Tag Archives: Robots

KEENON Humanoid Pours Drinks at GCS 2026, 100,000 Others Run Hotels

Posted on by

The robotics industry has a curious reputation problem. The machines getting the most attention, walking bipeds that do backflips, aren’t the ones driving real business value. By 2030, professional service robots are projected to account for $90 billion of a $161 billion global market, growing at 24.6% annually. That makes them the fastest-growing segment in robotics, which is a fact that barely makes the news.

KEENON Robotics has known this for a while. Founded in 2010, the Shanghai-based company didn’t wait for the hype cycle; it built the service delivery robot category from the ground up. Today, with over 100,000 units shipped across more than 60 countries, KEENON holds the number-one global market share in commercial service robots for the third consecutive year, according to IDC 2025.

Designer: KEENON Robotics

At Global Connect Show 2026, the star of KEENON’s booth was the XMAN-R1, a wheeled humanoid robot that makes popcorn, pours drinks, and hands out snacks. It’s the kind of demo that stops foot traffic, and it’s meant to. Underneath the theatrics, though, is a robot packing 275 TOPS of AI processing power, dual 7-DoF arms, and precision dexterous hands built for human-level manipulation.

What the demo doesn’t show is how much work went into teaching a robot to grab a cup. KEENON estimates that a single action of that kind requires at least 1,000 data points. A full coffee-making sequence demands over 20,000. That gap between what looks effortless and what it actually costs computationally is one of the clearest explanations of where physical AI sits right now.

KEENON is remarkably candid about this. Their own assessment puts the current “mind age” of humanoid robots at roughly three years old, which, if you think about it, explains a lot about why they move so deliberately. True general-purpose humanoid deployment is still at least five years out by their estimation. The “Model T” framing they use is apt; these are early machines, not finished ones.

That candor is also what makes KEENON’s established product line feel more credible. While the XMAN-R1 gets the headlines, KEENON’s delivery and cleaning robots have been running inside hotels, restaurants, hospitals, airports, and casinos across more than 600 cities globally. Their DINERBOT T10 can carry up to 40 kg, fits through a 59 cm passage, and operates for up to eight hours on a single charge.

A good example of what that looks like in practice is Shangri-La Hong Kong, where KEENON runs six different robot types across eight units within a single hotel. Delivery bots handle contactless room service; cleaning robots run scheduled cycles through lobbies and corridors; logistics carriers shuttle linens and supplies behind the scenes. None of this required the hotel to restructure its operations around the robots.

Part of why that integration works comes down to a deliberate design choice. KEENON chose not to make its service robots look human. The compact, rounded bodies, soft voices, and animated screen faces are intentional, because how a robot looks determines whether people trust it. Western audiences carry Terminator-shaped anxieties; Asian audiences grew up with characters like Doraemon. The design has to work for both.

The XMAN-R1 at the Global Connect Show is KEENON’s way of signaling where the product line is heading. Alongside it, the company also offers the bipedal XMAN-F1, a full-body humanoid with 43 degrees of freedom that takes the same task-driven approach. Both run on KEENON’s KOM 2.0 platform and are designed to work alongside existing robots like the DINERBOT T10 rather than replace them.

What KEENON actually has over the wave of humanoid startups entering the market is something harder to replicate than a robot body. With 100,000 units running across 70+ countries, the company has been accumulating proprietary operational data at a scale most competitors haven’t even started to approach. Every delivery, every corridor, every cup poured feeds back into that picture, one data point at a time.

The post KEENON Humanoid Pours Drinks at GCS 2026, 100,000 Others Run Hotels first appeared on Yanko Design.

Every Robot You’ll Ever Own Has 3 Separate Brains: Nvidia VP explains how AI Thinks at BEYOND Expo 2026

Posted on by

A robot on a factory floor may look self-contained, but Deepu Talla says its intelligence is distributed across a hidden chain of machines. At BEYOND Expo 2026, the NVIDIA executive broke robotics down into a deceptively simple formula: three computers. One handles the heavy lifting of training the robot brain, another tests that brain in simulation, and a third lives inside the physical robot, making decisions in real time.

