RoboHorizon Robot Magazine - AI you can touch

Genesis AI's Eno Robot Ditches the Head, Aims for an 'iPhone Moment'

Robot King — Wed, 17 Jun 2026 00:00:00 +0000

In a humanoid robotics field currently obsessed with churning out metal doppelgängers, Genesis AI has broken cover from stealth with a staggering $105 million seed round and a refreshingly contrarian playbook. The firm has just pulled the silk off Eno, a general-purpose robot that is quite pointedly trying not to look like a person. By ditching the head, opting for wheels over legs, and wrapping the whole affair in a seamless, appliance-like chassis, Genesis is betting that the robot you’ll actually invite into your home won’t look like a sci-fi extra, but rather a piece of high-end Scandinavian furniture.

The design philosophy behind Eno is one of “essentiality and intention”—which is a sophisticated way of saying it’s built for the job, not for winning a beauty pageant. Rising from a nimble wheeled base, its articulated torso can telescope its height and reach before folding down into a compact footprint for storage. You won’t find any exposed motors, messy cabling, or even a single visible screw hole. For those who need a bit of feedback, an optional chest screen can be fitted to display the robot’s “intent,” providing a “cognitive interface” so you can grasp what it’s up to without having to peer into a pair of cold, unblinking camera-eyes.

The real wizardry, however, is located at the end of the arms. Genesis AI has kitted Eno out with proprietary dexterous hands that the company claims mirror human form and function, allowing it to tackle fiddly tasks with millimetre precision. This manual dexterity is orchestrated by GENE, the company’s “robotics-native AI brain,” designed to ensure the hardware and software sing from the same hymn sheet. This “full-stack” approach—building everything from the neural architecture down to the physical finger joints—is what Genesis believes will give it the edge in an increasingly crowded market.

Why does this matter?

While heavyweight rivals like Tesla, Figure, and Agility are burning through billions trying to master the art of the two-legged stroll, Genesis AI is making a savvy side-bet: that wheels are cheaper, safer, and far more sensible for the flat-floored environments where robots will actually work. Co-founder and CEO Zhou Xian argues that the road to mass adoption lies in creating unobtrusive, functional “appliances” rather than uncanny mechanical humans. Armed with a war chest from big-hitters like Eric Schmidt and Xavier Niel, this well-funded startup isn’t just building another gadget; it’s challenging the very DNA of what a helpful robot should be. If they’re right, this could be the “iPhone moment” the industry has been waiting for. We can expect to see the first units rolling out to industrial partners by the end of 2026.

ABB Taps Bionic Hands to Give Its Robots a Human Touch

Robot King — Tue, 16 Jun 2026 00:00:00 +0000

For decades, industrial robots have been the heavy lifters of the factory floor—brilliant at lugging car chassis around, but absolutely rubbish at picking up a scotch egg without turning it into breadcrumb confetti. ABB Robotics reckons the solution isn’t more code, but a bit of human biology. The automation giant has announced a clever collaboration with PSYONIC to use its sensor-packed Ability Hand—a piece of kit usually reserved for bionic prosthetics—to finally teach robots what a delicate touch actually feels like.

The plan is as elegant as it is unorthodox: take the same bionic hand used by hundreds of amputees in their daily lives and mount it onto an ABB GoFa cobot. This creates a direct pipeline, feeding a deluge of real-world touch, pressure, and grip data from human-led movements straight into the robot’s learning model. The goal? To train a new generation of “physical AI” capable of handling the messy, unpredictable objects that have stumped traditional automation for years.

“Dexterous manipulation is ultimately a data challenge as much as a hardware challenge,” explained Dr. Aadeel Akhtar, Founder and CEO of PSYONIC, in the official announcement. The Ability Hand, which has already been put through its paces in research by the likes of NASA and Meta, is one of the most advanced prosthetics on the market, boasting haptic feedback that allows users to “feel” their surroundings. By pairing this human-tested hardware with the industrial precision of ABB’s GoFa robot, the partnership aims to translate human finesse into reliable robotic performance.

Why does this matter?

This is a full-frontal assault on one of the biggest remaining hurdles in the world of robotics: the “irregular object” problem. Most factory grippers are little more than basic, blunt pincers. Teaching a robot to handle fragile, soft, or oddly shaped items could blow the doors wide open for automation in agriculture, e-commerce fulfilment, and food processing—sectors that, until now, have had to rely almost entirely on human hands.

The International Federation of Robotics (IFR) estimates that advanced gripping tech could slash engineering time by up to 30%, but the real prize is the market expansion. By essentially crowdsourcing dexterity data from prosthetic users, ABB and PSYONIC might have just stumbled upon the ultimate shortcut to giving robots a much-needed human touch.

Satellite AI Has Its 'I See' Moment, No Humans Needed

Robot King — Tue, 16 Jun 2026 00:00:00 +0000

In a breakthrough that feels both inevitable and properly sci-fi, an Earth-observation satellite has, for the first time, tracked down its target entirely on its own. The mission, which took place in April aboard Loft Orbital’s YAM-9 spacecraft, represents the first recorded deployment of a vision-language model (VLM) in orbit. It’s a move that effectively emancipates the satellite from its constant reliance on human analysts back on the ground. This isn’t just a bit of clever code; it’s a seismic shift in what space-based sensors are actually capable of.

The satellite was running Google DeepMind’s Gemma 3 model, an AI specifically tailored for “edge” environments where processing power is at a premium—like, for instance, a metal box screaming through the vacuum of space at several kilometres per second. The demonstration was powered by an NVIDIA Jetson Orin AGX GPU and orchestrated by a software package from NASA’s Jet Propulsion Laboratory. Rather than the traditional, painstaking process of beaming terabytes of raw imagery back to Earth for harried analysts to comb through, YAM-9 was fed natural language queries—such as “identify infrastructure around railway hubs”—and the onboard AI handled the triage, flagging only the relevant data.

miért fontos ez?

This demonstration effectively transforms satellites from “dumb” cameras into proactive, autonomous scouts. By processing data at the source, it bypasses the monumental data logjam that usually plagues satellite operations. More profoundly, it paves the way for what Loft’s Head of AI, Paul Lasserre, describes as “always-on, patrol layers in space.” Instead of tasking a satellite to simply snap a photo, operators can issue persistent commands like, “Monitor this border and alert me the moment you see something dodgy.” it is the first step toward a future where our space infrastructure isn’t just collecting data, but actively making decisions.

This Robot Toilet Comes to You, Then Cleans Itself for $13,000

Robot King — Sun, 14 Jun 2026 00:00:00 +0000

In the ever-expanding universe of “robots designed to tackle the jobs you’d really rather avoid,” a new contender has rolled onto the scene. Chinese tech firm Yueban has pulled the wraps off the Xiaoban, an intelligent toilet robot that takes the “smart home” concept to its logical—if slightly surreal—conclusion: it drives itself right to your bedside. It’s a complete reversal of the age-old “trip to the loo”; now, the loo comes to you.

Unveiled at the 2026 Shanghai International Elderly Care Expo, the Xiaoban is far more than just a glorified commode on wheels. Once the user has finished their business—and yes, it’s engineered to handle the lot—the device kicks into gear with a full bidet wash and warm air dry. To keep things pleasant, it employs a heavy-duty odour-neutralising system, featuring a foam shield and activated charcoal filters to ensure the room stays fresh. In China, it’s hitting the shelves at ¥28,999, which is roughly £10,500. International pricing is still very much up in the air.

