SSPAI

Category: AI

  • Ditching the Remote Is the First Step Toward a Robot’s Sense of Life: The First Full-Web Hands-On of Vbot “Big Head”

    Ditching the Remote Is the First Step Toward a Robot’s Sense of Life: The First Full-Web Hands-On of Vbot “Big Head”

    Editor’s note: The following text version is a polished, written rearrangement based on the video’s verbatim transcript. AI tools were used in the process, and the images are screenshots from the original video. The goal is to help you quickly browse the key information. It is recommended to watch the original video alongside this article for a more complete picture.

    This is Vbot’s first product, “Big Head.”

    I’ve been especially excited about this episode, because it’s going to be different from what we’ve done before. I’ll start with the physical design of Big Head itself, and move earlier into its mobility performance; then I’ll cover the hardware specifications that many people care about, explain why it doesn’t need a remote control, and what kind of perception capabilities it has. Later on, I’ll also talk about quite a few standout features that rarely appear in past robot reviews, many of which are highly practical. Big Head performed far beyond my expectations, and I didn’t anticipate this episode becoming so long—there are simply too many details worth talking about. I hope you enjoy it. This might well be the beginning of the next generation of home robots.

    Appearance

    Let’s start with the design. As a To-C (consumer-oriented) product, Big Head stands out from other small robot dogs that lack a head: with its head design, it is noticeably taller when standing and looks much more like a real puppy. Its standing dimensions are 61 cm × 34 cm × 57 cm, and it weighs about 14 kg.

    The body makes extensive use of curved lines, giving it a cuter and more approachable feel. The aluminum alloy on the outer sides of the thighs is treated with a powder-coating process, so scratches and bumps are not very noticeable, and there’s no need to be overly protective during long-term use.

    In terms of details, Vbot’s Big Head features larger knee joint limiters. The underside of the body is slightly raised, so when it lies down, your hand won’t get caught between the thigh and the lower leg, nor will it be pressed under the body (either would be very painful). The mass-production version adds soft material wrapping between the joints and the body, with no exposed joints on the outside, so there are no safety hazards.

    After a few days of use, I found that not only do humans like Big Head—it also attracts some small animals, mainly dogs, which become very curious about it. Alpacas and deer tend to keep a bit of distance. Whether pigeons come over doesn’t seem to have much to do with Big Head; as long as there’s food, they’ll fly in. When Big Head is still, everything feels peaceful and calm; once it starts moving, the pigeons take off.

    I think its “human-friendly” and “animal-friendly” qualities come partly from its design, and partly from the fact that it runs very quietly—there’s almost no audible noise.

    As a home robot dog, Big Head supports silent foot pads and is suitable for indoor use. But if it could only be used indoors or only on flat terrain, then there wouldn’t be much point in it being a quadruped robot at all. So next, let’s talk about its mobility performance.

    Mobility Performance and Navigation

    As you can see, this is a wild IROS obstacle course site. Any robot dog that can decisively traverse terrain like this is no simple machine. The following is a video segment of “Big Head” moving through complex terrain, showcasing several of its capabilities: first, its motion control does not require mode switching; when encountering relatively high steps, it can adjust its leg-lifting height within just a few frames to ensure passability. This detail is only noticeable in slow motion—at normal speed, you’d hardly catch it, only feeling that it walks well and that its leg adjustments are extremely fast. Everyday steps and curbs are all handled smoothly as well.

    Its cruising speed can reach 14 km/h, with a top speed of 18 km/h. Overall, its mobility performance is excellent. Its disturbance rejection strategy is not overly aggressive: instead of using extremely high step frequency, it keeps its adjustments restrained and stabilizes its posture as much as possible while prioritizing user safety. In this kind of usage scenario, “Big Head” also has to account for the possibility of being messed with at close range by “rowdy kids (me).” From a safety standpoint, this design choice is very reasonable.

    Next, let’s take a look at the overall hardware configuration of “Big Head.”

    First, battery life—something I care a lot about, because carrying a dead robot home is both awkward and troublesome. “Big Head” offers 3 to 6 hours of runtime with a 594 Wh battery. The battery is not swappable, but that isn’t really a problem: the large capacity means you can head out without worrying about changing batteries, and there’s essentially no range anxiety. For charging, it supports both a Type-C port and a wireless “dock-and-charge” station, with a maximum charging power of up to 240 W, taking about 2.5 hours for a full charge.

    In terms of perception, “Big Head” is equipped with a 16-beam LiDAR and a stereo depth camera system, supporting 1080p video transmission. It uses the DiGuA Robotics S100P chip, providing up to 128 TOPS of AI computing power. Many of the features I’ll talk about next are closely tied to these perception capabilities and hardware specs.

    Lead, follow, navigate, tow—some of these functions are fun, some are special, and some are almost unheard of, yet they are all simple to use. The four-microphone array can detect the direction of sound and provide head-movement feedback—“Big Head” really is listening to me.

    I give the command: “Big Head, Big Head, start leading.”

    The leading function is a lot like walking a dog: “Big Head” walks in front, autonomously perceiving obstacles, finding its way, and planning routes. When it reaches an intersection, it chooses a path on its own; if I want to take a different one, I just tug the leash slightly from the side to guide it, and it will switch to the route I choose.

    These visuals show the world as “Big Head” sees it: perception results, route planning, depth information, its understanding of the road, and a top-down view. You might already recognize shades of autonomous driving here. “Big Head” knows where the road is and where it can go—but it doesn’t yet know when to stop. If I want it to halt, for example to tie my shoes or greet someone, I just pull the leash from behind.

    To resume leading, I don’t even need to give another voice command—just give “Big Head” a gentle nudge on the backside. The overall interaction feels very natural, a real-world version of “cyber dog-walking.”

    It’s worth noting that the leading function does not require positioning systems or high-precision point clouds. It relies solely on its onboard LiDAR and vision to perceive the environment. This means “Big Head” can explore and unlock new maps with me even in unfamiliar places.

    Now let’s look at following. I think this may currently be the most fully realized follow mode in the robotics industry, truly turning “Big Head” into a “little sidekick.” This uses an included accessory: a UWB beacon.

    I give the command: “Big Head, Big Head, follow me.”

    With obstacle avoidance enabled, “Big Head” locks onto the beacon, tracks its target precisely, and doesn’t drift or lose the subject the way vision-only solutions sometimes do. But even that alone would only count as “pretty good.”

    Vbot also offers an expansion board accessory, which can be mounted on “Big Head’s” back with hand-tightened screws. A basket can be installed on top for carrying items. “Big Head” supports a payload of 10 kg, with a maximum of 12 kg. The structure is solid and also supports quick release, making it easy to carry things or remove the module—already impressive.

    The expansion board also includes a standard 1/4-inch screw mount commonly used in the imaging industry, allowing you to attach action cameras, 360-degree cameras, camera gimbals, or even smartphone gimbals for follow-shot filming. I personally love this feature.

    There’s another great design detail: the UWB beacon has a built-in microphone. You can press its button and speak directly to “Big Head,” avoiding wind noise or other ambient sounds interfering with command recognition.

    Navigation is a feature I’ve rarely experienced on consumer robot dogs—it’s more common in industrial-grade products, mainly for scheduled inspection tasks. With “Big Head,” you first map the navigation area; usually walking it twice is enough. After mapping, you can tap a location on the map in the mobile app to name it, then give the command: “Big Head, Big Head, take me to the restaurant.”

    Each navigation run may involve different people and changing street scenes, yet “Big Head” still performs very well, showing a certain level of generalization ability. Night navigation is no problem either. And it doesn’t get boring along the way—every so often, “Big Head” draws attention to itself and reports the navigation progress. For me, the most convenient part is that there’s no need to set the robot’s position or orientation on the map beforehand; you can start navigating right away. It’s very user-friendly for ordinary users.

    Towing mode is for situations where you need to move faster or when the environment and obstacles are more complex: you simply pull “Big Head” along. I also made a point of confirming that its head structure has been reinforced, so using it as a leash attachment is reliable. The two degrees of freedom in the head (rotation and pitch) showed no issues at all during several days of use.

    After seeing all these features, I think you can understand why “Big Head” doesn’t need a remote control. Still, a remote option exists as a safety backup, tucked away in the secondary menu of the mobile app. It supports one-handed operation and offers plenty of movements to try, so I won’t go into detail here.

    But remote control doesn’t stop at remote control—there’s also a particularly interesting mode: the Avatar proxy. I can join this world in the form of “Big Head,” remotely control the robot itself, and experience extremely low network latency. With a new perspective and a different identity, I can run outdoors, greet passersby, and approach small animals. You have to realize: a person in their twenties or thirties chasing pigeons on a lawn looks pretty strange to others—that’s the kind of self-restraint adults impose on themselves.

    Remotely controlling a robot dog, however, is different. If I want, I can lie on the couch and chase pigeons until “Big Head” is almost out of battery. I can greet puppies and, within safe limits, interact with children. In truth, every adult was once a child—only a few still remember it. At this moment, I am one of those few.

    While playing, I once asked, “Can we turn on ‘Big Head’s’ microphone for live voice chat?” The answer I got was: “Absolutely not!” On second thought, that makes perfect sense. If you came across such an adorable robot dog and it suddenly spoke in the deep voice of a grown man, the sense of mismatch would be overwhelming—like when Disney’s LinaBell character head falls off and the staff inside greets you in a thick Beijing accent. The fairy tale would shatter in the most terrifying way.

    Of course, you can now type what you want to say (or use speech-to-text) and have “Big Head” speak it in its own voice. You really should try this—it gets some hilarious reactions.

    I believe many people can find a kind of pure, genuine, age-independent joy in “Big Head,” or in things that seem childish on the surface. I haven’t felt this way in a long time.

    There’s also a feature called “Dance to the Beat”: “Big Head” can recognize audio waveforms picked up by its microphone and generate movements according to the rhythm of the music, focusing on hitting the beat. In the past, whether robot dogs danced on beat largely depended on how accurately you pressed the play button for the music. Now, letting “Big Head” listen and generate beat-matched movements on its own solves that problem.

    A Sense of Life

    If the features mentioned earlier already show “Big Head’s” playfulness and technical autonomy, then what comes next is what fascinates me the most about it: its sense of life.

    Where does this sense of life come from? I see unpredictable autonomous decision-making, and movements that align with biological intuition.

    Here is an example of “Big Head’s” chain of thought. The command was: “Let Big Head look for something it finds interesting.” No one knows what it will choose, so every time I’m eager to see what kind of response it gives.

    The most surprising instance was when “Big Head” noticed an advertisement display. It wanted to get a closer look: first walking over, then lifting its head to look up. That head-raising motion matched biological instinct perfectly, but I had no expectation of it at all—seeing it for the first time was genuinely stunning. “Big Head” knows what it is interested in, knows that tall things need to be looked at by raising its head, and can even recognize that the image in the ad is a robot dog, judging it to be one of its “own kind.”

    Another time, many people stood in front of “Big Head” and asked it to find someone it found interesting. No one knew who it would pick, but being chosen by a robot dog is truly a delightful experience. “Big Head” would actively walk up and interact with the person it was interested in.

    Its spatial-agent capabilities allow “Big Head” to complete complex, multi-step tasks in the real world, such as going to find someone to get a bottle of water and then bringing it back. You might think that’s not particularly hard—but this is only the beginning.

