Tag Archives: Artificial Intelligence

An Ex-Alibaba Exec Spent 12 Years Building the Smart Glasses that Google Couldn’t

Posted on by

The story of Google Glass is a well-worn legend in Silicon Valley. It was a product so far ahead of its time that it became a cultural phenomenon and then a punchline, a symbol of technological overreach and social awkwardness. The project was ultimately shelved, a high-profile monument to a future that arrived too early. It was a public retreat, an admission that the world was not ready for a computer on its face, or perhaps that the computer was not ready for the world.

As that chapter closed, another one was just beginning, thousands of miles away. An executive from Alibaba, inspired by the initial audacity of Google’s idea, decided to take a different approach. Instead of chasing hype, he would chase utility. Instead of prioritizing features, he would prioritize weight and comfort. For twelve years, his company, Rokid, worked to solve the very human problems that Google had overlooked, and in 2026 that long bet looks less like a moonshot and more like a roadmap.

Designer: Rokid

That roadmap now has a new center of gravity. Following Google’s latest Gemini updates at I/O, Rokid says it is bringing Gemini Flash 3.5 to its smart glasses, pushing the company deeper into what it calls agentic AI. The phrase matters because it signals a shift away from voice assistants that answer one question at a time and toward systems that can hold context, respond faster, and handle more layered tasks through simple voice commands. Rokid is framing the glasses as a place where conversational AI can stay present, useful, and continuous rather than trapped inside a phone screen.

That ambition sits on top of an unusually broad AI strategy. Rokid has spent the last year positioning its glasses as an open ecosystem rather than a single-model device, supporting ChatGPT, Qwen, DeepSeek, and Gemini across different products and regions. In Asia, the company has already built an AI Agent Store and says it has received more than 3,000 submissions for agentic workflows, with over 400 approved and published. The international push comes next, and that is where the latest Gemini integration becomes more than a feature update. It becomes a bridge between Rokid’s regional momentum and its global pitch.

The hardware story still matters because smart glasses live or die by whether people will actually wear them. Rokid’s 2025 display-equipped glasses carried one of the most memorable specs in the category: 49 grams for a full-function AI and AR device with display. That number gave the company a clean answer to the oldest question in wearable tech, which is how much computation can disappear into something that still feels like eyewear. According to Rokid’s own materials, that product also helped it raise more than $6 million and move into global mass production by December, giving the company proof that its ideas could leave the demo stage.

This year’s bigger mainstream play is Rokid AI Glasses Style, a different kind of product aimed at lowering the barriers that have kept smart eyewear niche for so long. Style is display-free, voice-centric, and starts at $299. At 38.5 grams, it is even lighter than the 49-gram model, and Rokid presents that reduction as part of a larger balancing act between comfort, battery life, and functionality. The frame is designed like premium eyewear, with titanium alloy hinges, liquid-silicone nose pads, and a classic D-shaped silhouette. Underneath that familiar form is a dual-chip architecture, with one chip handling low-power always-on tasks and another managing AI and imaging workloads.

Rokid clearly wants to win on openness, but it also wants to win on practicality. One of the strongest parts of the press material is its prescription-first approach, which treats vision correction as core infrastructure rather than a niche add-on. Style supports prescriptions up to ±15.00D, covering myopia, astigmatism, presbyopia, progressives, and functional lens options like photochromic and blue-light filtering. Users can upload prescriptions online and receive custom lenses in about 7 to 10 days. That sounds mundane compared to AI buzzwords, but it may be one of the most important adoption levers in the entire category. Smart glasses cannot become everyday objects if they still behave like specialty gadgets.

The other major throughline is accessibility. Rokid has been consistent here, both in the visit materials and in the press kit. The company is working with Google on accessibility-focused solutions for users with vision and hearing impairments, and its broader messaging keeps returning to a principle it phrases simply: leave nobody behind. For blind and low-vision users, Rokid positions audio-based AI glasses as digital eyes, and it has attached a small subsidy to purchases made for visually impaired users. That choice gives the company a more grounded social purpose than most wearable launches, which often stop at lifestyle language and creator features.

Those creator features are still part of the package. Style includes a 12MP Sony sensor, 4K capture, open-ear audio, and a triple-format imaging system designed for 3:4, 4:3, and 9:16 shooting. Rokid’s pitch is obvious and smart: content should be ready for Instagram, TikTok, or YouTube the moment it is captured, without cropping or post-editing. The glasses also support voice interaction in 12 languages and translation in 89, while adding head gestures and AI shortcuts for hands-free control. Nod to answer a call, shake your head to end it, ask for help in your own language, and keep moving.