It is a framework that helps explain why robotics has moved so slowly, and why the field suddenly feels ready to accelerate. In language that cut through the usual keynote fog, Talla argued that AI in the physical world plays by harsher rules than chatbots or image tools. A text model can be 95 percent right and still be useful. A robot moving through a warehouse, a street, or a hospital has to perform with a completely different standard. In human terms, it is a little like splitting intelligence into learning, dreaming, and reacting, then assigning each function to a different machine.

That first machine is where the robot’s intelligence is forged. Talla described it as the computer used to train the robot brain, the heavy compute layer where models absorb data, patterns, and behaviors at massive scale. This is where a machine learns how the physical world works, long before it ever enters one. If that sounds abstract, the second computer makes it easier to picture. This is the simulation layer, the place where a robot rehearses reality in a safer, faster, cheaper environment, running through scenarios again and again until its behavior becomes reliable enough to trust.

The third computer is the one that actually lives inside the robot. It is the real-time brain, the system that has to perceive the world, make sense of it, and respond instantly. This is where Talla’s argument becomes especially sharp. In digital AI, a model can get close and still be useful because a human can smooth over the rough edges. In robotics, the rough edges are where accidents happen. A machine moving through a factory, a roadway, or a hospital has to work with a far tighter tolerance for error, because the physical world offers fewer second chances.

That is also why NVIDIA sees robotics as far bigger than a niche category. Talla pointed out that almost 80 percent of the world’s GDP sits in physical industries like manufacturing, logistics, retail, and transportation. These are sectors where intelligence has to leave the screen and interact with objects, spaces, and people. NVIDIA’s role, in his telling, is to provide the underlying architecture for that shift. The company may not build robots itself, but it wants to supply the stack beneath them, from training infrastructure and simulation tools to the compute that powers action on the edge.

The post Every Robot You’ll Ever Own Has 3 Separate Brains: Nvidia VP explains how AI Thinks at BEYOND Expo 2026 first appeared on Yanko Design.

FutureWave’s Furny home robot talks to you by moving, not by talking

Posted on by

Most home robots ask a lot from the room they inhabit. They arrive with screens, speakers, wake words, and personalities, all requiring acknowledgment from whoever happens to be nearby. The interaction model is fundamentally borrowed from smartphones: alerts, prompts, and responses delivered through layers of interface. The result is a machine that demands attention in a space that already has more than enough competition for it.

Brussels-based studio Futurewave took a different position with Furny, a domestic robot concept presented at the last Milan Design Week 2026. Rather than building something with a face, a voice, and a screen, the team asked whether a robot in the home could communicate through posture and movement alone, the way furniture communicates presence and purpose without saying anything at all.

Designer: Futurewave

The answer is a furniture-sized object with a movable head that expresses itself entirely through physical behavior. When something happens nearby, the head tilts. When the robot is ready to act, it orients toward the task. When it’s waiting, it recedes into a posture that reads as neutral, almost still. The timing, direction, and intensity of each shift are calibrated to communicate specific states: focus, readiness, attention, and reaction. There are no pixels involved in any of it.

Deliberately avoiding humanoid proportions was a foundational decision. Furny doesn’t mimic the way a person or animal moves. The gestures it uses are abstract enough to feel designed rather than imitated, which makes them easier to read in context without triggering the uncanny valley that tends to follow robots built on biological templates. The visual restraint also helps it belong in a room. It reads as an object with behavior, rather than a character out of place.

The research behind the project draws on work in expressive movement design for non-anthropomorphic robots, a field that looks at how physical states and intentions can be conveyed through spatial behavior without resorting to screens or voice. Furny’s head doesn’t speak for it. The way the body holds itself does. The robot signals what it’s about to do before it does it, which is a meaningful distinction from machines that simply act and leave the explanation for an app notification afterward.

Futurewave also built Furny within a manufacturable framework, which separates it from most conceptual robot work. The project integrates industrial design, embedded electronics, and software-controlled motion systems in a way that points toward practical production rather than exhibition only. That framing is important because the most interesting thing about Furny isn’t the movement vocabulary itself but the argument it makes about what a domestic robot is supposed to be.