    In past reviews, we were used to using a product to gaze into the future. This time, I hope you can look at the present instead: using today’s existing technologies, Vbot has already delivered a large number of innovative experiences for everyday users, listening to user needs and polishing the details thoroughly. Many things that usually require DIY modifications already have official solutions among the original accessories. For me, “Big Head” has almost no barrier to entry.

    Even though I understand what hardware and technologies are behind these features, I still find myself immersed in the “magic” they create. Sometimes “Big Head” leads the way ahead of me; sometimes I have a little sidekick at my side. A huge number of its movements and expressions are designed by animators, to the point that it can even give you the illusion of a cartoon character—and you never once need to think about “where the remote control is.”

    Epilogue

    The past few days with “Big Head” have felt more like the future—yet they’re happening right now.

    This robot dog is already far beyond the level of a mere “early taste” product. I genuinely want to have “Big Head” with me for the New Year, and I’d be happy to keep it by my side for a long time to come.

    Alright, thank you for reading all the way to the end. I hope you enjoyed this episode, and that you liked seeing all the different sides of “Big Head” in the review. Thanks to Vbot for creating such an amazing product—this was a truly satisfying review to make. Thanks as well to the Vlight Club at Beijing Forestry University for their support, which made it possible for us to complete this episode smoothly. There’s more robot review content currently in production, so stay tuned. See you in the next video.

  • SSPAI Morning Brief: OpenAI announces the launch of ChatGPT Health, and More

    SSPAI Morning Brief: OpenAI announces the launch of ChatGPT Health, and More

    Morning Brief

    1. OnePlus unveils the Turbo 6 smartphone series
    2. Aqara releases the U400 smart door lock with UWB support
    3. Lenovo showcases the Legion Pro Rollable concept device
    4. RayNeo presents the X3 Pro Project AR smart glasses with eSIM support
    5. Krono launches a black-and-white e-ink phone–style e-reader
    6. Sony introduces the new Hyperpop series shells and controllers for PS5 Slim
    7. TCL releases the P50 mobile router and BE36 Wi-Fi router
    8. MSI showcases multiple new products
    9. The world’s first ultrasonic kitchen knife, the C200 Ultrasonic, makes its debut
    10. OpenAI announces the launch of ChatGPT Health
    11. Rumors You Can Just Glance At

    OnePlus unveils the Turbo 6 smartphone series

    On January 8, the OnePlus Turbo 6 series was officially released, consisting of two models: the Turbo 6 and the Turbo 6V.

    The OnePlus Turbo 6 adopts a minimalist design language with a strong emphasis on in-hand feel, and is available in silver, black, and teal color options. It is powered by the fourth-generation Snapdragon 8s processor, paired with OnePlus’s self-developed “Fengchi Gaming Kernel.” According to OnePlus, this combination can reduce gaming power consumption by around 8%, lower kernel load by approximately 15.24%, and improve scheduling efficiency by about 35.46%.

    In terms of display, the OnePlus Turbo 6 features a 1.5K-resolution 165Hz screen, supporting three refresh-rate modes at 165Hz, 144Hz, and 120Hz. Select games also support native 165 fps gameplay, including titles such as Call of Duty: Mobile and Honor of Kings. For imaging, the Turbo 6 is equipped with a Sony 50-megapixel main camera with OIS optical image stabilization, ultra-clear Live Photos, portrait algorithms, and AI-powered photo editing features.

    For battery life, the OnePlus Turbo 6 series is fitted with a 9000mAh battery, supporting 80W fast charging and bypass charging technology, as well as up to 27W wired reverse charging. In other areas, the OnePlus Turbo 6 supports IP66 / IP68 / IP69 / IP69K dust and water resistance, runs ColorOS 16, and offers features such as one-tap Flash Notes, AI real-world conversations, Game Fluid Cloud, and a variety of gaming assistance tools. It also supports interoperability with the Apple ecosystem. Pricing for the OnePlus Turbo 6 starts at RMB 2,099, or RMB 1,784.15 after national subsidies.

    OnePlus also introduced the Turbo 6V model alongside it, available in white, blue, and black. It is powered by the Snapdragon 7s Gen 4 processor and features a 144Hz eye-care display, while also packing a 9000mAh battery. The Turbo 6V starts at RMB 1,699, or RMB 1,444.15 after national subsidies. Source

    Aqara releases the U400 smart door lock with UWB support

    At CES 2026, Aqara officially unveiled the Smart Lock U400. The product supports ultra-wideband (UWB) technology, enabling hands-free automatic unlocking through high-precision positioning with an iPhone or Apple Watch. When an authorized user approaches the door from outside, the lock can automatically unlock. By combining ToF distance measurement with AoA trajectory detection, the system reduces false triggers and enhances security to guard against relay attacks. In addition to UWB, the lock also supports unlocking via Apple Home Key, passcodes, fingerprints, and physical keys. It supports Matter over Thread and carries an IP65 water-resistance rating. The Smart Lock U400 is currently available in the United States, Canada, Australia, New Zealand, and Singapore. Source

    Lenovo showcases the Legion Pro Rollable concept device

    At CES 2026, Lenovo unveiled the Legion Pro Rollable concept gaming laptop. Built on the Legion Pro 7i platform, the device is equipped with an NVIDIA GeForce RTX 5090 graphics card and a high-end Intel Core Ultra processor. It features a 16-inch OLED display by default, which can be horizontally expanded via a rollable mechanism to 21.5 inches or 24 inches. Lenovo positions it as a premium mobile device for professional esports players, offering three display modes tailored to different sizes: focused training, tactical observation, and competitive play.

    According to Lenovo, the Legion Pro Rollable adopts a dual-motor tension-controlled rollable OLED solution to ensure stability and reliability during screen extension and retraction. The device also demonstrates a range of AI-powered gaming assistance features, including key visual element recognition, information magnification, cursor tracking, AI guidance, and dynamic lighting effects. The product remains in the concept stage, with no official release date or pricing announced. Source

    RayNeo presents the X3 Pro Project AR smart glasses with eSIM support

    At CES 2026, RayNeo officially unveiled the X3 Pro Project eSIM AR smart glasses. The product adopts a binocular full-color optical engine design, which the company says delivers a 3D spatial visual experience equivalent to a 43-inch display. On the hardware side, it is powered by the Qualcomm Snapdragon AR1 computing platform and features a built-in RayNeo AR application virtual machine, supporting apps such as WeChat, Douyin, and Bilibili.

    The X3 Pro Project eSIM AR smart glasses also integrate an eSIM communication module, allowing users to make calls, engage in real-time AI conversations, perform live translation, and stream online content without needing to connect to a smartphone or Wi-Fi network. Source

    Krono launches a black-and-white e-ink phone–style e-reader

    Dutch company DuRoBo unveiled the Krono monochrome E-Ink reader at CES 2026. The device adopts a smartphone-like form factor and features a built-in AI assistant. Krono is available in black and white color options, with dimensions of 154 × 80 × 9 mm and a weight of 173 grams. On the front, it is equipped with a 6.13-inch Carta 1200 monochrome E-Ink display, with a resolution of 824 × 1648, a pixel density of 300 PPI, and adjustable dual-tone front lighting.

    On the left side of the body, Krono includes a “smart dial” that allows users to turn pages and adjust settings through rotation or pressing. In terms of hardware, the device features built-in speakers and Bluetooth connectivity, an unspecified octa-core processor, 6 GB of RAM, and 128 GB of storage. It runs Android 15 out of the box and integrates the Libby AI assistant. The Krono is priced at €239.99. Source

    Sony introduces the new Hyperpop series shells and controllers for PS5 Slim

    Sony PlayStation unveiled a new color lineup, The Hyperpop Collection, on January 8, introducing three high-saturation color options for the DualSense controller and PS5 console covers. The collection includes Techno Red, Remix Green, and Rhythm Blue, with an overall aesthetic clearly leaning toward an RGB-inspired visual style.

    The collection will go on sale on March 12, with pre-orders opening on January 16. Sales channels include PlayStation Direct and selected partner retailers, depending on regional availability. The Hyperpop DualSense controller has a suggested retail price of USD 84.99. The Hyperpop-series PS5 console covers will be available for a limited time in select markets, with a suggested retail price of USD 74.99. It is worth noting that these covers are compatible only with the PS5 Slim, and do not support the original PS5 or the PS5 Pro. Source

    TCL releases the P50 mobile router and BE36 Wi-Fi router

    TCL unveiled two new networking products at CES 2026—the TCL 5G Mobile Wi-Fi P50 and the TCL Wi-Fi Router BE36—covering a wide range of scenarios from outdoor travel to home networking.

    The TCL 5G Mobile Wi-Fi P50 is designed for business travel and outdoor use. It supports both mmWave and Sub-6 full-band 5G, with peak download speeds of up to 3.97 Gbps, while Wi-Fi 7 delivers peak speeds of up to 5.8 Gbps. The device can connect up to 64 devices simultaneously. It features a built-in 5,000 mAh battery that provides around 12 hours of continuous use and can also function as a power bank to charge smartphones. A 2.4-inch touchscreen displays real-time data usage and network status, and a Gigabit Ethernet port is included for wired connections.

    Aimed at home users, the TCL BE36 router offers dual-band Wi-Fi 7 connectivity with speeds of up to 3.6 Gbps, using a 5 GHz 3T3R configuration to significantly improve signal coverage. It supports EasyMesh R6, making it easy to build a seamless whole-home Mesh network. With support for up to 512 connected devices, it is well suited for smart homes with a large number of IoT devices. On the security side, the router includes anti-malware, anti-tracking, and brute-force attack protection. For gamers, the BE36 integrates game acceleration features that intelligently optimize traffic to reduce latency. In terms of connectivity, it also provides a 2.5 G Ethernet port. Source

    MSI showcases multiple new products

    MSI showcased several new products at CES 2026, including the MPG CoreFrozr AP15 and AP17 dual-tower air coolers, as well as the MPG 272QRF X36 27-inch QHD 360 Hz monitor.

    Among them, the MPG CoreFrozr AP15 features a segmented digital display dubbed “Digi-Display.” The screen is divided into four sections, allowing real-time display of key system information such as CPU temperature, system load, and fan speed. The lower-left corner also integrates a weekly calendar display.

    In terms of performance, the AP15 is optimized for AMD Ryzen X3D series processors. To address heat buildup from high-performance gaming CPUs, it uses a solid copper base paired with six heat pipes. Positioned as a higher-end model, the MPG CoreFrozr AP17 is equipped with a 6-inch LCD screen that, in addition to system monitoring, supports fully customizable content, enabling users to tailor visuals to match their system’s overall aesthetic. For cooling, the AP17 uses a nickel-plated copper base and increases the number of heat pipes to eight to improve thermal conductivity. Both the AP15 and AP17 feature detachable display modules. Source

    Also on display was the MPG 272QRF X36 G-SYNC Pulsar gaming monitor. This model features a 2560 × 1440 resolution Rapid IPS panel with a 360 Hz refresh rate, a 0.5 ms GtG response time, 400 nits of SDR brightness, and 500 nits of HDR brightness, while supporting 10-bit color (8-bit + FRC). The monitor includes a built-in ambient light sensor that automatically adjusts color temperature and brightness.

