Tag Archives: Robots

This palm-sized robot dog has a Nokia battery and a customizable LEGO-compatible back

Posted on by

Petoi has been shrinking Boston Dynamics down to desk size for almost a decade now. Dr. Rongzhong Li started the company in his dorm lab in 2016, after a video of his open-source robot cat prototype went viral. That cat became Nybble, a laser-cut wooden kitty, and it was followed by Bittle, a palm-sized robot dog that ended up in classrooms, research labs, and a surprising number of home offices. Petoi says it has shipped more than 30,000 robots to over 60 countries, all running its open-source OpenCat framework. Each new model has gotten smaller, smarter, and easier to build, which brings us to the newest pup in the litter.

Meet Quaddle, a palm-sized robot dog that walks, glides, spins, and does backflips using just four servos. That’s half what Bittle uses in its legs and roughly a third of what most quadrupeds need, thanks to a spring-loaded linkage Petoi calls the MinDoF leg, which lets one motor drive both the shoulder and the knee. It weighs about 6 ounces, snaps together without screws, and takes LEGO bricks on its back. You can teach it tricks by physically moving its legs, chat with it, or code it in anything from beginner blocks to ROS. Prices start at $179, which is pretty reasonable for a dog that never needs walking.

Designer: Petoi

How one servo runs an entire leg

Each of Quaddle’s legs hangs off a single servo, with a spring and a four-bar linkage doing the job a second motor normally would. As the servo swings the upper leg, the linkage folds the knee up and straightens it again, turning one rotating shaft into a proper step. Petoi calls this the MinDoF leg, short for Minimal Degrees of Freedom, a degree of freedom being one independent way a joint can move. The spring clips onto different anchor points, changing how stiff or bouncy each leg feels. In Petoi’s tabletop tests, Quaddle walks about one body length per second, turns 90° in a single step cycle, and climbs onto a ledge as tall as its own knee.

Walking machines driven by a single crank go back to Pafnuty Chebyshev, the Russian mathematician whose 1870s linkage turned rotation into a surprisingly lifelike stride. Theo Jansen’s wind-powered Strandbeests on Dutch beaches run on the same basic idea, and Quaddle feels like their tiny, battery-powered cousin. Petoi added one more trick by fitting small bearings to the tips of the feet. Angle the legs just right and those bearings become wheels, letting Quaddle glide across a smooth desk at around three body lengths per second without a single extra motor. Fewer motors also means less weight, less cost, and less battery drain, which is how the whole dog comes in at roughly 6 ounces.

Teach it tricks by hand, then teach it in code

On the Buddy and Scout models, every leg servo reports its position back to the robot, so you can teach Quaddle a move by simply grabbing its legs. Tap to start recording, guide the legs through whatever silly dance you have in mind, and tap again to have it play the motion back. There’s no app or code involved, and Petoi says you can teach it a basic walk in about ten seconds. Recorded moves can be pulled into Petoi’s Skill Composer software for fine-tuning and shared with other Quaddle owners. For a kid, this is probably the fastest route from “I wonder if” to a robot actually doing the thing.

Buddy and Scout split their thinking across two ESP32-S3 chips, one for motion and balance and one for voice chat and smart home control. A physical slide switch cuts power to the AI chip entirely, which is great for privacy and for days when your robot dog gets a bit too chatty. Coding starts with drag-and-drop blocks, moves to Python or MicroPython, and eventually reaches the Arduino C++ firmware. Bigger projects can mount a Raspberry Pi Zero on its back or plug into ROS, the software framework most research robots run on. A browser-based simulator runs the same code as the real robot, so you can test ideas on a laptop or even a phone with no robot in sight.

LEGO bricks, a Nokia battery, and three Quaddles to pick from

The top of Quaddle’s shell takes LEGO bricks, and Petoi will publish printable shells and mounts for anyone with a 3D printer. Petoi’s own demos have turned it into a walking LEGO AT-AT, a waddling biped called the NanoDuck, and a rover on Mecanum wheels, which use angled rollers to slide sideways. Power comes from a swappable Nokia-format phone battery, the same BL-5C shape that lived inside Nokia’s famously indestructible early-2000s handsets. Quaddle ships with the bigger 1,800 mAh BL-10C cell, good for roughly 1 to 1.5 hours of walking depending on the model. Plug it into USB-C while you code and it charges as it runs, so marathon debugging sessions are fair game.

Builder PLUS ($179) handles walking and coding, Buddy ($219) adds AI chat and teach-by-hand, and Scout ($269) adds a sensor head that detects gestures, light, distance, and people. All three are priced at $144, $175, and $215, with deliveries starting in December. For context, Petoi’s own Bittle X lists at $279, Unitree’s Go2 starts around $1,600, and Sony’s Aibo launched in the US at $2,900. Petoi spent a decade shrinking big robots down to desk size, and Quaddle shows the smartest way to shrink one further is to give it fewer parts.

The post This palm-sized robot dog has a Nokia battery and a customizable LEGO-compatible back first appeared on Yanko Design.

This Robotic Pavilion Concept Would Change Its Layout Only When Empty

Posted on by

A public pavilion can give people somewhere to gather, but returning to it usually means finding the same spaces and following the same paths. Self-Assembling Robotic Environments explores a more flexible alternative: mobile architectural sections that could rearrange themselves between visits. This unbuilt public art concept would pause whenever someone enters, letting people explore one arrangement before the space changes for another group.

Each section combines a floor, wall, and roof in a large C-shaped profile, like a little shelter with its sides left open. Two versions stand about 10 and 14 feet tall, both four feet wide. Place several together, and those simple shapes could become passages or covered pockets. Separate them again and the same pieces would offer a different place to wander.

Designer: Michael Jantzen

Of course, rearranging those sections wouldn’t mean asking someone to wheel them into position. Each floor would contain its own computer, sensors, motors, wheels, and battery, allowing the entire unit to travel forward, backward, or sideways. The wall and roof would move with it. The robotics aren’t proposed as equipment that builds the installation and then leaves; they’d remain part of the architecture.

Rather than following one permanent floor plan, the modules would use artificial intelligence to decide on their positions through majority agreement. The proposal doesn’t explain how they would evaluate possible arrangements, though. A gap between two units could become a passage, while several facing one another might define a gathering spot. The same pieces could give visitors different ways to use the space.

The roof would do more than shelter whoever stands beneath it. Solar cells on top would recharge the battery inside the floor, keeping the proposed power supply within each moving section. It’s a sensible place to collect sunlight, but the project doesn’t specify how much energy movement would require. Keeping these large pieces in motion isn’t as simple as putting solar panels overhead.