All of this adds up to a company trying to define smart glasses less as a futuristic accessory and more as the next natural interface for AI. That is the real continuation of the Google Glass story. Google proved the cultural shock of putting a computer on your face. Rokid is trying to prove the quieter part, that wearability, prescription support, open AI access, and contextual software are what turn a provocative idea into a daily habit. The original dream never disappeared. It just needed lighter frames, better timing, and a company patient enough to spend twelve years building the version people might finally keep on.

The post An Ex-Alibaba Exec Spent 12 Years Building the Smart Glasses that Google Couldn’t first appeared on Yanko Design.

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

Posted on by

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

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

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

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

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

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

Apple said ‘AI’ exactly 28 times at WWDC 2026. Google mentioned it nearly 100 times at I/O.

Posted on by

By the end of this year’s big tech keynotes, one comparison stood out more than any product demo. Apple said “AI” 28 times at WWDC 2026. Google said it nearly 100 times at I/O 2026. Same industry, same race, same obsession, but two very different instincts about how to sell the next phase of computing.

Google’s keynote reflected the current rhythm of the AI industry, loud, relentless, and eager to stamp the term onto everything in sight. Apple’s presentation moved differently. It kept circling back to what people could actually do with the technology, how private it would be, and where it would fit into everyday routines. That softer framing may frustrate people who want Apple to move faster and compete harder. It may also be exactly why Apple’s pitch feels easier to absorb at a moment when audiences are already saturated with AI promises.

AI fatigue is real, and it has been building for a while. After years of keynotes, product launches, and press releases leading with the same two letters, the word has started to lose its grip on audiences. What once signaled breakthrough capability now signals marketing effort. When a company says “AI” 100 times in a single presentation, the listener stops hearing a technology and starts hearing a strategy. The signal becomes noise, and somewhere in that noise, the actual products get harder to see.

Apple’s approach at WWDC 2026 worked around that problem by reframing the conversation entirely. Instead of leading with technology, it led with moments. Siri finding a friend’s new address buried in a weeks-old message thread. A photo being reframed after the fact, as if you had stepped to the right before pressing the shutter. A restaurant bill split with Apple Cash by pointing a camera at it. These are small things, but they are the kind of small things that people actually think about during their day. Anchoring the keynote to those moments gave the technology a human scale that raw AI talk rarely achieves.

The branding reflects the same thinking. Apple calls it “Apple Intelligence,” a label that keeps the company name front and center while quietly sidestepping the overcrowded AI conversation. It is a deliberate choice, and it shows. Google’s keynote was structured around the technology itself, its power, its speed, its range. Apple’s keynote was structured around the people using it. That difference in framing shapes how audiences receive the same underlying capability, and Apple’s version is considerably easier to trust.

Privacy played a central role in building that trust. Apple returned to on-device processing and Private Cloud Compute repeatedly throughout WWDC, not as a footnote but as a feature. At a time when public concern about how AI companies handle personal data is growing steadily, that emphasis lands differently than it might have a few years ago. Google builds powerful models and serves them at enormous scale. Apple builds careful models and makes a point of telling you where your data goes and where it stays. For a meaningful portion of consumers, that distinction matters more than benchmark scores.

None of this means Apple is winning the AI race on capability. Google’s models are more powerful, more publicly accessible, and more deeply woven into the daily workflows of people around the world. Gemini’s reach across Search, Gmail, YouTube, and Android gives Google a distribution advantage that Apple’s ecosystem, for all its loyalty, cannot easily match. If the competition were judged purely on technical ambition and model performance, Google’s 100 mentions would feel earned.

But technology keynotes are not judged purely on technical ambition. They are judged on how they make audiences feel, what they make people want, and whether they leave the room energised or overwhelmed. On those terms, Apple’s 28 mentions of “AI” accomplished something that Google’s near-100 did not. They kept the word rare enough to mean something. Every time Apple said it, there was a feature attached, a privacy assurance nearby, and a use case grounded in daily life. The word carried weight because it was not being used to fill space.

The larger irony is that Apple may be the company best positioned to benefit from a backlash it did not entirely create. Google, Microsoft, Meta, and others have spent years flooding the conversation with AI language, and the fatigue that has followed is a byproduct of their own enthusiasm. Apple watched, built quietly, and showed up at WWDC 2026 with a keynote that treated restraint as a product decision. Whether that restraint reflects genuine strategic confidence or simply a capability gap dressed up in good marketing is the question the next few years will answer. For now, 28 versus 100 tells a story that Apple’s communications team could not have scripted better.

The post Apple said ‘AI’ exactly 28 times at WWDC 2026. Google mentioned it nearly 100 times at I/O. first appeared on Yanko Design.