    The display supports VESA 100 × 100 mm mounting and offers a range of ports, including one DisplayPort 1.4, two HDMI 2.1 ports, three USB-A 5 Gbps ports with KVM support, and a 3.5 mm audio jack. It incorporates NVIDIA G-SYNC technology, delivering variable refresh rate support and up to four times effective motion clarity, ensuring precise aiming and clear visuals even in fast-paced games. The monitor is currently available on domestic e-commerce platforms, priced at 4,999 yuan. Source

    The world’s first ultrasonic kitchen knife, the C200 Ultrasonic, makes its debut

    Seattle Ultrasonics unveiled the world’s first ultrasonic kitchen knife, the C200 Ultrasonic, at CES 2026. The core idea behind the product is to use high-frequency micro-vibrations to reduce cutting resistance, allowing the blade to penetrate ingredients more easily while also minimizing food sticking to the knife.

    The operation is straightforward: a button on the handle activates the ultrasonic vibration. Once turned on, the blade shows no visible shaking, but the user can feel a subtle buzzing sensation in the hand, similar to a faint electric current. When cutting high-density or starch-rich ingredients, its performance is noticeably superior to that of a conventional kitchen knife. The product is priced at USD 399. Source

    OpenAI announces the launch of ChatGPT Health

    OpenAI has announced the launch of ChatGPT Health, a dedicated health consultation area within ChatGPT that operates independently from the main chat experience. Users can ask health-related questions in this section and receive reference-oriented interpretations based on their personal data, but the service does not provide medical diagnoses or treatment recommendations.

    In terms of functionality, ChatGPT Health supports integration with multiple health services, including Apple Health, MyFitnessPal, and Peloton. It can read users’ exercise, sleep, and activity data, and can also incorporate medical records to help users understand lab results and prepare for medical appointments—again, without offering specific diagnostic or treatment advice. Regarding privacy and security, OpenAI states that the feature uses multi-layer encryption, runs as a separate space, and that related data is not used for model training by default.

    Currently, ChatGPT Health is available to users in select regions via a waitlist. The medical record integration feature is temporarily limited to the United States, with plans to gradually expand availability to the web and iOS platforms. Source

    Rumors You Can Just Glance At

    • According to a report by Beijing Youth Daily, China’s National Medical Products Administration has recently approved Apple’s “Mobile Pulse Rate Atrial Fibrillation Indication Recording Software” as a Class II imported medical device (registration). The software corresponds to the Atrial Fibrillation History feature on Apple Watch, which analyzes users’ pulse rate data and records potential signs of atrial fibrillation. Apple told Beijing Youth Daily reporters that it is currently advancing follow-up procedures in accordance with relevant regulations and processes, and looks forward to officially making the feature available to users in mainland China in the near future. Source
  • SSPAI Morning Brief: NVIDIA Unveils DLSS 4.5, Dell Launches XPS 14/16, and More

    SSPAI Morning Brief: NVIDIA Unveils DLSS 4.5, Dell Launches XPS 14/16, and More

    Morning Brief

    1. NVIDIA unveils DLSS 4.5, the Rubin platform, open models, and more
    2. Dell launches XPS 14/16 laptops
    3. HP introduces the keyboard-style mini PC Eliteboard G1a
    4. Intel releases Core Ultra 300 series AI processors
    5. LEGO launches smart bricks
    6. Rumors You Can Just Glance At

    NVIDIA unveils DLSS 4.5, the Rubin platform, open models, and more

    At CES 2026, NVIDIA announced DLSS 4.5, which introduces Dynamic Multi-Frame Generation, a new 6× Multi-Frame Generation mode, and a second-generation Transformer model applied to DLSS Super Resolution. The new DLSS Super Resolution addresses issues related to temporal anti-aliasing (TAA) and earlier super-resolution models in areas such as lighting, detail, and shadows. NVIDIA claims it improves image quality across all presets (including Quality and Balanced modes), with the most noticeable gains in Performance and Ultra Performance modes. All GeForce RTX users can try it by updating to the beta version of the NVIDIA App. DLSS Dynamic Multi-Frame Generation and the 6× Multi-Frame Generation mode will be available to RTX 50 series gamers in the full release scheduled for January 13. Source

    Comparison of edge detail with DLSS 4.5 Super Resolution

    The RTX Remix Logic update for classic games is expected to roll out later this month via the NVIDIA App. Building on existing RTX Remix features, the Remix Logic update introduces a new visual logic system that allows RTX Remix mods to respond visually to real-time in-game events. It provides mod developers with more than 900 configurable options, enabling dynamic effects triggered by various in-game events, and supports 165 classic games. Source

    Game scenes remade using RTX Remix Logic

    NVIDIA also announced four G-SYNC Pulsar monitors developed in collaboration with Acer, AOC, ASUS, and MSI. These displays feature variable refresh rate backlight strobing technology, delivering over 1000 Hz of effective motion clarity to further enhance motion sharpness and visibility. They also include built-in light sensors that intelligently adjust color temperature and brightness. Source

    The first four G-SYNC Pulsar monitors to launch

    NVIDIA ACE AI technology is expanding further. Total War: PHARAOH introduces a dynamic AI advisor powered by NVIDIA ACE to help players learn the game’s many systems and mechanics. Meanwhile, the AI teammate PUBG Ally in PUBG: BATTLEGROUNDS now supports long-term memory, allowing it to remember players’ past performance and in-game interactions and reference previous events in its responses. PUBG Ally will enter testing for PUBG: BATTLEGROUNDS Arcade users in the first half of this year, with support for English, Korean, and Chinese.

    NVIDIA also announced a series of AI upgrades for RTX GPUs and laptops, including native NVFP4 / FP8 precision support in ComfyUI; integration of RTX Video Super Resolution to accelerate 4K video generation; NVIDIA NVFP8 quantization for the LTX-2 audio-video generation model; AI Blueprints for generating 4K AI video from 3D scenes in Blender; and accelerated inference performance for small language models such as Ollama and llama.cpp. Source

    For developers and enterprise users, NVIDIA showcased the DGX Spark desktop AI supercomputer based on the NVIDIA Grace Blackwell architecture, featuring an ultra-cooperative design. The Rubin AI platform—an integrated six-chip system—was announced to have entered full mass production. NVIDIA also fully opened access to a suite of open models trained on its own supercomputers, including Clara for healthcare, Earth-2 for climate science, Nemotron for reasoning and multimodal AI, Cosmos for robotics and simulation, GR00T for embodied intelligence, and Alpamayo for assisted driving. Related physical AI open-source models, frameworks, infrastructure, and multiple robots also made their debut at CES. Source

    NVIDIA Rubin

    Dell launches XPS 14/16 laptops

    After nearly a year of retiring the XPS brand name, Dell unveiled the XPS 14 and XPS 16 laptops at CES 2026, targeting “enthusiasts who pursue ultimate performance.”

    The XPS 14/16 feature an all-aluminum chassis and an ultra-thin design, equipped with three USB-C/Thunderbolt 4 ports and 10W speakers. The 14-inch model weighs 1.36 kg, while the 16-inch version weighs 1.65 kg. Base configurations include a 1920×1200 IPS display, 16 GB LPDDR5X memory, 512 GB SSD storage, a 70 Wh battery, and an Intel Core Ultra 5 325 processor. The display supports a 1–120 Hz refresh rate and VRR technology. Both the XPS 14 and XPS 16 can be configured with lighter, high-resolution OLED display options, and higher-end versions powered by the Core Ultra X9 388H processor will be available later.

    Selected configurations of the XPS 14/16 went on sale on January 6, with the XPS 14 starting at $2,049 and the XPS 16 starting at $2,199.99. Entry-level and high-end configurations are expected to launch in February. In addition, an XPS 14 version preinstalled with Ubuntu 24.04 is scheduled to arrive later this year. Source

    HP introduces the keyboard-style mini PC Eliteboard G1a

    At CES 2026, HP unveiled the Eliteboard G1a mini PC, which adopts the appearance of a standard office keyboard. The device is just 12 mm thick, with the processor, memory, storage, speakers, cooling fan, and multiple ports all integrated inside the keyboard. Users simply need to connect the keyboard to a display and can then use it with the included Bluetooth mouse.

    Eliteboard G1a

    The Eliteboard G1a is available in two versions: one with a fixed USB-C cable and another with a detachable USB-C cable. It supports daisy-chaining up to two 4K displays. Inside, it is powered by AMD Ryzen AI 300 series processors, with configuration options ranging from AI 5 350 to AI 7 370 Pro. Memory support goes up to 64 GB of DDR5 via dual SODIMM slots, and storage capacity reaches up to 2 TB. The Eliteboard G1a also offers an optional fingerprint recognition module and includes a replaceable 35 Wh battery, with HP estimating around 3.5 hours of battery life. In terms of connectivity, the detachable-cable version caters more to conventional peripheral usage, while the fixed-cable version retains an additional USB4 port for external devices.

    Finally, the keyboard itself uses a membrane switch structure with 2 mm key travel, and the entire unit weighs just 750 grams. The Eliteboard G1a is expected to go on sale in March this year, with pricing to be announced closer to launch. Source

    Intel releases Core Ultra 300 series AI processors

    At CES 2026, Intel announced the Core Ultra 300 series processors, codenamed Intel Panther Lake. Built on Intel’s 18A (2nm) process, the high-end models in this lineup feature up to 16 CPU cores, 12 Xe cores, and an NPU delivering up to 50 TOPS. According to Intel, this results in up to a 60% boost in multi-threaded performance, a 77% increase in gaming performance, and battery life extended to as much as 27 hours.

    Full SKU lineup of the Core Ultra 300 series

    In addition, Intel revealed edge processors designed specifically for embedded and industrial environments. These processors offer wide-temperature support, enabling more stable performance in both higher and lower temperature conditions. They also deliver significant advantages in key AI workloads, including up to a 1.9× improvement in large language model performance, a 2.3× increase in video analytics efficiency, and a 4.5× boost in vision–language–action model throughput.

    Laptops powered by the Core Ultra 300 series began pre-orders as early as January 6, with global availability starting on January 27. Additional laptop models are expected to roll out gradually in the first half of 2026. Edge computing systems equipped with the third-generation processors are scheduled to launch in Q2 2026. Source

    LEGO launches smart bricks

    At CES 2026, LEGO unveiled its Smart Bricks. In appearance, the Smart Brick is indistinguishable from a classic 2×4 LEGO brick, but inside it houses a miniature computer. Unlike previous LEGO Mario components that relied on external batteries, the Smart Brick uses a custom ASIC chip, allowing for a compact design with support for wireless charging. The Smart Brick is equipped with inertial sensors, light sensors, and an NFC reader, enabling it to detect movement and tilt gestures, and to recognize nearby new light panels or minifigures embedded with Smart Tags. It can form a Bluetooth mesh network, allowing bricks to sense each other’s position and orientation.

    LEGO Smart Brick

    According to LEGO’s demonstrations, when a player moves a LEGO spaceship, the Smart Brick emits real-time engine sounds; if a vehicle flips over, the audio switches to a crash sound; and when Emperor Palpatine sits on his throne, the system automatically plays the “Imperial March.” LEGO said the Smart Brick supports firmware updates via a smartphone app and that its capabilities will continue to expand in the future. This technology is expected to inspire more creative combinations—for example, using a duck-quacking tag on a helicopter model to create a playful “duck helicopter.”

