Are you using a desk mouse like some kind of… normal computer user? Why, beg the heavens? For you could be using an air mouse, of your very own creation! [Misfit Maker] shows the way. Check out what he made in the video below.
An air mouse is a mouse you use in the air—which creates at least one major challenge. Since you’re not sliding along a surface, you can’t track the motion by mechanical friction like a ball mouse or by imaging as in an optical mouse. Instead, this build relies on a gyroscope sensor to track motion and translate that into pointer commands. The build relies on an ESP32-C3 as the microcontroller at the heart of things. It communicates with an MPU6050 gyroscope and accelerometer to track motion in space. It then communicates as a human interface device over Bluetooth, so you can use it with lots of different devices. The mouse buttons—plus media control buttons—are all capacitive touch-sensitive, thanks to an MPR121 touch sensor module.
There’s something neat about building your own tools to interface with the machines, almost like it helps meld the system to your whims. We see a lot of innovative mouse and HID projects around these parts.
AI has changed the game. Retail Express offers next generation tools that help retailers buy smarter, keep suppliers happy, and create a customer experience like no other.
Standing out in today’s fast paced retail environment requires more than just meeting customer demands. It requires agility, precision, and the ability to outpace the competition, even when dealing with the constant balancing act of pricing, inventory, promotions, and managing supplier relationships. Retail Express’ retail management solutions have AI at their heart, and offer a truly modern way for retailers to gain an advantage in a fast-advancing market.
Our modular platform brings algorithmic retailing to life, helping you take advantage of centralised planning, analytics and insights which enable smarter, faster and more profitable decisions. Optimise every part of a go-to-market strategy through AI tools which aid with dynamic pricing, supplier collaboration, and managing promotional activity. Overcome the barriers holding your business back: retail’s big questions remain, but the answer is now at hand.
Selling smarter by using AI
Take the complexities of driving revenue growth, not just a task of balancing regular and promotional sales but one of truly understanding customers’ needs and desires. AI offers retailers insights and foresight supported by the truth: real-time analytics can draw from a retailer’s data, from their competitors’ activities, and from the market at large to eliminate guesswork and paint a precise picture of pricing, timing, product selection and sales channels.
Retail Express can help you act fast. Revenue driving dynamic pricing tools help capture sales at peak demand or strategically adjust to increase sales volumes. Suppliers are equally happy: with AI, retailers can bolster joint business planning, offering vendors insights into trade funding, the effectiveness of promotions they’re involved in, and opportunities for negotiation. AI tools essentially unlock brand new revenue streams, boost existing revenue opportunities, and reduce the risk of unnecessary discounting.
Maximising margins with advanced insights
Remaining competitive on the knife edge of modern operating margins is incredibly difficult. Algorithmic retailing using AI offers opportunities to improve margins through efficiency, automation, and smarter decision-making, all while ensuring business wide strategic consistency. It’s a path to joined up thinking which offers a leg up to every area of a retail business.
Safely automating price adjustments enables a dynamic response to fluctuations in cost, demand, or tricky moves from competitors; demand forecasting not only helps build a promotional calendar with impact, but it also allows retailers to do so without the fear of over- or under-stocking. AI retailing tools work with real-time sales data and external factors like seasonality and market trends to help keep everything optimal, improving supply chain efficiency and vastly reducing waste.
Making the right moves to improve customer experience
Customers are the lifeblood of any business, and modern shoppers have high expectations. Whether they know it or not, customers look for personalised, seamless and efficient shopping experiences - and AI can help retailers understand their customers perfectly to deliver exactly the right experience. AI can help curate advanced assortments, and offer the foresight to prepare for rising demand and meet customer expectations head on.
In fact, Retail Express can help you discover everything from the broad to ultra-granular. Insights into customer purchase history help with personalised promotions that deliver against promotion targets. Targeted incentives increase conversion rates, removing the need for mass discounting, while demand forecasting tools remove key customer frustrations like out-of-stock items - improving trust and keeping customers coming back. and reducing the potential of missed sales.
Taking the next step
The future of retail is powered by AI. And not just the future: the data, analytics and insights of advanced artificial intelligence are already helping many of the world’s largest retailers to drive greater revenue, protect their margins, and enhance the experience for their customers. Those that have embraced AI have gained a significant competitive advantage. Don’t get left behind - Retail Express is ready to help you integrate your business with the next generation of retailing tools.
Retail Express is a leading provider of merchandising solutions and services for retail, wholesale and consumer packaged goods (CPG) manufacturers. It uses its deep industry understanding and expertise to provide business solutions that meet the evolving needs of merchandising and category management departments delivering improved productivity and enhanced financial results.
Through its AI powered end-to-end Intelligent Merchandising™ solution, Retail Express addresses the complex problems of advertising, marketing, promotions and pricing in retail, providing one version of the truth across the organisation and departments.
Retail Express operates out of Leeds, UK, across Europe, North America and Australia.
À Barcelone, les habitants ont pu tester une innovation futuriste en matière de transport. Un minibus autonome, développé par Renault en partenariat avec WeRide, circule actuellement dans le centre-ville. Sans chauffeur, ce véhicule suit un itinéraire précis de 2,2 km. Ceci en desservant quatre arrêts stratégiques. L'objectif est simple : démontrer que les bus autonomes peuvent s'intégrer en toute sécurité dans la circulation urbaine.
D'ailleurs, ce n'est pas la première fois que ce prototype est présenté. Il avait déjà été dévoilé à Roland-Garros l'an dernier. Cependant, sa mise en situation réelle à Barcelone permet d'évaluer concrètement son efficacité dans un environnement urbain. De plus, d'autres essais sont prévus à Valence, en France, ainsi qu'à l'aéroport de Zurich. Ces initiatives marquent une avancée importante pour le développement des transports autonomes en Europe.
Une technologie avancée au service des usagers
Ce minibus électrique impressionne par ses capacités techniques. Doté de 10 caméras et de huit lidars, il détecte son environnement et réagit aux obstacles en temps réel. Grâce à ces capteurs, il freine avant de changer de voie et s'adapte à la densité du trafic. De plus, il peut parcourir jusqu'à 120 km sans recharge et atteindre une vitesse de 40 km/h. Une autonomie suffisante pour répondre aux besoins des trajets urbains.
Ainsi, les passagers ont pu profiter d'une expérience inédite, gratuite et sécurisée. Pour certains, voir ce bus circuler seul était presque irréel. Pau Cugat, un étudiant de 18 ans, a comparé ce testà une rencontre entre le passé et le futur. Son étonnement reflète bien l'impact de cette innovation sur le public. Les réactions des autres usagers ont d'ailleurs été diverses : entre fascination, curiosité et parfois scepticisme, chacun réagit différemment face à cette avancée technologique.
Alors que les États-Unis et la Chine investissent massivement dans la conduite autonome, l'Europe reste en retard sur ce marché. Renault veut donc accélérer la transition et prouver que cette technologie est viable pour les transports en commun du futur. Comme l'explique Patrick Vergelas, responsable des projets de mobilité autonome chez Renault, ce projet vise à préparer l'Europe à intégrer ces solutions innovantes.
Dans plusieurs métropoles, les véhicules sans chauffeur commencent déjà à faire partie du paysage urbain. Des taxis autonomes circulent à San Francisco, tandis que Tokyo teste également ces nouvelles technologies. Barcelone devient ainsi une vitrine pour cette évolution. Cela prouve que les minibus autonomes peuvent s'adapter aux contraintes des grandes villes européennes.
Une avancée prometteuse pour les transports publics
L'expérimentation se déroule sans incident, selon les autorités de Barcelone. Aucun accident n'a été signalé depuis le début des essais. Ce qui renforce la crédibilité du projet. De plus, l'absence d'émissions polluantes en fait une solution écologique pour l'avenir des transports urbains.
Cependant, l'adoption massive de ces bus autonomes nécessitera encore des ajustements, notamment en termes de réglementation et d'acceptation du public. Mais une chose est certaine : ce test à Barcelone montre que le futur des transports publics est en marche, et il pourrait arriver plus vite qu'on ne le pense.
There's a new twist in the hunt for dark matter, the invisible substance believed to make up 85 percent of all the mass in the universe: it may actually be way lighter.
In a study published in the journal Physical Review Letters, an international team of researchers propose a new form of the hypothetical substance that's lower in mass compared to other dark matter candidates, which could explain a mysterious phenomenon at the center of our Milky Way galaxy, in a region called the Central Molecular Zone (CMZ).
"At the center of our galaxy sit huge clouds of positively charged hydrogen, a mystery to scientists for decades because normally the gas is neutral," said study co-lead author Shyam Balaji at King's College London in a statement about the work. "So, what is supplying enough energy to knock the negatively charged electrons out of them?"
"The energy signatures radiating from this part of our galaxy suggest that there is a constant, roiling source of energy doing just that," Balaji added, "and our data says it might come from a much lighter form of dark matter than current models consider."
While scientists have extensive evidence that dark matter exists, determining what it is and where it resides remains one of the biggest questions in physics, with theories ranging from parallel universes to primordial black holes. But one of the original and still leading explanations for dark matter is that it comprises a type of nearly undetectable particles called Weakly Interacting Massive Particles, or WIMPs.
As the moniker suggests, WIMPs don't interact strongly with other particles, including light, but are massive enough that they would clump together in the way that astronomers have observed galaxies and other cosmic structures do — which dark matter, through its gravitational influence, is hypothesized to govern the shapes of from the shadows.
But perhaps there's more than one form of it that's less massive than WIMPs, as the researchers suggest. The thinking goes that in the extremely dense environment of the CMZ, these lighter dark matter particles would be constantly colliding and destroying each other upon impact, and subsequently releasing energy. This is a process known as annihilation, and the energy it liberates would then ionize nearby hydrogen gas.
Conversely, the researchers argue that WIMPs and other proposed dark matter particles like axions don't undergo enough annihilation to ionize hydrogen to the extent observed in the CMZ. Nor can the phenomenon be explained by cosmic rays, powerful beams of energetic particles that zip throughout the universe at nearly the speed of light.
"The biggest problem this model helps solve is an excess of ionization in the CMZ," Balaji told Space.com. "Cosmic rays, the usual culprits for ionizing gas, don't seem to be strong enough to explain the high levels of ionization we observe."
There's still a lot of work to be done before the idea gains more steam. But if the theory holds true, Balaji says, we'd have an "entirely new way" to study dark matter rather than just its gravitational influence; now, we could observe the ionization it causes in gases.
"Dark matter remains one of the biggest mysteries in physics, and this work shows that we may have been overlooking its subtle chemical effects on the cosmos," Balaji told Space.com.
A researcher has made a puzzling discovery while analyzing observations taken by NASA's James Webb Space Telescope.
While analyzing images for the telescope's Advanced Deep Extragalactic Survey (JADES), Kansas State University associate professor of computer science Lior Shamir found that out of the 263 galaxies examined, two thirds of them rotated clockwise, while only a third rotated counterclockwise, as detailed in a paper published in the journal Monthly Notices of the Royal Astronomical Society.