A black perimeter would mark the robots’ territory, containing their movements within a circular arena. Sensors would watch for people or objects entering that space, pausing every module until the area is clear. You wouldn’t be expected to dodge a moving wall while exploring. The intended sequence is straightforward: the pieces rearrange when empty, then hold their positions while visitors are inside.

That rule still needs engineering before anyone could walk in. Detecting a visitor isn’t the same as stopping safely, and the project doesn’t establish braking distances or structural performance. Raised floor sections also need an accessible way in. These are questions for a working installation, rather than reasons to dismiss the idea, but the concept drawings don’t settle them on their own.

For visitors, the pause could make staying put part of the interaction. Sit inside a sheltered corner for a while, and that particular arrangement would remain available until everyone leaves. A later group might find a different route or gathering spot. Someone lingering to finish a conversation would therefore be doing something rather unusual: keeping a version of the architecture from disappearing yet.

The post This Robotic Pavilion Concept Would Change Its Layout Only When Empty first appeared on Yanko Design.

A $599 Robot Dog That Fits on Your Desk and Works Like LEGO Technic

Posted on by

A robot dog usually arrives with a trick and a lot of wiring. Q8botOne takes the opposite approach, putting its core electronics onto a central circuit board that also serves as its backbone. It fits on desks, but its exposed workings make it feel less like a miniature pet and more like a machine you can take apart, understand, and teach to move.

Its rectangular body doesn’t bother with a face, tail, or other attempts at animal charm. Four parallel-linkage legs do the expressive work instead, folding and extending as the robot changes pace. At about 4.7 inches long and 2.8 inches wide, Q8botOne is small enough to roam across a worktable, provided you clear a path between the coffee cup and your laptop.

Designer: ZeroWire Robotics

That open, mechanical character comes from doing away with the tangle of internal cables. Eight motor modules and two batteries plug into the main printed circuit board, which doubles as a structural spine. Fewer internal connections should mean fewer loose wires to chase when something goes wrong, although adding optional sensors or a camera can still bring cables back into the picture.

Those eight DYNAMIXEL actuators give the joints feedback, so the machine can respond to more than a simple instruction to move. The developers say it can jog, jump, and carry up to twice its weight, though those performance claims still need independent verification. For someone learning how a walking robot balances, watching a gait change on a tabletop beats an animation.

There’s a simple way to get it moving without writing code first. Power the included controller through USB-C, and its joystick and buttons can steer the robot wirelessly. Connect that controller to a computer, and it becomes a bridge to a desktop interface, an Xbox controller, or your own software. Beginners can start by driving, then try changing how the legs step.

The basic robot leaves off a built-in camera and motion sensor, which makes its design less cluttered but also means it won’t navigate the room by sight. Mounting holes and expansion connectors let tinkerers add those parts later. That turns an empty patch above the body into space for the next experiment, rather than a promise that every feature comes ready to use.

That invitation extends past plug-in accessories. The hardware plans, firmware, and assembly instructions are openly available, so someone can study the board, alter the chassis, or build a new controller around the same ideas. Its printed nylon legs and bearing-supported joints keep the mechanism legible. Instead of sealing the moving parts away, Q8botOne leaves something for a student or hobbyist to investigate.

The assembled Q8botOne is listed at $599; a $199 kit omits the eight actuators, which you’d need to supply separately. Production and delivery are still planned, so the finished machine and its claimed strength need a closer look when units arrive. For now, its most interesting move is architectural: making the circuit board carry the robot, while leaving enough room for people to make it their own.

The post A $599 Robot Dog That Fits on Your Desk and Works Like LEGO Technic first appeared on Yanko Design.

Wing’s Walmart drones are coming to Denver and Seattle — and they won’t drop your parcels in the pool

Posted on by

Lindsey Austen of Richmond, Texas, ordered cat food, light bulbs, and shoe insoles, then watched a drone hover over her backyard, open its hatch, and drop everything into her pool and hot tub. She had not realized drone delivery was selected at checkout, which might be the most 2026 sentence ever written. Her video of the splash went around the internet, and Amazon called the whole thing rare. I bring it up because that was an Amazon Prime Air drone, and it has very little in common with the Wing aircraft headed for Denver and Seattle. Same sky, completely different philosophy about where a package should end up.

Wing and Walmart announced on October 5 that Denver and Seattle join their drone network in 2027, part of a plan covering 270-plus Walmart stores. Wing’s Hummingbird hovers around 23 feet up and lowers the box on a tether, so a pool is only in play if you tell it to aim there. It cruises at up to 60 mph, and the bigger version carries up to five pounds (Wing’s own numbers). Walmart sells the whole thing as groceries in as little as 30 minutes, with Wing claiming a 23 minute average. I wanted to know what that promise looks like over your street, and what it means for delivery drivers.

Designer: Wing for Walmart

The Hummingbird is a hybrid, which means rotors for vertical lift and a fixed wing for cruising, so it launches like a drone and travels like a tiny airplane. Wing’s two versions weigh 14 and 24.3 pounds at max takeoff, per FAA filings, and the heavier one hauls up to roughly five pounds of payload. Both come out of Wing’s Aircraft Library, a parts-bin approach where shared components get remixed into different sizes. Wing built the bigger one after finding about 30 percent of deliveries needed multiple flights, according to The Register. Neither lands in your yard. They hover, lower the box on a tether, and release it at ground level.

Compare that with Amazon’s MK30, which weighs about 78 pounds empty (per DroneXL’s comparison table) and, according to Axios, comes down closest to the customer. Wing’s bet is the opposite: keep the heavy part high and put only a cardboard box within reach of your dog. It also skips a parachute, which Zipline’s 55 pound aircraft carries, so the safety case rests on being small and light. I love that as industrial design logic, because every ounce shaved is a risk shaved. Parachute research like the kirigami parachute concept chases safety by adding something, while Wing chases it by subtracting.

Wing’s own presentation puts average flight time at five minutes or less, which implies most of that 23 minutes happens on the ground. In the air, FAA filings authorize up to 400 flights per nest per day across 368 nests, which DroneXL tallies to 147,200 flights daily on paper. Wing has had a crash too: a drone on a return flight went down and caught fire in Kemah, Texas on March 1, with a power line suspected and nobody hurt. Part 108, the FAA’s routine beyond-line-of-sight rule, has sat at the White House since July 10. Meanwhile, Castle Rock and Parker froze hub permits, though towns control the parking lot and the FAA controls the sky.