    The LEGO Smart Brick will officially go on sale on March 1, 2026. The first wave of Smart Brick sets will be led by the Star Wars series and will include three kits: Darth Vader’s TIE Fighter priced at $70 (473 pieces), Luke’s X-Wing Fighter priced at $100 (584 pieces), and the Death Star Duel scene priced at $160 (962 pieces). Each set includes at least one Smart Brick and a corresponding Smart minifigure. It is worth noting that due to the hardware cost of the Smart Brick, these sets offer fewer pieces and smaller model sizes than traditional sets at the same price point. Source

    Rumors You Can Just Glance At

    • At CES 2026, Samsung showcased an OLED foldable display said to have no visible crease. The panel uses a laser-perforated metal plate that disperses pressure during folding, resulting in a crease-free visual appearance. The technology is expected to appear in the Galaxy Z Fold 8 in the second half of 2026, or in Apple’s foldable device. Source
    • iPolish demonstrated a press-on nail patch at CES 2026 that can change color through electrochemical methods. Source
    • On January 5, SanDisk announced the launch of a new gaming SSD brand, SANDISK Optimus, intended to replace the WD_BLACK and WD Blue lines inherited from Western Digital prior to the split. Source
  • SSPAI Review | Apps Worth Watching Recently

    SSPAI Review | Apps Worth Watching Recently

    Welcome to this episode of Pai Review. You can use the table of contents to quickly jump to the sections you’re interested in. If you’ve discovered other apps worth checking out, or topics you’d like us to cover, feel free to join the discussion in the comments.

    New Apps Worth Paying Attention To

    Although SSPAI has long been dedicated to discovering and introducing quality apps across platforms, there are still many excellent apps—impressive in design, functionality, interaction, and overall experience—that we haven’t yet covered. They might be long-standing apps, or newly released ones, and we’ll introduce them to you here.

    Bauhaus Clock: A New macOS Screensaver Option

    • Platform: macOS
    • Keywords: screensaver, clock

    @PlatyHsu: Screensavers already sound like a concept from a bygone era. Today, their role is more about visual decoration and temporary screen masking. Especially on larger external displays, a good-looking screensaver tends to draw quite a bit of attention. Perhaps because macOS already comes with a rich set of built-in screensavers—and because the screensaver development framework has long been neglected—it’s been rare in recent years to see truly good third-party screensavers (those cash-grab listings on the App Store don’t count).

    Bauhaus Clock is the first genuinely good macOS screensaver I’ve discovered in a long time. If you’ve used iOS 6, you probably still remember the Swiss railway clock in the Clock app back then. Bauhaus Clock follows a similar idea: its core function is simply to display a realistic Bauhaus-style clock on the screensaver screen.

    Displaying a clock isn’t difficult in itself, but Bauhaus Clock gets the details remarkably right. From the metallic sheen of the hands, to the shadows cast on the dial, to the recessed luminous coating on the markers, everything is rendered with great fidelity. There are three movement styles to choose from: quartz (one tick per second), mechanical (one-quarter tick per step), and digital (smooth movement). In quartz mode especially, the developer even recreated the subtle wobble of the hands as they move.

    The dial comes in 13 color themes. When switching themes, the markers, numerals, and hands all change to matching colors along with the dial. I especially recommend Glacier blue, Pistachio green, and Lavender purple—color combinations that make you want to steal them outright for a blog theme.

    At night, Bauhaus Clock automatically switches to dark mode along with the system theme. The effect resembles a wall clock after the lights go out: luminous markers glow softly with a grainy texture against a pure black background (14 colors available), faintly outlining the hands.

    The other “small touches” in Bauhaus Clock are equally well executed. For example, none of the controls in the settings panel use standard system styles; instead, they’re custom-drawn, with highlights and shadows that feel distinctly skeuomorphic. Another detail is that the clock markers subtly jitter at intervals to prevent burn-in (even though Macs don’t currently use OLED displays—perhaps this is for users with OLED external monitors, or simply preparation for future OLED MacBooks).

    If there’s one downside… anything with “Bauhaus” in its name is rarely cheap, and Bauhaus Clock is no exception. At $19, its price is already on par with many indie apps. The developer spends quite a bit of space in the FAQ explaining why it’s worth the price—no need to read it to guess the reasons: meticulous craftsmanship, free updates, never discounted, and so on. At this price, it’s hard for me to strongly recommend it. If it perfectly matches your aesthetic, or can serve as visual material (à la MKBHD), then this clock is genuinely beautiful and worth what it costs. Otherwise, saving the money might already buy you a month’s worth of emotional value from ChatGPT.

    Bevel: An All-in-One Health Assistant That Covers Both Fitness and Nutrition

    • Platform: iOS / watchOS
    • Keywords: health data, fitness training

    @Vanilla: Bevel is a health app on iOS that also offers an Apple Watch companion, enabling integration across health monitoring and workout training. Recently, Bevel made its core features free to use, keeping only Bevel Intelligence as paid, unlockable content. As a health app, Bevel offers an extremely comprehensive feature set, covering most of our needs related to fitness and overall health. Let me walk you through it.

    The Bevel app is divided into four tabs: Home, Journal, Fitness, and Vitals. As the name suggests, the Home tab serves as a dashboard, aggregating data such as stress, energy, nutrition, and various health metrics. The Journal tab allows us to manually log health-related behaviors such as caffeine intake, mood, added sugar, ketogenic diet, hydration, alcohol, device usage in bed, and pre-sleep meals. It also displays automatically recorded data, including steps, strength-training time, and cardio duration.

    The Fitness section presents workout-related data such as monthly views, activity summaries, effort output, muscle balance, and training frequency. This is also where you can add, manage, and sync strength-training modules. The Vitals tab shows metrics like VO₂ max, HRV baseline, resting heart rate baseline, body weight, and body fat percentage.

    That said, my favorite features in Bevel are tucked away behind the plus button in the bottom-right corner. Tapping it opens an add menu, where functions are mainly split into two categories: nutrition and training. For nutrition, you can log each meal in multiple ways, including text input, image import, taking photos, scanning, or direct search. For training, you can automatically generate templates based on your needs, view existing templates, or log individual activities.

    Here I want to focus on Bevel’s strength-training template feature. Whether you’re using automatically generated templates or creating your own, you can choose exercises from a large movement library and add them to your templates. Each exercise comes with animated demonstrations and detailed guides, clearly showing primary and secondary muscle groups. Once a workout starts, the Apple Watch client syncs automatically, and you can record the weight and reps for each set at any time. Doesn’t this moment remind you of the Strong app?

    As a paid feature, Bevel Intelligence can chat with you in real time, offer personalized suggestions based on your condition, and even generate nutrition and training plans directly for you, which is very convenient. If you’re looking for a health or fitness tracking app, Bevel is well worth trying—it offers most features for free and is exceptionally feature-rich.

    Bevel is available for free on the App Store. All features except the AI functions are free to use, with a subscription price of RMB 328 per year.

    Qie Ting: Luo Yonghao Launches an AI Book Explanation Tool

    • Platform: iOS
    • Keywords: audiobooks, AI narration

    @Snow: Last week, Luo Yonghao unveiled an AI audiobook app incubated by his team, called Qie Ting, at his so-called “Tech Spring Festival Gala.” Unlike traditional audiobooks that focus on vividly reading texts aloud, Qie Ting relies on an AI engine to first understand the full content and then reinterpret it, using scripts of 30,000–40,000 words or 1–2 hours of audio to help you absorb information efficiently.

    Qie Ting carries forward the refined, light-skeuomorphic design language of Smartisan OS. Its main bookshelf screen currently includes content from more than 10,000 books. If you’re not sure where to start, you can narrow things down using the category buttons at the top, or jump to the rankings page and explore included titles from lists such as Douban Hot Picks or recommendations from The Beijing News.

    As mentioned at the beginning, Qie Ting’s core feature is not “audiobooks” in the sense of word-by-word narration, but rather book “explanations.” By tapping the “Play Book Explanation” button, you can access key points distilled by the AI engine, along with supplementary background and extended information. It’s essentially the book equivalent of film or TV explainers: the AI creator helps condense the core storyline, organize the structure and logic, cite references to broaden context, and even convey its own interpretations and takeaways. Because this kind of secondary creation is constrained in length, it can only present certain facets of the original work and cannot fully reproduce it. For this reason, I don’t think Qie Ting can completely replace reading the original book. I’d recommend using it more as a way to filter interests or as a supplementary follow-up. Qie Ting currently does not support directly reading e-books; if an explanation sparks your interest, you can tap the in-app purchase link to buy the corresponding physical book via JD.com.

    Each explanation in Qie Ting is available in both text and audio formats. You can choose to read the text in full screen, or have it read aloud by an AI voice. The AI narration supports basic features such as fast forward and rewind, playback speed adjustment, sleep timers, shuffle, and continuous playback. You can choose from four voice models: Xiao Wen (male), Xiao Su (female), and their respective “Late Night” versions. Compared with the standard voices, the late-night versions are deeper and more soothing, making them especially suitable for listening before sleep. You can also optionally add background music to the AI narration to further enhance the listening atmosphere. Unfortunately, the celebrity voice models and timbre-cloning features demonstrated at the tech event have not yet been rolled out in this version, so overall vocal expressiveness falls slightly short of expectations.

    In addition, Qie Ting currently does not support importing or analyzing local books. Given that the existing library is still relatively limited in scope, it may not cover everyone’s reading needs. That said, considering its relatively low paywall—free users have no limits on reading volume, with differences mainly in audio quality and ads—it works well as a supplementary reading tool.

    You can download Qie Ting for free from the App Store.

    FunKey: Bringing Realistic Mechanical Keyboard Typing Sounds to the Mac

    • Platform: macOS
    • Keywords: typing sounds

    @化学心情下2: Beyond the clear sense of rhythm that provides positive feedback while typing, the mechanical sounds produced by a mechanical keyboard can also bring auditory pleasure to the act of input—whether your colleagues or roommates sharing the space find it pleasurable is perhaps another matter. On Mac keyboards, the relatively “short” key travel inevitably makes some users who are already accustomed to mechanical keyboards feel that the experience lacks a certain “punch.” Fortunately, a developer has finally noticed this “problem” and released a mechanical keyboard sound effect app for Mac users: FunKey.

    Although it may sound a bit like an absurd case of overengineering, after actually buying and using it, I found FunKey to be quite entertaining. The app doesn’t just offer mechanical keyboard sound effects—it even includes mouse click sounds. You could say that if you can think of it, FunKey probably provides it.

    In FunKey, there are more than a dozen mouse sound effects alone, which you can freely choose from according to your preferences. As for mechanical keyboard sounds, FunKey also offers a wide variety, with the settings menu describing the general tonal characteristics of each option. There are simply too many variations to describe one by one, so if you’re interested, it’s best to try them out yourself.

    You can download FunKey from the Mac App Store. It is priced at RMB 22.

    Anx Reader: A Cross-Platform Reader Integrated with Large AI Models

    • Platform: iOS / iPadOS / Android / Windows / macOS / HarmonyOS
    • Keywords: reading, AI translation

    @大大大K: I’ve recently been reading some English novels, and quite often I still need to rely on Chinese translations for reference. Manually creating bilingual files is too much trouble, while the built-in translation tools in most reading apps feel overly mechanical and fail to capture the “spirit” of literary works. I later stumbled upon Anx Reader, an open-source and free app that supports integration with large AI models. It not only translates quickly, but also enables a range of Q&A features.

    Anx Reader supports true cross-platform use and runs on Windows, Android, iOS, macOS, and HarmonyOS. The app mainly adopts a Material 3 design style, with some controls on iOS featuring a liquid-glass-like effect. Anx Reader also offers dedicated optimizations for E-Ink devices: when E-Ink mode is enabled, the app switches to black text on a white background and adapts to the refresh behavior of e-ink screens.