This challenges the assumption that any given universe would have half of them spinning one way, with the rest spinning counter to that, according to a press release about the discovery.
"It is still not clear what causes this to happen, but there are two primary possible explanations," said Shamir in a statement. "One explanation is that the universe was born rotating. That explanation agrees with theories such as black hole cosmology, which postulates that the entire universe is the interior of a black hole."
The findings add credence to an existing, Russian doll-like theory called "Schwarzschild cosmology," which suggests that our galaxy is trapped within a black hole, which in turn is located inside another universe.
As Space.com reports, this would imply that other observed black holes could be wormholes, otherwise known as Einstein-Rosen bridges, to other universes, which are unobservable to us due to the black holes trapping light within them.
"I think that the simplest explanation of the rotating universe is the universe was born in a rotating black hole," University of New Haven theoretical physicist Nikodem Poplawski, who champions the theory that we're surrounded by doorways to other universes and wasn't involved in the research, told Space.com. "A preferred axis in our universe, inherited by the axis of rotation of its parent black hole, might have influenced the rotation dynamics of galaxies, creating the observed clockwise-counterclockwise asymmetry."
"The discovery by the JWST that galaxies rotate in a preferred direction would support the theory of black holes creating new universes, and I would be extremely excited if these findings are confirmed," he added.
But Shamir's findings still leave the possibility that the Milky Way's own rotation could have influenced the galaxies' unusual distribution of spin rotation.
Since the Earth rotates around the center of the Milky Way, researchers expect light from galaxies rotating in the opposite direction to be brighter, causing the discrepancy in the JADES observations, Shamir suggests.
In other words, the velocity at which the Milky Way rotates may be influencing our celestial measurements, which had previously been considered negligible.
"If that is indeed the case, we will need to re-calibrate our distance measurements for the deep universe," he said in the statement. "The re-calibration of distance measurements can also explain several other unsolved questions in cosmology such as the differences in the expansion rates of the universe and the large galaxies that according to the existing distance measurements are expected to be older than the universe itself.”
Biohybrid robots work by combining biological components like muscles, plant material, and even fungi with non-biological materials. While we are pretty good at making the non-biological parts work, we’ve always had a problem with keeping the organic components alive and well. This is why machines driven by biological muscles have always been rather small and simple—up to a couple centimeters long and typically with only a single actuating joint.
“Scaling up biohybrid robots has been difficult due to the weak contractile force of lab-grown muscles, the risk of necrosis in thick muscle tissues, and the challenge of integrating biological actuators with artificial structures,” says Shoji Takeuchi, a professor at the Tokyo University, Japan. Takeuchi led a research team that built a full-size, 18 centimeter-long biohybrid human-like hand with all five fingers driven by lab-grown human muscles.
Keeping the muscles alive
Out of all the roadblocks that keep us from building large-scale biohybrid robots, necrosis has probably been the most difficult to overcome. Growing muscles in a lab usually means a liquid medium to supply nutrients and oxygen to muscle cells seeded on petri dishes or applied to gel scaffoldings. Since these cultured muscles are small and ideally flat, nutrients and oxygen from the medium can easily reach every cell in the growing culture.
With the help of mRNA technology proven effective during the COVID pandemic, researchers are now closer than ever to creating viable cancer vaccines.
In an interview with Wired, Lennard Lee, an oncologist with the United Kingdom's National Health Service (NHS) working on mRNA cancer vaccines, says he believes the groundbreaking research may prove to be a "silver lining" in the brutal COVID-19 pandemic.
Before COVID, as Lee told the magazine, "cancer vaccines weren’t a proper field of research."
"Pretty much every clinical trial had failed," the NHS oncologist said. "With the pandemic, however, we proved that mRNA vaccines were possible."
As with mRNA COVID vaccines, the logistics of these potential new cancer inoculations work by "giving the body instructions" to fight troublesome cells, as Lee detailed, ultimately providing the immune system with a how-to manual on fighting cancer.
"Going from mRNA Covid vaccines to mRNA cancer vaccines is straightforward," he told Wired. "Same fridges, same protocol, same drug, just a different patient."
Instead of the one-size-fits-all approach taken with the widespread usage of mRNA COVID jabs, however, these new cancer vaccines will be personalized for each individual cancer patient.
"In the current trials," Lee elucidated, "we do a biopsy of the patient, sequence the tissue, send it to the pharmaceutical company, and they design a personalized vaccine that’s bespoke to that patient’s cancer."
"That vaccine is not suitable for anyone else," he recounted to the magazine. "It’s like science fiction."
According to Lee, breakthrough cancer vaccine innovation came on the heels of the UK's rapid infrastructure-building during the COVID pandemic, which saw the country "open and deliver clinical trials" much faster than anyone would have expected.
As COVID began winding down in 2022, Lee and his colleagues set up the Cancer Vaccine Launch Pad, a post-pandemic pet project that segued mRNA research into the arena of oncology. Not long after, "the dominoes started falling very quickly" as that project and others around the world rapidly progressed towards cancer vaccines. One NHS trial seeking to stop skin cancer from coming back was completed a year early — something that's "completely unheard of," Lee said.
The NHS oncologist told Wired that the results from that trial should come out by the end of this year or the beginning of 2026. If it was successful, Lee told Wired, he and his team "will have invented the first approved personalized mRNA vaccine — an impressive feat indeed, especially this soon after the technology was deployed at scale during the pandemic.
Quest's highly anticipated "Passthrough Camera API" is now available for all developers to experiment with, though they can't yet include it in store app builds.
The new capability was announced at Meta Connect 2024 in September as coming this year. Now it's here, as an experimental release for Quest 3 and Quest 3S. That means any developer can experiment with it, and even distribute APKs using it on platforms like SideQuest, but they can't yet include it in Meta Horizon Store apps. Meta has taken this approach for new APIs multiple times in the past, and typically makes the feature available for use in store apps within a few months at most.
Select developers have had early access to experiment with the capability for a while now, and Meta will host Niantic, Creature, and Resolution to discuss it at GDC next week.
What Is Passthrough Camera Access?
While headsets like Quest 3 use cameras to let you see the real world, until now only the system software got raw access to these cameras. Third-party developers could use passthrough as a background, sure, but they didn't actually get access to it. They instead got higher-level data derived by the system, such as hand and body skeletal coordinates, a 3D mesh of your environment with bounding boxes for furniture, and limited object tracking capabilities. That meant they couldn't run their own computer vision models, which severely limited the augmentation capabilities of these headsets.
The exception was that on visionOS 2, Apple gives enterprise companies raw access to Vision Pro's passthrough cameras for non-public internal apps, but this requires a special licence from Apple and is restricted to "in a business setting only".
For the "Passthrough Camera API" to work, you as the user need to grant the app permission to access your headset cameras, just as you would the microphone. If granted, the app gets access to the forward-facing color cameras, including metadata like the lens intrinsics and headset pose, which it can leverage to run custom computer vision models.
Examples of how apps could use this include scanning and tracking QR codes, detecting a game board on a table to add virtual characters and objects to it, detecting physical objects for enterprise guide experiences, or integrating the visual AI functionality of cloud-hosted large language models (LLMs). Developers are only limited by which real-time computer vision models can run on the XR2 Gen 2 chipset performantly, or which cloud-hosted image models they're willing to pay for.
Meta software engineer Roberto Coviello's QuestCameraKit samples.
The passthrough camera stream is provided to the app with up to 1280×960 resolution per camera at 30FPS, with a stated latency of 40-60 milliseconds. That means it isn't suitable for tracking fast moving objects, such as custom controllers, nor for discerning fine features like small text.
Technically, at a base level, there is no specific Meta Quest Camera Passthrough API, nor is it an extension to OpenXR. Developers do need to request a Horizon OS specific Headset Cameras permission, but otherwise Quest's passthrough camera access leverages Android's existing Camera2 API to also return the headset pose, obtained with OpenXR, and the Camera2 API is what developers of custom engines, or source code for Unreal or Godot, use for it. This also means the same code should work on Google's upcoming Android XR platform, set to debut in Samsung's standalone headset, with only the permission request being different.
For Unity, developers can easily access the cameras through Unity's WebCamTexture API, which is how they already access phone, tablet, and PC cameras and webcams in the engine. A limitation here, however, is that Unity's WebCamTexture API only supports one camera at a time, not both.
Walkthrough from Meta software engineer Roberto Coviello.
Interested developers can find Quest passthrough camera access documentation here: Unity / Native Android.
Meta has published five official Unity samples on GitHub: CameraViewer, CameraToWorld, BrightnessEstimation, MultiObjectDectection, ShaderSample. Meta software engineer Roberto Coviello has separately published QuestCameraKit on GitHub, a collection of five further samples: Color Picker, Object Detection with Unity Sentis, QR Code Tracking with ZXing, Frosted Glass Shader, and OpenAI vision model.
In decades past, as the effects of climate change slowly became undeniable, some looked to the super-rich — the billionaires with enough cash to really make a splash — for solutions.
By the early 2020s, billionaires had positioned themselves as the masters of climate change policy, taking advantage of their great fortunes to become indispensable to environmentalism.
Now, however, many of those same billionaires are pulling support at an alarming rate. And Bill Gates — Microsoft founder, sixth richest man in the world, and alleged sex pest — is the latest among them.
New reporting by Heatmap is signaling the end of a "major chapter in climate giving," as Breakthrough Energy — Gates' climate change nonprofit — has locked the doors on its policy and advocacy office, laying off dozens of employees throughout Europe and the US.
Breakthrough's lobbying was central to advancing climate policy through legislation championed by the Biden administration, including the Inflation Reduction Act, the CHIPS Act, and the bipartisan Infrastructure Law.
Though the billionaire's for-profit green energy investments at companies like Arnergy and Mission Zero Technologies remain in place, Breakthrough's belt-tightening will very likely end the nonprofit's grant writing efforts. That's a major blow to climate nonprofits, and further evidence that, for all their feel-good bluster, the mega-rich never forget their bottom line.
Ever since billionaire real estate mogul Donald Trump won his second presidential election, tech barons like Mark Zuckerberg, Jeff Bezos, Sundar Pichai, and of course Elon Musk have made no bones about shedding their progressive skin and embracing the new administration.
Gates, too, is cozying up to the returning president. In early January, the Microsoft founder spent three hours dining with his fellow billionaire, telling the Wall Street Journal he was "frankly impressed" by Trump's grasp on the issues dear to him.
Though many no doubt feel betrayed by what seems like a sudden rightward turn, billionaires like Gates have always behaved like wolves in sheep's clothing, prioritizing their fortunes above all.
For example, Gates was heavily involved in establishing the Global Fund, a privately-funded rival to the World Health Organization. While the Global Fund did improve global vaccination rates, the cost of basic medicines skyrocketed thanks to his introduction of for-profit actors into global health efforts — another sector made to rely on the generosity of billionaires.
