By CHRIS McGOWAN This image of Jupiter from NASA’s James Webb Space Telescope’s NIRCam (Near-Infrared Camera) shows stunning details of the majestic planet in infrared light. (Image courtesy of NASA, ESA and CSA) Special effects have been used for decades to depict space exploration, from visits to planets and moons to zero gravity and spaceships – one need only think of the landmark 2001: A Space Odyssey (1968). Since that era, visual effects have increasingly grown in realism and importance. VFX have been used for entertainment and for scientific purposes, outreach to the public and astronaut training in virtual reality. Compelling images and videos can bring data to life. NASA’s Scientific Visualization Studio (SVS) produces visualizations, animations and images to help scientists tell stories of their research and make science more approachable and engaging. A.J. Christensen is a senior visualization designer for the NASA Scientific Visualization Studio (SVS) at the Goddard Space Flight Center in Greenbelt, Maryland. There, he develops data visualization techniques and designs data-driven imagery for scientific analysis and public outreach using Hollywood visual effects tools, according to NASA. SVS visualizations feature datasets from Earth-and space-based instrumentation, scientific supercomputer models and physical statistical distributions that have been analyzed and processed by computational scientists. Christensen’s specialties include working with 3D volumetric data, using the procedural cinematic software Houdini and science topics in Heliophysics, Geophysics and Astrophysics. He previously worked at the National Center for Supercomputing Applications’ Advanced Visualization Lab where he worked on more than a dozen science documentary full-dome films as well as the IMAX films Hubble 3D and A Beautiful Planet – and he worked at DNEG on the movie Interstellar, which won the 2015 Best Visual Effects Academy Award. This global map of CO2 was created by NASA’s Scientific Visualization Studio using a model called GEOS, short for the Goddard Earth Observing System. GEOS is a high-resolution weather reanalysis model, powered by supercomputers, that is used to represent what was happening in the atmosphere. (Image courtesy of NASA/Goddard Space Flight Center Scientific Visualization Studio) “The NASA Scientific Visualization Studio operates like a small VFX studio that creates animations of scientific data that has been collected or analyzed at NASA. We are one of several groups at NASA that create imagery for public consumption, but we are also a part of the scientific research process, helping scientists understand and share their data through pictures and video.” —A.J. Christensen, Senior Visualization Designer, NASA Scientific Visualization Studio (SVS) About his work at NASA SVS, Christensen comments, “The NASA Scientific Visualization Studio operates like a small VFX studio that creates animations of scientific data that has been collected or analyzed at NASA. We are one of several groups at NASA that create imagery for public consumption, but we are also a part of the scientific research process, helping scientists understand and share their data through pictures and video. This past year we were part of NASA’s total eclipse outreach efforts, we participated in all the major earth science and astronomy conferences, we launched a public exhibition at the Smithsonian Museum of Natural History called the Earth Information Center, and we posted hundreds of new visualizations to our publicly accessible website: svs.gsfc.nasa.gov.” This is the ‘beauty shot version’ of Perpetual Ocean 2: Western Boundary Currents. The visualization starts with a rotating globe showing ocean currents. The colors used to color the flow in this version were chosen to provide a pleasing look. (Image courtesy of NASA/Goddard Space Flight Center Scientific Visualization Studio) The Gulf Stream and connected currents. (Image courtesy of NASA/Goddard Space Flight Center Scientific Visualization Studio) Venus, our nearby “sister” planet, beckons today as a compelling target for exploration that may connect the objects in our own solar system to those discovered around nearby stars. (Image courtesy of NASA’s Goddard Space Flight Center) WORKING WITH DATA While Christensen is interpreting the data from active spacecraft and making it usable in different forms, such as for science and outreach, he notes, “It’s not just spacecraft that collect data. NASA maintains or monitors instruments on Earth too – on land, in the oceans and in the air. And to be precise, there are robots wandering around Mars that are collecting data, too.” He continues, “Sometimes the data comes to our team as raw telescope imagery, sometimes we get it as a data product that a scientist has already analyzed and extracted meaning from, and sometimes various sensor data is used to drive computational models and we work with the models’ resulting output.” Jupiter’s moon Europa may have life in a vast ocean beneath its icy surface. (Image courtesy of NASA/Goddard Space Flight Center Scientific Visualization Studio) HOUDINI AND OTHER TOOLS “Data visualization means a lot of different things to different people, but many people on our team interpret it as a form of filmmaking,” Christensen says. “We are very inspired by the approach to visual storytelling that Hollywood uses, and we use tools that are standard for Hollywood VFX. Many professionals in our area – the visualization of 3D scientific data – were previously using other animation tools but have discovered that Houdini is the most capable of understanding and manipulating unusual data, so there has been major movement toward Houdini over the past decade.” Satellite imagery from NASA’s Solar Dynamics Observatory (SDO) shows the Sun in ultraviolet light colorized in light brown. Seen in ultraviolet light, the dark patches on the Sun are known as coronal holes and are regions where fast solar wind gushes out into space. (Image courtesy of NASA/Goddard Space Flight Center Scientific Visualization Studio) Christensen explains, “We have always worked with scientific software as well – sometimes there’s only one software tool in existence to interpret a particular kind of scientific data. More often than not, scientific software does not have a GUI, so we’ve had to become proficient at learning new coding environments very quickly. IDL and Python are the generic data manipulation environments we use when something is too complicated or oversized for Houdini, but there are lots of alternatives out there. Typically, we use these tools to get the data into a format that Houdini can interpret, and then we use Houdini to do our shading, lighting and camera design, and seamlessly blend different datasets together.” While cruising around Saturn in early October 2004, Cassini captured a series of images that have been composed into this large global natural color view of Saturn and its rings. This grand mosaic consists of 126 images acquired in a tile-like fashion, covering one end of Saturn’s rings to the other and the entire planet in between. (Image courtesy of ASA/JPL/Space Science Institute) The black hole Gargantua and the surrounding accretion disc from the 2014 movie Interstellar. (Image courtesy of DNEG and Paramount Pictures) Another visualization of the black hole Gargantua. (Image courtesy of DNEG and Paramount Pictures) INTERSTELLAR & GARGANTUA Christensen recalls working for DNEG on Interstellar (2014). “When I first started at DNEG, they asked me to work on the giant waves on Miller’s ocean planet [in the film]. About a week in, my manager took me into the hall and said, ‘I was looking at your reel and saw all this astronomy stuff. We’re working on another sequence with an accretion disk around a black hole that I’m wondering if we should put you on.’ And I said, ‘Oh yeah, I’ve done lots of accretion disks.’ So, for the rest of my time on the show, I was working on the black hole team.” He adds, “There are a lot of people in my community that would be hesitant to label any big-budget movie sequence as a scientific visualization. The typical assumption is that for a Hollywood movie, no one cares about accuracy as long as it looks good. Guardians of the Galaxy makes it seem like space is positively littered with nebulae, and Star Wars makes it seem like asteroids travel in herds. But the black hole Gargantua in Interstellar is a good case for being called a visualization. The imagery you see in the movie is the direct result of a collaboration with an expert scientist, Dr. Kip Thorne, working with the DNEG research team using the actual Einstein equations that describe the gravity around a black hole.” Thorne is a Nobel Prize-winning theoretical physicist who taught at Caltech for many years. He has reached wide audiences with his books and presentations on black holes, time travel and wormholes on PBS and BBC shows. Christensen comments, “You can make the argument that some of the complexity around what a black hole actually looks like was discarded for the film, and they admit as much in the research paper that was published after the movie came out. But our team at NASA does that same thing. There is no such thing as an objectively ‘true’ scientific image – you always have to make aesthetic decisions around whether the image tells the science story, and often it makes more sense to omit information to clarify what’s important. Ultimately, Gargantua taught a whole lot of people something new about science, and that’s what a good scientific visualization aims to do.” The SVS produces an annual visualization of the Moon’s phase and libration comprising 8,760 hourly renderings of its precise size, orientation and illumination. (Image courtesy of NASA/Goddard Space Flight Center Scientific Visualization Studio) FURTHER CHALLENGES The sheer size of the data often encountered by Christensen and his peers is a challenge. “I’m currently working with a dataset that is 400GB per timestep. It’s so big that I don’t even want to move it from one file server to another. So, then I have to make decisions about which data attributes to keep and which to discard, whether there’s a region of the data that I can cull or downsample, and I have to experiment with data compression schemes that might require me to entirely re-design the pipeline I’m using for Houdini. Of course, if I get rid of too much information, it becomes very resource-intensive to recompute everything, but if I don’t get rid of enough, then my design process becomes agonizingly slow.” SVS also works closely with its NASA partner groups Conceptual Image Lab (CIL) and Goddard Media Studios (GMS) to publish a diverse array of content. Conceptual Image Lab focuses more on the artistic side of things – producing high-fidelity renders using film animation and visual design techniques, according to NASA. Where the SVS primarily focuses on making data-based visualizations, CIL puts more emphasis on conceptual visualizations – producing animations featuring NASA spacecraft, planetary observations and simulations, according to NASA. Goddard Media Studios, on the other hand, is more focused towards public outreach – producing interviews, TV programs and documentaries. GMS continues to be the main producers behind NASA TV, and as such, much of their content is aimed towards the general public. An impact crater on the moon. (Image courtesy of NASA/Goddard Space Flight Center Scientific Visualization Studio) Image of Mars showing a partly shadowed Olympus Mons toward the upper left of the image. (Image courtesy of NASA/Goddard Space Flight Center Scientific Visualization Studio) Mars. Hellas Basin can be seen in the lower right portion of the image. (Image courtesy of NASA/Goddard Space Flight Center Scientific Visualization Studio) Mars slightly tilted to show the Martian North Pole. (Image courtesy of NASA/Goddard Space Flight Center Scientific Visualization Studio) Christensen notes, “One of the more unique challenges in this field is one of bringing people from very different backgrounds to agree on a common outcome. I work on teams with scientists, communicators and technologists, and we all have different communities we’re trying to satisfy. For instance, communicators are generally trying to simplify animations so their learning goal is clear, but scientists will insist that we add text and annotations on top of the video to eliminate ambiguity and avoid misinterpretations. Often, the technologist will have to say we can’t zoom in or look at the data in a certain way because it will show the data boundaries or data resolution limits. Every shot is a negotiation, but in trying to compromise, we often push the boundaries of what has been done before, which is exciting.”

There's a new GPU champion in town.

Quantum computing enters new territory with multimode encoding, shrinking systems and slashing power demands.

ZDNET's key takeaways The OnePlus 13 is a snappy, nearly no-compromise phone that starts at $899. A Snapdragon 8 Elite, paired with a 6,000mAh battery and 80W fast charging, is a recipe for endurance success. IP69 is almost excessive, but you'll appreciate it when least expected. $999.99 at Best Buy apr / 2025Over at OnePlus' website, both OnePlus 13 models are on sale for $50 off, and each purchase comes with a free gift. Options include a OnePlus Nord Buds 3 Pro and a Sandstone Magnetic Case.It's not often that I review a smartphone in the first few calendar weeks and feel confident in calling it a "Phone of the Year" contender. But when I tested the OnePlus 13 back in January, that's precisely what happened.Whether Google finally launches a Pixel Pro Fold with a flagship camera system this summer, or Apple releases a thinner iPhone in the fall, the OnePlus 13 will likely still be on my mind when the year-end nominations are due.Also: I changed 10 OnePlus phone settings to significantly improve the user experienceThere's a lot going for the latest flagship phone, from the more secure (and reliable) ultrasonic fingerprint sensor to the IP69 rating to the 6,000mAh Silicon NanoStack battery. It's also one of the first phones in North America to feature Qualcomm's new Snapdragon 8 Elite chip, which promises improvements to performance, efficiency, and AI workloads.I tested the OnePlus 13 alongside my iPhone 16 Pro Max and Google Pixel 9 Pro XL to see exactly how the Android phone stacked up against one of the best phones from 2024. In a few ways, the OnePlus 13 falls short, but in many ways, it puts the iPhone and Pixel to shame.When I first unboxed the OnePlus 13 and held it in my hand, my reaction was audible. Allow me to geek out here: The slightly curved glass, the slimness of the phone, and the overall appearance made my then-four-month-old iPhone look and feel outdated. It's as if OnePlus made the iPhone 17 Air before Apple did.However, what sells the OnePlus 13 design for me is the new Midnight Ocean color, which flaunts a vegan-leather backing that makes the phone visually distinctive and more comfortable to hold than its glass-only predecessors. The texture isn't as rough and grippy as actual leather, though, so I'd be interested in seeing how it ages over the year. (April update: The textured backing is holding up well, save for a few dark spots on the corners, likely caused by the phone rubbing against my palms.) Kerry Wan/ZDNETIf you were hoping the first major Android phone of 2025 would feature Qi2 wireless charging, I have good news and bad news. While the OnePlus 13 doesn't have an in-body Qi2 charging coil, meaning MagSafe (and similar) accessories won't