The clever bit happens once the user is safely back in their chair or tucked into bed. Guided by a sophisticated sensor suite including LiDAR and lasers, the Xiaoban navigates its way to a dedicated docking station that hooks into the home’s existing plumbing. It then extends a disposal arm, flushes its contents through a low-noise grinder to prevent any “dodgy” blockages, and initiates a rigorous 360-degree self-sanitising cycle. To finish the job, it tops up its own clean water tank and recharges its battery, ready for the next call of duty.

Why is this important?

Beyond the initial “only in the future” shock value, the Xiaoban represents a serious attempt to tackle one of the most difficult challenges in elder care: preserving personal dignity while reducing the immense physical and emotional strain on carers. For a price tag that rivals a decent second-hand car, it offers a fully automated, contactless solution to an essential but sensitive part of daily life. While the cost will be a barrier for many, it signals a shift towards a future where robotics provide genuine autonomy for an ageing population, handling tasks that are as vital as they are unpleasant.

Your Robot Knows You Are Upset, It Just Might Not Care

Robot King — Sat, 13 Jun 2026 00:00:00 +0000

It turns out that a robot offering “sincere apologies” with the perfect hint of digital contrition after fumbling your morning brew is still just a robot that’s drenched your keyboard in scalding coffee. We are entering an era where our automated colleagues are being programmed with social graces, but a fascinating new study suggests that all the politeness in the world can’t mask basic incompetence.

Researchers are increasingly obsessed with the “squishy” science of human-robot interaction (HRI). They’ve realised that as robots migrate from the factory floor into our living rooms and offices, brute mechanical force isn’t enough. They need to get us. A study recently published in IEEE Robotics and Automation Letters dives headfirst into this challenge, training a collaborative robot to read human emotions—not just by scanning a face, but by interpreting the entire context of a situation. The results are a sobering, and frankly hilarious, reality check for anyone who thinks an empathetic robot is the final frontier.

Training a Bot to Read the Room

The research, led by Seung Chan Hong during his time at the University of Melbourne, decided to bin the tired, old methods of emotion detection. Instead of merely analysing a static facial expression—which can easily mistake a furrowed brow of concentration for a flash of anger—the team employed a Vision Language Model (VLM). Think of it as a cousin to ChatGPT, but with eyes.

They trained the VLM by showing it videos of human-robot handovers and asking human volunteers to label the emotions on display. Crucially, these volunteers could see the full picture: the fumbled object, the slight wince, the impatient drumming of fingers. This context-rich training paid off. When pitted against a conventional AI system that relied solely on facial analysis, the VLM performed significantly better, achieving a 0.86 similarity score to human observers, compared to the older model’s 0.77.

“I think [the VLM] was able to align with what human observers were seeing a lot better, because it wasn’t just looking at the person’s face for a brief amount of time, but seeing the whole scene,” Hong noted in an interview with IEEE Spectrum.

The Flawless Apology for a Flawed Performance

Here’s where things get interesting. The team designed an experiment involving 40 volunteers. Each participant had to work with the VLM-powered robot, which was programmed to deliberately mess up. After the inevitable blunder, the robot would offer one of two apologies: a generic, pre-scripted line or an “emotionally adaptive” apology tailored to the human’s perceived frustration.

The results were stark: people vastly preferred the robot that could read their annoyance and tailor its “I’m sorry” accordingly. A resounding 31 out of 40 participants favoured the emotionally attuned response. It seems a personalised apology acts as a potent “social lubricant.”

But here is the punchline. When asked about their trust in the robot, participants’ ratings plummeted across the board, regardless of how nicely the robot apologised. The uncomfortable truth is that a robot can be as sensitive as a poet, but if it can’t do its one job, we’re not going to trust it. As Hong bluntly puts it, an apology “cannot repair the trust lost by the robot failing its physical task.”

Not a Mind Reader, Just a Good Guesser

The study unearthed another critical limitation. While the VLM was a decent mimic of a third-party human observer, its emotion-guessing skills took a nosedive when compared to what the volunteers actually felt (according to their self-reported data).

This reveals a fundamental gap between perceiving outward social cues and understanding internal feelings. The VLM could spot a frown and a slumped posture and correctly infer “unhappiness,” but it couldn’t grasp the nuances of disappointment, frustration, or the sense of betrayal a user might be feeling internally. “While the VLM is a good observer of outward social cues, it isn’t a mind reader,” Hong explained.

This work serves as a vital reality check for the robotics industry. While the quest for emotionally intelligent machines that can seamlessly integrate into our lives is a noble one, it cannot come at the expense of fundamental reliability. Before we build a robot that can offer a shoulder to cry on, let’s first make sure it doesn’t spill the tea in the first place. You can read the full paper, “Can Robots Read Your Mind? A User Study on Inferring Human Emotions in HRI,” in IEEE Xplore.

Bosch Bets Billions on Brains, Not Brawn, for Humanoid Robots

Robot King — Fri, 12 Jun 2026 00:00:00 +0000

In a move that perfectly encapsulates the “selling picks and shovels during a gold rush” mantra, German engineering titan Bosch has announced it is diving headfirst into the humanoid robotics arena. But don’t expect a Bosch-branded butler anytime soon; the company isn’t building the robots themselves. Instead, it’s positioning itself as the indispensable architect of their internal organs.

At the Bosch Connected World (BCW) 2026 summit in Berlin, the company unveiled a high-stakes strategy to capture a “business worth billions” by supplying the critical components that will serve as the guts of the next generation of bipedal machines.

Bosch is pitching itself as the premier provider of the “brain and nervous system” for modern robotics. To spearhead this, the firm has launched a dedicated subsidiary, Robert Bosch Robotics GmbH, tasked with industrialising new robotics solutions at scale. This is bolstered by the newly minted Bosch Robotics Center China (BROC), designed to fast-track physical AI development in the world’s most aggressive robotics market. The plan is simple: leverage Bosch’s legendary manufacturing muscle to churn out high-precision electric motors, sophisticated servo drives, and its open ctrlX AUTOMATION platform via its Bosch Rexroth arm.

The real ace up Bosch’s sleeve, however, is its total dominance in microelectromechanical systems—better known as MEMS sensors. These microscopic components are what give a robot its “proprioception”—the sense of touch and balance required to handle a delicate wine glass without shattering it. With market analysts at Yole Group predicting the MEMS sector will top £15 billion ($19.2 billion) by 2030, Bosch is sitting on a goldmine.

Stefan Hartung, Chairman of the Bosch board of management, put the sheer scale of the challenge into perspective: “A human being has four million touch sensors. If we were to build robots with that same level of sensitivity, four years’ worth of total global sensor production would barely cover the needs of just 12,500 robots.”

Rather than going it alone, Bosch is playing the role of the ultimate collaborator. The company is currently working with German scale-up Neura Robotics to refine cognitive bots and has teamed up with several high-profile startups, including the UK’s own Humanoid, to help transition their ambitious prototypes into mass-market realities.

Why does this matter?

Bosch’s pivot is a massive vote of confidence in the humanoid dream. By sidestepping the eye-wateringly expensive and risky business of building a finished robot, Bosch is making a calculated bet on becoming the “Intel Inside” of the robotics age.