Geniatech’s latest AI-powered industrial ARM box PC integrates the NXP i.MX 8M Plus processor with Kinara’s Ara-2 AI accelerator, offering up to 40 TOPS of AI computing power. This fanless, industrial-grade solution is designed for edge AI applications, providing an alternative to high-cost desktop AI workstations. It supports large AI models with billions of parameters, making it a powerful tool for on-device inference and processing without relying on cloud-based APIs. The device is showcased at Embedded World 2025, emphasizing its capability to run models like DeepSeek and LLaMA while maintaining data privacy. For more details, visit: https://www.geniatech.com/embedded
The system-on-chip (SoC) combines a quad-core ARM Cortex-A53 processor with an additional Cortex-M7 core, ensuring efficient processing for AI workloads. The integrated neural processing unit (NPU) within the i.MX 8M Plus delivers 2.3 TOPS of AI acceleration, while Kinara’s Ara-2 extends this capability to 40 TOPS, significantly boosting inference speed. This setup allows developers to deploy AI applications directly at the edge, reducing latency and operational costs.
With up to 16GB of LPDDR4X RAM for the Kinara Ara-2 AI accelerator and 8GB to 16GB LPDDR4 for the i.MX 8M Plus, the system supports memory-intensive AI tasks. Storage options include 32GB eMMC, expandable to 256GB, ensuring ample space for large AI models and datasets. Video processing capabilities include 1080p60 HEVC, H.265, and VP9 decoding, alongside HDR ISP support for high-quality imaging applications.
The device’s connectivity features include dual Gigabit Ethernet, Wi-Fi, 4G LTE, and GPS support, making it adaptable to various industrial and AI-driven IoT environments. Video output is handled through HDMI and LVDS, with additional interfaces including USB 3.0, RS232, CAN, and MIPI CSI for camera integration. This wide range of I/O options enables deployment in smart surveillance, industrial automation, and other AI-driven edge computing applications.
Unlike traditional AI computing solutions requiring high-end GPUs, Geniatech’s embedded system provides a cost-effective and energy-efficient alternative. With its easy-to-install software and over-the-air update capabilities, the system ensures continuous improvements and access to the latest AI models. The focus on local processing enhances privacy by keeping data on the device rather than relying on cloud services.
As AI adoption continues to grow, Geniatech is actively exploring future developments, including support for larger AI models, potentially reaching 32 billion parameters. This commitment to scalability positions the company as a key player in industrial AI, addressing the increasing demand for edge AI computing solutions.
Check out all my Embedded World 2025 videos in this playlist: https://www.youtube.com/playlist?list=PL7xXqJFxvYvjgUpdNMBkGzEWU6YVxR8Ga
This video was filmed using the DJI Pocket 3 ($669 at https://amzn.to/4aMpKIC using the dual wireless DJI Mic 2 microphones with the DJI lapel microphone https://amzn.to/3XIj3l8 ), watch all my DJI Pocket 3 videos here https://www.youtube.com/playlist?list=PL7xXqJFxvYvhDlWIAxm_pR9dp7ArSkhKK
Join https://www.youtube.com/charbax/join for Early Access to my videos and to support my work, or you can Click the “Super Thanks” button below the video to send a highlighted comment!
Today, I want to discuss one of the XR technologies that has been getting more hype lately: smartglasses. I want to start my analysis with what I’ve seen at CES and then go beyond that and discuss what I envision for the future of this technology.
Smartglasses at CES
In the XR area (and beyond), smartglasses were one of the most popular technologies at CES. There were so many smartglasses and technologies related to smartglasses (e.g. waveguide systems) that I couldn’t try them all. For instance, it’s a pity that I’ve not been able to try the Halliday glasses. But still, I managed to get my hands on a few interesting devices.
Lightweight glasses
The first lightweight (AI) smartglasses that I tried there have been Rokid Glasses, the latest device by Rokid, one of the leading companies in AR. These are very lightweight smartglasses that have speakers, a monochrome green waveguide display for notifications and small texts, and a camera to shoot photos and videos. They have a companion app through which you can manage them and thanks to this you can also have connectivity to an AI agent and other AI services. Here you can see some photos I’ve taken of the product:
I liked the fact that Rokid glasses were very lightweight and also stylish. This is because the glasses are built in collaboration with Swedesh eyewear manufacturer BOLON… which is one of the brands of the EssilorLuxottica group.
These glasses look pretty cool. My face is not as cool as the glasses, though
They were also doing their job pretty well: I tried to have a conversation with one of the Rokid employees there, with her speaking Chinese and I speaking English (and also a bit Chinese), and I could see the translation of what she was saying written in green in front of my eyes.
You can see the glasses in action in this video by Tyriell Wood
The companion app also allowed me to speak with an AI assistant and also had some fitness-oriented features that I had not tried. But I have tried to shoot some photos and videos: they are recorded on the glasses and then can be moved on the phone through the companion app. The quality of the recorded media is not fabulous, but it’s ok. All the time, the screen was just green, but the color was quite vivid (the glasses have like 1000 nits), so the text was very readable.
Through-the-lenses of Rokid Glasses
I liked the glasses by Rokid: they were lightweight, fashionable, the screen was readable, and they were able to do few things and do them well. They were not perfect, but good for the current status of the technology.
Specifications of Rokid AR glasses
Fast forward a few days, the last glasses that I tried at CES were the ones by LAWK, which is another Chinese brand. They featured a display showing notifications and green text and they allowed me to speak with an AI, have live translation between English and Chinese, and shoot photos and videos. If this sounds familiar to you, it is because you’ve read the same things in this article, 10 lines ago. Long story short, many of the smartglasses were just clones the one of the other, just with a different design. LAWK ONE, the device currently available on the website of the company, is anyway much bulkier than the Rokid glasses, also because it is targeted at people doing sports, like cycling on a bike.
LAWK ONE glasses
I can confirm that also in this case, the glasses were working, with the translation service doing its job. I was not a big fan of the look and feel, though, especially because the frames were pretty big.
Me wearing the LAWK ONE smartglasses
LAWK had also a new model of glasses that were as lightweight as Rokid’s and I could put them on my face and see they were pretty comfortable, even if not as stylish as the ones designed by BOLON for Rokid.
LAWK’s more lightweight glasses
I could not turn them on because the guy at the booth told me “They have the same features as the others you tried, no need to turn them on”. So my review on them will basically be “Trust me, dude”.
They work. Probably.The companion app for LAWK smartglasses
I’ve already written an article about my hands-on with the Ray-Ban Meta glasses. But summarizing my experience with them: I’ve found them stylish and comfortable, the speakers were loud and clear, the videos and photos taken were good, and I was intrigued by the potentialities of the AI features. Even if they had no display, they were still able to deliver a lot. Still, the demo with them was about… photos and videos, live translation, and AI.
Me being very classy with my suit and the Ray-Ban Meta glasses
There were some private rooms by Google and Samsung at CES and surely they were demoing Android XR to close partners. Unluckily I’ve never been able to try an Android XR device. But from what I’ve read in the various magazines when they broke the news about this new operating system, the demos of the Android smartglasses prototypes were about live translation, AI, photos, and videos. I guess you’re not surprised by it.
The only official image about Project Moohan (Image by Google)
I also had a quick hands-on with the TCL RayNeo X3 Pro. What impressed me about these glasses is that the display had colors. It was not green, it was RGB and could show also 2D icons of the various applications. And the visuals were also pretty bright: with 2500 nits they should theoretically work also outdoors. The FOV was the usual small one typical of the smartglasses. And the processor was a Qualcomm Snapdragon AR-1. All of this was in a form factor which was still rather small and comfortable to wear. I’m not surprised that many journalists attending MWC praised this device (e.g. in this article or this other one). The fun thing is that if you read the articles from MWC, you will discover that the main use case that the journalists have tried was AI translation from Chinese to English. Ah and there were also photos and videos of course.
TCL RayNeo X3 Pro
I liked the use of RGB colors on the display because it allowed the glasses to have an interface that was prettier. And the glasses were not much bigger than the competition. But they were more expensive.
Also these glasses didn’t look that bad on my face, after all
Since most smartglasses were basically the same thing, after a while I even stopped trying them. I read so many translations from Chinese to English that I became fluent in Mandarin.
Ah, I’ve also shot some pictures at the Vuzix booth, let me share them with you:
Vuzix booth at CESOf course I’ve also taken a selfie with the Vuzix glasses!
My vision for lightweight smartglasses
I have mixed feelings about the hype for smartglasses. On one side, I understand it: thanks to the advancements in technology, it is now finally possible to have a tech wearable that looks cool on your face. You can wear smartglasses and not looking like a weird dork, but just a person wearing glasses. I pretty enjoyed trying them and taking selfies with them. I also think that they are useful because they will allow us to connect with AI in our everyday tasks: when we wear these glasses, the AI can see what we are seeing and suggest us what to do. This can be very useful, and potentially also life-changing.
On the other side, I think we should be cautious. First of all, the hype for smartglasses has stemmed from the success of Ray-Ban Meta, but it seems people are not considering that Ray-Ban had huge merits in this success story. Ray-Ban Meta are Ray-Ban glasses, so they are cool, they are stylish, they make you wear a famous brand. They are distributed by Essilorluxottica in all its glasses shops, so you can enter a shop to buy some sunglasses, and you can come out with sunglasses that beyond being cool sunglasses can also shoot photos and videos. Ray-Ban Meta is enhanced eyewear, the other devices I have tried are tech products. Yes, technically they are the same, but they are sold in a way that makes a difference. Plus Ray-Ban Meta is a product, a finished and polished one: many demos of the other smartglasses I tried had issues, while Ray-Ban Meta worked like a charm. Plus the Meta showcase booth was amazing and also had a very cool case for every one of the glasses. That’s why even if Meta glasses lacked a screen that instead the other glasses were having, I still consider them a better product than the other smartglasses I’ve tried.
Trying on Meta Glasses and their AI features
Then I think we should talk about the use cases. I don’t wear glasses, so to make me put something in my face for hours, you should give me a strong reason. And live translation from Chinese to English is not one: as much as I travel a lot to China, I don’t need it every day of my life. I’m someone who travels a lot, but most people travel less than me, so translation services are pretty useless to them. No one needs to do translations every day. Taking photos and videos from your point of view is nice. Regarding fitness, I don’t think I would ever run with glasses not made for sports. Long story short, I don’t understand why I should buy them. This is the same problem that smartwatches had in the beginning, until they found their purposes, like in the healthcare and fitness sectors. I don’t think we have a killer use case for smartglasses yet.
You may counter my argument by saying that if there are no clear use cases, then why are people buying the Ray-Ban Meta? Well, people buying a Ray-Ban Meta are entering an eyewear shop because they already want to buy glasses. So they already have a need. And if they can choose between cool Ray-Ban glasses and cool Ray-Ban glasses with extra features, they buy the second one, of course. This is different than waking up one day and going to the Rokid website and buying smartglasses. Hugely different. To do that, I have to feel the need to buy smartglasses, and currently, I don’t have it. Sure, we’ll get there, it is a matter of time, I’m just saying that TODAY I’m not as hyped about smartglasses as other people are, because I don’t see a reason why someone who doesn’t need glasses should wear them every day. But I’m surely positive about the future.