attach directly to the back of the device, OnePlus has embedded magnetic guides within its protective covers, enabling users to take advantage of the accessories so long as the OnePlus 13 is encased. It's a burdenless workaround, but one that hopefully won't be necessary with the next model.For what it's worth, since publishing this review, several other Android phones have been released, including the Samsung Galaxy S25 Ultra, Nothing Phone 3a Pro, and Motorola Razr Ultra -- none of which feature Qi2 wireless charging.For years, one aspect that's held OnePlus phones back is the water and dust resistance rating, or lack thereof. With the OnePlus 13, the company is finally taking a stronger stance on the endurance standard, certifying the phone with an IP69 rating. It's a step above the IP68 ratings we commonly see on competing devices, and allows the OnePlus 13 to withstand high-pressure, high-temperature water jets and humidity changes.Also: 5 habit trackers on Android that can reveal your patterns - and motivate you to changeIn practice, this means the OnePlus 13 can function properly even if you leave it in your washer and dryer, dishwasher, or a pot of boiling soup. The IP69 rating feels very much like a flex, but it's a benefit that users will appreciate when they least expect it. Kerry Wan/ZDNETPowering the device is a Qualcomm Snapdragon 8 Elite chip that, from my months of usage, has some noticeable strengths and weaknesses. For day-to-day usage, such as bouncing between productivity apps, definitely not scrolling through TikTok, and taking photos and videos, the processor handles tasks gracefully. It helps that OxygenOS 15, based on the latest version of Android, has some of the smoothest animations I've seen on a phone.Also: I found a Bluetooth tracker for Android users that functions better than AirTags (and it's cheaper)But once you fire up graphics-intensive applications like Adobe Premiere Rush and Honkai Star Rail, you'll notice some stuttering as the higher heat development leads to throttling performance. This isn't a dealbreaker, per se, as the nerfs are only apparent when you're using the device for a prolonged time.I've actually been using the OnePlus 13 quite liberally, as the 6,000mAh Silicon NanoStack battery has kept my review unit running for at least a day and a half per charge. That's unseen with any other mainstream phone in the US market, and I fully expect more manufacturers to adopt silicon batteries for their greater energy density. If not that, copy the 80W fast charging or 50W wireless charging; they're quite the revelation. Kerry Wan/ZDNETOn the camera front, the OnePlus 13, with its triple camera setup (50MP wide, ultrawide, and telephoto), has been a reliable shooter throughout most of my days. While the Sony LYT-808 sensor isn't on par with the one-inch sensors I've tested on international phones like the Xiaomi 15 Ultra, it does an excellent job of capturing details and finishing the output vividly. If you're a fan of sharp, bright, and slightly oversaturated imagery (read: more colorful than how the actual subject appears), then the OnePlus 13 will serve you well.Also: The best Android phones to buy in 2025Where the camera sensors fall short is in post-processing and AI-tuning features. For example, the phone leans heavily on computational photography to contextualize details when taking far-distance shots. This sometimes leads to images with an artificial, over-smoothing filter. But when the backend software works, it can reproduce details that you probably didn't think you'd capture in the first place.ZDNET's buying adviceFor a starting price of $899, the OnePlus 13 delivers some seriously good value -- possibly the best of all the major flagship phones I've tested so far this year. The company has improved the device in almost every way, from the design to the performance to its accessory ecosystem. I just wish OnePlus offered more extensive software support, as the OnePlus 13 will only receive four years of Android OS updates and six years of security updates. Samsung, Google, and Apple offer at least seven years of OS support. If you can shoulder the shorter promise of longevity, this is one of the easiest phones for me to recommend right now. Why the OnePlus 13 gets an Editors' Choice award We awarded the OnePlus 13 an Editors' Choice because it nails all the fundamentals of a great smartphone experience while leading the market in some regards, such as battery and charging, durability, and design. The specs this year are noticeably improved compared to its predecessor, the OnePlus 12, with a faster processor, lighter build, larger battery capacity, and a more capable camera system. Most importantly, the OnePlus 13 starts at $899, undercutting its closest competitors like the Google Pixel 9 Pro XL and Samsung Galaxy S25 Ultra. Show more When will this deal expire? As per OnePlus, this offer will end on June 8, 2025.However, deals are subject to sell out or expire at any time, though ZDNET remains committed to finding, sharing, and updating the best product deals for you to score the best savings. Our team of experts regularly checks in on the deals we share to ensure they are still live and obtainable. We're sorry if you've missed out on a deal, but don't fret -- we constantly find new chances to save and share them with you on ZDNET.com.  Show more What are the tariffs in the US? The recent US tariffs on imports from countries like China, Vietnam, and India aim to boost domestic manufacturing but are likely to drive up prices on consumer electronics. Products like smartphones, laptops, and TVs may become more expensive as companies rethink global supply chains and weigh the cost of shifting production.Smartphones are among the most affected by the new US tariffs, with devices imported from China and Vietnam facing steep duties that could raise retail prices by 20% or more. Brands like Apple and Google, which rely heavily on Asian manufacturing, may either pass these costs on to consumers or absorb them at the expense of profit margins. The tariffs could also lead to delays in product launches or shifts in where and how phones are made, forcing companies to diversify production to countries with more favorable trade conditions. Show more This story was originally published on January 7, 2025, and was updated on June 1, 2025, adding information for a new June discount.Featured reviews

Parametric design is a transformative approach to product development that integrates interconnected parameters to enhance a product’s performance and adaptability. This approach maximizes the relationships between parameters like color, size, and material by defining and adjusting them to improve design results. In contrast to conventional design systems, parametric design encourages creativity by making it possible to create adaptable and dynamic solutions that are suited to changing requirements. The use of sophisticated algorithms makes it easier to create intricate patterns and structures, changing the way engineers and designers approach architectural, industrial, and urban planning projects. Parametric design is based on systems that are naturally based on parametric principles and are inspired by biological and natural patterns. The Flip chair, inspired by the natural beauty of wind-blown grass, combines asymmetry and craftsmanship to create a unique seating experience. Designers focused on the “flip” as the central element, allowing the bent wood to flow through gradual transitions, and infusing the piece with vitality. The challenging process of steam-bending thick wood was repeated multiple times, ensuring that each chair is handmade, one-of-a-kind, and exhibits a distinctive form, making it an elegant yet functional addition to any space. Developed with the support of the