Hardware News

Humanoid Robot Genesis Eno Introduced, Unlike a Human The French startup Genesis AI, with former Google CEO Eric Schmidt as one of its investors, introduced the humanoid robot Eno, which doesn't quite resemble a human. Genesis AI explained the discrepancy by stating that Eno was developed "based on human capabilities" rather than appearance. The humanoid robot can be headless and legless, move on a wheeled platform, and fold like a deck chair. This is because, according to Genesis AI, "humanoid robots don't have to look like humans". In practice, the developer notes, Eno is capable of serving as a universal robot, rather than a machine designed for just one task, such as folding laundry. The resemblance to humans is retained, however, in the design of the upper limbs—they were developed in "precise accordance with the shape and functions of human hands" so that the robot can use tools and objects already created for humans. Genesis AI plans to begin production and deployment of Eno with its clients by the end of 2026—initially, the robot will be used in manufacturing, laboratories, and logistics, and later in hospitals, hotels, and the consumer sector. Development of "additional execution options" is also underway. If you value what we do, you can support our projects on Boosty. With love, NSWTL.

NASA has upgraded the quantum laboratory on the ISS — experiments with cold atoms will reach a new level After the upgrade, NASA resumed operations of the Cold Atom Lab quantum laboratory aboard the International Space Station. It is the only orbital facility in the world for experiments with ultracold atoms. In the microgravity conditions of the ISS, cooled atoms exhibit their quantum properties for longer, allowing scientists to delve deeper into the quantum world and our world with their help. The equipment, roughly the size of a mini-fridge, operates in an automated mode and is controlled from Earth. The new scientific module was delivered to the ISS on April 11 as part of a commercial cargo mission, and on May 8, astronaut Jessica Meir completed the installation of the new equipment, including optical fibers and associated instruments. The Cold Atom Lab cools atoms to nearly absolute zero (down to -273 °C) — a temperature range where conventional concepts of matter cease to apply. Under such conditions, atoms of rubidium or potassium, pre-evaporated by heating, form a Bose-Einstein condensate: a collective quantum state of many atoms, considered the fifth state of matter alongside solid, liquid, gas, and plasma. In this state, the atomic cloud behaves not as a set of individual particles, each with its own quantum state, but as a large quantum object, which is extremely convenient to study compared to individual atoms. Moreover, in microgravity conditions, the wave properties of particles, including the atomic cloud, manifest for longer. This is why the installation allows for the most detailed study of effects related to the wave nature of matter, ultra-precise time measurements, gravity, and motion. Technically, the experiment begins by heating metal strips of rubidium or potassium to about 400 °C to obtain an atomic gas in a vacuum chamber. Then lasers, tuned to strictly defined frequencies, slow down the atoms, removing their vibrational energy and thus cooling the cloud. After laser cooling, the gas is captured in a magnetic trap, and a series of further manipulations brings the atomic cloud almost to a standstill. Microgravity maximizes the time of its existence. On Earth, such clouds quickly fall under the influence of gravity, whereas in orbit, they can be studied longer, at lower temperatures, and with larger quantum waves. The current upgrade is the fourth since the Cold Atom Lab was delivered to the ISS in 2018. Among the main innovations are a new magnetic trap that allows changing the shape of quantum gas clouds and testing various properties of atomic systems, as well as redesigned metal sources of atomic gas. In addition to direct experiments with atoms, NASA is fundamentally testing the readiness of quantum technologies for space operations. In the future, such developments could form the basis of wave interferometers for studying fundamental physics, navigation, time synchronization, and gravitational sensing of Earth, the Moon, and other planets in the Solar System. If you value what we do, you can support our projects on Boosty. With love, NSWTL.

The US will lift the ban on importing Chinese drones, but only toy ones The US Federal Communications Commission (FCC) announced that it intends to allow the import of new models of toy Chinese drones into the country. Professional UAVs used in work will remain banned. In December, the FCC announced a ban on the import of all new models of foreign-made drones and critical components, including products from Chinese companies DJI and Autel, as they pose an unacceptable threat to US national security, according to the agency. Regarding the new FCC initiative, the commission stated that it decided to act based on the Pentagon's conclusion that there is no national security threat from "simple, low-risk toys" that lack "organic capabilities in terms of range and flight duration, payload sensors, communication, data collection, and storage" inherent in traditional drones. The commission presented a list of criteria that a device must meet to be recognized as a toy drone: its weight must not exceed 150 g, the flight range must be within line of sight at a distance of no more than 100 m, it must lack network functions or the ability to connect to a network, it must not have photo or video cameras, as well as sensors capable of data collection, and the flight time must not exceed 10 minutes. Drones already imported into the country will be able to receive critical software updates at least until the end of 2028. The issue of banning the import of Chinese equipment from a group of manufacturers is being considered separately after their new models were banned from import and sale in 2022. The FCC also proposes to prohibit American telecommunications companies from establishing connections with Chinese telecommunications companies deemed a national security threat. Such Chinese companies will lose the ability to own data processing centers in the US. If you value what we do, you can support our projects on Boosty. With love, NSWTL.