Five pounds is a hard ceiling, so drones chase the small stuff first, and Walmart says customers use them for forgotten ingredients and cold medicine. DoorDash insists Dashers stay central as it launches its own drone service, while a Wing operations director has described drones as skipping recruiting, insurance, commissions, and no-shows. Concepts like the DJI Express UAV already imagine cargo hopping between cities without a dedicated driver, so the direction of travel is clear. I’d guess the first gigs to feel the squeeze are the quick, small, suburban runs. The Hummingbird’s whole trick is delivering small things gently, and Austen’s hot tub is a reminder that picking the drop spot is half the job.

The post Wing’s Walmart drones are coming to Denver and Seattle — and they won’t drop your parcels in the pool first appeared on Yanko Design.

MagSafe Drone Lets You Attach An iPhone For Perfect Hands-free Selfies And Videos

Posted on by

Flying Gimbal Drone concept with an iPhone mounted on it

Google’s Add Me is one of the stranger things a phone has ever done. Introduced on the Pixel 9, it takes a photo of your family, then hands the phone to somebody in that photo, who takes a second shot of you standing where the gap was. The phone stitches the two together and produces a group picture that never existed, of a moment where everybody was never actually in the room at once. It works remarkably well. It is also, if you think about it for longer than four seconds, deeply odd, a family portrait assembled in post like a Marvel crowd scene.

Flafa proposes the analog answer to the same question. A concept from Kaiyuan Liu and four classmates at the Guangzhou Academy of Fine Arts, it’s a small quadcopter that carries your phone into the air so the camera can hold its own position while every human stays put. The phone snaps onto a magnetic pad below four fully enclosed rotors, hangs face-out, and records while the drone flies itself. It won its category at the 2025 European Product Design Award. Same problem, opposite philosophy: Google removes the photographer with software, Flafa removes them with lift.

Designers: Kaiyuan Liu, Qiushi Wang, Yan Wang, Shiqi Liu & Xiaoyi Chen

Family posing in front of a drone with an iPhone mounted on it

Selfie drones have failed so many times that the category has a graveyard. AirSelfie tried in 2017. Snap’s Pixy launched in 2022, died four months later, and was recalled in 2024 because the batteries caught fire. They all lost for the same reason. A drone light enough to fly indoors cannot carry a camera that beats the phone already in your pocket, so you buy a flying camera, fly it twice, and quietly retire it to a drawer. Flafa declines that fight entirely by bringing no camera at all.

Mounting an iPhone onto a Drone Gimbal using MagSafe

Two small sensor windows sit on the front propeller shrouds, giving the drone its own eyes for navigation while your phone dedicates itself to recording. That division of labor is the smartest thing in the whole project. Most flying cameras ask one sensor to handle obstacle avoidance, subject tracking, and image capture simultaneously, and it shows in the footage. Here the quadcopter worries about the ceiling fan and the phone worries about your kid’s face, which is roughly how you’d split the job if you had two people.

Somebody on this team understood that a drone meant to hover near a toddler cannot look like drone. The renders show warm oat cream, soft radii, matte surfaces, and not one visible screw, putting Flafa closer to a Muji humidifier than to anything DJI has shipped. Consumer drones are almost universally styled for a man in sunglasses on a cliff, all black plastic and angular vents. That grammar makes sense in a camera bag. It makes none in a living room during a birthday party.

The mount deserves credit for paranoia. A magnetic puck holds the phone, and a hinged brace folds along the handset’s edge as mechanical backup, which suggests the team thought hard about a thousand dollars of glass becoming a lawn dart. That said, the physics remains the open question. An iPhone Pro Max weighs around 233 grams, heavier than an entire DJI Neo, and hanging that mass below the center of lift makes a pendulum. The project is called a flying gimbal, but the renders show a hinge rather than an actuated axis.

Student concepts are allowed to be ahead of their batteries. What Flafa gets right is the framing of the problem, and that part is free. Every family album has one person missing, and it’s whoever picked up the camera. Google’s answer is to invent them back in afterward. This one just sends the camera upstairs for a minute.

The post MagSafe Drone Lets You Attach An iPhone For Perfect Hands-free Selfies And Videos first appeared on Yanko Design.

Exoskeletons Always Ignored Your Knees, but Halo Doesn’t

Posted on by

Ask anyone who hikes regularly where their body actually gives out, and it’s rarely the climb up. It’s the way down. Your quads and hips can push you up a mountain just fine, but it’s your knees that absorb the pounding on the descent, on stairs, on the last mile of a long trail when your legs stop cooperating. For years, the exoskeleton industry has quietly ignored this. Every consumer exoskeleton I’ve come across, including Hypershell’s own X series, has focused almost entirely on the hips. That’s helped people hike longer and climb steeper grades, but it never addressed the part of the leg that actually fails first.

Hypershell Halo is the company’s attempt to fix that gap, and it’s the first exoskeleton I’ve seen that treats the hip and knee as one connected system rather than bolting a knee brace onto an existing hip rig. The Halo runs four motors total, two at the hips and two at the knees, and Hypershell says they’re coordinated through a single unified architecture instead of running as separate, half-communicating systems. That distinction matters more than it sounds. Earlier attempts at dual-joint support, including some of Hypershell’s own internal prototypes, were reportedly scrapped because the control systems couldn’t stay in sync with each other, which is a nice way of saying the hips and knees were fighting instead of working together.

Designer: Hypershell

What actually convinces me the Halo is worth paying attention to is the hardware restraint. A carbon-fiber and titanium frame weighing about 2.6 kilograms, four motors producing up to 1,490 watts combined, and a swappable 370-gram battery good for roughly 20 kilometers of assisted walking. It folds down to under 8 liters, which is small enough to strap to the outside of a daypack, and Hypershell claims you can get it on in about 20 seconds, even with a full backpack already on your back. None of these numbers alone are revolutionary. Put together, they suggest a company that understands the actual friction point for anyone considering strapping a robot to their legs: if it’s heavy, slow to put on, or awkward with gear you’re already carrying, it stays in the closet.

The AI layer is where Hypershell is clearly trying to differentiate from the growing pack of exoskeleton makers chasing the same idea. HyperIntuition 2.0 is supposed to read terrain and movement in real time and adjust assistance without you fiddling with a mode dial mid-hike. There’s also Shelly, a chatbot-style assistant you control through a smartwatch, which sounds almost gimmicky until you consider that most people don’t want to stop on a trail to change settings on a phone. Whether Shelly ends up being genuinely useful or just a marketing flourish is something I won’t know until someone puts it through a real weekend hike, but the instinct behind it, reducing friction between the wearer and the machine, is the right one.