The prevailing assumption has been that robots become more useful as they become more capable of mimicking human interaction. Furny pushes back on that. A robot that remains quiet when nothing’s needed, reads the room through its posture, and signals intention before acting doesn’t interrupt the household. It becomes part of it, the way a good piece of furniture does, present and purposeful without drawing attention to itself until the moment calls for it.

The post FutureWave’s Furny home robot talks to you by moving, not by talking first appeared on Yanko Design.

Argus Just Showed Up With 20 Eyes, 20 Legs, and No Rules

Posted on by

The moment you see Argus rolling across a college lawn, you feel a kind of awe that’s equal parts scientific admiration and mild existential discomfort. It doesn’t look like a robot. It doesn’t look like anything you’ve seen before, actually. It looks like a sea urchin crossed with a fever dream, or if you’ve spent any time on the internet in the last few years, it looks exactly like what happens when someone renders a biblically accurate angel and sends it out to navigate uneven terrain.

That’s not an exaggeration. The internet made the comparison almost immediately after Duke University’s General Robotics Lab unveiled Argus, and the parallel holds up. In the Book of Ezekiel, the ophanim, a type of divine being, are described as wheels covered in eyes, seeing in all directions simultaneously. Argus, named after the Greek mythological giant with a hundred eyes, does essentially the same thing, minus the divine mandate. It has 20 legs, each one telescoping and tipped with a camera, arranged at the vertices of a regular dodecahedron. No blind spots. No preferred orientation. No front or back.

Designer: Duke University General Robotics Lab

That last part is what keeps pulling me in, design-wise. We’ve spent decades building robots that mirror the logic of our own bodies: two legs, bilateral symmetry, a definitive forward direction. It made intuitive sense. We move front-to-back, so we assumed machines should too. Argus rejects that assumption entirely. The team at Duke built it around a principle they’re calling dynamic symmetry, which refers to how uniformly a robot can accelerate in any direction. Most robots are strongest and most efficient when moving the way they were designed to move. Argus has no such preference. It moves sideways, backward, forward, and diagonally with the same ease, which sounds like a minor technical distinction until you watch it roll through rough terrain, navigate around trees, and absorb collisions without losing its course. That’s when you realize how significant the gap is.

The design precedent here matters more than it might seem. Robotics has long borrowed from nature by mimicking the shapes that evolution produced: bipedal forms for humanoids, quadruped frames for terrain bots, insect geometries for swarm machines. But Argus is borrowing something different from nature. It’s borrowing from the radial logic of starfish and sea urchins, creatures that don’t have a front because every direction is equally valid. The Duke researchers describe Argus as an “existence proof,” a demonstration that a robot built for dynamic symmetry isn’t just theoretically interesting but practically deployable. Postdoctoral researcher Boxi Xia put it directly: “It produces a robot you can deploy in the wild, on uneven ground and in clutter, even in low-gravity settings.”

Low-gravity settings. That detail is doing a lot of quiet work in this conversation. The practical applications being discussed range from disaster response and search-and-rescue operations to planetary exploration, environments where the rules of conventional locomotion break down fast and all-directional agility becomes the difference between success and failure. A humanoid robot in a collapsed building still has to worry about which way it’s facing. Argus doesn’t.

I’ll admit the design is deeply strange to look at. It is not sleek. It is not elegant in any conventional sense. It doesn’t have the clean industrial confidence of Boston Dynamics’ machines or the deliberate anthropomorphism of recent humanoid models. It looks a little chaotic, frankly, like it was assembled by someone working from a very different set of aesthetic values, someone less interested in how the thing looks than in what the thing can do. And maybe that’s the point. Beauty in engineering doesn’t always wear the shape we expect. Sometimes it rolls across a lawn on 20 legs, sees absolutely everything, and changes the conversation entirely.

Argus is the kind of design that reminds you why robotics is still worth watching. Not because of what it looks like, but because of what it means for how we think about movement, perception, and the assumptions we’ve been quietly building into machines all along.

The post Argus Just Showed Up With 20 Eyes, 20 Legs, and No Rules first appeared on Yanko Design.