AI and Neural Interface Restore Speech and Work for Paralyzed Patient An artificial intelligence system and a brain-computer interface helped a paralyzed and speech-impaired patient with amyotrophic lateral sclerosis (ALS) regain speech and work full-time. The neurointerface equipment used by the patient is not new, but as part of the project, a group of scientists from the University of California, Davis, developed a unique machine learning method that helped translate the patient's brain activity into coherent speech with 92% accuracy. The patient, Casey Harrell, received a neurointerface implant in 2023. Researchers from the University of California, Davis, are part of a broad coalition of universities with the U.S. Department of Veterans Affairs, known as BrainGate. They work on neurobiology projects aimed at restoring speech, enabling patients to use computers, and in some cases, restoring motor functions. In Harrell's case, engineers attempted to turn experimental technology into a durable and practical solution capable of working outside the laboratory. In controlled trials, the system was able to synthesize phrases based on the patient's brain activity with 99% accuracy; outside the lab, in everyday use, he rated the system's accuracy at 92%. Traditionally, the operation of neurointerface systems required the presence of experts, but in this case, it was not necessary: in a scientific paper that was peer-reviewed back in July 2025, it is reported that over several years, the device operated for more than 3,800 hours — an average of five hours a day. "This is a life filled with dynamic activities, communication with friends, family, colleagues, and it allows me to communicate in a more natural way for me than any other technology I have encountered," admitted Casey Harrell. This was made possible by a software platform developed by engineers for controlling brain-computer interface devices, named BRAND (Brain-computer interface for Rapidly Adaptive Neural Decoding). AI algorithms convert activity in the ventral part of the precentral gyrus — the brain area responsible for facial, mouth, and jaw motor functions — into phonemes in English. Additional software algorithms match these phonemes with words, and words with sentences. The output is synthesized speech with very high accuracy, allowing Mr. Harrell to work full-time as an environmental advocate. The goal of this project is to prove that neurointerface systems are not just laboratory experiments. They compare the current state of this technological segment to the first pacemakers that appeared in the 1950s. Back then, they were connected to large batteries or even to a power outlet. Similarly, the patient is now connected to bulky computers, but in the future, the necessary equipment may become more compact. The BrainGate organization is accepting applications for participation in future research. If you value what we do, you can support our projects on Boosty. With love, NSWTL.

Servers on architectures other than x86 have captured almost half of the market Servers using x86-compatible processors from AMD and Intel currently account for just over half of all sales, according to analysts at IDC. The reason for the market restructuring is the boom in artificial intelligence. In the first quarter of 2026, revenue from servers with processors on architectures other than x86 reached $58.7 billion, representing a 107% year-over-year growth. This means that in the overall server market, the share of such equipment reached 47.9% in monetary terms. The growth in turnover of this type of server is likely driven by systems based on Nvidia AI chips, which use Arm processors. The global server market during this period reached $122.6 billion, which is 30.4% more than the previous year. Investments in AI infrastructure by major cloud providers are "developing at scales that show no signs of stabilization"; the rest is characterized by analysts as a segment without acceleration, and there is a supply shortage here, largely due to investments in AI infrastructure. In the near future, a restraining factor for the industry's development will be the shortage of components, primarily DRAM and NAND memory chips, although the order portfolio remains strong. Revenue from x86 servers amounted to $63.9 billion, indicating a decline of 2.9%. Servers with GPU-based accelerators brought manufacturers $68.9 billion, almost 25% more year-over-year; revenue from other servers with accelerators showed a growth of 122%, reaching $17.7 billion. This category represents AI systems configured with FPGA or ASIC components rather than GPU-based ones. Infrastructure development is no longer limited to hyperscaler resources — government initiatives in the field of sovereign AI have entered the game, among others. In the segment of systems without accelerators, the situation turned out to be challenging. Revenue here declined, basic demand remains high, and many corporate clients have to contend with inflated component prices. According to IDC analysts, supply normalization is expected from the beginning of 2027: as new chip manufacturing plants are commissioned, an increase in production capacity will become apparent. Over the past two decades, servers outside the x86 segment accounted for less than 10% of revenue, with a significant portion going to IBM, which remained the last supplier of proprietary equipment. Oracle lost interest in Sun, and companies like HPE decided they could not support the business associated with exotic architectures. If you value what we do, you can support our projects on Boosty. With love, NSWTL.