7 Biggest AI Ideas That Came Out of BEYOND Expo 2026

Posted on by

The youngest person at BEYOND Expo 2026’s AI Hack Day was nine years old. That little fact, shared by co-founder Dr Lu Gang, actually says more about the state of AI than any big product launch. It means the tools are getting simple enough that you don’t need a PhD to build something interesting; you just need a good idea. The rest of the expo in Macao seemed to prove his point. You had 30,000 people and almost 800 companies all focused on a single question: what happens when AI stops being just software and gets built into actual, physical things?

It turns out the answer is a mix of things we expected and some we definitely did not. BEYOND Expo 2026 ended up giving us a pretty clear map of where this is all heading, with seven key ideas showing up over and over again. We saw everything from humanoid robots that are finally ready for production to underwater drones that can get around without GPS. Some of this was easy to see coming, but other parts showed that the tech has crossed a real line. These are the ideas that give us a solid picture of an AI that now has weight, form, and real-world impact.

1. Humanoid Robots Are Finally Getting Real

The most obvious trend on the floor was the sheer number of robots walking around. This wasn’t just one or two companies showing off a flashy prototype. The BEYOND Best of Innovation awards list was packed with names like AI² Robotics, DEEPRobotics, LimX Dynamics, and Pudu Robotics. Seeing that many different companies all get recognized for building functional, legged robots at the same event is a major signal. The hardware is clearly getting to a point where it’s reliable enough to be taken seriously.

What’s interesting is that the conversation is shifting from engineering to application. Companies were talking about humanoids for specific jobs in industry, retail, and even in the home. This tells you the focus is moving past the basic challenge of just making them walk without falling over. The new problem to solve is what they should actually do all day. BEYOND Expo made it feel like we’re at the very beginning of a real manufacturing race, not just a science fair.

2. Smart Glasses Found a Form Factor That Works

Smart glasses have been the “next big thing” for about a decade, but this year felt different. We saw new AI-powered glasses from iFlyTek and METLEN, and companies like Even Realities, Mobvoi, and XREAL all picked up innovation awards for their own takes on wearable displays. The key here is convergence. While each product has its own features, they’re all starting to look and feel like something a normal person might actually wear. They are lighter, the displays are better, and the battery life is getting there.

This isn’t another Google Glass moment where the tech was impressive but the product was awkward and socially weird. The new wave of smart glasses is being designed with more specific uses in mind, from on-the-fly translation to providing subtle notifications or acting as a personal design agent. The on-device AI is powerful enough to handle these tasks without being constantly tethered to a phone, which is the breakthrough that might finally make them stick.

3. Flying Vehicles Are Becoming Actual Products

For years, eVTOLs, or electric vertical take-off and landing aircraft, have been staples of futuristic concept videos. At BEYOND Expo, they started to look like real products. Aerofugia showed up with what it called its first production aircraft and, just as importantly, a production eVTOL battery. Wefly also got an innovation award, adding to the sense that this category is moving out of the lab and onto the launchpad.

The word “production” is what matters here. It signals a shift from speculative design to engineering with a supply chain. AI is the invisible engine driving this progress, handling the incredibly complex calculations needed for flight stability, power management, and autonomous navigation. This is the part of the “digital to physical” story where AI isn’t just a feature; it’s the core technology that makes a whole new category of hardware possible.

4. AI Is Getting Personal and Medical

While robots and flying cars grabbed a lot of attention, some of the most interesting AI was designed to be much closer to home, and even part of the body. The expo featured things like Zdeer’s bone conduction hearing aid and Ulike’s optical beauty devices. In the startup competition, one of the finalists was an “emotion-sensitive hugging bear,” and others included smart jewelry and wearables designed to be stylish.

This points to a quieter, more intimate side of the AI hardware boom. These aren’t just gadgets; they’re devices that interact with our bodies and our health. A hearing aid that uses AI can learn and adapt to a person’s specific hearing profile in different environments. A wearable that senses emotion is a step toward technology that responds to our mental state. It’s a reminder that the most impactful physical AI might be the kind that disappears completely into our daily lives.

5. The One-Person Company Is the New Unicorn Hunt

One of the most forward-thinking ideas came from Dr Lu Gang himself. He said that this year, the expo deliberately focused on “one-person companies” and individual programmers. He believes these tiny operations have the potential to become unicorns because AI tools have become such a powerful force multiplier. When the youngest hacker at your event is nine, it proves that the barrier to entry for building something real has dropped through the floor.

This is a structural shift in how tech companies might get built. The old model of needing a big team and millions in venture capital just to get a product off the ground is being challenged. With powerful AI handling coding, design, and operational tasks, a single motivated person can now build and launch something that would have taken a whole department just a few years ago. It suggests a future where the startup landscape is much more dynamic and accessible.