At $2,299 for the Origin Edition, the Halo isn’t cheap, and it’s priced well above the $999 to $1,999 range of Hypershell’s hip-only X models. That premium buys you the first product in this category actually built around full-leg coordination rather than partial support, and for anyone who has skipped a hike because their knees couldn’t handle the descent, that might be the entire pitch. I don’t think this device turns casual walkers into trail athletes overnight, and I’d be skeptical of any brand claiming it does. But for the growing number of people managing knee wear, recovering from injury, or just getting older and less willing to accept stairs as an enemy, a machine that finally treats the knee as seriously as the hip feels less like a novelty and more like an overdue correction.

Pre-orders are open now for the US, Canada, Europe, the UK, and China, with shipping expected in November. Whether the Halo lives up to its own architecture claims once it’s on real trails and real staircases is still an open question, but it’s the first exoskeleton pitch in a while that’s made me want to find out.

The post Exoskeletons Always Ignored Your Knees, but Halo Doesn’t first appeared on Yanko Design.

Smart Glasses, Robot Dogs, AI Tamagotchis, and Desk-Sized Supercomputers: How IFA 2026 Redefined Robots

Posted on by

The creator economy reshaped who makes content, and then quietly reshaped what tools are worth building. When millions of people are shooting, editing, scanning, and streaming from wherever they happen to be, the market for professional-grade hardware that works without a full production crew gets very real, very fast. That’s the underlying logic behind a lot of what’s in this roundup, running from portable LiDAR scanners and robotic camera arms all the way to ambient AI screens that sit on your desk and quietly keep track of everything.

Companionship, it turns out, is harder to engineer than most consumer tech companies have been willing to admit. Hengbot’s Sirius is a robot dog that owners can train like a real one, using a Creator Studio that requires no coding. Takway AI’s Sweekar hatches from an egg and grows through four life stages, shaped entirely by how its owner treats it. Mind with Heart Robotics built AnAn, a panda-shaped companion that has been piloted in eldercare and dementia care settings in Hong Kong and Shanghai. These are serious attempts at a genuinely hard question: what does it look like when AI lives with you, rather than just working for you? Here’s a look at some such brands and their innovative products at the Global Connect Show at IFA 2026.

INMO GO3: When Your Glasses Speak 98 Languages So You Don’t Have To

Smart glasses have had a rough decade of almost. Google Glass made everyone nervous at restaurants. The Ray-Ban Metas got people excited again, but a camera and a speaker do not a smart display make (not to mention the PR disaster they’ve been having this past year). INMO has been quietly chipping away at the real version of this category for a while now, with the Air series handling light office and audio-visual entertainment, and the GO series zeroing in on translation. The GO3 is that third iteration, and the brief is tighter and more useful than anything the brand has shipped before: a lightweight pair of glasses combining a discreet monochrome display with real-time AI assistance, in a form factor that resembles regular eyewear.

The translation engine handles 77 source languages and 98 target languages, with offline support for nine of them when you’re somewhere the internet has opinions about itself. The engineering detail that deserves more attention, though, is the battery system. Magnetic battery replacement takes approximately five seconds, and removed batteries automatically recharge in the case, enabling uninterrupted use throughout the day. The dual-eye green Micro-LED display with diffractive waveguide optics hits up to 1,500 nits at 640×480 resolution, staying visually discreet enough that nobody beside you would know you’re reading meeting notes. A built-in camera privacy cover physically shields the camera for more comfortable use in public or social settings, which is one of those small decisions that says a lot about how seriously a company has thought through the real-world context its product actually lives in.

Hengbot Sirius: The Robot Dog That Actually Learns New Tricks

When we covered Sirius on Yanko Design last year, we noted that it sat in a gap the robot dog market had left frustratingly empty for years, somewhere between Boston Dynamics-level industrial machines that cost more than a car and toy-grade STEM kits that shuffled around like arthritic spiders. Hengbot filled that gap, and the Kickstarter campaign backed that read up with over $308,000 raised. At IFA, the story moves forward: the focus is on how continued development in voice interaction and motion control is making Sirius easier to communicate with and more capable in physical movement. You tell it to sit, move forward, or shake hands, and Sirius translates these instructions into physical actions, letting people interact with the robot without relying entirely on conventional controls.

The motion engineering is where it gets genuinely interesting. Rather than releasing physical force abruptly, Sirius can reduce it progressively, which allows certain movements and physical responses to feel smoother and less rigid while reducing sudden loads on the mechanical system. If it falls over, it can perform a recovery sequence and return to a standing position, recovering from common physical interruptions without requiring the user to reposition it manually. Packed into a body weighing just over a kilogram, with a removable 2,250 mAh battery, 30W PD charging, and developer support spanning Python, C, C++, and ROS 2 interfaces, Sirius is targeting consumers and creators in equal measure, which is a harder pitch to land than it sounds, and one Hengbot is getting noticeably better at making.

Rokid Glasses: The AI Glasses That Let You Choose Your AI

If the INMO GO3 is purpose-built around translation and teleprompters, the Rokid Glasses are playing a broader game entirely. Rokid has been in this space since 2014, long before smart glasses were a mainstream conversation, and we’ve covered the brand’s journey on Yanko Design from its early AR gaming hardware all the way through to a full profile on how it’s building the glasses that Google couldn’t. The Rokid Ai Glasses Style Pack, announced at CES 2026 and covered here in March, brings that decade-plus of development into what is comfortably the most ambitious feature set in the display-less glasses category. At just 38.5 grams, in a TR90 frame with titanium alloy hinges and a 12MP Sony IMX681 camera with electronic image stabilization, the hardware alone makes a strong case. The Power Capsule that comes bundled in the Style Pack delivers 6+ charges and gets the glasses back to 80% in 20 minutes, so battery anxiety largely exits the conversation.

Rokid is the world’s first manufacturer to natively support Google’s Gemini on a pair of AI smart glasses, with the Ai Glasses Style integrating both OpenAI’s ChatGPT and Google Gemini to answer almost any question. That’s not just a spec flex, it’s a genuine design philosophy – the glasses function as an open platform rather than a locked ecosystem, which is exactly the kind of thinking that will matter as AI models continue to evolve fast. The Agent Store, a centralized platform where developers can create and share AI apps for Rokid’s smart glasses, has already attracted more than 3,000 submissions for new agent workflows, with more than 400 agents published and available to users. Real-time translation in 89 languages, object and scene recognition, live transcription, and posture monitoring round out a feature set that makes these feel less like a gadget and more like a platform you happen to wear on your face.