Japanese Scientist Solves Long-standing EUV Lithography Problem — New Optics to Significantly Reduce Advanced Chip Production Costs A scientist from the Okinawa Institute of Science and Technology (OIST) proposed a radical simplification of the optical scheme of EUV lithography — a key technology used today to produce the most advanced microchips. Professor Tsumoru Shintake developed a version of a high-NA EUV system with linear geometry, where the photomask, projection optics, and silicon wafer are aligned on a single axis, unlike current industrial scanners. According to the author's calculations, the proposed optical architecture could help print elements sized 2–3 nm and reduce equipment costs by 3–4 times, which today are estimated at hundreds of millions of euros per machine. This also promises to lower the cost of producing chips even more advanced than today, ultimately reducing expenses for supporting the most sophisticated AI models and data centers. As is known, EUV lithography uses extreme ultraviolet radiation with a wavelength of 13.5 nm. Such light cannot pass through ordinary lenses as it is simply absorbed by them, so the entire optical system is built on reflection from multilayer mirrors and operates in a vacuum. In the standard ASML lithographic scanner scheme, the beam hits a mirror photomask with the pattern of the future circuit, then the projection optics reduce and focus the image onto the silicon wafer. Increasing the numerical aperture, or NA, allows capturing a wider range of angles, which enhances resolution but simultaneously complicates the optics (see the figure below) and increases distortions. In developing the new optical architecture for EUV scanners, Professor Shintake primarily aimed to reduce the so-called "3D mask effects" — distortions arising from the three-dimensional structure of the mirror EUV mask and the angle at which light falls on it. Researchers could not solve this problem in the 1990s when they began designing EUV scanners, so the commercially viable option turned out to be an extremely complex off-axis optical scheme implemented in all modern ASML EUV scanners. The scientist proposed a linear focusing system consisting of two optical components, each containing a pair of concave and convex mirrors. Modeling showed that multiple reflections between mirrors with precisely calculated profiles and strictly defined distances between them can mutually compensate for some distortions while maintaining high numerical aperture and image quality. Unlike current complex high-NA solutions, this scheme should be significantly simpler to manufacture and adjust. However, industrial application is still far off. The calculations assume ideal mirrors — with 100 percent reflection and no defects, whereas real EUV optics lose some energy with each reflection and require extremely precise surface processing. The next step will be assembling a physical prototype: the team has already begun developing the corresponding EUV equipment. If the approach is experimentally confirmed, it could reduce the cost of producing high-density memory and logic microchips, decrease the number of technological operations, and lower the energy consumption of computations, which is especially important given the increasing load from AI and data centers. If you value what we do, you can support our projects on Boosty. With love, NSWTL.

Samsung showcased key components for next-generation XR devices Samsung Display has unveiled new microdisplays based on RGB OLEDoS (OLED on Silicon) technology, designed for use in augmented reality (AR) and mixed reality (MR) devices. The presentation took place at the Augmented World Expo 2026. A prototype of smart glasses with a miniature 0.62-inch RGB OLEDoS display is attracting particular attention in the wearable AR device segment. According to Samsung, it provides a clear image necessary for functions such as real-time translation, navigation, and weather information display. Simultaneously, a larger 1.3-inch model for mixed reality headsets and AR glasses was introduced, capable of reaching a brightness level of 40,000 cd/m2. RGB OLEDoS features a simplified panel structure compared to competing microdisplay systems and helps reduce the weight of devices. According to the manufacturer, this technology offers high energy efficiency, accurate color reproduction, and excellent brightness performance, making it an optimal choice for lightweight wearable electronics. Although Samsung Display has not confirmed a connection between the presented developments and future Samsung Electronics products, industry insiders report work on two models of smart glasses. It is expected that the device codenamed Haean will feature microLED displays, while another model, presumably called Glasses, will focus on cameras and voice control without its own screen. If you value what we do, you can support our projects on Boosty. With love, NSWTL.