6. Knowing How to Tell a Story Is a Technical Skill

With 800 companies all showing off impressive technology, just having a good product wasn’t enough. Kun Gao, the founder of Crunchyroll, made this point at the closing ceremony. He advised founders that they have to learn how to tell a compelling story to win over investors and partners. This wasn’t just abstract advice; it was happening live at the “Fund at First Pitch” competition, where over 300 startups were trying to get noticed.

This is a crucial idea for anyone in design or product development. In a crowded market, the clarity of your vision is just as important as the quality of your code or the cleverness of your engineering. Being able to explain who your product is for, what problem it solves, and why it matters is a design skill. It’s what separates a cool piece of tech from a real business, and BEYOND Expo put that challenge front and center.

7. AI Is Going Underwater, Literally

Probably the most unexpected idea at the expo was seeing AI get good at swimming. Zero Zero Robotics, known for its flying drones, launched the HOVERAir AQUA, an underwater drone. Another company, OrcaTech, also won an innovation award for its marine technology. This might seem like a niche category, but the technical challenge is enormous and says a lot about how capable AI has become.

Underwater is one of the hardest environments for autonomous tech to operate in. GPS doesn’t work, visibility is often terrible, and communication is extremely limited. For a drone to navigate, identify objects, and perform tasks on its own down there, its onboard AI has to be incredibly sophisticated. It proves that physical AI is not just conquering our cities and skies; it’s expanding into the most remote and difficult parts of our world.

The post 7 Biggest AI Ideas That Came Out of BEYOND Expo 2026 first appeared on Yanko Design.

This $99 USB-C Hub Also Runs GPT-5, Gemini, and Claude at the Same Time to AI Transcribe Your Meetings

Posted on by

Put the DockOrb A1 on a conference table without context and someone will reach for it expecting a scroll wheel. The gray brushed-aluminum slab, the gently rounded corners, and two physical buttons in familiar left-right symmetry on the top face read entirely as peripheral hardware. What the device actually does is listen, think, and report. Powered by OpenAI GPT-5, Google Gemini 2.5 Pro, and Anthropic Claude Sonnet 4, DockOrb A1 is a professional AI meeting and desktop assistant. The label on the box says meeting assistant; the object in the hand says otherwise.

The category it operates in has been filling up quickly. Plaud built a card-thin wearable that magnetically clips to your phone. HiDock shaped a USB-C hub into a ChatGPT-powered meeting stenographer, and we covered that launch here at YD. DockOrb’s A1 lands somewhere between those two worlds, combining a fully functional multiport dock with a multi-model AI engine, 100W PD, and 4K@60Hz HDMI output. Unlike either of those predecessors, it handles display output and power delivery in the same housing, making the desk real estate argument for a single device considerably more loaded.

Designer: DockOrb

Click Here to Buy Now: $99 $149 ($50 off) Hurry! Hurry, only a few left!

The mouse silhouette is instantly familiar, and anyone who’s ever used a computer will be able to navigate the DockOrb intuitively. Two buttons arranged horizontally on a flat top surface is the correct solution for a device that needs to be operated with a single press in a meeting context, no fumbling, no menus, no distraction. With a dedicated AI button and real-time processing, DockOrb A1 analyzes ongoing discussions and provides actionable suggestions and insights, helping teams improve collaboration and make decisions more efficiently without interrupting the meeting flow. LED status is handled by a single indicator, white for idle and blue for active capture, readable from across a conference table without breaking eye contact with whoever is speaking. The problem is that all of this correct ergonomic logic is housed inside a form that the product world has spent two decades teaching people to recognize as a pointing device.

Rather than anchoring to a single AI model, the A1 integrates with Esteno, an advanced AI fusion-processing software platform. Esteno integrates multiple advanced AI models, including OpenAI GPT-5, Google Gemini 2.5 Pro, and Anthropic Claude Sonnet 4. Each model is optimized for tasks such as speech-to-text, summarization, contextual reasoning, and insight generation. By intelligently routing tasks to the most suitable model, the system delivers efficient, flexible, and high-quality meeting intelligence across different use cases. That architectural approach is genuinely unusual in this category, where most competitors commit to a single backbone and build their entire brand identity around it.

Plaud’s card-thin approach to meeting intelligence, at 2.9mm thick and MagSafe-compatible, is built on the premise that the recorder travels everywhere with you, riding on the back of your phone. The A1 has no such intention, operating through USB power without a built-in battery, with a compact design, dedicated recording button, and AI activation key for stable and simple meeting operation. In exchange for that fixed-desk commitment, it handles 4K video output at 60Hz over HDMI and 100W power delivery over USB-C, turning the dock into the single device your entire workstation routes through. After transcription and analysis, DockOrb A1 automatically generates structured meeting reports highlighting key decisions, action items, and follow-up tasks, which can be exported directly in PDF, Excel, or PowerPoint formats. Getting a properly formatted, structured report out of a recorded conversation without manual reformatting is a genuine subtraction from the post-meeting to-do list, and it’s the kind of output that separates a real workflow tool from a novelty recorder.