This “arms dealer” strategy allows them to avoid a direct scrap with vertically integrated giants like Tesla or Figure. Instead, by providing the foundational hardware and software—from the tactile sensors to the motion control guts—Bosch is aiming to set the industry standard. It’s a shrewd, lower-risk play that could see Bosch quietly powering the entire robotics revolution without ever having to put its own name on a chassis.

Neura Robotics Secures Up to $1.4bn in Huge Funding Round

Robot King — Thu, 11 Jun 2026 00:00:00 +0000

German robotics and AI powerhouse Neura Robotics GmbH has just sent shockwaves through the industry, announcing a gargantuan Series C funding round of up to $1.4 billion (£1.1 billion). The company claims this is the largest-ever investment for a full-stack robotics firm, a move that reportedly catapults Neura’s valuation to a cool $7 billion. The backing comes from a formidable line-up of global heavyweights, including Amazon, NVIDIA, Qualcomm Technologies, and Bosch.

This massive injection of capital is earmarked for the global rollout of Neura’s “cognitive robots” and humanoids, the expansion of its Neuraverse software ecosystem, and an audacious manufacturing scale-up. The goal? To have millions of robots rolling off the assembly lines by 2030. Founded in 2019 in Metzingen, Germany, Neura isn’t just selling a dream; it already boasts a staggering order book worth over $1 billion. The funds will also bankroll the launch of NEURA Gyms—vast, real-world training grounds where its AI-driven machines can hone their skills.

The investor list reads like a “who’s who” of the tech and industrial sectors, but one name provides a fascinating curveball: Tether, the issuer of the world’s most dominant stablecoin. While the interest from chip giants like NVIDIA and industrial stalwarts like Schaeffler is par for the course, Tether’s lead role signals a pivot into the physical world. The crypto titan intends to integrate its wallet development kit into Neura’s platform. The vision? Robots that can autonomously transact and settle payments for the work they perform. It seems the future of the workforce involves machines that don’t just do the heavy lifting, but also handle their own invoicing.

Why does this matter?

This colossal round is a thumping vote of confidence in “Physical AI”—the conviction that the next great leap for artificial intelligence isn’t happening on our screens, but in the physical world. While American and Chinese firms have largely dictated the robotics narrative of late, this investment establishes Neura as a heavily armed European contender in the race for embodied AI.

The carefully phrased “up to $1.4 billion” suggests a nuanced deal likely tied to specific performance milestones rather than a simple blank cheque. Nevertheless, it hands Neura a formidable war chest to take on the biggest names in the humanoid space. The real test now lies in execution: transforming billions in venture capital into millions of reliable, working robots capable of functioning beyond the sterile confines of a lab. The race to build the world’s robotic workforce has just hit overdrive.

This Robot Dog Can Literally Smell Danger

Robot King — Thu, 11 Jun 2026 00:00:00 +0000

While the rest of the robotics world remains fixated on teaching machines to see, walk, and babble, Deep Robotics is following its nose—literally. At the “Future of Smell” exhibition hosted by Germany’s Dresden University of Technology, the firm showcased a modified version of its Lite3 quadruped, equipped with an “electronic nose” that effectively turns the robot into a high-tech bloodhound for the digital age.

The souped-up Lite3 navigated a cluttered exhibition hall with poise before pausing at a designated test point to take a “deep breath.” Its onboard e-nose system captured airborne scent signals, which were then instantly processed by AI algorithms to pinpoint specific chemical signatures. The demonstration was, by all accounts, a triumph; German researchers on-site were reportedly heard exclaiming, “It really smelled it!” This shift transforms the robot from a mere “camera on legs” into a sophisticated sensory platform capable of sniffing out dangers that remain invisible to the human eye.

The hardware itself is no slouch. By mounting the electronic nose onto the agile Lite3 base, Deep Robotics has created a scout that can operate for up to two hours, carry a 7.5 kg payload, and tackle 40-degree inclines. This level of mobility allows the bot to venture into hazardous, cramped, or high-risk environments where humans—and their actual noses—fear to tread, such as locating gas leaks or identifying the source of industrial pollution.

Why does this matter?

Giving a robot a sense of smell is a massive leap for “embodied AI.” It moves the needle from passive observation to active environmental interrogation. In an industrial setting, this means a robot could autonomously patrol a factory floor and detect a faint chemical leak long before it becomes a full-blown catastrophe. In a search-and-rescue scenario, it could potentially identify the chemical signature of a survivor trapped under rubble. By fusing mobility with an artificial olfactory system, Deep Robotics has created a “closed-loop” of perception, analysis, and early warning. It’s a timely reminder that in the world of safety and inspection, the most critical threats are often the ones you can’t see, but can definitely smell.

Forget Humanoid Hype: ICRA 2026 Reveals Global Robot Wars

Robot King — Tue, 09 Jun 2026 00:00:00 +0000

While the internet remains mesmerized by the latest humanoid robot performing a clumsy pirouette or folding a t-shirt with the agonizing slowness of a tectonic plate, the adults in the room have moved on. At the IEEE International Conference on Robotics and Automation (ICRA) 2026 in Vienna—a gathering of thousands of the world’s sharpest robotics minds—one of the most critical conversations had nothing to do with flashy tech demos. Instead, it was a brutal dose of geopolitical reality.

The panel, titled “Robots for All” in a Fragmented World: Competing Global Visions and Shared Futures from Europe, Asia, and the United States, served as a stark reminder that the next great robotics race won’t be won in the lab. It will be decided by industrial strategy, social trust, regulatory savvy, and the unglamorous, hard graft of actual, large-scale deployment. The message was clear: spectacular prototypes are all well and good, but the real prize is economic sovereignty.

The Myth of the Monolithic Robot Future

The current discourse is dangerously dominated by the idea of a universal humanoid—a one-size-fits-all solution for factories, hospitals, and homes. It’s a compelling narrative for investor pitch decks but a poor reflection of reality. A key insight from the panel, delivered by Hesheng Wang, a Chair Professor from Shanghai Jiao Tong University, cut through the noise: “‘For all’ cannot mean one robot, one pathway, or one vision.”

This single sentence dismantles the simplistic notion of a global robot monoculture. The future of robotics is being forged in the crucible of regional ambitions, shaped by vastly different economic structures, demographic pressures, and cultural values. What works in China’s state-driven industrial ecosystem won’t simply copy-and-paste into Japan’s precision-engineering and socially integrated model. What gets funded in Silicon Valley won’t necessarily survive the rigours of Europe’s regulatory landscape.

East Asia: Three Distinct Paths to Dominance

The panel provided a fascinating breakdown of the divergent strategies within East Asia, a region often mistakenly viewed as a single, monolithic tech bloc.