One last thing about use cases and usability: I think one big issue with these glasses is that they are not programmable. Apart from a few ones, like Brilliant Labs Frame, most of these glasses just work with their companion app and deliver the features implemented by the manufacturer (which means the translation between Chinese and English…). I wonder when these glasses, and in particular Meta Ray-Ban, will allow developers to create applications for them. This would be good for developers who can so have a new source of revenue in a growing market and would be good for manufacturers because developers could envision new use cases, possibly not related to translation. This could be a good boost for the ecosystem. That’s why I was pretty intrigued by the idea of AugmentOS to offer an SDK to develop your application once and let it run on different smart glasses.
XREAL, Lenovo, and the virtual screens
Beyond the AI smartglasses, there is another category of smartglasses that is kinda popular and it is the one delivering one or multiple virtual screens to the user. At CES I tried a few devices in this sense, one being the glasses from Lenovo, that were connected to a gaming console…
They look a bit like sunglasses for policemen
…. and another one being the XREAL One Pro, which can be connected both to your phone and your PC. XREAL was one of the booths with the most visitors in the XR area, and for a reason: in these years they managed to establish themselves as one of the best brands for what concerns stylish AR glasses and smart glasses.
XREAL One Pro is a very interesting device: it is quite lightweight (even if not as much as the AI smartglasses) and can show you a big virtual version of your laptop screen. Through the buttons on the frames, you can configure a few options and for instance, decide if you want this virtual screen always attached to your eyes (0 DOF) or to stay fixed in a position in front of you (3 DOF). You can also decide if to keep the classical aspect ratio of your display or have an ultra-wide one. The colors of the display were pretty crisp and the text of the virtual screen in front of me was very readable. These glasses have two clear use cases: one is media consumption, so watching Netflix on a big screen in front of you; and the other one is productivity, which is letting you do your work on a big screen, which is great, especially for people working with multimedia.
Xreal One Pro glasses
My friend Tyriel Wood, who attended the CES with me, told me that he likes this kind of device and they are already pretty useful for him. We are at a stage where they can already be used in productivity. After my hands-on, I’m almost convinced about this piece of hardware, with my three only problems being the FOV, the connected hardware, and the eye fatigue.
XREAL did a great job in making the FOV as large as possible with its 57°. But still, when I was looking at the ultrawide virtual screen, I felt I could not see the whole screen from my glasses, but some tiny lateral parts were missing. We need more FOV so that I don’t have to turn my head to see the different portions of the virtual screen.
I look pretty cool with the XREAL One Pro on
Regarding the connected hardware, my problem is finding a setup that lets me work on the go. For me, it would be ideal if I could just carry some smartglasses, a small keyboard with a touchpad (like the ones of tablets), my phone, and be able to work from everywhere (e.g. the planes, the buses, etc…), without having to carry the big weight and dimensions of my laptop. Having a widescreen on my desk would be good, but instead of buying an XREAL One Pro, I could buy an extra monitor for much less money. But if a future evolution of XREAL One Pro could give me a working station from everywhere, without having to take with me a big bag every time, I would insta-buy it. In fact, I loved the demo of the XREAL Air 2 Ultra because they made me try the glasses with a keyboard and a phone.
This is how I imagine I could work in the future
Regarding eye fatigue, I can not comment on the long-term usage of these glasses, because I’ve tried them only for a few minutes, but I wonder how would I feel after having worn them for 12 hours a day. If I had to make a bet, I would say they stress the eyes more than a standard display, but I can not be sure of that until XREAL sends me one (XREAL people, if you are reading this, send me your glasses!)
In any case, I think this type of device is already pretty nice and useful for some use cases. If the FOV was bigger, it would be even better.
XREAL Air 2 Ultra
Ted Schilowitz trying XREAL Air 2 Ultra glasses
The last type of device I want to talk about is the XREAL Air 2 Ultra, which is 6 DOF glasses. It’s good to see that XREAL is back to doing 6 DOF glasses and I have to say the device is pretty good. The glasses can show 3D objects in the environment around you, with bright colors and a decent FOV. They work being connected to the phone, so they can not render very heavy scenes. I’ve tried also the hand tracking and found it to be ok, but not as advanced as the one from Meta or Ultraleap.
The road to AR devices
Meta Orion glasses (Image by Meta)
I wanted to mention my brief hands-on with the XREAL Air 2 Ultra because I think that in the end, 6DOF AR glasses are the endgame for all the devices I have described in this article. AI smart glasses, and glasses for virtual screens, are all simplified versions of glasses that take care of a specific use case for an affordable price. But the final mission is having a device that can do all that these glasses can do, and even more: of course, I’m talking about 6DOF glasses that can understand the environment around us and render both 3D and 2D objects.
Unluckily, the technology today is not ready to make the AR glasses of our dreams, and in fact, the most advanced glasses we know about, Meta Orion, cost more than $20K to manufacture. But I’m a big believer we’ll get there, and in the meanwhile, all the various smartglasses that are being sold will be useful to find use cases for which people want to put some glasses on their face and to make wearing tech glasses more socially acceptable. Hopefully, this period will also be useful in understanding how to guarantee privacy to the users who are wearing glasses with cameras, but I’m not sure this will happen, unluckily.
A fun moment at CES
Since I like to always add a touch of humor to my posts, let me tell you something weird that happened when I visited the booth of a Chinese manufacturer of smartglasses. After I had tried the device, including the usual photo-taking and AI translation, I asked the guy at the booth. “What’s the price of this?” and he answered something like “58 grams” I was pretty confused, so I asked again “What’s the price?” And he answered this time. “Europe, USA” I was getting pretty confused… it was like a slot machine that every time was given me a random sentence in English in return, so I asked again. “What’s the price, the cost, money?” And he started looking at the sky as if he hoped the Gods may tell him the right answer. Considering that he stood still like this for 5 seconds, I guess the Gods were busy doing something else. In the end, I had really enough so I went into full Chinese mode and asked “多少钱?” to which he answered me 400$ and then said a lot of things speaking in Chinese superfast I couldn’t understand (and to which I was tempted to answer with “58 grams”).
I have a question for this guy: I understand that ...
This week, Vuzix Corporation noted a major uptick in its hardware sales and adoption. This follows a period in 2024 when 2023’s third-quarter figures dropped due to lower sales of the M400 product in key areas like Asia.
However, Xander is significantly reordering Vuzix Shield AR smart glasses that are helping to continue an upward tick for the veteran’s enterprise XR hardware vendor, which appeared to kick into action following the Q2 drop.
The Vuzix Shield provides a hardware framework to enable Xander to drive support and increase customer interest in its XanderGlasses solution, a patient care solution that allows users with auditory disorders to leverage real-time informatory AR captions to support everyday tasks.
The XanderGlasses solution leverages Vuzix brand AR smart glasses technology to enable the focused solution, with Alex Westner, Co-Founder and CEO of Xander, noting:
With 48 million people in the U.S. experiencing hearing loss, a number that continues to rise, XanderGlasses empower individuals to ‘see’ what others are saying. Our team has worked closely with individuals at all stages of hearing loss to develop a simple and reliable solution that meets real-world needs.
Xanderglasses transforms the Vuzix Shield hardware into a device ready for this specific healthcare need. Xander gained recognition for its product as the CES 2024 Innovation Award Honoree in the category of Accessibility & Aging Tech.
“As hearing loss affects a growing segment of the global population, including veterans, those exposed to prolonged noise, and ageing individuals, innovative solutions like XanderGlasses can make a significant impact,” remarked Paul Travers, President and CEO of Vuzix.
Vuzix Surges in the AR Market
The news comes during a much-needed period of growth for Vuzix. Travers noted that “by combining Vuzix Shield smart glasses with Xander’s powerful embedded voice captioning software, we are addressing a significant need,” creating further demand for its hardware portfolio, “we look forward to supporting Xander as they scale their solution this year and beyond.”
Last year’s Q2 financial report included total revenues decreasing by a staggering 77 percent to approximately $1.1 million. In comparison, the previous year’s revenue was about $4.7 million. Additionally, Vuzix reported an annual gross loss of $0.3 million for Q2 2023.
Since this staggering return, Vuzix has experienced a strong upturn, showing the drop was a mere blip. Moreover, interest in AR smart glasses continues to rise.
Vuzix has significantly expanded its collaboration with the U.S. Department of Labor’s Occupational Safety and Health Administration (OSHA), which broadens the deployment of its M400 workplace AR smart glasses within the governmental organization.
Additionally, Vuzix introduced AugmentOS, a specialized operating system designed for immersive human-computer interfaces in workplace settings. This operating system is now available for Vuzix’s product line, enabling both end-users and developers to utilize products such as the Z100 smart glasses, which incorporate AI functionality.
In 2025, Vuzix is experiencing notable success, and with the AR smart glasses market booming, the company is gaining impressive momentum.
At Embedded World 2025, Geniatech unveiled a groundbreaking 31.5-inch ePaper display, marking a significant advancement in ePaper technology. This large-format display integrates a specialized algorithm to enable smooth motion and fast refresh rates, making it one of the first ePaper screens capable of displaying video content. The development of this technology has been highly anticipated for decades, as traditional ePaper has struggled with slow refresh rates. Now, with Geniatech’s innovation, dynamic content can be displayed with impressive clarity and fluidity. https://www.geniatech.com/epaper-digital-signage/
One of the standout features of this display is its ability to render partial refreshes, allowing for rapid content updates without requiring a full-screen refresh. This makes it ideal for applications such as digital signage, public information displays, and outdoor advertising, where frequent updates are needed. The technology behind this display leverages FPGA-based processing to optimize refresh performance, overcoming the limitations of conventional ePaper solutions.
The reflective nature of ePaper ensures excellent visibility in various lighting conditions, including direct sunlight, making it particularly suitable for outdoor applications. Unlike traditional LCD or OLED screens, which rely on backlighting, this display utilizes ambient light to create high-contrast images, ensuring readability while reducing power consumption. This ultra-low power requirement makes it a strong candidate for energy-efficient digital signage solutions.
Geniatech’s approach to ePaper also focuses on sustainability. With minimal energy use and a display design engineered for longevity, it aligns with industry efforts to reduce electronic waste and power consumption. Businesses looking to implement digital signage with lower environmental impact will find this solution appealing, as it offers both performance and efficiency without compromising on readability.
The sharpness and responsiveness of this display make it particularly effective in scenarios where static and motion content need to coexist. For example, retail environments can use it for dynamic pricing updates, real-time promotional content, and interactive wayfinding displays. Similarly, transportation hubs could deploy it for up-to-date scheduling and route information.
Geniatech anticipates that this 31.2-inch ePaper display will enter mass production within the next two months. The combination of video playback capability, energy efficiency, and outdoor readability positions it as a transformative solution for businesses and organizations seeking innovative display technologies. As more industries explore ePaper for dynamic content, this breakthrough could reshape the landscape of digital signage.