Mind With Heart Robotics AnAn: The Companion Robot That Sits With You, Not For You

The conversation around companion robots has been building for a while on Yanko Design, from Sony’s Aibo to Hengbot’s Sirius to Sweekar, and a pattern is becoming clear: the more a robot is designed to perform, the less it feels like company. Mind With Heart Robotics took a different path entirely, and AnAn is the result. A panda-shaped companion robot built from soft, lifelike materials with full-body tactile sensing, AnAn is designed for companionship and elderly support first, and everything else second. The live demonstrations at IFA show AnAn’s full-body tactile sensing system responding to touch through gentle movements and emotional feedback, along with proactive interaction, gentle reminder functions, and its ability to learn a user’s interaction habits over time through long-term memory. That last part matters more than the spec sheet: a companion that actually remembers you is a fundamentally different product from one that resets. Mind With Heart Robotics develops emotionally intelligent companion robots across multiple form factors designed for emotional support, elderly companionship, children’s well-being, and institutional care, guided by its mission to Heal The World, backed by a patent portfolio of more than 30 filings and 18 patents granted to date.

IFA Berlin marks AnAn’s European debut, appearing at IFA’s official Robots on the Runway showcase, following its recognition as a CES 2026 Innovation Awards Honoree in the Artificial Intelligence category. The company is also debuting its Duncan Series companion robots in Europe, which are designed for neurodivergent children as well as typically developing children in early education. So the product family spans elderly care on one end and early childhood development on the other, which is a more considered and humane product brief than most robotics companies have attempted. AnAn goes on sale in Europe in September 2026 at retail pricing between $1,000 and $1,500, with distributor agreements already signed in Germany and Belgium, a completed production run of 100 units with samples available for immediate shipment, and a dedicated European repair and after-sales center in Belgium supporting customers across the region. The B2B version includes secure dashboards tracking behavioral trends and interaction patterns for care providers, meaning the product has been thought through at the institutional level too, for care homes, therapy settings, and clinical environments where the robot needs to report back as well as show up. Soft on the outside, seriously engineered on the inside.

XGrids Lixel CyberColor + PortalCam: The World Isn’t Flat, and Now Your Camera Knows It

3D Gaussian Splatting sounds like something a physics PhD invented to win an argument at a dinner party, but it has quietly become one of the most consequential breakthroughs in spatial computing in years. The short version: rather than building a 3D scene out of polygons the way traditional CGI does, 3DGS represents a space using millions of tiny colored blobs, or “gaussians,” that together produce photo-realistic environments you can actually navigate through in real time. XGrids has been building its entire product stack around this technology, and the two products at IFA represent opposite ends of the same pipeline. The PortalCam is billed as the first handheld device that captures reality as interactive 3D environments using 3DGS technology, combining LiDAR, a four-camera array, and automated processing to create photorealistic spatial models you can walk through, edit, and share across platforms. At under 900 grams, it can scan 100 square metres in 10 minutes at walking speed. The LiDAR system fires 856,000 points per second across a 180-degree azimuth by 180-degree elevation field of view, while the four-camera array pairs two fisheye lenses with two front-facing cameras, each fisheye covering 200 degrees. That’s a lot of spatial data coming in very fast, from a device you can carry in a backpack.

The Lixel CyberColor platform, or LCC, is where all that captured data becomes something usable. It combines Multi-SLAM precision with optimized 3D Gaussian Splatting to create measurement-grade, photorealistic, lightweight 3D models, bridging the gap between technical accuracy and immersive visualization for professionals across industries. The compression story is particularly worth noting: a breakthrough 90% compression technology dramatically reduces file sizes without compromising quality, enabling smooth playback on cloud, edge, and mobile. That solves the problem that has always made spatial content impractical outside of specialized production environments, the files are simply too large to move around and too heavy to stream. Ready-made integration with Unity, Unreal Engine, WebGL via Cesium and Three.js, mobile apps, and XR applications means the output plugs directly into workflows that architects, game developers, and filmmakers are already using. And then there’s the surprise announcement at the show: LCC Scan, a new iOS application that enables users to transform ordinary photos and videos into detailed 3DGS models directly on their iPhones. The professional pipeline and the consumer entry point, both at IFA, in the same week. That’s a confident move.

Sweekar: You Don’t Train This Robot… You Nurture It.

Hengbot’s Sirius is a genuinely impressive piece of engineering, a robot you put through its paces, teach tricks, and watch execute a controlled landing from an elevated surface. It rewards the tinkerer. Sweekar, from Takway AI, is asking a completely different question: what if the robot didn’t perform for you, but grew because of you? Not every AI companion needs to be a workhorse. Sometimes the whole point is that it just shows up, and keeps showing up. Sweekar is an AI pocket pet inspired by the iconic Tamagotchi, but rebuilt as a living, evolving AI companion. It hatches from a physical egg, grows through four life stages, and develops a personality shaped by the relationship you build with it. The growth mechanic has real engineering behind it: every interaction, feeding, talking, cleaning, picking it up, generates XP, and when Sweekar levels up, it physically grows a few millimetres taller, driven by a precision stepper motor, plays a special evolution animation, and unlocks new behaviors, new language, and new personality dimensions. That’s a stepper motor doing emotional labor, which is a sentence that shouldn’t work but somehow absolutely does.

The part that actually separates Sweekar from everything else in this category is the memory system. Most AI companions have the same gap: sessions reset, context vanishes, and the relationship never actually deepens. Users report the same frustration: it feels like starting over with a friendly stranger every single time. Sweekar’s long-term emotional memory retains the owner’s voice, daily habits, shared stories, favorite things, and small victories, and surfaces them naturally in future conversations. The longer you raise it, the more it sounds like it genuinely knows you. Voice recognition is trained specifically to the primary owner, and the personality develops through an MBTI-inspired behavioral framework, meaning two people who buy the same unit on the same day will have fundamentally different companions very soon, because every person raises things differently. At 89 grams it fits in a pocket, runs on a phone hotspot so it travels with you, and from Level 30 onward, Sweekar cannot die. The emotional investment compounds for years. What you build together is yours and it keeps going. That last detail is the one that sticks. Someone actually engineered emotional longevity into the product brief, and that’s worth paying attention to.