Michigan Built a $4,000 Robot You Can Rebuild from Scratch

Posted on by

The Ship of Theseus is one of philosophy’s most enduring thought experiments: if you replace every plank of a ship, one by one, is it still the same ship? Researchers at the University of Michigan decided that rather than debating the question in a classroom, they’d build it. And then they’d unbolt it, swap the legs, and build it into something else entirely.

TROT (The Robot of Theseus) is a 10-kilogram, four-legged robot whose entire identity rests on impermanence. Its limbs unbolt. Its leg configurations swap between a two-link hopper, a three-link knee, and a three-link elbow orientation. You can rebuild the entire body plan over an afternoon and walk away with something that moves more like a gazelle than the dog-sized quadruped it started as. Same chassis. Same motors. About $4,000 in 3D-printed brackets and off-the-shelf parts. Its backdrivable motors even recover energy as they’re driven backward, mimicking the way tendons store and release force in a running animal.

Designer: University of Michigan

That $4,000 figure is worth sitting with. For context, Boston Dynamics’ Spot runs closer to $75,000. But TROT isn’t a budget Spot. It’s a different idea entirely. Where Spot is optimized for a fixed body plan and real-world deployment, TROT is optimized for being taken apart. It’s an experiment in the value of non-permanence, and that’s a much more interesting design brief than “make it do more things.”

The team, led by assistant professor Talia Moore, designed TROT to help biologists ask questions that physical animals can’t easily answer. What makes a cheetah fast isn’t just muscle. It’s also leg length, segment ratios, and joint geometry. Isolating those variables in a living animal is nearly impossible. But with TROT, you can swap out a femur extension, flip the knee orientation, and run the same locomotion test again the same afternoon, with consistent hardware and no ethical review board required. The robot has been used to compress roughly 60 million years of evolutionary locomotion variation into weeks of lab data. That’s the actual scientific utility, not a metaphor.

What tends to get under-reported in the science coverage is the design language itself. TROT’s visual aesthetic isn’t cleaned up or consumer-ready. You can see the 3D-print layer lines, the exposed wiring, the actuators bolted directly to the brackets. It looks like something built to be understood rather than admired, and I think that’s intentional. The exposed construction is a form of communication. It tells anyone looking at it: this is not precious. Change it. That’s a genuinely rare posture for a piece of research hardware.

The open-source dimension also runs deeper than posting a GitHub repo. The team released the CAD files, not just the control code. That’s a meaningful distinction. Code describes behavior; geometry describes intent. Sharing the brackets and print files means a biology lab at a smaller institution can reproduce TROT without needing a dedicated robotics engineering team. The knowledge transfer is embedded in the shape of the parts, and that changes who can participate in this kind of research.

TROT didn’t arrive alone. The first quarter of 2026 brought a quiet cluster of modular robotics research: Northwestern’s terrain-adapting writhers, a self-configuring quadruped paper in PNAS, and Nature’s SoftRafts, all landing within roughly eight weeks of each other. Robotin debuted a modular home robot ecosystem at CES 2026. Analysts have put the modular robotics market on track for $18.94 billion by 2029. None of this is coincidental. The field has been asking whether modularity in robotics could move past novelty. Q1 2026 looks like the answer arriving.

Most robots are designed to be finished. They ship in a fixed form, and any change is a cost: a repair, a retool, a failure. TROT is designed around the opposite logic. Its value increases each time a limb is swapped. Its usefulness is inseparable from its willingness to be reconfigured. Whether a robot that constantly changes its parts stays the same robot is still a philosophical question. Whether that approach produces better science, and better design thinking, is looking less and less like a question at all.

The post Michigan Built a $4,000 Robot You Can Rebuild from Scratch first appeared on Yanko Design.

Michigan Built a $4,000 Robot You Can Rebuild from Scratch

Posted on by

The Ship of Theseus is one of philosophy’s most enduring thought experiments: if you replace every plank of a ship, one by one, is it still the same ship? Researchers at the University of Michigan decided that rather than debating the question in a classroom, they’d build it. And then they’d unbolt it, swap the legs, and build it into something else entirely.