Check out all my Embedded World 2025 videos in this playlist: https://www.youtube.com/playlist?list=PL7xXqJFxvYvjgUpdNMBkGzEWU6YVxR8Ga
This video was filmed using the DJI Pocket 3 ($669 at https://amzn.to/4aMpKIC using the dual wireless DJI Mic 2 microphones with the DJI lapel microphone https://amzn.to/3XIj3l8 ), watch all my DJI Pocket 3 videos here https://www.youtube.com/playlist?list=PL7xXqJFxvYvhDlWIAxm_pR9dp7ArSkhKK
Join https://www.youtube.com/charbax/join for Early Access to my videos and to support my work, or you can Click the “Super Thanks” button below the video to send a highlighted comment!
Google is reportedly set to acquire Canada-based eye-tracking startup AdHawk Microsystems Inc., something that would strengthen the company’s ongoing foray into XR headsets and glasses.
As reported by Bloomberg’sMark Gurman, Google is allegedly acquiring AdHawk for $115 million, according to people with knowledge of the matter.
The deal is said to include $15 million in future payments based on the eye-tracking company reaching performance targets. While the acquisition is purportedly slated to conclude this week, a deal still hasn’t been signed, leaving some room for doubt. Furthermore, should the deal go through, the report maintains AdHawk’s staff will join Google’s Android XR team.
This isn’t the first time AdHawk has flirted with an acquisition by a key XR player. In 2022, Bloomberg reported the company was in the final stages of an acquisition by Meta.
Notably, AdHawk is best known for its innovations in eye-tracking, which replaces traditional cameras with micro-electromechanical systems (MEMS), which is said to result in faster processing and reduced power consumption—two things highly prized by AR and smart glasses creators today.
Image courtesy AdHawk Microsystems Inc.
Its flagship product, the MindLink glasses, is a research-focused device that is meant to connect eye movements with neurological and ocular health, human behavior, and state of mind, the company says on its website. Additionally, the company offers its camera-free eye-tracking modules for researchers working with VR devices, such as Meta Quest.
While neither Google nor AdHawk have commented on report, Google is ramping up its XR division to compete with the likes of Meta and Apple.
In December, Google announced Android XR, marking a decisive shift for the company’s XR efforts, as the company is bringing a ‘full fat’ version of Android to headsets for the first time, which not only includes XR-specific apps but also the full slate of Android content. Android XR is ostensibly set to debut on Samsung’s Project Moohan mixed reality headset, which still has no release date or price.
Then, in January, Google announced the acquisition of a number of HTC’s XR engineers, a deal amounting to $250 million. At the time, Google said HTC veterans would “accelerate the development of the Android XR platform across the headsets and glasses ecosystem.”
In addition to supporting its Android XR software efforts, the acquisition of a novel eye-tracking startup would also prove valuable in the company’s internal XR hardware efforts, which has been nothing short of fragmented over the years.
Un actionnaire minoritaire d'Ubisoft lance un appel à la protestation contre la gestion actuelle de l'entreprise, après plusieurs années de résultats décevants et d'incertitudes financières.
Après avoir fait sensation auprès du public et des artistes à Las Vegas, le concept spectaculaire de la Sphere s’apprête à évoluer pour s’adresser à un public plus large et à diversifier les expériences proposées.
Constituant une véritable révolution technologique, l’Intelligence Artificielle (IA), et plus particulièrement l’IA générative, impacte désormais la gestion de contenu au sein des entreprises, redéfinissant notamment les pratiques des secteurs du luxe et du retail. En s’intégrant progressivement aux plateformes de gestion de contenu, ces technologies offrent des opportunités inédites pour optimiser les processus internes et enrichir l’expérience client. Décryptage.
Ce n’est pas un hasard si l’IA générative fait désormais partie intégrante des stratégies digitales des entreprises du luxe et du retail. Louis Vuitton, Dior ou encore Estée Lauder font partie des premiers à s’être emparés de cette technologie, l’intégrant notamment dans le développement d’agents conversationnels au service de l’expérience client.
For years, businesses, governments, and researchers have struggled with a persistent problem: How to extract usable data from Portable Document Format (PDF) files. These digital documents serve as containers for everything from scientific research to government records, but their rigid formats often trap the data inside, making it difficult for machines to read and analyze.
"Part of the problem is that PDFs are a creature of a time when print layout was a big influence on publishing software, and PDFs are more of a 'print' product than a digital one," Derek Willis, a lecturer in Data and Computational Journalism at the University of Maryland, wrote in an email to Ars Technica. "The main issue is that many PDFs are simply pictures of information, which means you need Optical Character Recognition software to turn those pictures into data, especially when the original is old or includes handwriting."
Computational journalism is a field where traditional reporting techniques merge with data analysis, coding, and algorithmic thinking to uncover stories that might otherwise remain hidden in large datasets, which makes unlocking that data a particular interest for Willis.
Tesla's stock tumbled by almost 16 percent on Monday, driven by widespread pessimism over the EV maker's plummeting sales numbers worldwide.
Wall Street analysts lowered their price targets significantly, expecting the stock to slide even further. Tesla's losses have been so staggering over the last couple of weeks, the company has wiped out all of its gains since president Donald Trump was elected in November. Its stocks are currently down over 55 percent since hitting a record high in mid-December.
The carmaker's CEO Elon Musk has felt the hurt too, wiping out over $120 billion of his net worth since reaching an all-time high late last year. Today alone, Tesla's sliding stock has cost the entrepreneur a whopping $12 billion.
Meanwhile, Musk has taken to his social media platform X-formerly-Twitter in a desperate attempt to drum up excitement for his embattled carmaker.
While Tesla's stock was getting hammered, Musk was retweeting posts that gushed over Tesla being the "best car" and tales of the company's plagued driver assistance feature saving them from a crash.
Anger over his endorsement of far-right extremism and two Nazi salutes during Trump's inauguration celebration has begun to spill over, with waves of protests and vandalism hitting Tesla dealerships across the country.
The man to blame for the outpouring of negative sentiment has since tried to play the victim card, while senselessly blaming other billionaires for Tesla's disastrous stock performance.
"Heartfelt thanks to everyone supporting Tesla," Musk wrote, "despite many attacks against our stores and offices."
Tesla's massive slide was accompanied by a widespread and extended outage of Musk's social media platform X-formerly-Twitter.
Without offering any evidence, the billionaire blamed a "massive cyberattack."
"We get attacked every day, but this was done with a lot of resources," Musk tweeted. "Either a large, co-ordinated group and/or a country is involved."
Considering the sheer number of enemies Musk has made, it's not exactly a stretch to imagine somebody had it out for the platform.
"What we've been seeing is consistent with what we've seen in past denial of service attacks, rather than a configuration or coding error in the platform," Netblocks director Alp Toker told the BBC, referring to a type of attack that takes down websites by overwhelming them with traffic.
Other social media users suggested Musk may have orchestrated the outage to distract from the company's woes.
It's not just Tesla and Musk feeling the hurt. US stocks plunged today overall, the worst day for the Nasdaq and Dow since 2022, as Yahoo Finance reports.
Trump's trade war has sparked widespread uncertainty and market pessimism. Over the weekend, the president refused to rule out the possibility of a recession, describing the economy as going through a "period of transition."
And Musk is happily doing his bidding, leading mass layoffs of federal government workers. The billionaire has been accused of stuffing his pockets and using his newfound influence in the White House to secure plush governmentcontracts for his businesses.
But Tesla investors aren't impressed with his performance, accusing Musk of being far too distracted by his government rampage. A poll by retail stock-trading platform StockTwits conducted earlier this month revealed that 60 percent of respondent investors said that "Musk's White House focus" is "hurting Tesla," as Business Insider reports.
"We think shareholders have legitimate concerns about Elon Musk being spread too thin and it's become clear he's now spending more time on DOGE than anything else," CFRA Research senior equity analyst Garrett Nelson told BI.
"The truth of the matter is the company should be run by somebody who isn't so political and can bring the brand back, if that's possible," Tesla investor Ross Gerber told the outlet last month.
Tesla's disastrous day leaves plenty of questions unanswered. Is this just a temporary blip, or are investors looking for a more permanent readjustment?
Musk has bet big on the carmaker's plans for rolling out a robotaxi service. But given the company's financials, Tesla appears to have far bigger fires to put out before it can deliver on that promise.
If you’ve ever watched a sci-fi movie where doctors use futuristic technology to see inside the human body in real time, then Medivis’ augmented reality surgery might feel like it’s straight out of the future. But it’s real—and it’s happening now.
Using Microsoft’s HoloLens 2, this groundbreaking technology overlays 3D MRI scans onto a patient’s body, allowing surgeons to see tumors, blood vessels, and other critical structures before they even make an incision. It’s like having a holographic X-ray that comes to life right in front of you. Watch the video and see how augmented reality is making surgery safer and smarter!
if you’re in Brussels and looking for something to do this Thursday night, come out to my book talk!
…As always, the searchable archive of all past issues is here. Please please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay support access for all AND compensation for awesome ChinAI contributors).
Context: There’s been a lot of noise about Chinese firms racing to adopt DeepSeek, but can Chinese AI chip companies support the deployment of DeepSeek models? This AItechtalk article (link to original Chinese) analyzes the capabilities of Chinese AI chip outfits (e.g., Cambricon, Moore Threads, Enflame, etc) to support inference services for DeepSeek models. One crucial distinction: whether the chips support a distilled version of DeepSeek1 — a smaller-sized model with a few billion or dozens of billions of parameters, which reduces deployment costs — or the full-parameter version of DeepSeek (with a parameter count as high as 671 billion).
Key Takeaways: With nearly 20 Chinese chip companies rushing to announce that their products could support DeepSeek’s models, one important point of clarification is needed: can your chips the full-parameter version or just the distilled version?
Running the full-powered version of DeepSeek requires more than just one server rack of 8 cards, presenting challenges for Chinese chip firms that struggle to run multiple high-performance servers in parallel. On these interconnection problems, senior AI chip engineer Jack relays, “It will be difficult to do it, and there may be no end in sight to successfully adapting to the full-parameter version of DeepSeek.”
In contrast, Chinese AI chips face no issues with supporting the distilled version of DeepSeek. And perhaps we shouldn’t be so quick to discount the distilled versions. “I once thought that adapting a distilled version of DeepSeek model was not very valuable, and many engineers also preferred the full-blooded version of DeepSeek, but now my thoughts have changed.” Bo Lin, who has more than 20 years of experience in the chip industry, said, “The distilled version of the model can meet the chat needs of ordinary users, which is of great significance to the dissemination of AI.”
Despite the accuracy limitations, Jack also stated that distilled models can significantly boost the capabilities of edge AI: “With a distilled version of DeepSeek, for example, a particular application scenario that could only deploy a 7B model before can now achieve the performance of a 14B model.”