National Taiwan Craft Research and Development Institute (NTCRI), the Flip chair seamlessly blends traditional craftsmanship with modern parametric design. The result is a chair that not only offers aesthetic appeal but also carries a story of innovation and meticulous workmanship. Its organic texture and thoughtful design make it a statement piece that is ideal for those seeking a functional yet artistic element to complement their home or office decor. What are the Applications of Parametric Design? Parametric design enables the creation of customized product designs by allowing users to adjust shape, size, and functionality based on specific inputs. This approach combines various algorithms and parameters to create and refine designs that enhance the functionality and aesthetics of the product. Designers begin by defining key parameters, which can then be adjusted to optimize efficiency and performance. CAD software and similar tools are commonly used in this process and by utilizing parameters, the design process becomes more controlled and enables innovative solutions. Multiple design options can be explored quickly which heps in reducing time and waste, and making parametric design a powerful tool for innovation. Located in Jakarta’s business district, the Stalk Tree-Hugger, designed by RAD+ar, integrates five existing tall trees with a parametric fabric structure that creates an engaging play of light and shadows. This innovative design combines a restaurant and lounge, offering a unique spatial experience while blending seamlessly with the surrounding greenery. The shifting shadows from the trees foster an intimate, inviting atmosphere, while the dynamic tensile structures, swaying with the wind, create a nature-centric ambiance. The 750-square-meter bar restaurant showcases minimal intervention in nature, providing a versatile space for commercial activities. The lightweight steel-timber thatch roof offers both shade and support to the parametric fabric, transforming into a lantern at night, enhancing the cityscape and event ambiance. Designed by Antonius Richard and the RAD+ar team, the project emphasizes reducing environmental impact while maintaining the commercial value of the land. With constant movement and changing lighting, the design creates a relaxing, tranquil space, where visitors can observe and interact with nature. The Generico Chair, created by Marco Hemmerling and Ulrich Nether, exemplifies the power of generative algorithms in computational or parametric design. This innovative process allows the software to optimize the design to meet specific parameters and results in a lightweight yet strong structure. The chair’s Voronoi-inspired design reduces material usage significantly while maintaining strength and flexibility in the backrest. The careful application of generative design principles ensures the chair’s efficiency without compromising its functionality. Designed for both performance and comfort, the Generico Chair retains its ergonomic qualities despite its reduced volume. The chair is 3D printed to accommodate the complexities of its generative design, which introduces specific manufacturing constraints. The result is a visually captivating piece with a skeletal charm that blends creativity and precision. Its unique form showcases how software-aided design can produce functional and aesthetically pleasing furniture pieces while adhering to sustainability, precision, and material efficiency goals. How does Parametric design bring innovation to product design? Parametric design, powered by CAD software, achieves a harmonious balance between complexity and simplicity, streamlining the processes of product creation and modification. By defining specific parameters, designers can make adjustments with ease, minimizing waste and improving efficiency. This highly versatile approach adapts to various products, such as adjusting the height and tilt of a chair or modifying the color, angle, and brightness of lighting design. Therefore, parametric design promotes flexibility, providing a wide range of solutions to complex challenges. By establishing parameters at the outset, significant time is saved, enabling refinements during the design phase before creating a physical prototype. It excels in managing intricate shapes and patterns while opening new possibilities for innovation and creativity. John Mauriello, the designer behind the Coral Lighting Collection, articulates his creations as moments frozen in time. These lamps are the result of complex algorithms simulating natural growth, each one capturing a unique form by pausing the simulation at a precise point. The collection features three distinct designs namely Timor, Sargasso, and Celebes, each inspired by different types of coral. Mauriello’s deep connection to the ocean, fostered through his experience as a surfer, is reflected in his tribute to coral’s inherent beauty and vibrant life. The lamps are meticulously designed to be visually appealing both when illuminated and when turned off. Crafted from a white, ceramic-like material, they are 3D printed in the USA using sustainable methods that enable the recycling of any waste materials. The integrated LED lights interact with the lamp’s uneven cross-sections to produce dynamic lighting effects of brightness and shadow. The manufacturing process is eco-conscious and allows any waste material that is produced during the design process to be recycled. A Guide to Parametric Design Software Parametric modeling is a computer-aided design (CAD) tool used to create and manipulate parametric models. Several software programs incorporate advanced algorithms for parametric design, combining computer software with AI to offer innovative design solutions. These sophisticated algorithms enable an interactive design process, allowing designers and engineers to input specific constraints directly within the software. This approach minimizes errors and ensures the creation of accurate digital models. Furthermore, the integration of virtual reality enhances the experience, providing immersive and interactive opportunities to visualize and refine concepts. Designers can explore complex geometries and design intricacies, enabling the creation of 3D shapes and models with greater precision and flexibility. CATIA’s Visual Scripting app, part of the 3DEXPERIENCE Platform, makes parametric design easier and more accessible. This no-code tool lets designers, engineers, and architects create complex designs without needing to write code. Using a simple, node-based interface, users can define parameters and generate detailed 3D models effortlessly. The tool’s Capture & Reuse feature boosts efficiency by allowing users to share and reuse design algorithms across different projects. By combining CATIA’s CAD software with a visual approach to parametric and generative modeling, Visual Scripting enhances creativity and speeds up design iterations. Its smooth integration with the 3DEXPERIENCE platform provides a streamlined workflow, making algorithmic design a built-in feature. Beyond traditional modeling, it also enables mesh manipulation and the creation of custom lattice structures—perfect for 3D printing in industries like aerospace, furniture, and jewelry design. Parametric design is transforming the product design industry by fostering creativity, improving efficiency, and enabling customization. Its seamless integration with digital manufacturing and sustainability efforts makes it an essential tool for modern designers. As technology continues to evolve, parametric design will remain at the forefront, driving ethical and innovative advancements in product development.The post Parametric Design And a Guide to Creating Adaptable, High-Performance Products with Algorithms first appeared on Yanko Design.

