding as the anchor since tea operation data is easy to collect because everyone already knows how to do laundry. PI 0.5 focused on generalization and they got a result that surprised even the team. Training across roughly 100 home environments was enough for the model to generalize to a 100 home it had never seen. They expected to need thousands of environments, maybe millions. Their current version pushes toward performance and reliability, targeting high success rates on laundry, coffee prep, and package assembly consistently. Locky is honest that this isn't solved yet, and they can't claim to be close, but the pace has been striking regardless. Philip Clark at Thrive Capital, who invested when the company was four people working in a living room, said that great investments come down to two things: people and timing. The team here includes worldclass researchers, and the timing is right. Go back even 10 years, and the underlying model capabilities simply weren't there yet. His take on their progress is that it's been two to three times faster than his most optimistic projections. He thought it would take 3 to 5 years to reach current capability. It took about 18 months. Lackey's framing for where the field is right now is the GPT2 moment, not GPT4, not GPT5. Signs of real life, genuine potential, but significant scaling still needed before it's useful for most people globally. For the next 1 to 3 years, he thinks enterprise level deployment is within reach with a broader consumer product wave to follow. His longer term vision is straightforward. Robots should handle the work humans don't want to do. the boring, dangerous, repetitive, meaningless tasks people do out of necessity rather than choice. Free people from that and they can spend their time on things that actually matter to them. It's a weird time to be watching this space, but it's a good weird. Also, if you want more content around science, space, and advanced tech, we've launched a separate channel for that. Links in the description. Go check it out. Anyway, that's it for this one. Let me know what you think about Unitre's new robot and whether you'd actually step inside one if you had the chance. Thanks for watching and I'll catch you in the next one.