China: The strategy can be summed up in one word: scale. Fuelled by tight integration between government, academia, and industry, China is building entire robotics ecosystems at a dizzying pace. The goal isn’t just to build robots; it’s to embed AI into the physical economy and establish national standards for everything from embodied intelligence to humanoid supply chains. With over two hundred humanoid startups in China alone, the sheer velocity is staggering, though it remains to be seen how many will find a sustainable market.
Japan: Following a different playbook entirely, Japan focuses on reliability, precision engineering, and deep social integration. With a long history of industrial automation and a rapidly ageing society, robotics is seen as a core solution for everything from high-end manufacturing to elder care. The Japanese approach may appear more measured, but it prioritises long-term stability and public trust over flashy, short-term stunts. With a robot density of 446 units per 10,000 employees, its methodical approach has already yielded one of the world’s most automated economies.
South Korea: The approach here is one of concentrated coordination and aggressive execution. Despite its smaller size, South Korea boasts the world’s highest robot density, with a staggering 1,220 robots per 10,000 employees. This is the result of clear national directives, such as the Master Plan for Intelligent Robots, and a relentless focus on delivery, particularly through its “K-Humanoid Alliance” and national AI projects.

Europe’s Gambit: Don’t Play Someone Else’s Game

Faced with China’s scale and America’s venture capital firepower, Europe could easily fall into the trap of trying to be a “slower China” or a “more regulated Silicon Valley.” According to Francesco Ferro, the newly appointed President of euRobotics and CEO of PAL Robotics, that would be a strategic blunder of the highest order.

Instead, Europe must lean into its unique strengths. The euRobotics Vienna Statement, unveiled at ICRA, outlines a strategy built not on imitation, but on a distinct European identity.

The three non-negotiable principles for Europe:

Robots should work with and for the people, not merely as replacements for them.

Sustainability is a fundamental design requirement, not an afterthought.

Solutions must be community-driven, rather than imposed from the top down.

These principles might sound less thrilling than a new LLM, but they are Europe’s secret weapon. They directly address the biggest hurdle to mass adoption: social acceptance. In a region with robust labour protections, high consumer standards, and an ageing population in desperate need of healthcare and agricultural solutions, building trust isn’t a PR exercise—it’s the entire business case.

The Unsexy Infrastructure of Success

The consensus from the panel, which also included insights from Oussama Khatib, Director of the Stanford Robotics Lab, was that the transition from a functioning prototype to a commercially viable product is a “valley of death” littered with brilliant failures. Why? Because success depends on an infrastructure of trust that most tech companies treat as an afterthought.

A robot can be a technical marvel yet fail spectacularly if workers see it as a threat, patients don’t trust it, insurers won’t touch it, or regulators create a decade of uncertainty. Social acceptance isn’t a marketing campaign; it’s a feature that must be baked in from day one, involving users, addressing liability, and proving a clear return on investment.

The hard truth is that the humanoid hype, while great for generating clicks, is a distraction. The real “killer app” for robotics won’t be a single, all-purpose machine. It will be a diverse ecosystem of specialised, reliable, and trusted systems solving urgent problems—from automating back-breaking tasks in agriculture to supporting caregivers in care homes. The winner of the global robotics race won’t be the one with the most-viewed YouTube demo, but the one who solves the most real-world problems, durably and economically. And that, as it turns out, is a much harder—and far more interesting—challenge.

1X Declares War on 'VLA Wrappers,' Launches World Model Lab

Robot King — Thu, 04 Jun 2026 00:00:00 +0000

In the frantic, eye-wateringly expensive race to build thinking machines that can navigate the physical world, a philosophical rift is widening into a canyon. On one side stand the pragmatists, intent on harnessing the colossal power of existing Large Language Models. On the other are the purists, who argue that true physical intelligence cannot be bolted on—it must be forged from the ground up. This week, the humanoid robotics firm 1X Technologies planted its flag firmly in the second camp, launching the 1X World Model Lab with a declaration that felt less like a press release and more like a manifesto.

“You can’t fine-tune your way to AGI,” declared 1X CEO Bernt Bornich in a pointed announcement. “And you definitely can’t fine-tune your way to robots that can operate in the physical world.” It was a direct shot across the bow of competitors who are enthusiastically adopting Vision-Language-Action (VLA) models—AI systems that essentially “wrap” a powerful model like GPT-4 with motor control capabilities. 1X is staking its future on a different, far more arduous path: embodied world models.

The Great Divide: Fine-Tuning vs. First Principles

To understand the gravity of 1X’s move, one has to appreciate the two competing doctrines currently battling for the soul of robotics.

The Vision-Language-Action (VLA) approach, championed by the likes of Figure AI, is very much the path of least resistance. The logic is seductive: take a multi-billion-pound foundation model that already “understands” language and vision, fine-tune it on a dataset of robot actions, and—Bob’s your uncle—you have a robot that can follow instructions. It’s a strategy that piggybacks on the immense progress (and staggering investment) in LLMs. The snag, critics argue, is that these models lack a genuine grasp of physics. They are sophisticated pattern-matchers, not physics engines. They might know from their training data not to drop a pint glass, but they don’t intrinsically understand that gravity will inevitably smash it to bits.

Then there is the World Model approach. This is the hard road. The goal is to build a foundation model that learns an internal, predictive simulation of reality. Before it ever learns a specific task like “pick up the apple,” it must first master concepts like space, motion, object permanence, causality, and the unforgiving laws of physics. Proponents believe this is the only route to true generalisation—the ability for a robot to act intelligently in novel situations it has never encountered in its training data.

Bornich’s stance is unequivocal. “The frontier is not better VLA wrappers,” he stated. “The frontier is embodied world models.”

1X’s All-In Bet and a Strategic Coup

The new 1X World Model Lab is the company’s answer to this challenge. Its mission is to build the most generalisable foundation model for humanoids from scratch. To lead this ambitious effort, 1X has poached Sam Sinha, a founding research scientist from the generative video AI darling Luma AI.

The hire is a masterstroke. Luma AI specialises in creating spookily realistic video models, a technology that is conceptually adjacent to building a world model that predicts future physical states. Sinha’s entire career has been spent at the frontier of scaling multimodal generative video. As he put it, for too long robotics has been treated as a “second-class citizen” in AI, with robot data being a “thin fine-tuning layer bolted onto a model.” The new lab aims to flip the script, treating embodied data as a first-principle ingredient.

1X’s strategy relies on a “data flywheel”—a virtuous cycle of learning:

The Foundation: Web-scale media, egocentric human videos, and simulation data.
The Nuance: Dexterous data harvested from remote-operated robots.
The Deployment: A fleet of NEO humanoids collecting real-world data on the fly.
The Result: The robot collects data, the model evolves, and the robot becomes more capable. Rinse and repeat.

An Alliance of World-Builders

1X is not alone in its philosophical conviction. The world model camp has some heavy hitters, even if they aren’t all building bipedal robots.

Tesla’s Full Self-Driving (FSD) system is perhaps the most famous real-world application of this concept. FSD relies on a “World Model” to predict the likely future actions of every car, cyclist, and pedestrian in its vicinity, running an internal simulation of plausible futures to inform its driving decisions. It isn’t just reacting; it’s anticipating.

AI luminary Yann LeCun, now leading AMI Labs after a storied career at Meta, has been a vocal proponent of world models for years. He argues that LLMs are “fundamentally incomplete” because they lack an internal model of how the world actually works. His work on Joint Embedding Predictive Architectures (JEPA) aims to build models that learn common sense by observing and predicting video—a core tenet of the world model philosophy.

The Road Ahead is Paved with Petabytes

1X’s move is a high-risk, high-reward gambit. Building a foundational world model from the ground up is an astronomically expensive and data-hungry endeavour. While the VLA camp gets a massive head start by standing on the shoulders of giants like Google and OpenAI, 1X is choosing to dig its own foundations.