Huawei’s AI capabilities have made a breakthrough in the form of the company’s Supernode 384 architecture, marking an important moment in the global processor wars amid US-China tech tensions.The Chinese tech giant’s latest innovation emerged from last Friday’s Kunpeng Ascend Developer Conference in Shenzhen, where company executives demonstrated how the computing framework challenges Nvidia’s long-standing market dominance directly, as the company continues to operate under severe US-led trade restrictions.Architectural innovation born from necessityZhang Dixuan, president of Huawei’s Ascend computing business, articulated the fundamental problem driving the innovation during his conference keynote: “As the scale of parallel processing grows, cross-machine bandwidth in traditional server architectures has become a critical bottleneck for training.”The Supernode 384 abandons Von Neumann computing principles in favour of a peer-to-peer architecture engineered specifically for modern AI workloads. The change proves especially powerful for Mixture-of-Experts models (machine-learning systems using multiple specialised sub-networks to solve complex computational challenges.)Huawei’s CloudMatrix 384 implementation showcases impressive technical specifications: 384 Ascend AI processors spanning 12 computing cabinets and four bus cabinets, generating 300 petaflops of raw computational power paired with 48 terabytes of high-bandwidth memory, representing a leap in integrated AI computing infrastructure.Performance metrics challenge industry leadersReal-world benchmark testing reveals the system’s competitive positioning in comparison to established solutions. Dense AI models like Meta’s LLaMA 3 achieved 132 tokens per second per card on the Supernode 384 – delivering 2.5 times superior performance compared to traditional cluster architectures.Communications-intensive applications demonstrate even more dramatic improvements. Models from Alibaba’s Qwen and DeepSeek families reached 600 to 750 tokens per second per card, revealing the architecture’s optimisation for next-generation AI workloads.The performance gains stem from fundamental infrastructure redesigns. Huawei replaced conventional Ethernet interconnects with high-speed bus connections, improving communications bandwidth by 15 times while reducing single-hop latency from 2 microseconds to 200 nanoseconds – a tenfold improvement.Geopolitical strategy drives technical innovationThe Supernode 384’s development cannot be divorced from broader US-China technological competition. American sanctions have systematically restricted Huawei’s access to cutting-edge semiconductor technologies, forcing the company to maximise performance within existing constraints.Industry analysis from SemiAnalysis suggests the CloudMatrix 384 uses Huawei’s latest Ascend 910C AI processor, which acknowledges inherent performance limitations but highlights architectural advantages: “Huawei is a generation behind in chips, but its scale-up solution is arguably a generation ahead of Nvidia and AMD’s current products in the market.”The assessment reveals how Huawei AI computing strategies have evolved beyond traditional hardware specifications toward system-level optimisation and architectural innovation.Market implications and deployment realityBeyond laboratory demonstrations, Huawei has operationalised CloudMatrix 384 systems in multiple Chinese data centres in Anhui Province, Inner Mongolia, and Guizhou Province. Such practical deployments validate the architecture’s viability and establishes an infrastructure framework for broader market adoption.The system’s scalability potential – supporting tens of thousands of linked processors – positions it as a compelling platform for training increasingly sophisticated AI models. The capability addresses growing industry demands for massive-scale AI implementation in diverse sectors.Industry disruption and future considerationsHuawei’s architectural breakthrough introduces both opportunities and complications for the global AI ecosystem. While providing viable alternatives to Nvidia’s market-leading solutions, it simultaneously accelerates the fragmentation of international technology infrastructure along geopolitical lines.The success of Huawei AI computing initiatives will depend on developer ecosystem adoption and sustained performance validation. The company’s aggressive developer conference outreach indicated a recognition that technical innovation alone cannot guarantee market acceptance.For organisations evaluating AI infrastructure investments, the Supernode 384 represents a new option that combines competitive performance with independence from US-controlled supply chains. However, long-term viability remains contingent on continued innovation cycles and improved geopolitical stability.(Image from Pixabay)See also: Oracle plans $40B Nvidia chip deal for AI facility in TexasWant to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with other leading events including Intelligent Automation Conference, BlockX, Digital Transformation Week, and Cyber Security & Cloud Expo.Explore other upcoming enterprise technology events and webinars powered by TechForge here.

What just happened? The Department of Energy has announced plans for a new supercomputer designed to significantly accelerate research across a wide range of scientific fields. The initiative highlights the growing convergence between commercial AI development and the computational demands of cutting-edge scientific discovery. The advanced system, to be housed at Lawrence Berkeley National Laboratory and scheduled to become operational in 2026, will be named "Doudna" in honor of Nobel laureate Jennifer Doudna, whose groundbreaking work on CRISPR gene editing has revolutionized molecular biology. Dell Technologies has been selected to deliver the Doudna supercomputer, marking a significant shift in the landscape of government-funded high-performance computing. While companies like Hewlett Packard Enterprise have traditionally dominated this space, Dell's successful bid signals a new chapter. "A big win for Dell," said Addison Snell, CEO of Intersect360 Research, in an interview with The New York Times, noting the company's historically limited presence in this domain. Dell executives explained that the Doudna project enabled them to move beyond the longstanding practice of building custom systems for individual laboratories. Instead, they focused on developing a flexible platform capable of serving a broad array of users. "This market had shifted into some form of autopilot. What we did was disengage the autopilot," said Paul Perez, senior vice president and technology fellow at Dell. The Perlmutter supercomputer at the National Energy Research Scientific Computing Center at Lawrence Berkeley National Laboratory. A defining feature of Doudna will be its use of Nvidia's Vera Rubin platform, engineered to combine the strengths of traditional scientific simulations with the power of modern AI. Unlike previous Department of Energy supercomputers, which relied on processors from Intel or AMD, Doudna will incorporate a general-purpose Arm-based CPU from Nvidia, paired with the company's Rubin AI chips designed specifically for artificial intelligence and simulation workloads. // Related Stories The architecture aims to meet the needs of the laboratory's 11,000 users, who increasingly depend on both high-precision modeling and rapid AI-driven data analysis. Jensen Huang, founder and CEO of Nvidia, described the new system with enthusiasm. "Doudna is a time machine for science – compressing years of discovery into days," he said, adding that it will let "scientists delve deeper and think bigger to seek the fundamental truths of the universe." In terms of performance, Doudna is expected to be over 10 times faster than the lab's current flagship system, making it the Department of Energy's most powerful resource for training AI models and conducting advanced simulations. Jonathan Carter, associate lab director for computing sciences at Berkeley Lab, said the system's architecture was shaped by the evolving needs of researchers – many of whom are now using AI to augment simulations in areas like geothermal energy and quantum computing. Doudna's design reflects a broader shift in supercomputing. Traditional systems have prioritized 64-bit calculations for maximum numerical accuracy, but modern AI workloads often benefit from lower-precision operations (such as 16-bit or 8-bit) that enable faster processing speeds. Dion Harris, Nvidia's head of data center product marketing, noted that the flexibility to combine different levels of precision opens new frontiers for scientific research. The supercomputer will also be tightly integrated with the Energy Sciences Network, allowing researchers nationwide to stream data directly into Doudna for real-time analysis. Sudip Dosanjh, director of the National Energy Research Scientific Computing Center, described the new system as "designed to accelerate a broad set of scientific workflows."