Following ISO and SOC data protection standards, DockOrb A1 secures recorded audio and AI-generated content through encrypted storage and processing, allowing users to export, archive, or delete files at any time while ensuring full control over their data. That’s pointed positioning in a market where corporate IT departments are increasingly skeptical about meeting audio being routed through third-party AI servers without accountability. Recordings, transcripts, summaries, and reports from multiple meetings can be stored and organized within a centralized memory archive, with AI-powered indexing and searchable meeting names, content, or dates, so teams can quickly retrieve past discussions, track long-term decisions, and build a continuously growing knowledge base. Built on a platform-independent architecture, DockOrb A1 processes audio from Zoom, Teams, Google Meet, mobile devices, and more, delivering consistent transcription, analysis, and structured outputs. Retrieving a specific discussion from three months prior becomes a search query rather than a manual scroll through unlabeled audio files.

The Kickstarter campaign prices the A1 at $89 for the Super Early Bird tier against an MSRP of $149. Shipping is targeted for August 2026, with production beginning the month prior. Plaud’s Note Pro retails at $169 on the market and handles no dock hardware whatsoever, making the A1’s value calculation sharper for anyone already planning to put a USB-C hub on their desk. The Esteno software platform tiers at $8 per month for Basic, covering 600 minutes of monthly transcription, and $15 per month for Pro, which adds 2,400 minutes, unlimited AI features, and priority processing. That’s a fully loaded meeting intelligence setup, dock and display output included, for a first-year cost that lands well under what most enterprise-grade transcription tools charge for software alone.

Click Here to Buy Now: $99 $149 ($50 off) Hurry! Hurry, only a few left!

The post This $99 USB-C Hub Also Runs GPT-5, Gemini, and Claude at the Same Time to AI Transcribe Your Meetings first appeared on Yanko Design.

PixVerse Just Made Product Videos as Easy as Writing a Brief

Posted on by

Product design has always been part craft, part communication. Getting a concept from sketch to client approval demands a level of visual storytelling that most designers simply haven’t had the budget or tools to manage on their own. Video production, in particular, has long been the step that gets quietly skipped, not because the ideas aren’t there, but because the process is expensive, slow, and complicated.

That’s a gap PixVerse has been working to close. Founded in 2023, the platform has grown to over 100 million users across 177 countries, powered entirely by proprietary models it builds in-house. At the iMpact Global Connect Show 2026, the company’s team walked through three distinct products that together make a compelling case for AI-generated video as a practical part of the design process.

Designer: PixVerse

The most immediately useful of the three, at least for most designers, is V6. It’s the platform’s flagship model, and the latest update improves camera movement, character performance across scenes, and physical object interaction in noticeable ways. More significantly, V6 can now generate a complete multi-shot short film with native audio from a single prompt, without any separate editing or sound production steps involved.

Think about what that actually means for a product designer. A 30-second product video typically means writing a brief, hiring a videographer, sourcing music, shooting, and editing over several days or weeks. With V6, a designer who can clearly articulate how a product should look, move, and feel in context can produce that same result from a prompt and a reference image in considerably less time.

That kind of speed has obvious advantages for solo designers and small studios. A freelancer can arrive at a pitch with three distinct video directions instead of three mood boards. A startup preparing a crowdfunding campaign doesn’t need a separate production budget for a launch video. An in-house team can test how a product reads in a real context before committing to a full-scale shoot.

The second product, C1, goes further by targeting actual film production pipelines. It combines a cinematic visual effects system, an industrial-grade action engine, and a storyboard-to-video feature in a single workflow, letting production teams convert static panel layouts directly into continuous video sequences. Reference-guided generation also keeps characters and scenes consistent across shots, which has historically been one of the harder problems for AI video to solve.

For designers, that matters most when a concept already lives as a sequence of moments rather than a single frame. A transportation designer communicating a user journey, a consumer electronics team mapping how a device gets picked up, handled, and put down, or a lifestyle brand building a product narrative around daily routines, all of them are telling stories that C1 is built to handle.

Then there’s R1, which doesn’t behave quite like any other AI video tool currently available. Rather than producing a fixed clip with a clear beginning and end, R1 generates a continuous, interactive visual environment that responds to user input as it runs. It’s less like watching a video and more like navigating a space that exists, evolves, and reacts, one that you can steer and share.

Users can build a personalized digital avatar from photos and enter these generated worlds alongside others in real time. During the demo, a shared environment called “Cat Takes Charge” had 118 users inside it at the same time, running continuously for over nine hours. Each participant could submit prompts into a live feed, with the AI realizing them as video within the shared space as they appeared.