TROT (The Robot of Theseus) is a 10-kilogram, four-legged robot whose entire identity rests on impermanence. Its limbs unbolt. Its leg configurations swap between a two-link hopper, a three-link knee, and a three-link elbow orientation. You can rebuild the entire body plan over an afternoon and walk away with something that moves more like a gazelle than the dog-sized quadruped it started as. Same chassis. Same motors. About $4,000 in 3D-printed brackets and off-the-shelf parts. Its backdrivable motors even recover energy as they’re driven backward, mimicking the way tendons store and release force in a running animal.

Designer: University of Michigan

That $4,000 figure is worth sitting with. For context, Boston Dynamics’ Spot runs closer to $75,000. But TROT isn’t a budget Spot. It’s a different idea entirely. Where Spot is optimized for a fixed body plan and real-world deployment, TROT is optimized for being taken apart. It’s an experiment in the value of non-permanence, and that’s a much more interesting design brief than “make it do more things.”

The team, led by assistant professor Talia Moore, designed TROT to help biologists ask questions that physical animals can’t easily answer. What makes a cheetah fast isn’t just muscle. It’s also leg length, segment ratios, and joint geometry. Isolating those variables in a living animal is nearly impossible. But with TROT, you can swap out a femur extension, flip the knee orientation, and run the same locomotion test again the same afternoon, with consistent hardware and no ethical review board required. The robot has been used to compress roughly 60 million years of evolutionary locomotion variation into weeks of lab data. That’s the actual scientific utility, not a metaphor.

What tends to get under-reported in the science coverage is the design language itself. TROT’s visual aesthetic isn’t cleaned up or consumer-ready. You can see the 3D-print layer lines, the exposed wiring, the actuators bolted directly to the brackets. It looks like something built to be understood rather than admired, and I think that’s intentional. The exposed construction is a form of communication. It tells anyone looking at it: this is not precious. Change it. That’s a genuinely rare posture for a piece of research hardware.

The open-source dimension also runs deeper than posting a GitHub repo. The team released the CAD files, not just the control code. That’s a meaningful distinction. Code describes behavior; geometry describes intent. Sharing the brackets and print files means a biology lab at a smaller institution can reproduce TROT without needing a dedicated robotics engineering team. The knowledge transfer is embedded in the shape of the parts, and that changes who can participate in this kind of research.

TROT didn’t arrive alone. The first quarter of 2026 brought a quiet cluster of modular robotics research: Northwestern’s terrain-adapting writhers, a self-configuring quadruped paper in PNAS, and Nature’s SoftRafts, all landing within roughly eight weeks of each other. Robotin debuted a modular home robot ecosystem at CES 2026. Analysts have put the modular robotics market on track for $18.94 billion by 2029. None of this is coincidental. The field has been asking whether modularity in robotics could move past novelty. Q1 2026 looks like the answer arriving.

Most robots are designed to be finished. They ship in a fixed form, and any change is a cost: a repair, a retool, a failure. TROT is designed around the opposite logic. Its value increases each time a limb is swapped. Its usefulness is inseparable from its willingness to be reconfigured. Whether a robot that constantly changes its parts stays the same robot is still a philosophical question. Whether that approach produces better science, and better design thinking, is looking less and less like a question at all.

The post Michigan Built a $4,000 Robot You Can Rebuild from Scratch first appeared on Yanko Design.

The Robots That Built Seoul’s Robot Museum

Posted on by

The Seoul Robot & AI Museum is the definitive parametric architecture reference of 2026, and it’s easy to understand why the design world keeps returning to it. Every few years, a building comes along that doesn’t just represent a movement; it is the argument. RAIM is that building right now, and the reason has less to do with how it looks and more to do with how it got made.

Opened in 2024 in the Chang-dong district of northeast Seoul, the museum was designed by Turkish studio Melike Altınışık Architects. From the street, it reads like something that landed rather than was built: a spherical, mirror-finish shell that catches the sky and refuses to look like any cultural institution you’ve encountered before. The facade is wrapped in 3,422 double-curved metal panels, each one a unique geometry, each one positioned according to a structural logic you can actually read from the outside. The gridded surface pattern isn’t decorative. It follows the structural steel grid concealed behind it, making the building’s skeleton visible through its skin. That level of architectural honesty is rarer than it should be.