In this tutorial, we implement the Agent Communication Protocol (ACP) through building a flexible, ACP-compliant messaging system in Python, leveraging Google’s Gemini API for natural language processing. Beginning with the installation and configuration of the google-generativeai library, the tutorial introduces core abstractions, message types, performatives, and the ACPMessage data class, which standardizes inter-agent communication. By defining ACPAgent and ACPMessageBroker classes, the guide demonstrates how to create, send, route, and process structured messages among multiple autonomous agents. Through clear code examples, users learn to implement querying, requesting actions, and broadcasting information, while maintaining conversation threads, acknowledgments, and error handling. import google.generativeai as genai import json import time import uuid from enum import Enum from typing import Dict, List, Any, Optional from dataclasses import dataclass, asdict GEMINI_API_KEY = "Use Your Gemini API Key" genai.configure(api_key=GEMINI_API_KEY) We import essential Python modules, ranging from JSON handling and timing to unique identifier generation and type annotations, to support a structured ACP implementation. It then retrieves the user’s Gemini API key placeholder and configures the google-generativeai client for subsequent calls to the Gemini language model. class ACPMessageType(Enum): """Standard ACP message types""" REQUEST = "request" RESPONSE = "response" INFORM = "inform" QUERY = "query" SUBSCRIBE = "subscribe" UNSUBSCRIBE = "unsubscribe" ERROR = "error" ACK = "acknowledge" The ACPMessageType enumeration defines the core message categories used in the Agent Communication Protocol, including requests, responses, informational broadcasts, queries, and control actions like subscription management, error signaling, and acknowledgments. By centralizing these message types, the protocol ensures consistent handling and routing of inter-agent communications throughout the system. class ACPPerformative(Enum): """ACP speech acts (performatives)""" TELL = "tell" ASK = "ask" REPLY = "reply" REQUEST_ACTION = "request-action" AGREE = "agree" REFUSE = "refuse" PROPOSE = "propose" ACCEPT = "accept" REJECT = "reject" The ACPPerformative enumeration captures the variety of speech acts agents can use when interacting under the ACP framework, mapping high-level intentions, such as making requests, posing questions, giving commands, or negotiating agreements, onto standardized labels. This clear taxonomy enables agents to interpret and respond to messages in contextually appropriate ways, ensuring robust and semantically rich communication. @dataclass class ACPMessage: """Agent Communication Protocol Message Structure""" message_id: str sender: str receiver: str performative: str content: Dict[str, Any] protocol: str = "ACP-1.0" conversation_id: str = None reply_to: str = None language: str = "english" encoding: str = "json" timestamp: float = None def __post_init__(self): if self.timestamp is None: self.timestamp = time.time() if self.conversation_id is None: self.conversation_id = str(uuid.uuid4()) def to_acp_format(self) -> str: """Convert to standard ACP message format""" acp_msg = { "message-id": self.message_id, "sender": self.sender, "receiver": self.receiver, "performative": self.performative, "content": self.content, "protocol": self.protocol, "conversation-id": self.conversation_id, "reply-to": self.reply_to, "language": self.language, "encoding": self.encoding, "timestamp": self.timestamp } return json.dumps(acp_msg, indent=2) @classmethod def from_acp_format(cls, acp_string: str) -> 'ACPMessage': """Parse ACP message from string format""" data = json.loads(acp_string) return cls( message_id=data["message-id"], sender=data["sender"], receiver=data["receiver"], performative=data["performative"], content=data["content"], protocol=data.get("protocol", "ACP-1.0"), conversation_id=data.get("conversation-id"), reply_to=data.get("reply-to"), language=data.get("language", "english"), encoding=data.get("encoding", "json"), timestamp=data.get("timestamp", time.time()) ) The ACPMessage data class encapsulates all the fields required for a structured ACP exchange, including identifiers, participants, performative, payload, and metadata such as protocol version, language, and timestamps. Its __post_init__ method auto-populates missing timestamp and conversation_id values, ensuring every message is uniquely tracked. Utility methods to_acp_format and from_acp_format handle serialization to and from the standardized JSON representation for seamless transmission and parsing. class ACPAgent: """Agent implementing Agent Communication Protocol""" def __init__(self, agent_id: str, name: str, capabilities: List[str]): self.agent_id = agent_id self.name = name self.capabilities = capabilities self.model = genai.GenerativeModel("gemini-1.5-flash") self.message_queue: List[ACPMessage] = [] self.subscriptions: Dict[str, List[str]] = {} self.conversations: Dict[str, List[ACPMessage]] = {} def create_message(self, receiver: str, performative: str, content: Dict[str, Any], conversation_id: str = None, reply_to: str = None) -> ACPMessage: """Create a new ACP-compliant message""" return ACPMessage( message_id=str(uuid.uuid4()), sender=self.agent_id, receiver=receiver, performative=performative, content=content, conversation_id=conversation_id, reply_to=reply_to ) def send_inform(self, receiver: str, fact: str, data: Any = None) -> ACPMessage: """Send an INFORM message (telling someone a fact)""" content = {"fact": fact, "data": data} return self.create_message(receiver, ACPPerformative.TELL.value, content) def send_query(self, receiver: str, question: str, query_type: str = "yes-no") -> ACPMessage: """Send a QUERY message (asking for information)""" content = {"question": question, "query-type": query_type} return self.create_message(receiver, ACPPerformative.ASK.value, content) def send_request(self, receiver: str, action: str, parameters: Dict = None) -> ACPMessage: """Send a REQUEST message (asking someone to perform an action)""" content = {"action": action, "parameters": parameters or {}} return self.create_message(receiver, ACPPerformative.REQUEST_ACTION.value, content) def send_reply(self, original_msg: ACPMessage, response_data: Any) -> ACPMessage: """Send a REPLY message in response to another message""" content = {"response": response_data, "original-question": original_msg.content} return self.create_message( original_msg.sender, ACPPerformative.REPLY.value, content, conversation_id=original_msg.conversation_id, reply_to=original_msg.message_id ) def process_message(self, message: ACPMessage) -> Optional[ACPMessage]: """Process incoming ACP message and generate appropriate response""" self.message_queue.append(message) conv_id = message.conversation_id if conv_id not in self.conversations: self.conversations[conv_id] = [] self.conversations[conv_id].append(message) if message.performative == ACPPerformative.ASK.value: return self._handle_query(message) elif message.performative == ACPPerformative.REQUEST_ACTION.value: return self._handle_request(message) elif message.performative == ACPPerformative.TELL.value: return self._handle_inform(message) return None def _handle_query(self, message: ACPMessage) -> ACPMessage: """Handle incoming query messages""" question = message.content.get("question", "") prompt = f"As agent {self.name} with capabilities {self.capabilities}, answer: {question}" try: response = self.model.generate_content(prompt) answer = response.text.strip() except: answer = "Unable to process query at this time" return self.send_reply(message, {"answer": answer, "confidence": 0.8}) def _handle_request(self, message: ACPMessage) -> ACPMessage: """Handle incoming action requests""" action = message.content.get("action", "") parameters = message.content.get("parameters", {}) if any(capability in action.lower() for capability in self.capabilities): result = f"Executing {action} with parameters {parameters}" status = "agreed" else: result = f"Cannot perform {action} - not in my capabilities" status = "refused" return self.send_reply(message, {"status": status, "result": result}) def _handle_inform(self, message: ACPMessage) -> Optional[ACPMessage]: """Handle incoming information messages""" fact = message.content.get("fact", "") print(f"[{self.name}] Received information: {fact}") ack_content = {"status": "received", "fact": fact} return self.create_message(message.sender, "acknowledge", ack_content, conversation_id=message.conversation_id) The ACPAgent class encapsulates an autonomous entity capable of sending, receiving, and processing ACP-compliant messages using Gemini’s language model. It manages its own message queue, conversation history, and subscriptions, and provides helper methods (send_inform, send_query, send_request, send_reply) to construct correctly formatted ACPMessage instances. Incoming messages are routed through process_message, which delegates to specialized handlers for queries, action requests, and informational messages. class ACPMessageBroker: """Message broker implementing ACP routing and delivery""" def __init__(self): self.agents: Dict[str, ACPAgent] = {} self.message_log: List[ACPMessage] = [] self.routing_table: Dict[str, str] = {} def register_agent(self, agent: ACPAgent): """Register an agent with the message broker""" self.agents[agent.agent_id] = agent self.routing_table[agent.agent_id] = "local" print(f"✓ Registered agent: {agent.name} ({agent.agent_id})") def route_message(self, message: ACPMessage) -> bool: """Route ACP message to appropriate recipient""" if message.receiver not in self.agents: print(f"✗ Receiver {message.receiver} not found") return False print(f"\n📨 ACP MESSAGE ROUTING:") print(f"From: {message.sender} → To: {message.receiver}") print(f"Performative: {message.performative}") print(f"Content: {json.dumps(message.content, indent=2)}") receiver_agent = self.agents[message.receiver] response = receiver_agent.process_message(message) self.message_log.append(message) if response: print(f"\n📤 GENERATED RESPONSE:") print(f"From: {response.sender} → To: {response.receiver}") print(f"Content: {json.dumps(response.content, indent=2)}") if response.receiver in self.agents: self.agents[response.receiver].process_message(response) self.message_log.append(response) return True def broadcast_message(self, message: ACPMessage, recipients: List[str]): """Broadcast message to multiple recipients""" for recipient in recipients: msg_copy = ACPMessage( message_id=str(uuid.uuid4()), sender=message.sender, receiver=recipient, performative=message.performative, content=message.content.copy(), conversation_id=message.conversation_id ) self.route_message(msg_copy) The ACPMessageBroker serves as the central router for ACP messages, maintaining a registry of agents and a message log. It provides methods to register agents, deliver individual messages via route_message, which handles lookup, logging, and response chaining, and to send the same message to multiple recipients with broadcast_message. def demonstrate_acp(): """Comprehensive demonstration of Agent Communication Protocol""" print("🤖 AGENT COMMUNICATION PROTOCOL (ACP) DEMONSTRATION") print("=" * 60) broker = ACPMessageBroker() researcher = ACPAgent("agent-001", "Dr. Research", ["analysis", "research", "data-processing"]) assistant = ACPAgent("agent-002", "AI Assistant", ["information", "scheduling", "communication"]) calculator = ACPAgent("agent-003", "MathBot", ["calculation", "mathematics", "computation"]) broker.register_agent(researcher) broker.register_agent(assistant) broker.register_agent(calculator) print(f"\n📋 REGISTERED AGENTS:") for agent_id, agent in broker.agents.items(): print(f" • {agent.name} ({agent_id}): {', '.join(agent.capabilities)}") print(f"\n🔬 SCENARIO 1: Information Query (ASK performative)") query_msg = assistant.send_query("agent-001", "What are the key factors in AI research?") broker.route_message(query_msg) print(f"\n🔢 SCENARIO 2: Action Request (REQUEST-ACTION performative)") calc_request = researcher.send_request("agent-003", "calculate", {"expression": "sqrt(144) + 10"}) broker.route_message(calc_request) print(f"\n📢 SCENARIO 3: Information Sharing (TELL performative)") info_msg = researcher.send_inform("agent-002", "New research paper published on quantum computing") broker.route_message(info_msg) print(f"\n📊 PROTOCOL STATISTICS:") print(f" • Total messages processed: {len(broker.message_log)}") print(f" • Active conversations: {len(set(msg.conversation_id for msg in broker.message_log))}") print(f" • Message types used: {len(set(msg.performative for msg in broker.message_log))}") print(f"\n📋 SAMPLE ACP MESSAGE FORMAT:") sample_msg = assistant.send_query("agent-001", "Sample question for format demonstration") print(sample_msg.to_acp_format()) The demonstrate_acp function orchestrates a hands-on walkthrough of the entire ACP framework: it initializes a broker and three distinct agents (Researcher, AI Assistant, and MathBot), registers them, and illustrates three key interaction scenarios, querying for information, requesting a computation, and sharing an update. After routing each message and handling responses, it prints summary statistics on the message flow. It showcases a formatted ACP message, providing users with a clear, end-to-end example of how agents communicate under the protocol. def setup_guide(): print(""" 🚀 GOOGLE COLAB SETUP GUIDE: 1. Get Gemini API Key: https://makersuite.google.com/app/apikey 2. Replace: GEMINI_API_KEY = "YOUR_ACTUAL_API_KEY" 3. Run: demonstrate_acp() 🔧 ACP PROTOCOL FEATURES: • Standardized message format with required fields • Speech act performatives (TELL, ASK, REQUEST-ACTION, etc.) • Conversation tracking and message threading • Error handling and acknowledgments • Message routing and delivery confirmation 📝 EXTEND THE PROTOCOL: ```python # Create custom agent my_agent = ACPAgent("my-001", "CustomBot", ["custom-capability"]) broker.register_agent(my_agent) # Send custom message msg = my_agent.send_query("agent-001", "Your question here") broker.route_message(msg) ``` """) if __name__ == "__main__": setup_guide() demonstrate_acp() Finally, the setup_guide function provides a quick-start reference for running the ACP demo in Google Colab, outlining how to obtain and configure your Gemini API key and invoke the demonstrate_acp routine. It also summarizes key protocol features, such as standardized message formats, performatives, and message routing. It provides a concise code snippet illustrating how to register custom agents and send tailored messages. In conclusion, this tutorial implements ACP-based multi-agent systems capable of research, computation, and collaboration tasks. The provided sample scenarios illustrate common use cases, information queries, computational requests, and fact sharing, while the broker ensures reliable message delivery and logging. Readers are encouraged to extend the framework by adding new agent capabilities, integrating domain-specific actions, or incorporating more sophisticated subscription and notification mechanisms. Download the Notebook on GitHub. All credit for this research goes to the researchers of this project. Also, feel free to follow us on Twitter and don’t forget to join our 95k+ ML SubReddit and Subscribe to our Newsletter. Asif RazzaqWebsite |  + postsBioAsif Razzaq is the CEO of Marktechpost Media Inc.. As a visionary entrepreneur and engineer, Asif is committed to harnessing the potential of Artificial Intelligence for social good. His most recent endeavor is the launch of an Artificial Intelligence Media Platform, Marktechpost, which stands out for its in-depth coverage of machine learning and deep learning news that is both technically sound and easily understandable by a wide audience. The platform boasts of over 2 million monthly views, illustrating its popularity among audiences.Asif Razzaqhttps://www.marktechpost.com/author/6flvq/Yandex Releases Yambda: The World’s Largest Event Dataset to Accelerate Recommender SystemsAsif Razzaqhttps://www.marktechpost.com/author/6flvq/Stanford Researchers Introduced Biomni: A Biomedical AI Agent for Automation Across Diverse Tasks and Data TypesAsif Razzaqhttps://www.marktechpost.com/author/6flvq/DeepSeek Releases R1-0528: An Open-Source Reasoning AI Model Delivering Enhanced Math and Code Performance with Single-GPU EfficiencyAsif Razzaqhttps://www.marktechpost.com/author/6flvq/A Coding Guide for Building a Self-Improving AI Agent Using Google’s Gemini API with Intelligent Adaptation Features