Why do Chinese AI chip companies trail Nvidia on this front? We’ve mentioned the issue of interconnections within and between server racks (Nvidia’s NVLink is a key strength here). Let’s get deeper into the details:
Chinese AI chips do not support FP8 data representations, which is a common method to reduce the memory footprint of AI applications. By contrast, in 2022, Nvidia’s H100 chip already supported FP8. Since Chinese AI chips only support FP16, deploying DeepSeek requires twice the storage and memory footprint, increasing the need for more cards.
Bo Lin, the chip industry veteran, is very blunt about the fact that the latest Chineese AI chips do not support FP8: “This shows that many people who make AI chips in China do not understand AI.”
From the article: “After DeepSeek exploded, we wanted to adapt it with a card from a domestic AI chip company.” Boyuan, a practitioner at a Chinese intelligent computing center, said, “But the reality is that if the (inference) performance of DeepSeek on an (Nvidia) A100 is 100 points, this domestic card only provides a few points of performance, and even if it is optimized, it only has a performance of around 10 percent that of the A100.”
I’ll conclude with some granular figures to keep an eye on going forward.
One useful indicator is the inference processing speeds of these chips (when running DeepSeek models, for instance). You want to get to at least 20 tokens per second to provide a satisfactory user experience; this results in a first word latency of 1-1.4 seconds.
AItechtalk learned that “current leading Chinese AI chip companies have only achieved 10 tokens/s…to adapt to the full-parameter version of DeepSeek.” Though, the piece also cites some reports by Chinese AI chip companies that they’ve achieved 25 tokens per second in intelligent computing centers in deploying the full-parameter version of DeepSeek.
Many of AItechtalk’s sources suggested that Chinese companies wouldn’t get to the 25 tokens/s mark until the end of the month.
Note: Jack, Bo Lin, and Boyuan in the article are all pseudonyms.
For The Verge, Justine Calma revisits America’s first generation of semiconductor manufacturers:
The industry employed many women and young people from Hispanic and Asian immigrant families who’d previously worked in canneries that were closing up shop as Americans started importing more fresh fruit. The factories offered a new kind of assembly line work that you didn’t need a degree or much training to land. But a lack of appropriate safety measures left workers vulnerable to a slurry of chemicals that posed dire health risks. Over the years, many of the workers had miscarriages, including Yvette.
Now, as the U.S. is building out new semiconductor manufacturing factories, CHIPS Communities United — a a coalition organizing for the safe and responsible implementation of the CHIPS Act — has “published a letter to semiconductor industry execs asking them to sign legally binding community benefit agreements when they build new fabs. They asked companies to replace chemicals that can cause cancer, miscarriages, birth defects, and fetal brain damage.”
The Intellectual [知识分子] did a helpful summary of Dalian University of Technology’s annual report on China’s R&D spending. The comparison of R&D intensity levels between China and G7 countries was especially useful. China sits in the middle of the G7 countries on this metric, with a R&D investment intensity of 2.54%, which trails the U.S., Japan, Germany, and the United Kingdom.
By Zeyi Yang, this is a portal-opening Wired article that profiles Candise Lin, a California-based social media influencer who “scours the Chinese internet looking for a new celebrity feud, the hottest meme, or perhaps a viral college dorm challenge, which she then translates into English and explains in a minute-long video.”
As Josef Burton, former US diplomat who follows Lin on Instagram, captures the significance of this portal, “China is presented as this completely othered place where no one jokes around, this censored, barren hell space that’s all hyper propaganda … But no, people joke around. Daily life exists. Memes exist.”
Robyn Mak’s Reuters commentary applies some of the key takeaways from my argument about China’s “diffusion deficit.” One interesting tidbit: “Last year, more than 60% of 500 small and medium sized (Chinese) enterprises polled are only in the ‘early’ stages of digitisation, using basic data management and IT applications.”
Thank you for reading and engaging.
These are Jeff Ding's (sometimes) weekly translations of Chinese-language musings on AI and related topics. Jeff is an Assistant Professor of Political Science at George Washington University.
Check out the archive of all past issues here & please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay for a subscription will support access for all).
Also! Listen to narrations of the ChinAI Newsletter in podcast format here.
From the article: “A distilled DeepSeek model uses the data generated by DeepSeek-R1 to fine-tune other models. The parameters range from a few billion to dozens of billions, such as DeepSeek-R1-Distill-Qwen-1.5B/7B/14B/32B, DeepSeek R1-Distill-Llama-8B/70B.”
Сybercrime has been growing, and hackers are invading your systems and networks to steal data, install malware, and more. So why do these individuals commit attacks in cyberspace? The way to gain a better understanding of the mindset of different types of hackers is to be able to better protect yourself as a business, as a government and as an individual.
In this article, we will look at the most common psychological profiles of hackers in general and try to determine what drives cybercriminals. We’ll look at financial motivation, ideology, ego, curiosity and boredom, revenge, and more. Knowing the diverse reasons they have will help strengthen cybersecurity strategies.
Financial and Material Gain
One of the most common motivations for hackers is financial or material gain. Cyberattacks provide opportunities to make money quickly and anonymously. Even basic ransomware campaigns that encrypt files can net thousands of dollars in cryptocurrency from a single business.
Selling stolen data is also lucrative - full identities and financial information often fetch over $1,000 per record on dark web marketplaces. The profits only increase for more sensitive data like healthcare records, intellectual property, or classified government information. This black market economy fuels many financially driven attacks.
Crime as a Service business model also eliminates the barriers to monetising cybercrime. Aspiring hackers now have the ability to rent the malware, tools or botnets needed to launch a DDoS attack or a card fraud scam. It is a ‘cybercrime gig economy’ where hackers do not require advanced technical skills to make money.
As IT security trends evolve, the rewards of cybercrime continue to grow while the risks remain relatively low. Proving that a cybercriminal operates across multiple jurisdictions and borders is hard for law enforcement to do. Many financially motivated hackers are betting correctly that a successful attack is worth taking the low risk of any consequences. The money is there, and the barriers to entry are low, so it is no surprise that many hackers are lured by the money.
Ideology and Espionage
Nation state hacking groups also carry out cyberattacks, but they are typically motivated by ideology, espionage or geopolitical interests instead of just profits. These state sponsored groups have resources, capabilities and targets that set them apart from traditional cybercriminals.
Groups linked to countries like China, Russia, Iran and North Korea routinely infiltrate foreign corporations, governments and critical infrastructure systems. The goal is to steal classified information for military and economic advantages.
Attacks related to government espionage have targeted nuclear power plants, electrical grids, government agencies, defense contractors and more. The rise of state sponsored hacking greatly expands the traditional scope of cybercrime.
Ideological hacktivists also attack to advance political agendas, albeit without the resources of nation-state groups. Anonymous and spinoff collectives like LulzSec have claimed responsibility for high-profile denial-of-service attacks, data leaks and website defacements against targets ranging CIA.
While these ideological attacks may not be as technically sophisticated or prolonged, they can still be highly disruptive. Distributed denial-of-service (DDoS) campaigns that overwhelm sites with junk traffic remain a popular tactic for political hacktivists.
Understanding the motives between state sponsored groups and hacktivists can better inform defenses for likely targets. Their attacks are often more focused on disrupting operations or manipulating public perception to advance ideological causes rather than just stealing data or demanding ransoms.
Ego and Fame
In an anonymous world, hacking can appeal to individuals seeking infamy, notoriety and ego boosts. Being the first to exploit a major vulnerability or compromise a high-value target earns significant respect and credibility amongst the cybercriminals underground.
Some hackers even incorporate ego and status into their public personas. Figures like Eugene Kaspersky and Kevin Mitnick achieved mainstream name recognition during their indictments for hacking-related crimes. Other hackers like Guccifer 2.0 maintain blogs to discuss their latest data leaks and taunt victims.
The media itself fuels this cycle by portraying cybercriminals as eccentric masterminds. Of course, the reality is often more unimpressive – most common attacks rely on simplistic methods, reused malware toolkits and vulnerable targets.
However, we have a psychological desire for recognition. Journalists on the receiving end of a hack may be tipped off in advance so that the attack can be hyped up. Or they might tweet about a major exploit on social media before releasing proof of concept code. These public theatrics all feed egos.
Also, in the development of malware and vulnerabilities, underground credibility comes into play. Exploits or 0-day attacks sell for premium rates on criminal marketplaces before software vendors patch the software. Elite technical skills are shown in the capability to compromise systems in ways that defenses can’t react to.
Many hackers undoubtedly get intrinsic satisfaction from overcoming complex security controls. Outsmarting Fortune 500 security teams feeds egos and reputations no matter the underlying motive.
Curiosity and Boredom
For less experienced hackers, curiosity and boredom can be big motivators, especially among younger demographics like students. This helps explain some opportunistic cyber-vandalism, such as website defacements.
In these cases, novice hackers often want to test their abilities more than cause real damage. Breaking into insecure websites feeds intellectual curiosity even if no data is actually stolen. It creates challenges to learn new techniques like SQL injection attacks, cross-site scripting and other web app exploits.
These hackers were curious to explore cybersecurity topics through both legal and illegal methods. On the one hand, many students participate in capture-the-flag competitions, security meetups and hackathons to experiment freely. However, some also turn to unauthorised penetration testing against websites or networks.
Although curious hackers don’t have malicious intent, what they do is very serious. If no data is changed, but there is unauthorised access, public trust and relationships will be shaken. However, there may be reasons to respond to hackers who are more curious than greedy or ideologically motivated.
For some hackers, more experienced, boredom also plays a role on the other end. When basic vulnerabilities such as SQLi or XSS get old and not as novel, then more advanced hackers can start targeting IoT devices, industrial control systems and other specialized victims.
Even though they may not have a financial payoff, these under-protected systems bring forth new challenges to stave off boredom. To compromise an industrial network or an embedded healthcare device is not the same expertise as is required for typical enterprise IT environments. Intellectual curiosity and technical skills are being targeted on operational technology and critical infrastructure feeds.
Of course, these attacks also carry graver damage potential, given the lack of monitoring and oversight. While the hacker's motivation may just be alleviating boredom, the implications spotlight the risks of under-secured networks.
Revenge
Revenge represents another common personal motivation behind cyberattacks. Disgruntled employees or angry spouses may seek to destroy data, leak documents, and disrupt operations at a specific organisation or individual that wronged them.
In one high-profile example, a Saudi Arabian hacker named OxOmar compromised over 15,000 Israeli credit card details before leaking them online. He posted, "Free Saudi's credit cards!". This attack followed similar data dumps from hacktivist groups like Anonymous.
In other cases, former employees turn to hacking tools for retaliation after being fired. Attackers with intimate knowledge of internal networks and systems can cause disproportionate damage through targeted sabotage. Even simple actions like deleting records, misconfiguring servers or wiping workstations demonstrate the security risks of insider threats.
Romantic partners (both former and current) also hack one another more frequently through spyware, location tracking and device monitoring. Physical abuse often extends to digital spheres to control and monitor victims during relationships or even after breakups.