Designer: Melike Altınışık Architects

The geometry didn’t come from sketching. Melike Altınışık and her team scripted the form parametrically, then reverse-engineered the entire envelope to make it buildable. That second part is where most parametric ambitions historically die. Double-curved panelization at this scale is the kind of thing that gets value-engineered into something flatter and sadder during construction documentation. But Melike Altınışık Architects designed specifically for fabrication from the start, using a methodology called DFMA (Design for Manufacture and Assembly), which meant the form and the production method evolved together rather than fighting each other.

The fabrication pipeline is where the story gets genuinely interesting. The panels were cut using laser CNC machines and welded using industrial robots. On-site, 3D scanning ensured alignment that human measurement couldn’t consistently achieve at that tolerance. What this unlocks, practically, is that double-curved metal panelization stops being a budget line reserved for landmark commissions and becomes something mid-scale cultural buildings can actually afford. Robot welding doesn’t get tired. It doesn’t accumulate small errors across 3,422 repetitions. The precision holds, and holding precision across a spherical envelope is a very different proposition from getting it right once.

Now layer in the subject matter of the museum itself. RAIM is dedicated to robotics and artificial intelligence. Its permanent exhibitions trace the evolution of AI from predictive fraud detection systems to generative models. Robots greet visitors at the entrance. The interior reads like a spaceship, with a vertical exhibition tunnel at the building’s center blurring the boundary between the physical and the technological. So when you consider that robots also assembled the facade above your head, the recursion is almost too neat. Architect Altınışık framed it clearly: the architecture is “both shelter and pedagogy.” The building doesn’t just house the argument. It makes it.

Parametric facades are having a genuine cultural moment in 2026, and it’s not limited to the usual European flagships. Studios in South Korea and India are pushing computational design into more projects, and the international awards circuit is beginning to reflect that geographic shift. The conversation has moved from “can parametric architecture actually be built?” to “what does it cost, and who controls the pipeline?” RAIM answers both questions at once, which is probably why it’s the reference point of record for this particular moment.

That shift is worth paying attention to. For decades, the most ambitious architectural geometries required either enormous budgets or a willingness to absorb serious construction risk. Robotic fabrication and CNC manufacturing are quietly changing that calculus. In Altınışık’s own words, “the division between design and construction is becoming obsolete. The parametric model becomes not just a design tool but a construction platform.” The next wave of museums and civic buildings won’t choose simpler geometry because they have to. They’ll choose the complex version because their fabricators can deliver it, and because, as RAIM proves, the building becomes a more interesting object for it. Seoul’s robot museum was built by robots. The next one might be anywhere.

The post The Robots That Built Seoul’s Robot Museum first appeared on Yanko Design.

Unitree’s transforming mecha robot is the closest thing yet to owning a real Transformer

Posted on by

Flying solo tied to a jetpack is a dream we have been savoring for quite some time now. And since movies like the Transformers, the idea of transforming Autobots has been another craze humanity is chasing for various applications. In China, robotics company Unitree has made the idea of humans piloting transforming ‘mecha’ robots a reality. The first glimpse of this is making rounds on the internet and has had netizens hailing this engineering breakthrough, which makes science fiction a reality.

Meet the Unitree GD01, the world’s first ‘production-ready manned mecha built for industrial use. It measures roughly 2.7 meters tall and weighs over 500 kg. The GD01 is the successor to Unitree robots such as the dog companion, but Unitree does not reveal many details about the robot.

Designer: Unitree

What we have as the source is a video doing the rounds on the internet, which shows how the GD01 can transform from a two-legged humanoid into a four-legged crawler. The machine with human-like legs and arms with hands is remotely controlled, but can also accommodate a person in its torso, who can control the transforming humanoid in the style of a mecha. Mecha is different from an autonomous robot as they are piloted from a cockpit inside. They have been popularized in Japanese anime, but it’s in China that they’re getting a realization for real-world applications.