For product designers, R1 opens up possibilities that a rendered video simply can’t replicate. Imagine walking a client through a simulated retail environment built around a new appliance, or letting a stakeholder explore a furniture concept in a living, reactive interior before a prototype even exists. It’s the kind of tool that starts to make spatial storytelling feel accessible at the concept stage, not just post-production.

What all three tools share is that they reward the same skill designers already rely on: clarity of intent. A well-constructed prompt isn’t a technical exercise; it’s a creative direction, not unlike a solid design brief. Companies integrating PixVerse into their workflows reportedly cut costs by 68% and finish work 57% faster than conventional production methods, a significant gain for teams of any size.

None of that requires a production background, and it doesn’t even require familiarity with video editing software. What it does require is the ability to describe a vision precisely, which is something designers do every single day across briefs, sketches, and presentations. PixVerse just moved video closer to the beginning of that process, somewhere between the first concept and the final approval, rather than as an afterthought at the very end.

The post PixVerse Just Made Product Videos as Easy as Writing a Brief first appeared on Yanko Design.

This 3D-Printed Macintosh Replica Is Actually a Voice AI Assistant

Posted on by

Smart speakers have become some of the most visually forgettable objects in modern homes. A cylinder, a puck, a fabric-wrapped drum, placed wherever the Wi-Fi is strong and largely invisible once the novelty wears off. They do their jobs well enough, but none of them look like they belong in a collection or on a desk that someone cares about. The hardware has always been purely functional, and the design has always shown it.

Alisher Ashimov approached the idea of a desk-based AI assistant from a completely different direction. Kira, his open-source project, takes its visual cues directly from the original 1984 Macintosh, a machine whose beige monolith silhouette is arguably the most iconic in personal computing history. The result is a voice-activated AI companion that looks more like a cherished collectible than a utility device.

Designer: Alisher Ashimov

The enclosure is 3D printed in a single recommended filament color: Light Khaki matte PLA, the closest approximation of that distinctive Apple beige. Rounded top corners, a recessed front panel, horizontal side vents, and a decorative floppy-drive-style slot below the display all reproduce the original’s proportions at pocket scale, somewhere around 80mm wide. A small four-color badge on the lower front panel adds the final recognizable touch.

Where the original Macintosh showed a desktop environment, Kira shows a face. The 1.5-inch OLED display renders two rectangular eyes and a small dash mouth, animating expressively in response to interaction. The wake word is “Hey, Kira,” and from there, a built-in microphone picks up questions while a 4Ω, 3W speaker delivers spoken answers through the sculpted housing. It handles everyday voice queries the same way any smart assistant does, just with considerably more personality sitting on the shelf.

The electronics are deliberately approachable. The core is a Seeed Studio XIAO ESP32-S3 Sense, a capable and compact microcontroller with built-in Wi-Fi, Bluetooth, and a microphone. The rest of the bill of materials, a speaker, amplifier, SH1107 OLED module, mini breadboard, and jumper wires, are available on Amazon for modest amounts. The 3D-printed enclosure is optimized to print in about three hours across two plates with minimal support material, and an assembly guide walks builders through wiring, assembly, and firmware flashing.

The software carries the same open-ended spirit as the hardware. Voice, language, the assistant’s character, and memory settings are all user-definable, which means Kira isn’t locked into a single personality or a single cloud service. Tinkerers can tune the firmware directly. Ashimov has published the files freely, with no commercial barriers between the design and anyone with a printer and an afternoon to spare.

The objects people choose to keep on their desks tend to say something about them. A tiny Macintosh-shaped AI assistant that you built yourself and tuned to your own preferences says rather a lot. It combines a piece of design history, a genuinely capable voice interface, and an honest invitation to understand exactly how the thing works, all in a form that most people will stop and ask about the moment they see it.

The post This 3D-Printed Macintosh Replica Is Actually a Voice AI Assistant first appeared on Yanko Design.

After 6 Years, Google Finally Remembered To Launch A New Smart Speaker, This Time with Gemini Built-in

Posted on by

The Nest Audio came out in September 2020. If you bought one that fall, you were probably still navigating pandemic grocery runs and wondering when offices would reopen. Nearly six years later, Google has finally shipped something new to put on your kitchen counter. The Google Home Speaker, now landing in June 2026 after a “Spring 2026” promise that tested the meaning of the word spring, is the company’s first new standalone smart speaker in half a decade. Six years is a long time in consumer electronics. Apple refreshed AirPods three times. Sonos launched and then partially broke its app and still found time to make new speakers. Google, meanwhile, treated the entire category like a parked car, leaving the Nest Audio to quietly collect dust while the company sprinted elsewhere.