The success of the 1X World Model Lab hinges on its ability to execute its data flywheel strategy at a massive scale. If it succeeds, it could create a formidable data moat and a generation of robots with a far more robust and generalisable intelligence than their VLA-powered counterparts. If it fails, it will serve as a cautionary tale of eschewing a pragmatic shortcut for an elegant but impossibly difficult ideal.

The battle lines have been drawn. Is the future of robotics a clever extension of the LLM revolution, or does it require a completely new beginning? The industry is now watching to see if 1X’s bold bet on building the world from scratch will pay off, or if they’ll find themselves stuck fine-tuning their balance sheets.

Helsing's New RX-1 Robot Dog is Built for War, But Starts in the Lab

Robot King — Thu, 04 Jun 2026 00:00:00 +0000

German defence AI heavyweight Helsing has decided its code is getting a bit lonely in the cloud. The firm, best known for its Centaur AI co-pilot that helps Saab’s Gripen fighter jets stay ahead of the pack, has officially stepped into the physical realm. Meet the RX-1: a beefy, four-legged robot engineered for speed and resilience in the kind of brutal outdoor environments that would wreck your average gadget. This robotic hound is the debut project from Area 9, Helsing’s shiny new advanced research wing, which feels very much like Europe’s answer to Lockheed’s Skunk Works or Google’s “Moonshot” factory.

The RX-1, which looks every bit the military-grade machine you’d expect, is being touted as a “sovereign European alternative” to platforms built across the pond or further east. In a pointed nod to tech independence, Helsing’s Chief Scientist, Antoine Bordes, stressed that the RX-1 is “designed in Europe and manufactured in Europe, including crucial components such as our in-house designed actuators.” In other words, they aren’t just bolting together off-the-shelf parts; they’re building the muscle as well as the brain.

For the time being, the RX-1 is a research platform rather than a front-line soldier. Helsing is shipping the first units to top-tier academic labs to kickstart research into the messy intersection of AI and autonomous hardware. The initial partners are two of Europe’s heavy hitters in the robotics world: Professor Marco Hutter’s lab at ETH Zürich and France’s national research institute, Inria. Hutter described the RX-1 as an “advanced, European-developed hardware platform enabling exciting field robotics research.”

Why should we care?

Helsing’s pivot from pure software to vertically integrated hardware is a massive strategic play. The company isn’t just churning out another robot dog; it’s building a full-stack robotics ecosystem with a sharp geopolitical edge. By creating a “sovereign” platform, Helsing is tackling Europe’s nagging reliance on foreign tech in sensitive defence sectors head-on. The end goal, according to the company, is to master autonomous systems that can navigate “unpredictable terrain—like, say, a debris-laden battlefield,” where sending a human is out of the question. The RX-1 is the first physical step towards giving Helsing’s AI brains a body of their own making.

OpenAI builds robots again after public breakup with Figure

Robot King — Mon, 01 Jun 2026 00:00:00 +0000

In a move that essentially says, “Right then, we’ll do it ourselves,” OpenAI CEO Sam Altman has confirmed the firm is officially wading back into the robotics fray. The new arm, predictably dubbed OpenAI Robotics, is already on a massive recruitment drive for world-class engineers to build machines capable of “helping people in the physical world.” This pivot comes just months after the very public, and rather abrupt, end of its partnership with humanoid specialist Figure.

Spearheading this ambitious reboot is Aditya Ramesh—a name that’ll ring a bell for anyone who’s ever used AI to conjure up a photorealistic image of an avocado-shaped armchair. As the mastermind behind DALL-E and a key player on the Sora video-generation team, Ramesh’s appointment as VP of Robotics is a dead giveaway. OpenAI isn’t just sticking a chatbot inside a tin man; they’re betting that their entire generative AI and world-simulation stack can solve the puzzle of embodied intelligence. Ramesh himself has confirmed the move, noting his aim is to “bring the intelligence of our video generation models into the physical realm.”

This isn’t OpenAI’s first rodeo, of course. Industry veterans will remember the original robotics team—the ones who famously taught a robotic hand to solve a Rubik’s Cube—before the project was unceremoniously mothballed in 2021. Back then, the company cited a lack of high-quality training data as the primary roadblock. Now, the subtext is clear: with world models like Sora, they can simulate reality so convincingly that the data drought is no longer a deal-breaker.

Why does this matter?

This isn’t just another R&D side-hustle; it’s a strategic U-turn that shifts the entire AI landscape. Following the messy, highly publicised split from Figure—where Figure claimed it had reached its own “major breakthrough” and needed to go it alone—OpenAI has clearly decided it’s done outsourcing its physical-world ambitions. This move puts Altman’s outfit in direct competition not only with its former flame Translation not available (en-gb) but with every other player in the increasingly crowded humanoid arena.

By putting a generative model wizard in charge, OpenAI is sending a clear message: the real hurdle in robotics isn’t the nuts and bolts, it’s the brain. They’re wagering that a truly intelligent, world-aware AI can overcome any mechanical shortcoming—a theory that’s about to face its ultimate test in the messy, unpredictable world of physical reality.

Unitree H2 Plus Humanoid Gains NVIDIA Jetson Thor Brain

Robot King — Mon, 01 Jun 2026 00:00:00 +0000

Just when you thought the humanoid robot arms race couldn’t get any more frantic, Unitree Robotics has decided to give its H2 model a serious cognitive overhaul. The newly unveiled Unitree H2 Plus is less of an incremental update and more of a full-blown brain transplant, thanks to the integration of NVIDIA’s monstrously powerful Jetson Thor superchip. This move firmly positions the H2 Plus as the high-end development platform of choice for the next generation of AI-driven robotics.

The H2 Plus itself is a formidable bit of kit even before you look under the bonnet. Standing at 6ft (1.82m) and weighing in at 70kg, it has a commanding physical presence. The robot boasts 31 degrees of freedom (DOF) in its main chassis, delivering impressive peak torque figures of 360 N·m in the legs and 120 N·m in the arms. But the real showstopper is the optional Dual SharpaWave Tactile Five-Finger Dexterous Hands. These add a further 22 degrees of freedom per hand, bringing the grand total to a dizzying 75 DOF. Each fingertip on these advanced manipulators is packed with over 1,000 tactile pixels, granting the robot a sense of touch with a force sensitivity as delicate as 0.005N.

The star of the show, however, is the NVIDIA hardware and software stack. The Jetson Thor system-on-module provides a staggering 2,070 TFLOPS of FP4 AI performance, powered by the cutting-edge Blackwell architecture GPU and a 14-core Arm CPU. This onboard supercomputer is purpose-built to run the entire NVIDIA Isaac platform, including the GR00T foundation models designed for general-purpose robotic tasks. In practice, this means developers can tap into NVIDIA’s exhaustive suite of tools for simulation (Isaac Sim), data capture (Isaac TeleOp), and deploying trained policies directly onto the hardware.

Why this matters

Unitree, a company that initially built its reputation on surprisingly affordable and agile quadrupeds, is now making an aggressive land grab in the high-performance humanoid arena. By partnering with NVIDIA to create an open reference design, Unitree isn’t just shifting units; it’s offering a complete, research-ready ecosystem. This significantly lowers the barrier for academic and commercial labs to get stuck in with top-tier hardware that’s already been harmonised with a powerful AI software stack. Rather than spending months wrestling with component integration, researchers can jump straight into developing complex skills and behaviours using GR00T foundation models. The H2 Plus is a direct shot across the bows of more vertically integrated players, betting that an open, powerful, and accessible platform will be the one to accelerate the entire field.