The durable alloy robot is designed to transport a person and to be used in high-risk and harsh environments. It can walk like a humanoid robot in its red and gray avatar. The demonstration video shows Wang Xingxing, founder and CEO of Unitree Robotics, climbing into the torso-mounted cockpit of the GD01 mecha to maneuver it. We can also see the robot knocking down a brick wall with its hand before transitioning into a four-legged robot.

According to initial information, the Unitree Robotics GD01 will start at $650,000, which easily makes it the most expensive humanoid in the competition. For comparison, the previous Unitree models like R1 only cost about $6,000. Of course, the price tag is owing to the functionality of the GD01, which can be used for various applications, “mainly aimed at changing the way we work,” the company notes.

In the video, we can see the GD01 walking on flat surfaces, but it can be assumed that it will be able to maneuver different terrains in the near future. China and Unitree are leading the way in the production of capable humanoids, which prompts us to make such assumptions. According to a report from research firm Omdia, China accounts for nearly 90 percent of global humanoid sales in 2025.’ Amid those sales, it is worth noting that Unitree alone has shipped upward of 5,500 humanoid robots in 2025, the South China Morning Post reports.

 

The post Unitree’s transforming mecha robot is the closest thing yet to owning a real Transformer first appeared on Yanko Design.

MIT Turned 12 Labubu Heads Into a Robot and It’s Watching You

Posted on by

Nobody told MIT grad students to build a rolling sphere covered in twelve Labubu faces. They did it anyway, and now the rest of us have to sit with that. The project is called Labububot, and it comes from three graduate students at the MIT Media Lab: Miranda Li from the Personal Robots group, and Jake Read and Dimitar Dimitrov from the What’s Taking Form group. Together, they took the internet’s favorite ugly-cute collectible, multiplied it by twelve, and fused everything into a single spherical body that rolls around following people through hallways. The official description calls it “one of the rarest monsters on Earth,” and that phrasing alone tells you everything about the tone the team was going for.

The design is not subtle. Twelve identical Labubu faces stare outward from every angle simultaneously. When the thing moves, it does so with that particular brand of slow, deliberate motion that robots somehow always use when they want to feel unsettling. The MIT team leans into every bit of that discomfort, which is exactly what separates Labububot from most social robotics research you’ll come across.

Design: MIT Media Lab

Social robots usually chase approachability. They get rounded edges, pastel palettes, and soft digital expressions designed to lower your guard on contact. The whole field runs on the logic that comfort builds connection, and most research in this space reinforces that assumption without questioning it much. Labububot rejects that premise entirely. It is meant to provoke a reaction before it earns one, and the reaction it tends to get first is somewhere between amusement and mild dread. That’s a deliberately chosen emotional space, and it works.

The Labubu connection makes this sharper than it might otherwise be. The original toy built its following on a very specific kind of ugly-cute tension. It’s not conventionally adorable. It has sharp teeth, wide eyes, and a design that sits right at the border of charming and unsettling. That’s precisely why it resonated. The blind-box format added a layer of collector obsession on top, and after BLACKPINK’s Lisa was seen collecting them, the whole thing escalated into cultural phenomenon territory fast. The fact that it already carried that complicated emotional charge before MIT ever touched it makes the robot version feel like the natural next step, even if nobody saw it coming.

Scaling that same energy up to twelve faces on a rolling robot body is not an accident. The MIT team is clearly aware of what they’re working with. The official framing pitches Labububot as a “playful critique of social robots” and poses a question worth taking seriously: what do the monsters we make reveal about the monsters we are? For a project built around a pop culture collectible, that’s a surprisingly direct line of inquiry. It doesn’t answer the question so much as roll it directly toward you and wait.

The timing adds another dimension. Labubu started as a toy, became a fashion accessory, turned into a resale market, and has now arrived at experimental robotics research inside one of the most prestigious institutions on the planet. That arc is completely absurd and also perfectly logical if you’ve been watching how internet culture compresses timelines. Trends don’t climb ladders sequentially anymore. They collide with things that have no business intersecting, and occasionally the collision produces something genuinely interesting. The path from blind-box collectible to MIT thesis statement is ridiculous, and also makes complete sense.