Navimow H5 Pro: The Robotic Mower That Finally Does the Edges

Robotic lawn mowers have gotten remarkably good at the easy 80% of the job and left everyone dealing with the same annoying leftover, the trim along fences, flower beds, and hedgerows that no autonomous mower could quite reach. Segway Navimow, the world’s No. 1 wire-free robotic lawn mower brand by retail sales volume for two consecutive years, built the H5 Pro specifically to close that gap, using a breakthrough EdgeMaestro extendable robotic arm for true zero-edge mowing. The arm pushes the cutting system out 7.7cm beyond the mower’s body, letting it trim flush against borders and obstacles instead of leaving that familiar uncut strip that sends homeowners out with a separate trimmer every weekend. It’s a genuinely clever mechanical solution to a problem that’s been treated as an acceptable limitation for years, mostly because nobody had bothered to engineer around it properly.

What makes the H5 Pro worth watching at IFA specifically is that this is Navimow’s first appearance at the show, after building its reputation almost entirely on the gardening and outdoor living expo circuit. That’s a notable shift in venue for a brand whose audience has traditionally been landscapers and homeowners rather than gadget crowds, and it says something about how robotic garden tech is increasingly being read as consumer tech rather than a niche horticultural category. The H5 Pro also carries a thermal sensor that protects hedgehogs and other small animals even in low-light conditions, a detail that fits Segway’s stated push toward mowing technology that’s thoughtful about the wildlife sharing the same lawn, not just efficient at cutting grass. Alongside the existing X4, H2, i2, and Terranox lines, the H5 Pro rounds out a lineup built for gardens of every size and complexity, and its arrival at a tech show rather than a gardening fair suggests Navimow is angling for a bigger, more design-conscious audience going forward.

QIVIKA Smart Scope: Photograph Saturn’s Rings, From Your Pocket

Smartphone camera hardware has had a wild few years. Periscope lenses pushed optical zoom past 5x, then 10x, then further, and the gap between a phone and a dedicated camera narrowed considerably. But there’s a ceiling, and the people who actually want to see lunar craters in detail, or pick out a bird species from 200 metres away, still had to choose between carrying dedicated optics that require setup and expertise, or making do with a phone camera that simply couldn’t bridge the distance. QIVIKA’s Smart Scope is a direct answer to that gap. Using a precision periscope optical architecture to fold the optical path into a compact form factor, it delivers 25x optical magnification and up to 3000mm equivalent focal length at approximately 300g, bringing high-performance long-range optics into a genuinely portable smartphone-based system. ED low-dispersion glass and multi-coated optics help preserve clarity and contrast, bringing distant subjects from lunar details to birds and wildlife within reach of a smartphone. That’s a meaningful spec for the price point, and the form factor is the real story: it clips onto your smartphone.

The hardware is only half the pitch. QIVIKA combines long-range optics with on-device AI composition. The AI analyzes the live view directly on the device and provides real-time framing assistance, helping users capture distant subjects more effectively, without cloud processing or an internet connection. This brings computational intelligence directly into the long-range imaging experience. The app splits into three intelligent shooting modes covering solar, explore, and deep sky use, with a Smart Sky Guide that overlays constellation lines and guides users to celestial targets with on-screen directional arrows, functioning essentially as GPS for the night sky. The birding side of the app is equally considered: AI identification supports 10,000+ bird species, while Life Lists, personal birding maps, community-shared locations, and observation records turn individual sightings into a growing personal exploration history. The whole experience is built around the loop of observe, capture, identify, remember, and explore, which means the product gets more valuable the more you use it, rather than plateauing once the novelty of the optics wears off. At $359 with a Kickstarter launch coinciding with IFA, it’s targeting exactly the kind of curious outdoor person who wants to look further without becoming an astronomer.

Yanhe Technology: The Solar Cell That Fits Inside Your Smart Glasses

Most of the products in this roundup are end products, things you hold, wear, scan with, or talk to. Yanhe Technology is something different. It’s the infrastructure layer that makes the whole category more interesting. The pitch is elegantly simple: perovskite solar cells thin and efficient enough to build directly into everyday devices, so the smartwatch on your wrist, the AI glasses on your face, or the smart lock on your front door can top itself up from whatever light happens to be in the room. The indoor low-light power solution activates at an illuminance below 20 lux, achieves power conversion efficiency of more than 30% under low-light conditions, and carries an indoor service life exceeding 10 years, supporting devices such as keyboards, mice, smart locks, remote controls, AIoT sensors, and electronic shelf labels while reducing reliance on batteries or power cables. The component thickness sits below 0.5mm with an areal density below 0.15g per square centimetre, which means it conforms to curved and arbitrary freeform surfaces, including smart wearables and automotive aerodynamic contours. A module with a 1.7 square centimetre aperture can initiate power generation at just 20 lux, which is dimmer than most indoor spaces ever get.

The manufacturing story is what separates Yanhe from a university lab with a press release. The company has built the world’s first 100MW fully automated production line for consumer-grade perovskite modules, supporting 24-hour lights-out manufacturing, consistent product quality, and volume delivery. Yanhe is among the world’s first perovskite companies to receive TÜV Rheinland certification in accordance with IEC 63163. Its products are now available in more than 50 countries and integrated into the supply chains of over 100 brands, with collaborations established with Lenovo, Samsung, and Xiaomi. Founded in 2024 by a team out of Peking University and Columbia, that’s a remarkable amount of ground covered in a very short time. The bonus detail that makes this feel genuinely sci-fi: alongside its consumer business, Yanhe is developing perovskite photovoltaic solutions for space applications, has completed an engineering prototype of a flexible roll-out solar wing, and has begun ground-based engineering validation, with a satellite model and a space photovoltaic module also on display at IFA. So yes, they’re powering your keyboard from room light and also building solar wings for satellites. Just a Tuesday for Yanhe, apparently.

ACEMAGIC F9A Series: When the Server Room Fits on Your Desk

Running a local LLM at 200-billion-parameter scale on a 1.3-kilogram box that sits next to your monitor sounds like a footnote in a spec sheet, but the ACEMAGIC F9A Mini Workstation makes it the headline. The chassis comes in two configurations: the F9A-395 powered by AMD Ryzen AI Max+ 395 with 128GB of max memory at 8000MT/s, and the F9A-PRO 495 with AMD Ryzen AI Max+ PRO 495 scaling up to 192GB at 8533MT/s. On the AI inference side, the 395 supports up to 96GB of shared memory on Windows for MoE models up to 200B parameters, while the PRO 495 pushes that to 160GB of shared memory and models up to 300B parameters. Both variants share the same aerospace-grade aluminum unibody with a fingerprint-resistant finish, a hidden bottom air intake for improved airflow, and 3,000-plus ventilation holes across both sides for sustained full-power CPU performance. Keeping that thermal headroom intact is a five-heat-pipe turbo fan with a 3D air duct, sustaining 140W peak and 120W TDP without throttling during long-running AI workloads. OCuLink via PCIe 4.0×4 adds flexible expansion for external GPUs, compute cards, data acquisition cards, and industry-specific devices through an expansion dock.