Sony LYTIA L910 Introduced — The First Mobile Sensor with LOFIC Architecture Sony has introduced the LYTIA L910, a camera sensor for smartphones designed to capture high-quality images in challenging lighting conditions. The component incorporates several image processing technologies that help produce realistic photos and videos while maintaining controlled energy consumption. This is also the first model in the Sony LYTIA family to use the LOFIC architecture. The Sony LYTIA L910 is a multi-layer CMOS sensor measuring 1/1.28 inches with an effective resolution of about 50 megapixels and a pixel size of 1.22 µm. It can provide a dynamic range of up to 100 dB in a single shot. Unlike traditional HDR methods that require multiple exposures, Sony's solution captures the necessary data in one shot, reducing blur and flicker when photographing moving objects or bright light sources. To achieve this result, Sony combined the LOFIC (Lateral Overflow Integration Capacitor) structure with Triple Conversion Gain HDR technology: the sensor reads information in one exposure with three different conversion gains, preserving details in both bright and dark areas of the image. Overexposure is minimized, noise in shadow areas is reduced, and images become more balanced with smoother tonal transitions. Sony mentioned the Ultra High Conversion Gain technology, which helps improve efficiency in converting charge to voltage: compared to the LYTIA 828 sensor, the level of random noise has been reduced by about 30%, which helps enhance image quality in low-light conditions, such as night cityscapes with LED lighting. The Sony LYTIA L910 sensor features an optimized design with low power consumption during image processing. It can record 4K video at 60 frames per second with a high dynamic range. It can even shoot video at full 50-megapixel resolution at 30 frames per second; when the resolution is reduced to 12.5 megapixels, the frame rate increases to 120 per second. Mass production of the Sony LYTIA L910 will begin in the summer of 2026. The first smartphones with this sensor are likely to be models from the Vivo X500 and Oppo Find X10 series. If you value what we do, you can support our projects on Boosty. With love, NSWTL.

Scientists Determine How Much More Transistors Can Be Reduced Researchers from the Korea Advanced Institute of Science and Technology (KAIST) have introduced a method for calculating the quantum limit of transistor reduction. Beyond a certain point, the quantum tunneling effect leads to an uncontrollable increase in current leakage. The method provides a quantitative understanding of ways to counteract these processes, allowing chip manufacturers to move towards the smallest possible transistor with awareness rather than guesswork. The method is based on the widely used density functional theory in quantum physics. The scientific community successfully uses it to model the electronic structures of material molecules. With the right approach, the theory is entirely suitable for a preliminary and fairly accurate assessment of the scaling limits of future semiconductor devices. This task becomes especially important against the backdrop of the industry's transition to so-called 2-nm processes, where the marketing name of a functional element no longer matches the actual physical sizes of transistor structures, and the elements themselves are approaching quantum-mechanical limitations. The main problem of miniaturization is that at very small element sizes, electrons begin to pass through energy barriers that classical physics would have stopped them from crossing. This phenomenon is known as quantum tunneling. For a transistor, it means an increase in parasitic leakage currents and a deterioration in current control between the source and drain. Experimentally determining such limits is extremely difficult: the contact area between the metal electrode and the semiconductor channel is atomically tiny, and its geometry and electronic structure are hard to control with sufficient precision. As a model to prove their method's effectiveness, the researchers used monolayer molybdenum disulfide (MoS2) — a two-dimensional semiconductor considered one of the candidates for the base materials of next-generation transistors. For MoS2, contacts with different metals, including scandium, silver, gold, and palladium, were calculated. Calculations were conducted for two architectural variants: with a top contact and with an edge contact. Modeling showed that the critical tunneling length is not a constant value: it depends on the work function of the metal (how easily an electron leaves the metal) and the geometry of the contact structure. In other words, the miniaturization limit can be shifted by selecting the electrode material and the method of connecting the metal with the two-dimensional channel. This is good news, offering hope for further reduction in transistor sizes. According to KAIST's calculations, with the optimal choice of metal and contact structure, the critical tunneling length can be reduced to less than 4 nm — real, not marketing. For n-type transistors, a top contact scheme with metals with a low work function proved promising, while for p-type, an edge contact with metals with a high work function was favorable. This does not mean the immediate emergence of mass-produced transistors of such sizes, but it provides engineers with a new design tool: instead of costly trial and error with prototypes, they can pre-assess contact resistance, tunneling leakage mode, and the ultimate scalability of 2D transistors at the atomic level. If you value what we do, you can support our projects on Boosty. With love, NSWTL.

Wear OS 7 debuted on Google Pixel Watch smartwatches Google has begun rolling out the Wear OS 7 update for the Pixel Watch line of smartwatches, introducing it alongside the release of the Android 17 mobile operating system. The new version focuses on enhancing energy efficiency and improving interaction with other connected user devices. Developers have optimized power consumption in Wear OS 7, providing up to a 10% increase in battery life compared to the previous version. The watches now support real-time event updates through the Live Updates feature, which displays various information, including sports match scores and workout progress, as a persistent icon on the watch face and a card in the notification feed. In the second half of 2026, certain models running Wear OS 7 will receive Gemini Intelligence features. Among them is a feature called Create My Widget, which allows users to create personalized dashboards using natural language requests. There is also automation of app multitasking, enabling Gemini to perform actions directly from the watch (such as booking a workout session or ordering food). Additionally, there will be a new Neural Expressive design and a Personal Intelligence feature that analyzes data from Google apps, including Gmail, and chat history. Multimedia controls will receive an audio output switch, allowing users to change the playback device directly from their wrist, such as through headphones, a smartphone, or home speakers. If you value what we do, you can support our projects on Boosty. With love, NSWTL.