Labububot will make its public debut this summer as a Grand Challenge finalist at the 2026 International Conference on Social Robotics in London. Moving from the controlled environment of an MIT hallway to a public conference floor is a meaningful shift. Real audiences bring expectations about what robots should look and feel like, and a twelve-faced Labubu sphere is going to challenge most of those expectations immediately.

Some people will read it as satire. Some will find it genuinely unnerving. A few will want to know if they can buy one. I’m not entirely outside that last group, which tells me the project landed exactly where it was supposed to. Labububot doesn’t ask you to like it. It just follows you down the hall until you decide how you feel.

The post MIT Turned 12 Labubu Heads Into a Robot and It’s Watching You first appeared on Yanko Design.

The Roomba Guy Just Built a Robot Pet You Might Actually Love

Posted on by

If you’ve ever watched your Roomba bump helplessly into a chair leg for the third time and thought, “I deserve better from my robots,” you’re not alone. And apparently, neither did Colin Angle. The co-founder of iRobot, the man who essentially put a hockey puck-shaped vacuum in millions of homes, left the company in 2024 with a new question rattling around in his head: what if a robot could actually feel like it cares about you? The answer is the Familiar, the first prototype from his new startup, Familiar Machines & Magic. And it is not your average robot.

Picture a creature somewhere between a soft-eared dog and a round, slightly abstract bear. It has four legs, huge paws, and doe eyes that make it immediately charming in a way that no Roomba ever attempted to be. It’s furry, expressive, and was designed with the help of former Disney Imagineers, which explains why it looks like it belongs in an animated feature rather than a tech showcase. The Familiar has 23 degrees of freedom, meaning it can wiggle its ears, tilt its head, and wag a small nub of a tail with the kind of fluidity that feels less mechanical and more… alive. Its coat is touch-sensitive, built specifically to encourage physical interaction between you and it.

Designer: Familiar Machines & Magic

It also doesn’t talk. That detail feels deliberate and, to me, very smart. Voice assistants have trained us to think of robots as tools we command. The Familiar is going for something completely different. It’s designed to read your tone of voice, your body language, your overall energy, and respond accordingly. Angle calls it “Consumer Physical AI,” and while the label sounds like something off a product white paper, the idea behind it is genuinely compelling.

The name itself is worth noting. A “familiar” in folklore refers to the supernatural animal companion of a witch or magical figure, a creature bonded to a person not through ownership but through genuine connection. Angle’s team chose that name deliberately, and I think it sets the tone for what they’re trying to build. The goal isn’t to sell you a novelty gadget. It’s to create a new kind of relationship between humans and machines, one built on trust, attentiveness, and something approaching care.

Now, I’ll be upfront: I have feelings about this. Part of me finds it genuinely beautiful as a design concept. The Familiar was clearly approached the way good industrial design should be, with deep thought about how an object makes you feel, not just what it does. The choice to make it animal-like rather than humanoid is interesting, too. There’s far less of the uncanny valley unease that tends to follow humanoid robots around, and more of the universal warmth that most people already extend toward animals.

But another part of me wonders about the emotional stakes here. We’re already watching people form attachments to AI chatbots. A touch-sensitive, furry, expressive robot that mirrors your emotional state is a much more potent version of that. Angle has said he wants it to feel like the machine actually cares about him. That’s a lovely vision. It’s also a design brief that puts enormous responsibility on the creators to get it right, because the flip side of emotional bonding is emotional dependence.

Still, I’d be lying if I said the Familiar didn’t make me curious in the best possible way. The prototype images are almost disarmingly sweet. It looks like something you’d want sitting on the couch next to you while you read, or settled quietly in the corner while you work. If any robot was ever designed to move through your life rather than just function within it, this might be it.

The Familiar is still in the prototype stage, with no confirmed price or release date. But as debut concepts go, it’s a strong one. Whether or not it ever makes it into our homes, it raises questions about what we actually want from the machines we live with. And those questions feel well overdue.

The post The Roomba Guy Just Built a Robot Pet You Might Actually Love first appeared on Yanko Design.