The connectivity list punches well above the 2-liter form factor. Dual 2.5GbE ports handle high-speed wired data transfer for cloud computing and external services, while Wi-Fi 7 delivers twice the bandwidth of Wi-Fi 6 with support for 4096-QAM, Multi-RU, and Multi-Link Operation. HDMI 2.1, dual USB4, and DP2.1 combine to support four simultaneous 8K@60Hz displays for multi-screen creative setups. A quad-DMIC microphone array provides 360-degree far-field voice recognition with a built-in noise-canceling mic chip and 2W x 2 DTS speakers, making it viable for meeting rooms and home AI hubs. Dual M.2 NVMe slots on PCIe 4.0×4 support up to 8TB with RAID 0 and RAID 1 configurations, and both SSDs get thermal pads and aluminum plates for cooling. The broader pitch ACEMAGIC is making is a clean one: plug-and-play from a standard outlet, no server room, quiet and energy-efficient, a one-time investment with no recurring costs, and local processing for data security.

GEEKOM Mini PCs: The Mini PC That Makes Your Tower Look Embarrassed

Mini PCs have had a quiet decade of steady credibility-building. What started as underpowered boxes for digital signage and light home-theater use has, thanks largely to AMD’s Ryzen and Intel’s latest mobile silicon, turned into a genuine alternative to full desktop builds for a surprising range of workloads. GEEKOM has been one of the more consistent players in that space, and we’ve covered the brand on Yanko Design before, noting that the case for a palm-sized PC has never been more compelling than right now, when the chips inside them have stopped making excuses. At IFA, GEEKOM is showing two products that sit at very different points on the performance spectrum, and together they make a pretty complete argument for the category. The A9 Mega is the AMD flagship, built around the Ryzen AI 9 HX 370 with its 12-core, 24-thread architecture, 50 TOPS of NPU performance, and up to 64GB of DDR5-5600 RAM. That NPU figure puts it squarely in Microsoft’s Copilot+ PC certification territory, meaning this is hardware built for the current wave of on-device AI workloads, local large language models, real-time transcription, and AI-assisted creative workflows, running entirely without a cloud connection.

The IT13 Max is the Intel counterpart, and it plays a slightly different game. Built on the Intel Core i9-13980HX, a 24-core, 32-thread processor with a 5.6GHz peak boost clock, it supports up to 96GB of DDR5 RAM and up to 4TB of NVMe storage across two M.2 slots. That memory ceiling is unusual for a device this size, and it’s what makes the IT13 Max interesting for professionals running virtual machines, large datasets, or memory-intensive creative applications who genuinely cannot justify a full tower but also cannot afford to compromise. Both units support triple 4K display output, meaning a three-monitor productivity setup from a box smaller than most external hard drives, which remains one of those specs that sounds made up until you’re actually looking at the back panel. GEEKOM backs both products with a three-year warranty, which is the unsexy detail that actually closes the deal for anyone considering replacing a desktop. The whole pitch across both machines is that raw performance and physical footprint stopped being a trade-off a while ago, and these are the receipts.

The post Smart Glasses, Robot Dogs, AI Tamagotchis, and Desk-Sized Supercomputers: How IFA 2026 Redefined Robots first appeared on Yanko Design.

Finally, Learning Robotics Doesn’t Cost a Five-Figure Lab

Posted on by

For most of the last decade, if you wanted to actually build and train a robot, you needed one of three things: a university lab, a five-figure budget, or a job at a tech company willing to foot the bill. The tools existed, but the entry price kept ordinary curious people locked out of the room. That gap between “I want to understand AI in the physical world” and “I can actually afford to try it” has been the quiet, unglamorous problem holding back a whole generation of hobbyist roboticists.

Microduck, the new pre-order from French robotics company Pollen Robotics and AI platform Hugging Face, closes that gap for $399. It is a 25-centimeter biped that walks, sits, crouches, picks small objects up with its beak, and gets back up on its own after a fall. It even roller-skates. Under a shell built with 15 motors, a camera, a depth sensor, and two IMUs, it runs on fully open-source software, which means you are not just buying a cute desk companion, you are buying a real, hackable platform for training reinforcement learning models on hardware you can hold in one hand.

Designer: Pollen Robotics

Pollen Robotics is not new to this space. Last year the team launched Reachy Mini, a small desktop robot built purely for face-to-face interaction, and more than 10,000 of them found homes with developers and tinkerers around the world. Microduck is what happened when that same team asked a different question: what if the robot did not stay on the desk at all, but actually moved through your house? That shift from stationary companion to mobile, fallible, get-back-up-and-try-again machine is, to me, the more honest representation of what learning robotics actually looks like. Real robots fall down. Good ones figure out how to stand back up.

I will admit the duck framing initially struck me as a marketing gimmick, the kind of cute mascot skin slapped onto serious tech to make it more shareable. But the more I read about the design choices, the articulated beak that doubles as a gripper, the roller-skate foot swap, the four soft colorways named Cream, Graphite, Lavender, and Sky, the more it felt deliberate rather than decorative. A humanoid robot falling over looks like failure. A duck robot falling over and waddling back up looks like a duck being a duck. That emotional permission to fail without embarrassment might be the smartest design decision in the entire product, because it lowers the psychological barrier to experimentation just as much as the price tag lowers the financial one.

Weighing in under 800 grams and priced before taxes and shipping, Microduck ships to North America, Europe, and the UK before Christmas 2026. Hugging Face CEO Clem Delangue called it part of an effort to “democratize physical AI and world models,” and whatever you think of Silicon Valley mission statements, the market response backed it up fast. Within roughly a day of pre-orders opening, demand reportedly pushed sales past the $26 million mark and the duck went to backorder. That is not a niche hobbyist purchase pattern. That is a signal that a lot of Americans, tech-curious but priced out of serious robotics until now, were simply waiting for an affordable, unintimidating way in.

There is a broader mood shift happening in how the country talks about AI right now, somewhere between exhausted and skeptical, tired of chatbots that talk in circles and promises that never quite land in the real world. Microduck feels like a small, tangible answer to that fatigue. You cannot hug a language model or watch it recover from a stumble. You can with this. For $399, families, students, and weekend tinkerers get a genuine on-ramp to physical AI, not a toy pretending to be smart, but a real machine willing to be clumsy in front of you while it learns.