While cyberattacks from nation states and cybercriminals dominate headlines, hacking tied to personal vendettas can be just as devastating. Understanding these motivations alongside safeguards against insider threats strengthens resilience.
Mental Illness and Disorders
Mental health disorders represent another potential factor behind malicious cyber activities. However, it is critical not to overgeneralise or make armchair diagnoses that further stigmatise conditions. Most individuals with mental illnesses are not hackers, and most hackers likely do not have these disorders.
Nonetheless, obsessive personality traits and neurodiverse conditions that manifest in social disorders, impulsivity or addictive behaviors can motivate certain attacks. Hacking may act as an outlet for aggression, obsession and lack of empathy in some cases.
Several infamous cybercriminals have shown potential symptoms of Asperger’s syndrome, narcissistic personality disorder or obsessive-compulsive disorder. Figures like Gary McKinnon and Michael Calce exhibited obsessive traits around technology and hacking from a young age.
Again, this motivation captures only a tiny subset of attackers. Speculating on mental health conditions among hackers should not reinforce inaccurate stereotypes. Additionally, these disorders typically interact with other motivations like curiosity rather than directly causing criminal behaviour alone. However, understanding how mental illnesses may remove inhibitions provides some insight into the mindsets of especially aggressive attackers.
Ethical Implications and Deterrence
Analyzing hacker motivations has ethical implications too. Curiosity-driven students may deserve school sanctions rather than criminal charges. Insider threats from former employees demand responses that are similar to traditional cybercrime.
The compulsive, addictive behaviors implicated in certain mental illnesses also raise issues of agency and consent. Incarceration may be less appropriate than rehabilitation programmes in these complex cases when hacking manifests from disorders rather than malicious intentions.
Incorporating motivational psychology into cybersecurity policies can shape better deterrence, though. For example, emphasising legal penalties may deter financially driven crime but have little impact on ideological hacktivists. Public awareness campaigns regarding ethics and consequences might deter students from hacking for curiosity but not profit-driven cybercriminals.
Understanding the motives behind attacks can inform security controls as well. Strong access management policies mitigate insider threats while anonymising payment systems hamper financially motivated hacking. Implementing controls to match likely adversary motivations boosts efficiency.
Research Limitations
However, there is a lack of even datasets and reliance on self-reporting in current research on hacker motivations. Most empirical studies have mainly relied on rather small samples of students in academic computing programs that are more likely to emphasise curiosity than criminality.
Moreover, most of the psychological assessments of cybercriminals are based on prosecuted cases. Hackers who find themselves in court documents may have very different traits than those who get away with it. Selection bias skews many empirical findings.
Another is survey data and interviews with anonymous hackers. Nevertheless, these studies again rely on honest self-reporting by unreliable narrators.
In reality, the motivations of an individual are likely to be numerous, many overlapping. Hackers’ mindsets are driven by curiosity, profit, ideology and other factors, some of which prevail over other factors, situationally. Granular differentiation between different hacker typologies would be possible to a greater extent if better-standardised assessments were available.
Conclusion
The motivations of hackers cover a wide spectrum from financial gain to ideology, curiosity, ego and others. Technical defenses attempt to block vectors, and combined with these vast psychological profiles, they strengthen the ability to prevent and respond.
There are very different incentives and goals for cybercriminals compared with state-sponsored groups, insiders, hacktivists, or script kiddies. The security teams can get a granular analysis of behavioral motivations and be able to implement targeted controls that reflect possible threats. Additionally, calibrated deterrence policies for compulsion or rational choice are made possible.
Of course, even with a robust behavioral profile of hacker psychology, the human element at the centre of these attacks makes prediction very difficult. The desire for money, change, mischief and mayhem is as enduring as technology changes. Knowing these motivators can help organisations keep up with developing tools and tactics on the cybercriminal underground, even as they change.
Specifically, they found hidden proprietary Bluetooth HCI (Host Controller Interface) commands used to read & write controller memory, and typically used for debugging. However, they could also facilitate supply chain attacks, the concealment of backdoors in the chipset, or the execution of more sophisticated attacks. Tarlogic initially called it a “backdoor”, but some disputed the claim (more on that later), and the company eventually issued an update downgrading it to a “hidden” feature:
We would like to clarify that it is more appropriate to refer to the presence of proprietary HCI commands—which allow operations such as reading and modifying memory in the ESP32 controller—as a “hidden feature” rather than a “backdoor.”
The use of these commands could facilitate supply chain attacks, the concealment of backdoors in the chipset, or the execution of more sophisticated attacks. Over the coming weeks, we will publish further technical details on this matter.
According to the researchers, bad actors could potentially infect not only the ESP32 chips themselves, but devices that connect to them through Bluetooth such as smartphones or even medical devices:
Exploitation of this hidden functionality would allow hostile actors to conduct impersonation attacks and permanently infect sensitive devices such as mobile phones, computers, smart locks or medical equipment by bypassing code audit controls.
That looks scary, especially since over one billion ESP32 devices are in the wild. Let’s have a closer look.
Tools used for discovery included a LibUSB-based Bluetooth device driver and Scapy sockets developed by Antonio Vazquez from Tarlogic, and ROM ELF documentation from Espressif. These allowed them to enable raw access to Bluetooth traffic, and after reverse-engineering work, they eventually discovered 29 undocumented HCI commands in the ESP32 Bluetooth firmware. Those commands can be used to read and write RAM and Flash, MAC address spoofing, and LMP/LLCP packet injection.
Demo code from TarlogicList of hidden ESP32 Vendor-specific HCI commands
The ESP32 Bluetooth security vulnerability has its own CVE (CVE-2025-27840) with a medium severity score of 6.8 points. Targlogic says they will provide more technical details later, and still need specific hardware that will allow them to implement advanced attacks.
What the researchers highlight (vendor-specific HCI commands to read & write controller memory) is a common design pattern found in other Bluetooth chips from other vendors as well, such as Broadcom, Cypress, and Texas Instruments. Vendor-specific commands in Bluetooth effectively constitute a “private API”, and a company’s choice to not publicly document their private API does not constitute a “backdoor”.
Backdoor claims aside, they also tried to assess whether this feature constitutes a security vulnerability. The short answer is that it depends. The longer answer explains that it’s not a vulnerability for customers who do not make a distinction between the privileges of the Host and the Controller. However, it would be for a customer who does not want a compromised userspace on the Host to automatically guarantee a compromised Bluetooth Controller firmware. In any case, they consider use of VSCs granting the capability to read and write memory, flash, or registers is a bad security design, but it impacts all Bluetooth vendors.
Australian startup Cortical Labs has launched what it's calling the "world’s first code deployable biological computer."
The shoe box-sized device, dubbed CL1, is a notable departure from a conventional computer, and uses human brain cells to run fluid neural networks.
In 2022, Cortical Labs made a big splash after teaching human brain cells in a petri dish how to play the video game "Pong."
The CL1, however, is a fundamentally different approach, as New Atlas reports. It makes use of hundreds of thousands of tiny neurons, roughly the size of an ant brain each, which are cultivated inside a "nutrient rich solution" and spread out across a silicon chip, according to the company's website.
Through a combination of "hard silicon and soft tissue," the company claims that owners can "deploy code directly to the real neurons" to "solve today's most difficult challenges."
"A simple way to describe it would be like a body in a box, but it has filtration for waves, it has where the media is stored, it has pumps to keep everything circulating, gas mixing, and of course temperature control," Cortical Labs chief science officer Brett Kagan told New Atlas late last year.
Whether it will actually prove useful remains to be seen, but Kagan is excited for scientists to get their hands on the tech.
"There's so many different options," he told Australian broadcaster ABC News, suggesting it could be used for "disease modelling, or drug testing."
"The large majority of drugs for neurological and psychiatric diseases that enter clinical trial testing fail, because there’s so much more nuance when it comes to the brain — but you can actually see that nuance when you test with these tools," Kagan told New Atlas. "Our hope is that we’re able to replace significant areas of animal testing with this."
For now, the company is selling the device as a way to train "biological AI," meaning neural networks that rely on actual neurons. In other words, the neurons can be "taught" via the silicon chip.
"The only thing that has 'generalized intelligence'... are biological brains," Kagan told ABC. "What humans, mice, cats and birds can do [that AI can't] is infer from very small amounts of data and then make complex decisions."
But the CL1 isn't about to disrupt the entire AI field overnight.
"We're not here to try and replace the things that the current AI methods do well," Kagan added.
Nonetheless, the approach could have some key advantages. For instance, the neurons only use a few watts of power, compared to infamously power-hungry AI chips that require orders of magnitude more than that.
Apart from selling the CL1, Cortical Labs is also looking to sell compute via the cloud, using its own assembled racks of the unusual computers.
In short, while it sounds like an exciting new take on conventional computers, Cortical Labs still has a lot to prove, especially when it comes to teaching neurons not unlike an AI.
"I know where it's coming from, because it is clear that these human neuronal networks learn remarkably fast," University of Queensland biologist and stem cell research specialist Ernst Wolvetang told ABC.
"At this stage I would like to reserve my judgement, because, learning Pong is one thing, but making complex decisions is another," he added.
The latest evidence suggests that Mars used to be a wet world covered in oceans, with astronomers uncovering not just icy remnants of this period, but signs of entire reservoirs of liquid water still lurking beneath the planet's arid surface.
But we're still far from having a complete picture of what the Martian climate looked like billions of years ago, before these oceans dried up.
Now, however, the chance discovery of pale yet unremarkable looking rocks by NASA's Perseverance rover suggest that the Red Planet was not only wet, but much warmer than once believed, according to a new study published in the journal Communications Earth & Environment.
"These rocks are very different from anything we've seen on Mars before," coauthor Roger Wiens, professor of Earth, atmospheric, and planetary sciences at Purdue University, said in a statement. "They're enigmas."
It's a discovery long in the making. The rocks, in the form of pebbles, were actually spotted the day that the Perseverance rover landed on the planet four years ago — but scientists had so much on their plate at the time that the objects went overlooked.
But these pale oddities just kept turning up — over 4,000 of them, in fact, the researchers said. And on a planet as unremittingly monochromatic as Mars, any deviation from the color palette could be significant. Thus, later on, when the team found larger versions of the ashen stones strewn above the bedrock in the Jezero crater, they decided to finally take a closer look.
To investigate, the team used the laser equipped on Perseverance's SuperCam instrument, the state-of-the-art camera that forms the WALL-Elooking head of the rover.
What they found was that these loose "float rocks" — so named because they come from somewhere else and not from the local bedrock — were composed of a suspiciously high amount of aluminum associated with a mineral called kaolinite. And here's the kicker: kaolinite, along with other uncovered minerals like spinel, typically form in the kinds of warm and wet environments that microbial lifeforms thrive in.
"On Earth, these minerals form where there is intense rainfall and a warm climate or in hydrothermal systems such as hot springs. Both environments are ideal conditions for life as we know it," Wiens explained. "These minerals are what's left behind when rock has been in flowing water for eons. Over time, the warm water leaches away all the elements except those that are really insoluble, leaving behind what we found on Mars."