Gatsby's $150 Humanoid Butler Will Clean Your Apartment, No Strings Attached

Robot King — Sat, 30 May 2026 00:00:00 +0000

The domestic robotics market has reached a fascinating crossroads: you can either let a tech firm scrub your floors for free in exchange for your data to train their future robot army, or you can cough up a premium for a humanoid to do the job with no strings attached. San Francisco-based startup Gatsby is betting on the latter, launching an on-demand humanoid cleaning service for a flat fee of $150 (roughly £120).

This move puts Gatsby—operating under its parent company West Egg Labs—in direct philosophical opposition to firms like Shift. As Shift Offers Free Cleaning, But Your Mess Is Training the Robot That Will Replace You , Shift offers complimentary cleaning precisely because your domestic mess is the perfect training ground for their AI. Gatsby, however, is pitching convenience and privacy as a luxury. For a price that’s competitive with a human cleaner in San Francisco, the company will dispatch a full-size humanoid robot to your flat to tackle the washing up, buff the surfaces, vacuum the floors, and even fold your laundry. The entire process is booked through an iOS app, with absolutely no human interaction required.

Interestingly, Gatsby isn’t interested in the “greasy” business of hardware manufacturing. The company describes itself as a “robot-agnostic” consumer platform, aiming to be the service layer that connects customers with high-end humanoids from makers like 1X, Figure, or Sunday. It’s a classic Silicon Valley platform play: let others fight the brutal hardware wars while you own the customer relationship. The company claims the routine parts of the clean are fully autonomous. However, its website also admits that “the harder parts are teleoperated by real humans to make sure everything’s done right”—a critical detail that seems to be missing from its privacy policy. When a company’s entire value proposition is “no strangers in your home,” the presence of a remote pair of eyes is a rather significant “gotcha.” RoboHorizon has reached out to Gatsby for clarification on their teleoperation policies and how they handle user data.

Why does this matter?

Gatsby’s launch signals a genuine schism in the emerging domestic robotics market. On one side, you have the “data-for-service” model, where consumers trade their privacy for a bit of help around the house. On the other, Gatsby is establishing a premium tier where privacy itself is the product. Their $150 price point isn’t meant to undercut the human cleaning market, but rather to match it while offering a different kind of value: no scheduling headaches, no last-minute cancellations, and no awkward small talk over a cup of tea.

The success of this “Uber for humanoids” model will depend entirely on execution and transparency. While the promise of a fully autonomous clean is the ultimate goal, the quiet mention of human teleoperators is a stark reminder that we’re still in the early innings of this technology. How Gatsby navigates the murky waters of privacy and the “uncanny valley” of remote human assistance will determine if it becomes the new gold standard for home automation or just another curious Silicon Valley experiment.

AGIBOT 2B World Model Tops Benchmark: Physics Trumps Pixels

Robot King — Fri, 29 May 2026 00:00:00 +0000

In a classic case of punching well above its weight, a nimble 2-billion-parameter world model from AGIBOT has just muscled its way to the top of the WorldArena benchmark. The newcomer, christened Genie Envisioner-Sim 2.0 (GE-Sim 2.0), is currently sitting pretty at the #1 spot, looking down on the bloated generative video engines that usually hog the limelight. It turns out that while making a cinematic video is all well and good, teaching a robot to fold a tea towel without a meltdown is a different kettle of fish entirely.

This isn’t just another flashy video generator for the “vibes.” GE-Sim 2.0 is a closed-loop physical simulator designed as a high-octane boot camp for actual hardware. It boasts “High-Consistency Multi-View Generation,” which is a fancy way of saying the robot’s head camera and wrist cameras actually agree on what they’re seeing—even if an object is tucked away in a blind spot or bouncing off a mirror. It’s this level of obsessive precision that separates a proper simulation from a digital hallucination.

To get this over the line, AGIBOT tackled three major simulation headaches. First up, a “Proprioceptive State Expert” decodes joint angles directly from video feeds, preventing the robot from spiralling into mechanical anarchy. Then there’s the “VLM-Based World Judge”—essentially an automated referee that scores simulation runs so human engineers can actually go home for dinner. Finally, using a clever distribution-matching distillation framework, they’ve managed to whittle down inference times, churning out a complex 25-frame multi-view sequence in a snappy 2.3 seconds.

Why does this matter?

Because the results are the real McCoy. Physical robots trained on GE-Sim 2.0’s refined synthetic data saw a whopping 15% boost in real-world success rates for tricky, contact-heavy tasks. This is a massive leap towards solving the “data drought” in embodied AI. While the rest of the industry is obsessed with visual polish, AGIBOT is building the practical, physical foundations that make robots genuinely useful. The days of robots just looking the part are gone; we’re entering the era where they actually do the job.

The project is open-source, so you can have a poke around the code yourself. Check out the code on GitHub or read the full paper on arXiv.

NASA $30bn Moon Base: A Brutal Test for Robots Before Mars

Robot King — Thu, 28 May 2026 00:00:00 +0000

Let’s be clear: NASA’s latest blueprint for a permanent lunar base isn’t some nostalgic “flags and footprints” exercise. This is about infrastructure, heavy industry, and staring down an environment so hostile it makes a mockery of our terrestrial engineering conceits. The agency, led by Administrator Jared Isaacman, has attached a staggering price tag to this ambition: $30 billion, 79 launches, and 73 landers over the next 11 years—all to establish a permanent human foothold on the rim of Shackleton Crater.

This isn’t some far-off sci-fi pipe dream. This is the official Moon to Mars architecture, a concrete roadmap for learning how to survive on another world. But before the first long-term residents can start moaning about the lack of decent high-street coffee, an army of robots will have to build their home. And they’ll have to do it while battling an enemy that Apollo astronauts learned to dread: dust. Not the benign fluff that gathers on your skirting boards, but a microscopic menace sharp enough to shred solid steel.

The Blueprint for a Lunar Beachhead

The grand strategy is divided into three aggressive phases. Phase One, running from now until 2029, is the robotic vanguard’s time to shine. It involves a steady cadence of commercial payload deliveries—up to 25 missions—to scout the terrain, stress-test technologies, and begin deploying the first bits of hardware. This is where NASA’s Commercial Lunar Payload Services (CLPS) initiative takes centre stage, with firms like Intuitive Machines, Astrobotic, and Firefly Aerospace acting as the solar system’s most ambitious delivery drivers.

Phase Two (2029-2032) is when the outpost actually starts to look like a base. This involves establishing “initial operating capability”—NASA-speak for setting up the power grid and dropping off the heavy kit. The centrepiece is a 40-kilowatt nuclear fission reactor; because when the lunar night plunges temperatures to -203°C for 14 Earth days, solar panels become little more than expensive ornaments. Finally, Phase Three (2032 onwards) aims for a “semi-permanent crew presence,” evolving into the first continuously inhabited human settlement on another celestial body.

The location, Shackleton Crater at the South Pole, is a strategic masterstroke. Its rim enjoys near-perpetual sunlight for power, while its permanently shadowed floor hides billions of years’ worth of frozen water ice—the solar system’s most precious resource for drinking, breathing, and refining rocket fuel.