    When it comes to reading features, Anx Reader’s standout strength lies in its AI support. There is a dedicated AI tab in the app’s navigation bar; tapping it opens a conversational mode where you can ask the AI to summarize, explain, or analyze a specific book, or to provide summaries and suggestions based on your overall reading activity within the app. You can also dive into the AI settings to define custom prompts for specific conversation types, enabling more targeted and relevant responses.

    During reading, you can tap the AI button in the top-right corner to have the AI generate a mind map and reading guide for the book. You can also use word selection search or full-text translation to assist with reading foreign-language materials. The AI-generated translations are more natural than those from typical translation tools, and they do a better job of conveying the colloquial expressions often found in novels.

    In addition, Anx Reader supports data synchronization via WebDAV, delivering a more seamless cross-platform experience. However, during synchronization, Anx Reader may upload your book files. If you’re using services that charge based on data usage, such as Nutstore, you’ll want to keep an eye on your bandwidth consumption.

    Anx Reader is currently available for free download from its official website and GitHub. Note that iOS users need to pay when downloading via the App Store, though the developer encourages users to install it for free on iOS devices through self-signing.

    App Updates You Shouldn’t Miss

    Beyond “new” apps, many familiar names on the App Store continue to iterate and roll out updates, adding more interesting and practical features. SSPAI aims to help you filter through App Store updates worth paying attention to, so you can quickly catch up on the latest developments from apps and their developers.

    Life Footprints 2: Annual Summary and 3D Globe Mode Added

    • Platform: iOS
    • Keywords: location tracking, annual summary

    @ElijahLee: The location-tracking app Life Footprints has released a major 2.0 update, introducing a 2025 annual summary, a new 3D globe mode in the trajectory overview, an ink-black map style, and custom time-range viewing.

    First up is the 2025 annual summary, which appears on the map on the app’s home screen. The summary consists of 10 pages, analyzing the 10 cities you visited in 2025, the total number of location points recorded throughout the year, the average number of points recorded per day, as well as various daily habits. At the end, you can generate a concise annual recap to save or share on social media. If you don’t see the annual summary entry in the app, you can tap the globe icon in the bottom-right corner to reprocess the data.

    In the trajectory overview, the new version adds a 3D globe mode that displays your global footprints on a three-dimensional Earth. Notably, this is a real-time, data-rendered 3D globe that you can freely rotate to explore. It also includes small delightful touches such as stars, airplanes, and meteors. Tapping the globe lets you enter full-screen mode, where a lively BGM soundtrack plays automatically. For users who travel across countries or continents, this feature is particularly appealing—your footprints are no longer flat data points, but transformed into a kind of spatial memory.

    The update also adds an ink-black map style, which—along with the turquoise green option introduced in the previous version—looks especially refined. Custom time-range viewing is available on the home screen: you can view trajectory statistics by default ranges such as a single day, a month, or a full year, or manually select start and end times to generate a trajectory map for a specific period. This feature is ideal for recording life stages such as travel or business trips, slicing your footprint data into meaningful time segments for easier reflection later on.

    You can download Life Footprints for free from the App Store. Most features are available at no cost, while a subscription unlocks exclusive profiles, trajectory analysis, sports modes, and more. The subscription fee is RMB 12 per year.

  • SSPAI Morning Brief: Clicks launches the Power Keyboard and Communicator phone

    SSPAI Morning Brief: Clicks launches the Power Keyboard and Communicator phone

    Morning Brief

    1. Clicks launches the Power Keyboard and Communicator phone
    2. Pebble releases the Round 2 watch
    3. DeepSeek publishes a paper introducing the efficient training technique mHC
    4. ASUS notifies channels of price increases starting today
    5. A new batch of works enters the public domain
    6. BYD surpasses Tesla to become the world’s top EV seller in 2025
    7. Rumors You Can Just Glance At

    Clicks launches the Power Keyboard and Communicator phone

    On January 2, tech startup Clicks Technology unveiled its first batch of new products following its physical keyboard cases.

    Among them, the Clicks Communicator phone is positioned as a “modern communication companion,” with a role similar to that of a Kindle alongside an iPad, intended primarily as a user’s secondary phone. The device runs Android, features a candybar design, and comes with a 4-inch display and a full physical keyboard. It weighs 170 grams and measures 12 mm thick. To reduce information overload, the system interface is deeply customized based on Niagara Launcher, abandoning the traditional app grid in favor of a minimalist message aggregation center. In terms of hardware, the Communicator is equipped with a 50-megapixel rear camera and a 24-megapixel front camera, while retaining a 3.5 mm headphone jack, a mute switch, and a MicroSD card slot. A signal-light button on the side changes color according to notification type. The device is powered by a 4000 mAh battery. Details about the processor and storage have not been disclosed. The Communicator starts at $499 (approximately RMB 3,487), with an early-bird pre-order price of $399, and is expected to ship later this year.

    The other new product, the Clicks Power Keyboard, is a two-in-one device combining a Bluetooth keyboard and a power bank. It supports MagSafe and Qi2 wireless charging standards and can magnetically attach to the back of a phone. The keyboard features a sliding design and, in addition to serving as a physical keyboard for phones, can connect via Bluetooth 5.4 to up to three devices such as tablets and smart TVs. The Power Keyboard weighs 180 grams, is 15.2 mm thick, and has a battery capacity of 2150 mAh, of which 500 mAh is allocated specifically for keyboard use. The product is priced at $109 (approximately RMB 790), with an early-bird price of $79, and will officially go on sale this spring.

    Pebble releases the Round 2 watch

    On January 2, smartwatch brand Pebble announced the Pebble Round 2. As part of the brand’s reboot plan, the Round 2 is a revival of the Pebble Time Round released in 2015, aiming to recreate the classic circular design that was once known as the “thinnest in the industry,” now updated with more modern technology.

    In terms of hardware design, the Pebble Round 2 addresses one of the biggest drawbacks of the original model—its overly wide bezels. While retaining an ultra-thin 8.1 mm stainless steel case, the new model features a 1.3-inch color e-ink display with a pixel density doubled from the previous generation to 283 ppi, along with backlight support. The Round 2 keeps physical side buttons, allowing users to rely on tactile, eyes-free operation in scenarios such as meetings.

    To achieve a battery life of 10 to 14 days, the Pebble Round 2 makes clear trade-offs in functionality. It retains only basic health features such as step counting and sleep tracking, and does not include a heart rate sensor, making it unsuitable as a dedicated fitness-tracking device. On the software side, it runs the open-source Pebble OS and is compatible with thousands of watch faces and apps. A built-in microphone supports voice input and replies, but due to limitations imposed by Apple’s system, this feature currently mainly targets Android users, with iOS support set to roll out first in the EU. Previously, Pebble released a low-cost AI smart ring with recording and transcription capabilities, and the company has stated that it plans to bring similar AI features to its watches in the future.

    The product is priced at $199 (approximately RMB 1,390) and is now available for pre-order on the official website, with shipments expected in May this year. For customers who previously pre-ordered the square-faced Pebble Time 2, the company is also offering the option to switch their reservation to the Round 2 while retaining their place in the queue.

    DeepSeek publishes a paper introducing the efficient training technique mHC

    On January 1, DeepSeek released a technical paper co-authored by its founder Liang Wenfeng, proposing a new deep learning architecture called “Manifold-Constrained Hyper-Connections” (mHC). The research aims to optimize computational efficiency, enabling larger-scale models to be trained at lower cost in environments with limited computing resources.

    The mHC technique is a further refinement of the “Hyper-Connections” architecture. Hyper-Connections were originally proposed by researchers at ByteDance in September 2024, with the goal of improving upon the mainstream ResNet (Residual Network) architecture invented by He Kaiming and others at Microsoft Research Asia. The approach was designed to address signal attenuation in deep networks—where information gradually weakens as it propagates through layers, preventing deeper layers from receiving effective signals—as well as model collapse, in which different inputs become increasingly similar or even converge after successive layers of processing.

    The DeepSeek team points out that while ByteDance’s approach increased network complexity, it overlooked the rapidly rising memory costs in large-model training, resulting in limited real-world scalability. In response, DeepSeek’s mHC approach introduces specific manifold constraints that force data flows to operate only along certain geometric trajectories (manifolds). This preserves the advantages of hyper-connections while successfully addressing memory and cost bottlenecks, achieving what the paper describes as “almost negligible computational overhead.” According to the paper, the DeepSeek research team conducted experiments on models with 3 billion, 9 billion, and 27 billion parameters, and the empirical results show that mHC offers better scalability than traditional architectures while maintaining stability during large-scale training.

    The paper was uploaded to arXiv under Liang Wenfeng’s personal account. Previously, Liang had personally published only DeepSeek’s most important technical papers, such as those related to the R1 and V3 models, and in this work he is listed as the final author. This move confirms that he remains deeply involved in core research and development, and is widely seen as a signal that the technical direction for DeepSeek’s next-generation core models has been set. Market expectations suggest that DeepSeek may follow the R1 release strategy and unveil a new large model before the Lunar New Year holiday in mid-February.

    ASUS notifies channels of price increases starting today

    According to VideoCardz, ASUS sent a “2026 Product Price Adjustment Notice” to its partners on December 30 last year, announcing that price adjustments will take effect starting January 5, 2026.

    The notice states that due to structural volatility in the global supply chain, multiple key components are facing severe cost pressures, particularly DRAM and NAND memory. These changes stem from “adjustments in global foundry capacity allocation,” “rising investment costs in advanced process nodes,” and “structural gaps in the industry caused by demand for AI computing power.”

    Although ASUS did not specify which models would be affected, it clearly stated that it would implement “strategic price adjustments for certain product portfolios.” As the effective date of the price increase comes just ahead of the opening of CES 2026, industry analysts believe that ASUS’s upcoming next-generation hardware—especially AI PCs and gaming devices expected to be unveiled at CES—will be the first to be impacted, potentially setting new market price benchmarks.

    ASUS emphasized that the adjustment is a “necessary decision made after long-term absorption of and responses to cost pressures,” with the goal of “ensuring stable supply and maintaining quality and service standards.” Regarding subsequent impacts, ASUS business representatives will proactively contact partners to provide detailed explanations and assist in planning the most appropriate response strategies or configuration recommendations.

    ASUS’s public relations department has confirmed the authenticity of the document, while clarifying to the media that it is an internal commercial communication intended solely for channel partners, rather than an official press release issued to the public.

    A new batch of works enters the public domain

    On January 1, 2026, the world marked its annual milestone for copyright expiration. Works by renowned figures such as Freud, Einstein, and Toynbee entered the public domain across jurisdictions including the United States, Europe, and China.

    In the United States, with the expiration of copyright protection for works published in 1930, books such as William Faulkner’s As I Lay Dying and Sigmund Freud’s Civilization and Its Discontents, as well as films like All Quiet on the Western Front, entered the public domain. Paintings by Piet Mondrian and Paul Klee, the original “Betty Boop” doll, and Disney’s earliest depiction of Pluto (then known as Rover) are also no longer subject to copyright restrictions, making derivative creations based on these early materials possible.