"It's fascinating," Wiens added. "It's unexpected on a cold, dry planet like Mars." Wiens suggests that the kaolinite discovery means that "a lot of the water is still there, on Mars, bound up in the minerals."
To follow up, the team is trying to determine where the rocks came from in order to study them "in place," which would allow them to test how they formed.
SID Display Week 2025 is coming to San Jose, CA, May 11–16. Last year’s event was a whirlwind of groundbreaking innovations. While I managed to share insights on Jade Bird Display’s MicroLED Compensation, I’ve been sitting on information and photos on many other companies from Display Week 2024. With this year’s show right around the corner, it’s time to dust off those notes and many photos and write about the display and optics technology I found at last year’s Display Week, which was so interesting that I want to go back again this year.
For this article, I’m going to cover the LCOS-related companies I met with at Display Week. I plan on writing three or four articles to cover everything I saw on MicroLEDs, OLEDs, and Optics from Display Week 2024. I learned about a lot of AR/MR-related developments at Display Week 2024 and why I am going to it again this year. As I discussed in SPIE AR/VR/MR 2025 Next Week (with comments on CES, Display Week, & AWE), each conference tends to cover different aspects of displays and optics for AR/VR/MR. I will be mixing in some information from these other conferences that pertain to technology, as shown at Display Week 2024.
Partnering with SID and Discount Code
I’ve partnered with SID to share my insights from Display Week — past, present, and future. If you’re planning to attend Display Week, SID has provided the code DW25KARL for a free exhibit hall pass — don’t miss the chance to explore the latest innovations shaping the future of displays.
LCOS is still the display of choice for full-color waveguide AR glasses designs
While (mostly green only) MicroLEDs seem to garner most of the attention these days, they are still expensive and relatively low resolution, and the roadmap to full-color still has some uncertainties. OLEDs, for physics reasons, don’t work with waveguides.
LCOS is still the display technology of choice for full-color, higher-resolution, and larger FoV (>30 °) waveguide AR glasses. The key reasons include:
Cost (relative to MIcroLED and DLP) and availability
Resolution and variety of resolutions and form factors
Optical design experience leading to size reductions
Full color for a little more than monochrome
Color is still a big problem for MicroLEDs
Brightness/efficiency with waveguides (due to relatively low étendue)
Much lower étendue than MicroLEDs
Field sequential color (FSC) results in a smaller pixel and thus a smaller device for better étendue. FSC can have color breakup due to eye movement.
LCOS’s étendue advantages (over MicroLED)
Something often overlooked when comparing LCOS to MicroLEDs is the issue of étendue. Most of the LED (large or small) output Lambertian (somewhat diffuse) light (for more on étendue, Lambertian, and related topics, see Collimation, Étendue, Nits (Background for Understanding Brightness). Waveguides can only accept highly collimated light, and their entrance pupils are relatively small. With MicroLEDs, the difference between the emitter size and pixel pitch allows for some collimation with microlens arrays (MLAs). However, with LCOS, the area of the illuminating LEDs sets the étendue limit. The area of the LCOS’s illuminating LEDs is much smaller than the area of the MicroLED displays. This results in LCOS with LED illumination coupling much more efficiently into waveguides.
MicroLED’s big efficiency advantage over LCOS is that MicroLED’s power consumption is roughly proportional to the Average Pixel Value (APV, also known as Average Pixel Lit = APL) of the whole image. With most LCOS designs, the whole display is illuminated regardless of the AVP. In many, if not most, AR applications, the AVP is likely to be less than 10%; otherwise, the display would block out the real world. However, there is a design dilemma of what to do if it is possible to display a high AVP image. A MicroLED with an AVP of 100% can consume and thus have to dissipate several times the power of an LCOS design for the same brightness. One approach, as used on many larger OLED displays, would be to limit the overall brightness based on content.
The chart below was created by Bernard Kress (of SPIE and Google and formerly a technical leader on Microsoft Hololens 1 &2). The charts show different types of display content and their typical APV/APL for different types of glasses/content. On the other axis, power usage is displayed. The chart gives a rough idea of the concept of power consumption versus content (APV/APL) at a high level.
A key point on the chart is where MicroLED (red line) and ordinary LCOS (cyan/light-blue line) cross at about 12% APL. For small APL, the advantage of MicroLEDs to turn off most of the pixels “wins,” but above ~12%, LCOS, with its étendue advantage, wins. Kress shows “local dimming,” which uses arrays of mini-LEDs to illuminate LCOS (green line), significantly moving the crossover point. How much local dimming helps is a function of a lot of factors, including the size of the mini-LEDs, the number of LEDs that are arrayed, and the location of the content. As Kress’s chart shows, with miniLED local dimming LCOS, the crossover moves to about ~3%.
Meta’ Zonal Illumination (= Local Dimming) Non-Emissive Displays for AR Glass of LCOS
Between Display Week 2024 and AR/VR/MR 2025, Meta has presented papers on LCOS, MicroLEDs, and Laser Beam Scanning (LBS) for use in AR. In other words, they cover everything, or as I have said on many occasions, “In Mixed Reality, if you can dream of it, Meta has tried it.” After all, Meta is spending about $1B per month on MR.
Fenglin Peng of Meta’s Reality Labs presented Zonal Illumination Non-Emissive Displays for AR Glass. They discussed what Meta calls “Zonal Illumination,” which is local dimming. They show 12 x 12 (144) dimming zones. What was shown was an R&D proof of concept. I suspect the 12 x 12 array of mini-LEDs is too large to be what is known as being “étendue-matched” to the entrance pupil of the waveguide. If the array is too large, then due to étendue, part of the light will not couple in and will be lost.
Avegant Spotlight (Local Dimming)
Avegant showed a more modest but aimed at a real product (compared to Meta’s research study) 3 x 3 segmented illumination at SPIE AR/VR/MR in January 2024. The diagram below combines several of Avegant’s concepts, including segmented diming and a “reflective waveguide” to eliminate the large PBS and LED illumination for a small form factor. Avegant said that the LED array is small enough to be étendue-match. However, the small number of segments will mean that it will only work well if the content is not spread over the whole display area.
Ultra High Brightness Color Sequential Front-lit LCOS by Himax (at Display Week 2024)
Ultra High Brightness Color Sequential Front-lit LCOS by Himax Yuet-Wing LI from Himax presented Himax’s front-lit LCOS to reduce the size of the LCOS projector engine. Their design uses a “polarized waveguide,” which is likely different from the “reflective waveguide” used by Avegant in their Spotlight design (discussed above). You should also note that the Himax illumination waveguide is against the panel, whereas in the Avegant design, the light passes through the projection optics to illuminate the panel.
FocaLight small LCOS engine (at CES and AR/VR/MR 2025)
On the subject of small LCOS engines, at both CES and AR/VR/MR 2025, a new company, Focal Light, showed an LCOS projector engine that they say is only 0.7cc (right). I don’t know how they achieved this size, but it is about half the size of the engines shown in the Himax presentation and Avegant’s engines.
Citizen Fine Device Myota Development Center – FLCOS and other LCOS Foundary (Display Week 2024)
Citizen, most famous for watches, has many different divisions/groups. Citizen Fine Devices (CFD) started developing FLCOS (ferroelectric-LCOS, also known as fast-LCOS) device manufacturing for Displaytech. This was sold to Micron, which, in turn, sold it to Citizen. So now Citizen sells the FLCOS devices that it manufactures. Citizen also provides foundry services to make Twisted Nematic (Tn) and Vertically Aligned Nematic (Van) liquid crystal LCOS designs for other companies.
CFD showed many different devices and applications for its FLCOS technology, which is used for both display and electronic shutters (lower left). CFD also makes Quarter Waveplates and high-speed shutters that could be useful components for use in Mixed Reality.
FLCOS has the advantage of being about 10X faster than Tn with similar cell geometries. The big downside of FLCOS is that when it is “DC-Balanced,” it produces a “negative image,” and thus, the illumination must be turned off about ½ the time. Tn and Van LCOS, when DC-balanced, produce a positive image, so the light does not have to be turned off while balancing. The DC balancing problem was a big problem for very bright projectors, but it is less of an issue for LED-illuminated small projectors in AR, where the LEDs can be driven harder for a shorter time period.
CFD claims FLCOS has an advantage over Tn and Van in terms of “cross-talk” (right), which I think is their term for “lateral fields.” Lateral fields occur when light and dark pixels are next to each other, which causes electric field lines to go between the adjacent pixels rather than between the pixel mirror and the ITO coating of the top glass. These lateral fields can cause adjacent pixels to bleed together, particularly when the values are very different.
Creal Using FLCOS (AR/VR/MR 2025)
Creal, which is developing a Light Field headset, is leveraging the switching speed of FLCOS (likely from CFD) to produce time-sequential light fields for its headset device. While Creal has not reduced the headset to its final form factor, it has been showing continuous progress in developing its technology. The photographs below are from their private room at AR/VR/MR 2025.
RaonTech LCOS (plus some OLED and MicroLED development) at Display Week 2024
Raontech’s main business is LCOS, and I see many companies using their panels. At SID Display Week 2024, they showed AR glasses made by Singularity, Geding, and Lumus using their LCOS devices.
The picture below shows their many LCOS panels, plus some work they have jointly developed with Micro-OLED and MicroLED companies, where RaonTech designed the CMOS backplane for controlling the pixels.
VitreaLab has developed a “quantum light chip” that routes laser light for illuminating either LCOS or small LCDs used in VR. Lasers output light with near zero etendue, which results in the light coupling very efficiently into waveguides or other optics with extremely low losses.
Their booth at Display Week 2024 included a demonstration of their technology using an LCOS device with a Digilens waveguide (below).
Meta’s laser routing “photonic integrated circuit,” with Zonal Illumination and using FLCOS (AR/VR/MR 2024)
In another example that proves my statement, “if you can dream of it, Meta has tried it,” Meta Labs, in a presentation at AR/VR/MR 2025, showed a similar photonic integrated circuit (PIC) for routing lasers to illuminate LCOS. At least superficially, it looks a lot like VitreaLabs PLC.
In Meta Lab’s design, the PIC illuminates the LCOS device with routed polarized laser light, which passes back through the PIC after the LCOS has modulated the light. Meta’s presentation goes on to discuss the concept of selective dimming with the PIC and the fact that FLCOS would be a good LCOS technology to use with their PIC illumination.
Conclusion
MicroLEDs get most of the attention these days in optical see-through (OST) mixed reality. However, LCOS involves a lot of “physics,” particularly when it comes to using waveguides. Companies are still innovating to make smaller and more efficient LCOS designs.
Additionally, Meta is spending over $1B/month with multiple teams of researchers. They cover all bases for displays and optics for mixed reality. At AR/VR/MR 2025, they presented papers showing MicroLEDs, Laser Beam Scanning (LBS), and LCOS/FLCOS, thus my “If you can dream it, Meta has tried it” saying.