The Real Boss: A Microscopic Grain of Terror

Glossy renders of gleaming habitats look brilliant on a pitch deck, but they conveniently ignore the single greatest engineering headache of a permanent lunar presence: regolith. Lunar dust is a total nightmare. Without wind or water to erode it, every single particle is a microscopic shard of glass and rock. It’s electrostatically charged, meaning it clings to everything it touches. During the Apollo missions, it chewed through layers of spacesuit fabric, gummed up delicate mechanisms, and caused equipment to overheat.

“We learned from Apollo that lunar dust can be less than 20 microns… very fine, abrasive and sharp, like tiny pieces of glass, making it more of a dangerous threat than just a simple nuisance.” - Sharon Miller, NASA Glenn Research Center

Now, imagine robotic systems designed to operate not for 75 hours, but for years on end. Every joint, seal, solar panel, and connector is a potential point of catastrophic failure. The chasm between a three-day Apollo jaunt and a permanent outpost is the engineering problem that keeps scientists up at night. This is where the real war will be fought, not by astronauts, but by robotic systems built for unprecedented durability and, crucially, autonomous self-repair.

Rise of the Robotic Grafters

Humans are fragile, expensive cargo. The dirty, dangerous, and repetitive graft of building “Moon Base Alpha” will fall to a new generation of space-hardened robots. We are looking at a robotic ecosystem far more sophisticated than anything we’ve ever launched.

Construction Bots: Autonomous rovers will be tasked with levelling the lunar terrain, winching modules into place, and piling up berms for radiation shielding. Companies like Astrolab and Lunar Outpost are already perfecting the Lunar Terrain Vehicles (LTVs) that will serve as the workhorses for both machines and humans.
Mining and Utility Drones: To tap into that vital water ice, NASA envisions a fleet of robotic systems, including hopping “MoonFall” drones inspired by the Mars Ingenuity helicopter, capable of diving into treacherous, pitch-black craters.
Nuclear Technicians: Deploying and maintaining a fission reactor on the Moon is a job you’d much rather delegate to something that doesn’t have DNA to worry about. The Fission Surface Power project is one of the most critical—and robot-dependent—pillars of the entire mission.

This robotic workforce won’t just be remote-controlled from a desk in Houston. The communication lag and the sheer complexity of the work demand high levels of autonomy. These machines will need to diagnose their own glitches, navigate lethal terrain, and collaborate with one another to get the job done.

The Real Prize: Mars

As audacious as a $30 billion Moon base sounds, it is effectively a dress rehearsal. NASA has been explicit: every scrap of technology and operational experience gained on the Moon is a direct stepping stone to putting boots on Mars. Learning to extract water, generate nuclear power, and build habitats in a vacuum just a few days’ flight from home is infinitely more sensible than trying to wing it on a planet that’s a six-month journey away.

The multi-planetary economy is no longer a sci-fi trope; it’s a line item in the budget. While legacy aerospace firms struggle to get capsules into low-Earth orbit, NASA is architecting a future where commercial heavy-lifters like SpaceX’s Starship act as the freight trains for a new industrial frontier. The first pioneers on this frontier won’t be made of flesh and bone. They’ll be made of metal and silicon, and their primary job is to survive the dust. If they can pull it off, humanity might just have a future beyond this pale blue dot.

Shift Offers Free Cleaning, But Your Mess Is Training the Robot That Will Replace You

Robot King — Thu, 28 May 2026 00:00:00 +0000

In a city where the “no such thing as a free lunch” mantra is practically written into the building codes, a new startup is betting that New Yorkers will happily settle for a free deep clean instead. Shift has officially touched down in New York City, offering professional flat cleaning at the unbeatable price of zero pounds. The catch—and in the tech world, there is always a catch—is that you aren’t just getting your skirting boards scrubbed; you’re providing the raw data to teach robots how to do it themselves.

The process is deceptively simple, if a little “Black Mirror” in practice. A vetted “Shift Operator” arrives at your home wearing a head-mounted device that records a first-person perspective of the entire cleaning process. They scrub, they hoover, they leave, and your wallet remains firmly shut. In exchange for the complimentary service, Shift walks away with a high-fidelity video dataset of a human performing complex, real-world tasks. According to the company, this footage is anonymised before being processed and licensed to AI and robotics firms to train the next generation of domestic androids.

Shift is tackling one of the biggest bottlenecks in embodied AI: the desperate need for high-quality, diverse, and messy real-world training data. While labs spend fortunes on sterile simulations, Shift has turned the entire city into a data farm, with your dust bunnies and dirty dishes serving as the curriculum. The company is already planning to expand its data-for-service model beyond cleaning to include handymen, repairs, and other errands on a global scale.

Why is this important?

Shift’s business model is a brilliantly pragmatic, if slightly dystopian, solution to the data acquisition problem. Instead of paying for data, they are bartering with a service that has tangible value, effectively creating a new kind of currency backed by clean countertops. The company, which appears to operate under the legal name GETTHESHIFT, INC., is making an explicit trade: your privacy for convenience.

While Shift assures users that footage is anonymised and never used for advertising, the service represents a new frontier in the “you are the product” economy. It is one thing to give up your search history, but another entirely to provide a POV stream of your unkempt bedroom. For now, early public sentiment seems positive, with many viewing it as a fair exchange for a tangible benefit. Ultimately, Shift is betting that for the price of a free cleaning, people are willing to let a robot-in-training learn from their lifestyle—one scrubbed toilet at a time.

X Square's Robot Will Clean Your Home for $22—With a Human Chaperone

Robot King — Tue, 26 May 2026 00:00:00 +0000

The dream of a robotic butler tidying up your cluttered flat just took a surprisingly pragmatic, if slightly unusual, leap forward in China. Shenzhen-based startup X Square Robot has unleashed its wheeled humanoid robots into the chaotic world of domestic cleaning, but this is no solo mission. For roughly 149 RMB (about £16), residents in Shenzhen and Beijing can book a cleaning service where a robot arrives accompanied by a professional human minder.

The service, launched in partnership with Chinese e-commerce giant 58.com, has reportedly been fully booked for weeks. The division of labour is clear: the robot—likely the company’s Quanta X2 model—handles the repetitive, structured graft like tidying shoes, clearing tables, and bagging up rubbish. Its human partner tackles the complex, deep-cleaning jobs that still require a bit of elbow grease and a keen eye, such as scrubbing the kitchen and bathroom. X Square has been refreshingly honest about the robot’s current teething problems, noting on social media that they “may move slowly, hesitate, and sometimes look a little clumsy.”

This isn’t just about sparkling floors; it’s a clever, large-scale experiment in training embodied AI. At its “Born to Bot, Bot to Family” event on 23 April 2026, the company promised its robots would be in real homes within 35 days—a bold claim it has now made good on. The entire operation is designed to feed X Square’s WALL series of foundation models with invaluable data from unpredictable, real-world domestic environments—the final frontier for general-purpose robotics.

Why is this important?

While competitors are still showing off flashy but brittle demos in sterile labs, X Square is stress-testing its AI in the wild and getting customers to foot the bill for the privilege. This human-in-the-loop model is a brilliant bit of lateral thinking that sidesteps the current reliability gap in robotics. It allows the company to deploy a 70%-capable robot today, rather than waiting a decade for a 99% autonomous one.