From large language models (LLMs) to reasoning agents, today’s AI tools bring unprecedented computational demands. Trillion-parameter models, workloads running on-device, and swarms of agents collaborating to complete tasks all require a new paradigm of computing to become truly seamless and ubiquitous. Silicon’s mid-life crisis AI has evolved from classical ML to deep learning to generative AI. The most recent chapter, which took AI mainstream, hinges on two phases—training and inference—that are data and energy-intensive in terms of computation, data movement, and cooling. At the same time, Moore’s Law, which determines that the number of transistors on a chip doubles every two years, is reaching a physical and economic plateau. For the last 40 years, silicon chips and digital technology have nudged each other forward—every step ahead in processing capability frees the imagination of innovators to envision new products, which require yet more power to run. That is happening at light speed in the AI age. As models become more readily available, deployment at scale puts the spotlight on inference and the application of trained models for everyday use cases. This transition requires the appropriate hardware to handle inference tasks efficiently. Central processing units (CPUs) have managed general computing tasks for decades, but the broad adoption of ML introduced computational demands that stretched the capabilities of traditional CPUs. This has led to the adoption of graphics processing units (GPUs) and other accelerator chips for training complex neural networks, due to their parallel execution capabilities and high memory bandwidth that allow large-scale mathematical operations to be processed efficiently. But CPUs are already the most widely deployed and can be companions to processors like GPUs and tensor processing units (TPUs). AI developers are also hesitant to adapt software to fit specialized or bespoke hardware, and they favor the consistency and ubiquity of CPUs. Chip designers are unlocking performance gains through optimized software tooling, adding novel processing features and data types specifically to serve ML workloads, integrating specialized units and accelerators, and advancing silicon chip innovations, including custom silicon. AI itself is a helpful aid for chip design, creating a positive feedback loop in which AI helps optimize the chips that it needs to run. These enhancements and strong software support mean modern CPUs are a good choice to handle a range of inference tasks. Beyond silicon-based processors, disruptive technologies are emerging to address growing AI compute and data demands. The unicorn start-up Lightmatter, for instance, introduced photonic computing solutions that use light for data transmission to generate significant improvements in speed and energy efficiency. Quantum computing represents another promising area in AI hardware. While still years or even decades away, the integration of quantum computing with AI could further transform fields like drug discovery and genomics. The developments in ML theories and network architectures have significantly enhanced the efficiency and capabilities of AI models. Today, the industry is moving from monolithic models to agent-based systems characterized by smaller, specialized models that work together to complete tasks more efficiently at the edge—on devices like smartphones or modern vehicles. This allows them to extract increased performance gains, like faster model response times, from the same or even less compute. Researchers have developed techniques, including few-shot learning, to train AI models using smaller datasets and fewer training iterations. AI systems can learn new tasks from a limited number of examples to reduce dependency on large datasets and lower energy demands. Optimization techniques like quantization, which lower the memory requirements by selectively reducing precision, are helping reduce model sizes without sacrificing performance.  New system architectures, like retrieval-augmented generation (RAG), have streamlined data access during both training and inference to reduce computational costs and overhead. The DeepSeek R1, an open source LLM, is a compelling example of how more output can be extracted using the same hardware. By applying reinforcement learning techniques in novel ways, R1 has achieved advanced reasoning capabilities while using far fewer computational resources in some contexts. The integration of heterogeneous computing architectures, which combine various processing units like CPUs, GPUs, and specialized accelerators, has further optimized AI model performance. This approach allows for the efficient distribution of workloads across different hardware components to optimize computational throughput and energy efficiency based on the use case. As AI becomes an ambient capability humming in the background of many tasks and workflows, agents are taking charge and making decisions in real-world scenarios. These range from customer support to edge use cases, where multiple agents coordinate and handle localized tasks across devices. With AI increasingly used in daily life, the role of user experiences becomes critical for mass adoption. Features like predictive text in touch keyboards, and adaptive gearboxes in vehicles, offer glimpses of AI as a vital enabler to improve technology interactions for users. Edge processing is also accelerating the diffusion of AI into everyday applications, bringing computational capabilities closer to the source of data generation. Smart cameras, autonomous vehicles, and wearable technology now process information locally to reduce latency and improve efficiency. Advances in CPU design and energy-efficient chips have made it feasible to perform complex AI tasks on devices with limited power resources. This shift toward heterogeneous compute enhances the development of ambient intelligence, where interconnected devices create responsive environments that adapt to user needs. Seamless AI naturally requires common standards, frameworks, and platforms to bring the industry together. Contemporary AI brings new risks. For instance, by adding more complex software and personalized experiences to consumer devices, it expands the attack surface for hackers, requiring stronger security at both the software and silicon levels, including cryptographic safeguards and transforming the trust model of compute environments. More than 70% of respondents to a 2024 DarkTrace survey reported that AI-powered cyber threats significantly impact their organizations, while 60% say their organizations are not adequately prepared to defend against AI-powered attacks. Collaboration is essential to forging common frameworks. Universities contribute foundational research, companies apply findings to develop practical solutions, and governments establish policies for ethical and responsible deployment. Organizations like Anthropic are setting industry standards by introducing frameworks, such as the Model Context Protocol, to unify the way developers connect AI systems with data. Arm is another leader in driving standards-based and open source initiatives, including ecosystem development to accelerate and harmonize the chiplet market, where chips are stacked together through common frameworks and standards. Arm also helps optimize open source AI frameworks and models for inference on the Arm compute platform, without needing customized tuning.  How far AI goes to becoming a general-purpose technology, like electricity or semiconductors, is being shaped by technical decisions taken today. Hardware-agnostic platforms, standards-based approaches, and continued incremental improvements to critical workhorses like CPUs, all help deliver the promise of AI as a seamless and silent capability for individuals and businesses alike. Open source contributions are also helpful in allowing a broader range of stakeholders to participate in AI advances. By sharing tools and knowledge, the community can cultivate innovation and help ensure that the benefits of AI are accessible to everyone, everywhere. Learn more about Arm’s approach to enabling AI everywhere. This content was produced by Insights, the custom content arm of MIT Technology Review. It was not written by MIT Technology Review’s editorial staff. This content was researched, designed, and written entirely by human writers, editors, analysts, and illustrators. This includes the writing of surveys and collection of data for surveys. AI tools that may have been used were limited to secondary production processes that passed thorough human review.