    In jurisdictions that apply a copyright term of “the author’s lifetime plus 70 years,” such as the European Union, the United Kingdom, and Russia, works by authors who died in 1955 have now reached the end of their protection period. This means that works by physicist Albert Einstein, Dale Carnegie—the author of How to Win Friends and Influence People—and Nobel Prize–winning writer Thomas Mann are no longer protected by copyright in these regions.

    In jurisdictions that apply a copyright term of “the author’s lifetime plus 50 years,” including China and most parts of Asia and Africa, works by authors who died in 1975 have reached the end of their protection period. This includes figures such as Chiang Kai-shek, historian Arnold J. Toynbee, Soviet composer Dmitri Shostakovich, and political philosopher Hannah Arendt.

    BYD surpasses Tesla to become the world’s top EV seller in 2025

    According to The New York Times, Tesla lost its position as the world’s largest electric vehicle manufacturer in 2025, being surpassed by BYD for the first time. Data released on December 31 show that Tesla’s annual deliveries fell 9% year on year to 1.64 million vehicles, with fourth-quarter sales plunging 16% due to policy impacts. In contrast, BYD’s annual sales of pure electric vehicles rose 28% to 2.26 million units, and the company achieved significant expansion across Asia, Europe, and Latin America thanks to its pricing advantage.

    Tesla’s sales decline was primarily driven by a sharp shift in U.S. policy. Despite Elon Musk’s strong support for Donald Trump in the 2024 election, after Republicans regained control of the White House and Congress, they swiftly repealed the federal EV tax credit of up to $7,500 and began rolling back clean air regulations. As the dominant player holding about 45% of the U.S. EV market, Tesla became the biggest casualty of these policy changes, with fourth-quarter deliveries dropping sharply from 496,000 a year earlier to 418,000 units.

    Beyond the deteriorating external environment, Tesla’s own lagging product updates were also a key factor. Its flagship Model Y has seen no major refresh since its launch in 2020, while the newly introduced Cybertruck has delivered disappointing sales. In the European market, Tesla’s sales have already been overtaken by Volkswagen. In addition, Musk has shifted the company’s focus toward robotaxis and humanoid robots, which have yet to generate meaningful revenue, and Tesla has fallen behind Alphabet-owned Waymo in the deployment of autonomous driving.

    Industry analysts predict that the U.S. EV market will remain sluggish in 2026, with a recovery unlikely until 2027, when more low-cost models priced below $30,000 enter the market. Despite weak vehicle sales, Tesla’s stock price remains near historical highs, as Wall Street continues to bet on its long-term potential in autonomous driving. Meanwhile, although BYD is unable to enter the U.S. market due to high tariff barriers, it has firmly established its leadership in the rest of the world.

    Rumors You Can Just Glance At

    • Recently, multiple Reddit users complained that Duolingo was displaying subscription ads through its real-time events feature, a practice prohibited by App Store review guidelines and one that could potentially lead to removal from the store. Duolingo later appeared to have stopped showing the ads.
    • According to official data from Stack Overflow, the site added only 3,862 new questions in December 2025, a figure that has fallen below the level seen in August 2008, shortly after the platform’s launch.
    • Internal Meta documents reviewed by Reuters show that in response to pressure from global regulators cracking down on scam ads on social media, Meta developed a set of countermeasures known as the “Global Playbook.” The documents reveal that rather than fully verifying advertisers’ identities or eliminating scams, Meta made it harder for regulators to discover violating ads by manipulating search results in order to pass compliance reviews. For example, in Japan, Meta teams found that regulators searched public ad libraries using specific keywords such as celebrity names to identify scam ads. Ahead of regulatory reviews, Meta selectively cleaned up search results related to those keywords, creating the impression that scam ads had been significantly reduced. In addition, when regulation was tightened in one region, algorithms would automatically divert scam ad traffic to other regions with looser oversight. Previously, Meta had faced tough questioning from governments in Japan, Singapore, and elsewhere over the proliferation of investment scams and AI-generated fake celebrity endorsement ads on Facebook and Instagram.
    • On the evening of January 3, Lei Jun revealed during a New Year livestream that Xiaomi’s original vehicle delivery target for 2025 had been set at 300,000 units, later raised to 350,000, and ultimately surpassed 410,000 units. He also announced that Xiaomi’s vehicle delivery target for 2026 is 550,000 units. During the four-hour livestream, Lei Jun and his engineering team completely disassembled a YU7. After the teardown, Lei Jun addressed a series of recent public controversies surrounding Xiaomi Auto. Regarding marketing claims such as “1,300 kilometers on a single charge” and “instant braking from 200 km/h,” Lei Jun said these statements originated from driving demonstration videos meant to showcase the SU7’s range and braking performance, with full videos available as evidence, but were taken out of context during dissemination. He also acknowledged that “fine-print marketing” is a bad industry practice, accepted the criticism, and said that since November last year he has required business teams to use larger, clearer text whenever possible.
  • SSPAI Morning Brief: Samsung Unveils Galaxy A17, NVIDIA completes a $5 billion investment in Intel, and More

    SSPAI Morning Brief: Samsung Unveils Galaxy A17, NVIDIA completes a $5 billion investment in Intel, and More

    Morning Brief

    1. The 2026 “National Subsidy” plan is unveiled
    2. Luo Yonghao hosts an annual technology innovation sharing event
    3. 2026 GAC Trumpchi M8 “Xiangwang” officially launched
    4. Xiaomi responds to reports of a loose Leica master zoom ring on the Xiaomi 17 Ultra
    5. Samsung releases the Galaxy A17
    6. TCL Note A1 NXTPAPER announced
    7. NVIDIA completes a $5 billion investment in Intel
    8. Rumors You Can Just Glance At

    The 2026 “National Subsidy” plan is unveiled

    On December 30, the National Development and Reform Commission (NDRC) and the Ministry of Finance released a notice on the implementation of large-scale equipment upgrades and consumer goods trade-in programs in 2026. The notice includes the following measures:

    • Support for home appliance trade-in programs. Individual consumers who purchase products meeting Level 1 energy efficiency or water efficiency standards in six categories—refrigerators, washing machines, televisions, air conditioners, computers, and water heaters—will receive a subsidy equal to 15% of the product’s sale price. Each consumer may receive a subsidy for one item per category, with a maximum subsidy of 1,500 yuan per item.
    • Support for new purchases of digital and smart products. Individual consumers who purchase products in four categories—smartphones, tablets, smartwatches or fitness bands, and smart glasses—will receive a subsidy equal to 15% of the product’s sale price. Each consumer may receive a subsidy for one item per category, with a maximum subsidy of 500 yuan per item.
    • Support for subsidies on new smart home product purchases. Specific subsidized categories and subsidy standards will be independently and reasonably determined by local governments based on local conditions.

    The 2026 “National Subsidy” plan further optimizes the scope of support. On the basis of continuing overall equipment upgrade policies, it expands to include areas such as adding elevators to old residential communities, elderly care institutions, firefighting and rescue, inspection and testing, and offline commercial consumption facilities. In terms of consumer goods trade-ins, the plan supports vehicle trade-ins, while home appliance trade-ins focus on six categories: refrigerators, washing machines, televisions, air conditioners, computers, and water heaters. It also supports new purchases of digital and smart products, including smartphones, tablets, smartwatches (bands), smart glasses, and smart home products (including age-friendly home products).

    The state has recently allocated in advance the first batch of 62.5 billion yuan in ultra-long-term special treasury bonds for 2026 to local governments, earmarked to support consumer goods trade-in programs and meet peak consumption demand during periods such as New Year’s Day and the Spring Festival. Source

    Luo Yonghao hosts an annual technology innovation sharing event

    On the evening of December 30, Luo Yonghao held the “Luo Yonghao’s Crossroads” Annual Tech Innovation Sharing Conference at the Shanghai West Bund International Convention and Exhibition Center. The event is widely seen as the return of his signature “Tech Spring Festival Gala” after a seven-year hiatus. At the conference, Luo introduced a range of innovative tech products, including the DJI Neo 2, Jike exoskeleton robot, Bambu Lab desktop 3D printer, Yingling panoramic drone, Qingxian ergonomic chair, and Like dishwasher, and also presented awards to outstanding video creators on the Douyin Featured platform. Source

    Image from the news source

    It is also worth noting that Thin Red Line’s app Qie Ting officially debuted at the event. Positioned as “an AI-era audio library for in-depth interpretation and analysis of great books,” Qie Ting is now available for download on major app stores. Source

    2026 GAC Trumpchi M8 “Xiangwang” officially launched

    On December 30, the 2026 GAC Trumpchi Xiangwang M8 was officially launched. A total of three variants are available, with official guide prices ranging from 309,900 to 389,900 yuan.

    The 2026 Xiangwang M8 features the brand’s “Yipin Qilin front fascia” design and comes equipped with electric sliding doors, a floating roof, and hidden B-, C-, and D-pillars. It also offers features such as a 192-line LiDAR, intelligent welcome projection, and more. Inside, the cabin uses soft-touch materials and wood-grain trim, and is fitted with a 12.3-inch dual-screen setup, the ADiGO intelligent system, and the HarmonySpace 5 HarmonyOS cockpit, along with a rear-seat entertainment screen. In terms of comfort, the second row offers five-star SPA dual zero-gravity seats, while the third-row power seats support freely adjustable backrest angles.

    For power, the 2026 Xiangwang M8 is equipped with a 2.0T plug-in hybrid system, delivering a combined maximum output of 300 kW and a combined peak torque of 645 Nm. It offers a CLTC-rated pure electric range of 248 km and a WLTC-rated combined driving range of 1,177 km. Source

    Xiaomi responds to reports of a loose Leica master zoom ring on the Xiaomi 17 Ultra

    On December 30, Xiaomi’s official smartphone Weibo account published “Xiaomi 17 Ultra Series: Answers to Questions from Users,” responding to recent online reports claiming that the “Master Zoom Ring on the Xiaomi 17 Ultra Leica Edition is loose.”

    According to the statement, the Master Zoom Ring on the Xiaomi 17 Ultra Leica Edition is designed with allowable movement in both the radial and axial directions. This is intended to prevent deformation of the phone’s outer frame in the event of a drop, as well as to avoid jamming caused by dust or liquid ingress. As a result, slight movement may be felt when shaking the Master Zoom Ring, and a faint sound may be heard when tapping it—both are considered normal phenomena. These do not affect the functionality of the Master Zoom Ring, nor do they compromise the device’s dust and water resistance. The statement also addressed topics such as optical zoom focal length selection, the so-called “clicking” sound from the lens, and product availability issues. Source

    Samsung releases the Galaxy A17

    On December 30, Samsung unveiled a new A-series smartphone, the Galaxy A17. It is powered by the Exynos 1330 processor, paired with 4GB of RAM and 128GB of storage, with support for microSD card expansion. The device is rated IP54 for dust and water resistance, and features a 6.7-inch 90Hz FHD display, a 5,000mAh battery, and 25W wired charging. In terms of imaging, it comes with a triple rear camera setup consisting of a 50MP main camera, a 5MP ultra-wide camera, and a 2MP macro camera, along with a 13MP front-facing selfie camera.

    Samsung promises six years of Android OS version updates and security patches for the Galaxy A17. It will be available in blue and black color options, and is set to go on sale first in the United States on January 7, 2026, with a price tag of $199. Source

    TCL Note A1 NXTPAPER announced

    On December 30, TCL unveiled the TCL Note A1 NXTPAPER, an e-reading device based on its NXTPAPER display technology. It features an 11.5-inch NXTPAPER Pure screen using an LCD panel, with a resolution of 2200 × 1440 and a 120Hz refresh rate. The device is equipped with an 8,000mAh battery, 256GB of storage, and support for TCL’s own T-Pen Pro stylus. It also includes eight built-in microphones, as well as a 13MP rear camera dedicated to document scanning.