Where was Google sprinting? Toward Gemini, mostly. The AI model has been grafted onto Search, Maps, Workspace, Android, Chrome, YouTube, and practically every other product in the portfolio with enough surface area to carry a chatbot. Google even announced the Googlebook at I/O 2026, a new category of premium Android laptops billed as the successor to the Chromebook and the Pixelbook, built, predictably, around Gemini Intelligence. When Google finally announced the new Home Speaker at its Made by Google event in October 2025, the device was framed almost entirely around its role as a Gemini endpoint. The speaker came back because Gemini needed somewhere new to live, and the kitchen seemed underserved.

Designer: Google

There was a time when the company’s smart home pitch felt like a real platform strategy, ambient computing, voice everywhere, helpful devices fading into the background. The original Google Home arrived in 2016 with a sense of ambition. It was a bet that Google could own the center of the connected home by making voice control feel natural, useful, and quietly omnipresent. Then came the Mini, the Max, the Hub, the Nest rebrand, and eventually the Nest Audio. After that, the energy drained out of the room. The category was never formally abandoned, but it entered that peculiarly Google state where a product remains alive enough to avoid a funeral and neglected enough to make users wonder whether anyone still remembers where the light switches are.

The new speaker itself looks perfectly pleasant. It is small, rounded, soft, and available in the sort of colors Google hardware teams always seem to get right, the kind that make every room look slightly more curated than it probably is. Google says it has 360 degree audio, faster processing for more fluid conversations, and a new light ring that signals when Gemini is listening, thinking, or responding. Fine. Great, even. The problem is that none of this arrives in a vacuum. Google has trained people to see its hardware launches through a second lens, one that asks a less flattering question: for how long is this category going to matter to the company?

That question hangs over almost every Google device that is not a Pixel phone. The company loves a fresh start, a new naming scheme, a reset button disguised as a vision statement. It also has a long history of treating hardware categories like experiments that can be deprioritized the minute a more interesting internal narrative comes along. Smart speakers spent years as a central piece of Google’s ambient computing story. Then Gemini became the story, full stop. Once that happened, every product had to justify itself in AI terms. Phones became Gemini phones. Search became Gemini search. The smart home became Gemini for Home. Laptops became Googlebooks. And now, after years of silence, the speaker has returned as a vessel for the new corporate religion.

There is a certain irony in that. Smart speakers were already one of the clearest examples of what AI in the home was supposed to feel like: conversational, contextual, present without demanding attention. Google had the hardware footprint. It had the installed base. It had a brand that, for a while, was practically synonymous with talking to your house. If the company had kept iterating steadily, this new moment could have felt like a natural evolution. Instead, it feels like a rediscovery. Google wandered away from the category long enough that its return carries a faint air of surprise, as if someone opened a closet at Mountain View headquarters and found an entire product line under a sheet.

Maybe the Google Home Speaker will be excellent. Maybe Gemini will finally make the smart speaker feel smarter than a kitchen timer with good branding. But this launch still lands as a reminder of how erratic Google’s hardware attention span can be. The company did not so much nurture this category back to health as remember it was still on the org chart. After nearly six years, Google has a new smart speaker, and the most Google part of that sentence is that it only happened once the device could be recast as AI infrastructure. The speaker is back on the counter. Whether Google stays in the room this time is the harder question.

Image Credits: 9to5Google

The post After 6 Years, Google Finally Remembered To Launch A New Smart Speaker, This Time with Gemini Built-in first appeared on Yanko Design.

Microsoft RTX Dev Box Has 1,000 Holes, All of Them Intentional

Posted on by

The economics of AI development have been quietly changing how developers think about their hardware. Cloud GPU bills compound fast when you’re iterating through a model dozens of times a day, and every fine-tuning run or inference loop on a remote server adds to a cost that has no natural ceiling. The push toward local AI compute isn’t just about performance. It’s about moving from a metered relationship with infrastructure to one you own outright and sit in front of.

Microsoft announced the Surface RTX Spark Dev Box at Build 2026 as its answer to that shift. It’s a compact mini PC powered by the NVIDIA RTX Spark superchip, the same ARM-based silicon debuting in the Surface Laptop Ultra, and it arrives on a developer’s desk already configured and ready to run serious AI workloads without touching a cloud endpoint.

Designer: Microsoft

The device’s most distinctive quality isn’t anything in the spec sheet. It’s the body itself, a 3D-printed anodized aluminum chassis perforated with exactly 1,000 vents arranged across its surface in a precise grid. Those vents are functional, the aluminum chassis doubles as the passive heatsink, managing a 100W sustained thermal envelope without a traditional cooling tower. They’re also a deliberate reference: 1,000 vents for 1,000 teraflops, or 1 petaflop, of AI compute. It’s a design that’s equally a statement and an engineering solution, and nothing else on a desk looks remotely like it.