Pollen Robotics has been upfront that Microduck will not solve robotics’ hardest problems, and it should not have to. What it offers instead is something rarer than a breakthrough: a starting point that is actually within reach.

The post Finally, Learning Robotics Doesn’t Cost a Five-Figure Lab first appeared on Yanko Design.

Meet the Robot Dog Built to Climb Your Stairs With Your Packages

Posted on by

I’ve lost count of how many times a delivery robot has shown up in a video, rolled confidently across a parking lot, and then just stopped at the base of a staircase like it hit an invisible wall. That has always been the real bottleneck in last-mile delivery, not the miles of open road but the last twenty feet of stairs, curbs, and uneven walkways leading to an actual front door. Robot.com’s new concept, R-dog, is the first design I’ve seen that treats that problem as the whole point rather than an afterthought.

R-dog is a four-legged robot with wheels built into its feet, plus a furry tail and a face designed to look approachable rather than robotic. Its four flexible legs bend at the knee, letting it climb stairs, step over thresholds, and cross the mixed, unpredictable terrain that trips up wheel-only robots, while the wheels themselves handle flat ground efficiently. It’s a hybrid design that finally matches the actual geography of a residential delivery, and it’s the kind of practical engineering that makes you wonder why every delivery robot before this one settled for wheels alone.

Designer: Robot.com

The cargo compartment doubles as the robot’s body, and it’s large enough to carry multiple packages at once, whether that’s stacked food orders sitting alongside standard parcels. Once R-dog reaches your door, an unattended auto drop-off feature completes the handoff without anyone needing to be home, then it moves on to its next stop. For dense urban blocks and residential neighborhoods with walk-ups, that’s a genuinely useful piece of infrastructure, not just a cute demo built for a press release.

Here’s where the story gets a little more complicated, and I think it’s worth sitting with rather than glossing over. R-dog isn’t only a delivery robot. It also runs on Robot.com’s R-ads platform, meaning it can carry branded advertising campaigns as it moves through neighborhoods and hold natural voice conversations with anyone who stops to interact with it. The design leans into this on purpose. R-dog’s look is built specifically to attract attention and encourage people to walk up and engage with it, and that furry tail is doing exactly the job it was designed to do.

I don’t think that’s a dealbreaker, but it does raise a fair question. Is this a delivery robot that happens to carry ads, or an advertising platform that happens to carry packages? One outlet covering the reveal asked almost the same thing, noting that R-dog “definitely could be both,” and admitting the answer isn’t settled yet. I appreciate that kind of honesty from the coverage, because pretending this is purely a delivery innovation would undersell what the company is actually building here.

On the technical side, R-dog runs on REMI, Robot.com’s fleet management platform, and its navigation comes from FieldAI’s Field Foundation Models, which let it operate in unpredictable real-world environments without needing the area pre-mapped first. That matters enormously for a robot meant to handle whatever staircase, curb, or oddly angled walkway it encounters next. Robot.com has been building robots since 2017 and currently has more than 500 deployed across the US, Canada, Dubai, and the MENA region, having completed over 2.5 million tasks. R-dog joins an existing lineup that includes the R-kiwi for campus deliveries, the R-cargo for industrial logistics, and the humanoid R-noid.

Commercial deployment is planned for 2027, and pricing hasn’t been announced. What I keep landing on is that R-dog solves the one problem that has quietly limited every delivery robot before it. Whether it ends up remembered as the design that finally cracked the last twenty feet, or as a very charming way to get people to look at ads, will depend entirely on how it’s actually deployed once it leaves the concept stage. For now, it’s the most convincing answer I’ve seen to a problem that’s been sitting in plain sight for years.

The post Meet the Robot Dog Built to Climb Your Stairs With Your Packages first appeared on Yanko Design.

Watti Is the Robotic Desk Lamp That Wants to Know What You’re Doing Before You Ask for Light

Posted on by

The desk lamp has been, essentially, a solved problem since Edison. A light source, a shade to direct it, and a switch. Every design evolution since then has adjusted the aesthetics or added a dimmer; the functional premise, that a lamp sits passively until pointed and switched, has remained largely untouched for more than a century. Watti is an attempt to challenge that premise at every level, simultaneously and on purpose.

At first glance, Watti reads as a robotic arm that happens to have a light on the end. Five independently motorized axes control the base, shoulder, elbow, neck, and head, each driven by a RobStride motor that reports its position back to the computer. That range of motion gives the lamp the articulation of a traditional Anglepoise-style design, except it can get there on its own with measured precision.

Designer: Nikolay Tyulkin

A ring of 45 individually addressable WS2812 RGB LEDs provides the actual light output, and a Raspberry Pi 5 running ROS 2 coordinates the whole system. The camera is what separates Watti from a sophisticated motorized fixture. An Orbbec Astra Mini S captures both color and depth data at 640×480 resolution and 30 frames per second, giving the system a three-dimensional view of whatever sits in front of it.

Those streams are visible in the companion software but not yet acted on in a meaningful way; the lamp can see the desk without fully understanding what’s happening there. That gap is precisely what the project is working to close. Hand tracking, automatic work-area illumination, and 3D scanning are all on the development roadmap, which would let the lamp decide where to point itself based on where hands are positioned and what is being worked on, without any manual adjustment.

Even before those features arrive, Watti is already more programmable than most dedicated lighting tools. Watti Studio, a browser-based visual editor, lets users compose scenes that coordinate motion and light on a shared timeline. A virtual model of the five-axis lamp renders in the editor so poses and LED configurations can be arranged, previewed, and refined without a physical unit present. When a scene is ready, it’s packaged and transferred to the Raspberry Pi, which runs it locally without a continuous connection back to the browser.

The prototype runs on an external 24 V, 10 A power supply and uses approximately $700 worth of parts. That puts it firmly in research and enthusiast territory for now. Plans to release hardware files, the bill of materials, electrical diagrams, the communication protocol, and example scenes would eventually make it a fully buildable project for others who want to take the concept further.

What Watti points toward is a lamp that treats lighting as a spatial and contextual problem rather than a manual adjustment. A light that can track where work is happening, adapt its position in real time, and coordinate color and intensity without interrupting the task at hand is a fundamentally different object than a lamp with a flexible neck. The premise sounds obvious once stated, which is usually a sign that someone should have built it sooner.

The post Watti Is the Robotic Desk Lamp That Wants to Know What You’re Doing Before You Ask for Light first appeared on Yanko Design.