    Unlike previous NXTPAPER tablets, the TCL Note A1 NXTPAPER is positioned as a note-taking and productivity assistant. It runs the Android operating system and supports AI-powered real-time translation and handwriting beautification. The device is currently live on Kickstarter and is scheduled to officially go on sale by the end of February 2026, priced at $549. Source

    NVIDIA completes a $5 billion investment in Intel

    According to regulatory filings disclosed on December 29, NVIDIA completed a $5 billion investment in Intel on December 26 by acquiring 214.7 million Intel shares through a private placement at a price of $23.28 per share.

    The transaction was first announced in September and is regarded as one of the most significant strategic shifts in the semiconductor industry in 2025. As a result, it drew the attention of the U.S. Federal Trade Commission, with the relevant approvals ultimately granted on December 18.

    In addition, under the agreement, Intel will provide NVIDIA with customized x86 CPUs for its AI infrastructure platform, and the two companies will also develop a new type of processor featuring integrated NVIDIA RTX graphics and NVLink technology, with transmission bandwidth reaching up to 1.8 TB per second. Source

    Rumors You Can Just Glance At

    According to a report from Calcalist, NVIDIA is planning to acquire the Israeli AI startup AI21 Labs for between $2 billion and $3 billion. Source

    Well-known leaker @OnLeaks released a hands-on video of a Galaxy S26 Ultra dummy unit. The video shows that, compared with the S25 Ultra, the overall design has not changed much, though the rear camera module has been slightly redesigned and appears to protrude more noticeably.

    Marvel Entertainment’s official YouTube channel has released the second teaser trailer for Avengers: Doomsday. Similar to the previous teaser, this one focuses on Thor and his adopted daughter Love from Thor: Love and Thunder.

  • SSPAI Morning Brief: Digital RMB to Enter the Era of Digital Deposit Currency, and More

    SSPAI Morning Brief: Digital RMB to Enter the Era of Digital Deposit Currency, and More

    Morning Brief

    1. The digital renminbi will enter the era of digital deposit currency
    2. Meta acquires Manus
    3. GOG co-founder acquires the GOG platform
    4. More than 20% of videos recommended by YouTube’s algorithm to new users are AI-generated junk
    5. ACCA cancels online exams to combat cheating with AI tools
    6. Fujifilm to begin shipping 40 TB magnetic tapes next January
    7. LG releases UltraGear evo monitors
    8. Rumors You Can Just Glance At

    The digital renminbi will enter the era of digital deposit currency

    On December 29, the People’s Bank of China released the Action Plan on Further Strengthening the Management and Service System of the Digital RMB and the Construction of Related Financial Infrastructure. At the institutional level, the plan clarifies that the digital RMB will move from the era of digital cash into the era of digital deposit currency. A new-generation measurement framework, management system, operating mechanism, and ecosystem for the digital RMB will officially come into effect on January 1, 2026.

    Under the new institutional arrangements, the digital RMB held by customers in commercial bank wallets will constitute account-based liabilities of commercial banks, marking the transition of the digital RMB from a cash-based 1.0 version to a deposit-currency-based 2.0 version. Banking institutions will pay interest on the balances of real-name digital RMB wallets in accordance with self-regulatory agreements on deposit interest rate pricing; banks will be able to independently conduct asset–liability management for digital RMB wallet balances; and deposit insurance will provide the same level of statutory security protection as for traditional deposits. Source

    Meta acquires Manus

    On December 29, Meta announced the acquisition of AI startup Manus. Since releasing its general-purpose AI agent Manus earlier this year, the company has processed nearly 147 trillion tokens and created over 80 million virtual computers. Following the acquisition, some Manus team members will join Meta, while Manus will continue to operate as an independent company based in Singapore, and its services will remain available for separate purchase. Source

    GOG co-founder acquires the GOG platform

    On December 29, GOG and CD PROJEKT co-founder Michal Kiciński fully acquired the GOG platform from CD PROJEKT. GOG is a game distribution platform known for its DRM-free philosophy, founded in 2007 by Marcin Iwiński and Michal Kiciński and officially launched in 2008. GOG will continue its partnership with CD PROJEKT, and CD PROJEKT RED titles will remain available on GOG. Michal Kiciński stated that he hopes to preserve GOG’s original vision—no forced bundling, no mandatory platform lock-in—so players can truly feel ownership of their games. CD PROJEKT will shift greater focus toward creating high-quality RPG titles. Source

    More than 20% of videos recommended by YouTube’s algorithm to new users are AI-generated junk

    According to a report by The Guardian on December 27, video-editing company Kapwing analyzed the top 100 YouTube channels across various countries—around 15,000 channels in total—and found that 278 of them publish only AI-generated junk content (“AI slop”). These channels have collectively amassed 63 billion views, 221 million subscribers, and generate an estimated $117 million in annual revenue. The study also tested recommendation feeds using brand-new accounts and found that among the first 500 videos recommended by YouTube, 104 were AI junk—about one-third of which were low-quality “brainrot” content. Source

    ACCA cancels online exams to combat cheating with AI tools

    According to the Financial Times, the Association of Chartered Certified Accountants (ACCA) announced that starting in March, online exams will be canceled. Except in special circumstances, candidates will be required to take assessments in person. ACCA introduced online proctoring during the COVID-19 pandemic to accommodate candidates with special needs, but with the rapid advancement of AI tools, cheating methods have increasingly outpaced ACCA’s ability to monitor and control them, ultimately leading to this decision. In recent years, accounting industries in multiple countries have already seen cheating scandals. The Financial Times also learned through interviews that some candidates have indeed photographed exam questions and sent them to AI chatbots to cheat during exams. Source

    Fujifilm to begin shipping 40 TB magnetic tapes next January

    Fujifilm has announced that it will begin selling the FUJIFILM LTO Ultrium 10 (40TB) data tape—LTO-10 (40TB)—starting in January 2026. The storage medium can hold 40 TB of uncompressed data, with a maximum capacity of up to 100 TB when compressed. Compared with the previous-generation LTO-10 (30TB), the recommended operating temperature range of LTO-10 (40TB) has been expanded to 15°C–35°C, and it can withstand up to 80% humidity at temperatures between 15°C and 25°C. According to Fujifilm, sales of magnetic tape as a storage medium have grown by around 10% annually since 2020. Source

    LG releases UltraGear evo monitors

    On December 26, LG announced its new high-end gaming monitor lineup, UltraGear evo, with physical units set to be showcased at CES 2026. The lineup includes three models: the 39-inch 5K2K (5120×2160) OLED monitor 39GX950B, the 27-inch 5K MiniLED monitor 27GM950B, and the 52-inch 5K2K 240Hz monitor 52G930B. The 39-inch and 27-inch models feature exclusive functions such as on-monitor 5K AI upscaling, AI-based scene and audio optimization. The 27-inch model is equipped with 2,304 local dimming zones, designed to minimize haloing effects commonly seen on high-resolution MiniLED displays. Source

    Rumors You Can Just Glance At

    • On December 29, Chery responded to Yicai and denied rumors of a partnership with Dreame to launch the Jetour Zongheng G700 in the first quarter of 2026. Source
    • Samsung published a post announcing plans to integrate Google Photos into its Samsung AI TV lineup, allowing users to view their photos directly on the TV. Source
    • Lenovo released an update for the Legion Go handheld, version 1.3.4.9, enabling the device to gain the Xbox full-screen experience feature that the ROG Ally supported at launch. Source
    • SoftBank announced on December 29 that it will acquire data center investment firm DigitalBridge for $4 billion. Source
  • SSPAI Review | Apps Worth Watching Recently

    SSPAI Review | Apps Worth Watching Recently

    Welcome to this edition of Pi Review. You can use the table of contents to quickly jump to the sections you’re interested in. If you’ve discovered other interesting apps or topics you’d like us to cover, feel free to join the discussion in the comments.

    New Apps Worth Paying Attention To

    While SSPAI has long been committed to discovering and introducing high-quality apps across platforms, there are still many excellent apps—impressive in design, functionality, interaction, and overall experience—that we haven’t yet covered. Some may be long-standing apps, others newly released. We’ll be introducing them to you here.

    Powo: Hang a Photo Wall on Your iPhone Home Screen

    • Platform: iOS
    • Keywords: photo wall, widgets

    @ElijahLee: Powo lets you turn a curated collection of photos into a desktop-style photo wall and place it directly on your iPhone Home Screen, so you can revisit your favorite memories without even opening the app.

    Creating a photo wall with Powo is very straightforward. The app offers five photo wall templates to choose from. Simply go to the Photos tab, select the images you want to “hang” on the wall, and you’re done. Each photo is framed with a Polaroid-style border, creating a realistic, skeuomorphic look.

    The photo wall also comes with a wide range of decorative tools. Beyond photos, you can freely add figurines, stickers, posters, captions, holiday elements, ticket stubs, and more. If you want to add your own custom decorations, that’s easy too. The customization feature allows you to select images from your system library and instantly turn full-size photos into stickers, vinyl records, CDs, and other decorative elements, all with impressively realistic results.

    Powo also offers powerful freeform editing and layout options. You can customize the background, colors, and border styles of your photo wall—whether you’re aiming for a felt-board look to suit close relationships, or a dark-toned style better suited for photography showcases. Photos themselves come in multiple styles, with mounting options like push pins or small wooden clips. Any photo or decorative element on the wall can be tapped to adjust its size and hanging angle, making the overall effect feel even more lifelike. If you’re not satisfied with how photos overlap, the app even provides layer-order controls.

    Once your photo wall is complete, you can add it to your iPhone Home Screen via a Powo widget and enjoy your memories without opening the app. Of course, you can also export the photo wall within the app to save it locally or share it on social media.

    Powo is available for free on the App Store. Most features are usable at no cost, but the free version is limited to creating one photo wall, with up to five photos per wall, and does not support customization features. Upgrading removes these restrictions, with pricing at CNY 3 per month, CNY 18 per year, or a one-time purchase of CNY 38.

    Mental Math: Practice Calculations Anytime, Anywhere

    • Platform: Android
    • Keywords: calculation, mathematics

    @Peggy_: After starting work, aside from routine lesson reviews, I’ve found myself using my brain far less than before. For a while, I turned to sudoku to pass the time and keep my mind active. The app I’m introducing today—Mental Math—can serve as another simple tool for mental exercise. Mental Math is a puzzle-style app whose core function is to generate arithmetic problems. Because the types of calculations, difficulty levels, and even the number of questions can all be freely customized, it works both as a brain-training game for adults and as a practical tool for elementary school students to practice their calculation skills.

    After opening the app, the first step is to set the practice mode. There are two options: a fixed number of questions or a timed session. The number of questions can be set from a minimum of 10 to a maximum of 50, while timed sessions range from 1 to 5 minutes. Once the mode is selected, you can further design the calculation types. Available operations include addition, subtraction, multiplication, and division. Tapping an operator selects it, and dragging the slider on the right adjusts the difficulty—the larger the number, the more challenging the problems become. Perhaps in order to keep the app lightweight and the difficulty manageable, Mental Math does not offer mixed-operation questions.