That petaflop is delivered by NVIDIA’s RTX Spark, which combines a 20-core Grace CPU with a Blackwell RTX GPU carrying 6,144 CUDA cores, connected via NVLink-C2C. The 128GB of unified memory shared dynamically between the processor and GPU is what separates this from a high-end gaming box. That memory ceiling is what makes loading a 120-billion-parameter model possible without partitioning it or shunting inference work to the cloud.

The software side ships pre-configured and aimed precisely at the developer who doesn’t want to spend time on setup. WSL2 with native GPU passthrough and full CUDA support comes pre-installed and ready to use, alongside Visual Studio Code, GitHub Copilot, and PowerShell 7. Windows settings are tuned specifically for development work rather than general consumer use, a small but meaningful distinction when your machine runs long overnight training jobs and needs stability rather than a live tiles grid.

Connectivity covers HDMI, Ethernet, USB-C, USB-A, and a headphone jack, nothing exotic, but a port set that covers what a desk-based development machine actually uses. The machine runs under 100W during intensive workloads, which means it can sustain training jobs and agentic pipelines without the kind of thermal throttling that eventually frustrates sustained use.

For a machine announced without a price, the Surface RTX Spark Dev Box is already doing a specific kind of work. It positions local AI inference as a fixed cost rather than a running expense, and it makes that argument in a chassis that doesn’t look like any other mini PC on the market. A 3D-printed aluminum grid covered in a thousand deliberate holes is an odd form for a developer tool, but it makes the machine’s purpose unmistakably legible from across the room. Availability is expected later in 2026 in the US through Microsoft’s online store.

The post Microsoft RTX Dev Box Has 1,000 Holes, All of Them Intentional first appeared on Yanko Design.

ASUS ProArt Just Closed the Gap Between a Laptop and a Workstation

Posted on by

Creative professionals have been carrying a compromise for years. The laptop powerful enough for serious work tends to be too heavy or too loud, and the one thin and light enough for a day bag can’t handle the work. Purpose-built workstations solve the performance side but solve nothing about portability. The gap between the two has been a persistent frustration, not a deliberate choice most people would make.

ASUS is addressing that directly at Computex 2026, where the ProArt P16, ProArt P14, and ProArt Mini PC were unveiled as the first ASUS devices powered by NVIDIA’s RTX Spark superchip. The same ARM-based chip combining a 20-core Grace CPU with a Blackwell RTX GPU and up to 128GB of unified memory runs across all three products, making the performance difference between a laptop and a desktop largely a matter of form factor rather than capability.

Designer: ASUS

The ProArt P16 and P14 are the portable entries, and they arrive 13% thinner and 16% lighter than the previous P16 generation. The P16 weighs 1.77kg at 12.9mm, and the P14 comes in at 1.48kg and 13.9mm. Both are CNC-manufactured in Nano Black and Neo White finishes, and carry 99.9Wh batteries for all-day runtime, a detail that matters when the work is intensive enough to drain power quickly. The machines don’t sacrifice weight for performance or performance for weight.

The display on both laptops is ASUS Lumina Pro OLED, calibrated to Delta E < 1 color accuracy, Pantone Validated, and certified for VESA DisplayHDR True Black 1000. Peak HDR brightness reaches 1,600 nits, which is more than three times what the previous ProArt generation could manage. A 120Hz variable refresh rate, 0.2ms response time, and an anti-reflection coating that cuts glare by 65% complete a panel that keeps color decisions accurate regardless of the lighting conditions a shoot or edit session happens to land in.

Under the hood, RTX Spark’s 1 petaflop of AI compute and unified memory pool change what locally processed work looks like. Rendering a 90GB-plus 3D scene, editing 12K 4:2:2 video, generating 4K AI video, or running a 120-billion-parameter language model locally are tasks that previously needed significantly bigger machines. Adobe is rebuilding Photoshop and Premiere specifically for RTX Spark to deliver 2x faster AI and graphics performance, and a three-month Creative Cloud subscription ships with the ProArt laptops.

The ProArt Mini PC extends the same logic to the desk. At 150 × 150 × 51mm, it fits anywhere a small speaker would and carries up to 128GB of unified memory, 10GbE wired networking, M.2 PCIe Gen 5 expansion, and up to 140W of thermal headroom for sustained demanding workloads. A single RTX Spark-powered box of that size, running AI renders or local large language models around the clock, is a genuinely different proposition for a small studio or home setup than what was available previously.

All three products sit within a broader ASUS ProArt ecosystem that integrates displays, peripherals, creator apps, and AI workflow software into a connected end-to-end experience. ProArt P16, P14, and Mini PC are expected to be available in fall 2026 in select regions, with additional configurations announced closer to launch.

The post ASUS ProArt Just Closed the Gap Between a Laptop and a Workstation first appeared on Yanko Design.