By charging the same price as a traditional cleaning service, X Square has created a powerful feedback loop: it generates revenue, gathers immense quantities of training data from diverse households, and gets the public used to having a robot in the spare room. It’s a refreshingly honest and practical approach to commercialising a technology that always seems to be “just around the corner.” The future of home robotics, it seems, will be a tag-team effort for the foreseeable future.

LimX Luna Ditches Chores for Choreography in Humanoid Dance-Off

Robot King — Mon, 25 May 2026 00:00:00 +0000

Just as we were getting used to the idea of humanoid robots as the ultimate warehouse drones, designed to stack boxes or weld car doors with joyless precision, Shenzhen-based LimX Dynamics has pirouetted onto the stage with a rather different ambition. Meet Luna, a full-size humanoid that seems less interested in manual labour and more concerned with mastering its catwalk strut and gymnastics routines. Unveiled in a live launch event, Luna is being pitched not as a factory workhorse, but as an interactive performer—a piece of “limitless art” in fluid motion.

While competitors are laser-focused on automating the world’s dull, dirty, and dangerous jobs, LimX is making a calculated bet on a different market: the stage. The humanoid market isn’t just about logistics anymore; it’s looking for a Vegas residency. And with the global market for entertainment humanoids projected to skyrocket from around £245 million ($310 million) in 2024 to nearly £6.1 billion ($7.8 billion) by 2034, it’s a gamble that could pay off handsomely.

Not Just Another Pretty Faceplate

Luna stands at a human-like 5ft 3in (160cm) and weighs 54kg, but the spec sheet quickly diverges from its more utilitarian cousins. It’s a sleeker, more elegant machine than the company’s previous general-purpose humanoid, Oli, which was built for more rugged, all-terrain research. Luna’s design language is deliberately softened with premium textile finishes, aiming for “approachable aesthetics” rather than an industrial, Terminator-lite vibe.

Under the bonnet, however, it’s all business. Luna is powered by a second-generation “Sys 0” motion engine with upgraded joint motors and 27 degrees of freedom, enabling it to pull off advanced manoeuvres like side flips and the Ilyushin spin. The company is boasting some serious performance tweaks, including a 30% reduction in joint surface temperature thanks to upgraded thermal management—critical for preventing a meltdown during an extended dance number. More impressively, an enhanced battery management system boosts stamina by 150%, offering about 4 hours of untethered performance. For those all-night raves, it can also be plugged in for 24/7 operation.

This isn’t the first bipedal robot from LimX Dynamics, a company founded in 2022 that has rapidly raised over £235 million ($296 million) from heavy-hitters like JD.com and Alibaba. Their portfolio includes the research-focused Oli and the modular TRON biped. Luna represents a strategic pivot, or at least a significant diversification, into a commercial vertical where aesthetics and performance are the main event.

The AI Choreographer and The Robot Army

What truly sets Luna apart is its brain, or rather, the software that makes it dance. LimX has equipped Luna with a suite of tools designed for people who wouldn’t know a line of C++ if it hit them. The “Video to Motion” feature allows a user to simply upload a video of a human dancing, and Luna’s AI analyses and replicates the movements. This, combined with kinesthetic “teach by showing” programming and an AI Task Editor, lowers the barrier to creating complex routines from a coding marathon to a simple natural language prompt.

Simply describe your scenario, and AI composes dance, motion, visual, and voice into complete tasks.

This ease of use is crucial, but the real showstopper is the swarm control. A single operator can synchronise and command a troupe of over 200 Luna units with millisecond precision. The implications are obvious: massive, visually stunning robotic performances at opening ceremonies, concerts, and theme parks. Forget a flash mob; we’re talking about a full-scale robot spectacle.

Of course, with a fleet of high-kicking robots comes the inevitable health and safety question. LimX has implemented a four-layer safety architecture that includes active fall mitigation (it goes into a “damped state” to soften the blow), external force sensing to halt operations upon impact, and the ever-reassuring big red hardware emergency-stop button.

A New Niche in the Humanoid Market

Luna isn’t here to scrap with Figure in a factory or Tesla’s Optimus in a lab. It’s carving out a new role for humanoids as brand ambassadors, performers, and interactive centrepieces. The public debut at the Taobao Influencer Festival in March 2026, complete with a red-carpet catwalk, underscored this strategy perfectly. This is a robot designed to be seen.

While no official pricing has been listed for the UK market, reports from its China release peg the starting price around £34,000 ($43,000), with an introductory offer for the first 100 units at roughly £30,000 ($38,000). This places it firmly in the professional-grade category—well above consumer toys, but potentially accessible for businesses in the events and entertainment space. For comparison, the more development-focused LimX Oli EDU platform is listed at nearly £47,000.

The launch of Luna signals a fascinating maturation of the humanoid market. As the underlying tech of bipedal locomotion and AI control becomes more robust, companies are beginning to specialise. Not every humanoid needs to be a jack-of-all-trades. Some can be specialists, and Luna is the first high-profile humanoid to specialise in the art of performance. The robot revolution might not just be about automating labour; it might also be about stealing the limelight.

CARA 2.0 Slashes Cost by 52%, Then Its Creator Abandons the Core Tech

Robot King — Sat, 23 May 2026 00:00:00 +0000

Maker and YouTuber Aaed Musa has returned with a major overhaul of his idiosyncratic rope-driven robot dog, and the stats are genuinely impressive. The new CARA 2.0, born from a university senior engineering project, has managed to slash the price tag of its predecessor by more than half, tumbling from $3,000 to a far more palatable $1,450 (roughly £1,150). It’s also undergone a serious diet, shedding 42% of its bulk to drop from 14.25kg to a nimble 8.26kg. This is the successor to the original CARA: The Rope-Driven Robot Dog Revolution that first turned heads with its clever, low-backlash capstan drives.

Developed alongside a team as a final-year capstone project, CARA 2.0 clocks a walking speed of 0.55 m/s and can lug a 6.8kg payload for about an hour. While the team’s ambitious “moonshot” goal was to break the sub-$1,000 barrier, a final build cost of $1,450 for a dynamic quadruped of this calibre is certainly not to be sniffed at. These savings were found by ditching pricey carbon fibre tubes in favour of 3D-printed structures and sourcing budget drone motors—which the team then had to painstakingly rewind by hand to triple their torque output.

Why does this matter?

Here’s the kicker: despite the project’s success, Musa is officially calling time on the very technology that gave the robot its name. CARA stands for “Capstans Are Really Awesome,” but after this latest build, Musa has concluded that while they might be awesome in theory, they are deeply impractical in the real world. In his project post-mortem, he confirmed he is retiring the design and won’t be revisiting capstan drives, lamenting that they “just aren’t very assembly-friendly.”

It’s a classic engineering lesson learned the hard way. A design can be a masterstroke on paper—offering zero backlash and high performance—but if it’s a total nightmare to build and maintain, it’s a dead end for practical use. Musa notes that for his next quadruped, he’ll be “using off-the-shelf actuators.” It’s a pragmatic, if slightly bittersweet, conclusion to a project that successfully built a better, cheaper robot dog, only to prove that its core innovation was a beautiful but flawed premise. You can dive into all the technical nitty-gritty on Musa’s project page.