    After completing the initial setup, you can enter the quiz interface. Once all questions are finished, Mental Math provides feedback on the session, including total time spent, accuracy rate, and a list of questions with their answers. As you complete more exercises, the app aggregates your results and assigns an overall star rating.

    Of course, if you’re just looking to do a few quick exercises out of boredom, you can choose Random Mode. The app will automatically generate 20 questions for you. However, since the difficulty in this mode is unpredictable, it’s not well suited for younger students.

    So the next time you’re bored and about to open TikTok out of habit, you might consider doing a few calculations instead to keep your mind active. Or if you have children in the lower or middle grades of elementary school who need regular arithmetic practice, Mental Math is also a solid option. For kids, short sessions during spare moments—paired with instant feedback—can be an effective way to train.

    You can download and try Mental Math via F-Droid. The app is completely free.

    TranscribeX: A Local-First Power Tool for Speech Transcription

    • Platform: macOS
    • Keywords: text extraction, speech transcription

    @化学心情下2: TranscribeX is a speech-to-text transcription tool for macOS. Compared with many transcription apps that rely on online AI models, TranscribeX’s biggest advantage lies in its fully local AI–based approach. All transcription is performed locally, meaning data is stored entirely on your device—ensuring a much higher level of data security and privacy.

    In terms of functionality, TranscribeX offers multiple ways to capture audio, including importing audio and video files, recording directly, hooking into recording-type applications, and importing from sources like YouTube. On the model side, it supports Apple’s built-in intelligence models as well as locally downloaded large models. Supported transcription models include Whisper, NVIDIA Parakeet, and others.

    To get started, you first need to set the default model and default transcription language under Transcription in the settings. Then return to the main screen and click Import to bring in the audio or video file you want to transcribe. In my case, I selected a two-hour YouTube video. From there, you simply wait in the transcription workspace until the progress bar completes.

    TranscribeX splits the audio into multiple time segments and displays the transcribed text by segment. This allows you to start organizing and reviewing the completed portions immediately while the rest of the transcription is still running, significantly improving overall processing efficiency.

    Once the transcription is finished, you can export the text in multiple formats, making it easy to continue working with it in other apps. You can also translate the transcribed text into other languages directly within the app.

    In testing, transcribing a two-hour audio file took nearly one hour. This may partly be due to the limited performance of my M1 MacBook Air. Even so, overall accuracy was quite high, and the results were well suited for further editing, organization, and summarization.

    You can download and use the basic features of TranscribeX for free from the official website. Purchasing the Pro version unlocks additional local models and a range of advanced features. Setapp users can access the paid version of TranscribeX directly.

    Artwall: Turn Your Favorite Album Covers into Wallpapers

    • Platform: HarmonyOS
    • Keywords: wallpapers, music albums

    @侧脸君: Artwall is a distinctive wallpaper app on the HarmonyOS platform, built around a clever idea—combining album artwork with wallpapers. With its unique image-processing algorithms, virtually any album cover can be transformed into a stunning phone wallpaper.

    The interface and features are extremely simple. Just enter a song title or artist name, tap to view the album-cover wallpapers, and download them. At the moment, Artwall does not support setting wallpapers directly within the app—you’ll need to manually go to your gallery to set the image as your wallpaper.

    Although Artwall doesn’t offer an option to choose wallpaper resolution, the resolution is tightly bound to the device being used. If you’re on a Huawei foldable device, the wallpaper resolution and layout saved for folded and unfolded states will differ accordingly.

    A good idea naturally sparks the urge to share—especially when it satisfies both “showing off your wallpaper” and “sharing your music taste” at the same time. Just five days after its first release, more and more screenshots of wallpapers generated by Artwall have already begun circulating on social media, highlighting users’ distinctive musical preferences.

    You can download Artwall from the HarmonyOS App Gallery. The app is completely free.

    Summary Expressive: An AI Summarization Tool with Bilibili Support

    • Platform: Android
    • Keywords: video summarization

    @大大大K: One of the most common uses of AI tools is solving the “TL;DR” problem. Whether it’s a text running thousands of words or images packed with information, AI can easily extract the key points. For video content, however, the mainstream approach still relies on subtitles or transcripts provided by video platforms, converting them into structured text for analysis. This is where a third-party tool like Summary Expressive comes in, feeding platform-generated transcripts to AI models.

    Summary Expressive supports multiple content formats, including plain text, web links, images, Word documents, and PDFs. Simply select a file within the app and the AI can quickly generate a summary—tasks that are trivial for modern AI models and don’t even require separate conversion tools. Video summarization is different, though: subtitles and transcripts embedded by video platforms aren’t always standardized like documents or images, which is why Summary Expressive is particularly useful here.

    Summary Expressive supports video links from YouTube and Bilibili. After pasting a video link, you can choose the desired length of the summary. In practice, longer summaries aren’t dramatically longer in word count, but they do provide more comprehensive coverage. Medium and short summaries tend to omit some details. The app also offers a text-to-speech feature for summaries, making it ideal for quickly reviewing video content while commuting.

    Summary Expressive supports major AI model APIs both domestic and international, including OpenAI, Gemini, and DeepSeek. If you want to summarize Bilibili videos, you’ll need to log in to your Bilibili account in the settings beforehand.

    You can now download Summary Expressive for free from GitHub and Google Play. The GMS version requires Google Play Services and downloads image-recognition models online, while the standalone version includes built-in models—choose whichever best suits your needs.

    App Updates You Shouldn’t Miss

    Beyond brand-new releases, many familiar faces on the App Store continue to iterate and evolve, adding features that are both interesting and genuinely useful. At SSPAI, we aim to help you filter through the updates that are truly worth your attention, so you can quickly stay up to date with the latest moves from apps and their developers.

    PhotonCam 1.30 Update: A More Refined Editing Experience for Pro-Level iPhone Photography

    • Platform: iOS / iPadOS
    • Keywords: mobile photography

    @Vanilla: iPhone photography enthusiasts may already be familiar with PhotonCam. While it’s not a household name, this third-party camera app stands out for its rich feature set, polished design, and well-thought-out interactions. At the same time, the developer actively embraces new system features and keeps pace with advances in imaging technology, allowing users to enjoy a more professional photography experience on the iPhone. Recently, PhotonCam rolled out version 1.30—its final major update of 2025—focused primarily on improving the photo-editing experience. Let’s take a look.

    First, PhotonCam has updated the HDR display behavior of the focus indicator. More specifically, when adjusting depth-of-field effects on HDR photos within PhotonCam, you can tap the screen to position the focus indicator. In the new version, the area covered by the focus indicator is now rendered with HDR effects as well, making it easier to accurately select the desired focus point.

    Second, the new version also refines the behavior of parameters such as fading and brightness adjustments, and introduces a setting that automatically falls back to SDR when the system crops HDR photos under strong lighting conditions. PhotonCam has also improved its launch speed, allowing you to enter shooting mode faster and capture moments more promptly. In addition, a new photo queue feature has been added: when shooting in burst mode or when photos require longer processing times (such as long exposures at night), PhotonCam displays a real-time counter on the gallery button in the lower left corner, indicating how many photos are currently being processed.

    Finally, PhotonCam has introduced several new Lock Screen widgets, including options for wide-angle, telephoto, and ultra-wide lenses. However, the developer has clarified that tapping these widgets will only open the app, rather than directly switching to the corresponding focal length. It’s unclear whether this is due to system permission limitations or other design considerations.

    That covers the main highlights of PhotonCam’s 1.30 update. If you’re looking for a full-featured third-party photography app on iPhone that balances both shooting and editing capabilities, PhotonCam remains a very solid choice. PhotonCam is available on the App Store under a “free download + in-app purchase” model, with subscription options priced at CNY 12 per month or CNY 68 per year, as well as a one-time lifetime purchase option for CNY 88.

    Controller for HomeKit 8.1: Introducing State Logic

    • Platform: iOS / iPadOS / macOS / watchOS
    • Keywords: smart home, HomeKit

    @Snow: Last week, Controller for HomeKit rolled out version 8.1, introducing an all-new States feature. Compared with the native Home app or Aqara’s “Scenes,” which apply a one-time batch change to device states, States enable Controller to support much more complex logical layering, allowing users to build far more flexible automation flows.

    Within Workflows, you can create all kinds of state logic. For example: when “Dog at Home” is active, enable the balcony door sensor and alarms; when “Deep Cleaning” is active, pause all motion-triggered lighting rules; when “Guests Visiting,” keep all living room lights permanently on; when “Watching a Movie,” silence all notifications and alarms. On the surface, each individual state may look similar to a scene, but the real power lies in the fact that States not only support logical stacking, they can also be used as triggers, conditions, steps, or stop events within a wide variety of workflows.

    For instance, if you’re about to do a deep clean after walking the dog, you can enable both the “Dog at Home” and “Deep Cleaning” states at the same time. This way, the lights won’t constantly change while you’re cleaning, and you’ll still receive notifications if your dog is causing trouble on the balcony. You can also use “Guests Visiting” as a workflow trigger: while turning on the lights to welcome visitors, the system can notify your family members and immediately lock the study door—preventing curious kids from wreaking havoc on your “collection.”

    Back in version 7.4, when Hub Mode was introduced, Controller already brought external control links to non-HomeKit users. With this update, the new States feature also supports external control URLs. This means you can use local Shortcuts, or third-party workflow tools like IFTTT, to trigger vacation modes via calendar events or quickly change states from non-Apple devices.

    After the 8.0 update, the developer restructured Controller’s pricing model. The original subscription / lifetime purchase option has been renamed Controller Essentials, with the lifetime price remaining unchanged. However, continued access to Hub Mode, Automate Workflows, and the new charting features now requires a Controller Plus subscription. Non-Essentials users must pay CNY 598 per year, while Essentials users can subscribe for CNY 228 per year, both with a 7-day free trial. The price barrier for Controller has clearly risen, so it’s recommended to try it out before deciding whether to commit to further payment.

    You can download Controller for HomeKit from the App Store.

  • SSPAI Morning Brief: Draft Rules Propose Timed Exit Reminders and Timely Intervention for AI Chat Addiction

    SSPAI Morning Brief: Draft Rules Propose Timed Exit Reminders and Timely Intervention for AI Chat Addiction

    Morning Brief

    1. Draft rules propose requiring AI chat services to issue timed exit reminders and intervene promptly in cases of addiction
    2. Xiaohongshu responds to a public “call-out” from the Lijiang Old Town Culture and Tourism Bureau
    3. Cyberspace Administration issues regulations to standardize the behavior of online influencer accounts
    4. Honor releases the WIN series gaming smartphones
    5. Plaintiff presents forged evidence bearing Doubao AI watermarks, receives a judicial reprimand
    6. Beijing arbitration case clarifies that dismissal due to a role being replaced by AI constitutes unlawful termination
    7. Rumors You Can Just Glance At

    Draft rules propose requiring AI chat services to issue timed exit reminders and intervene promptly in cases of addiction

    On December 27, the Cyberspace Administration of China released the Interim Measures for the Management of Anthropomorphic Interactive AI Services (Draft for Public Comment), soliciting public feedback. The draft defines “anthropomorphic interactive services” as products or services that use AI technologies to simulate human personality traits, thinking patterns, and communication styles, and engage in emotional interaction with humans through text, images, audio, video, and other forms.

    Beyond basic content compliance, the draft proposes that anthropomorphic interactive services must not provide false promises that seriously affect user behavior or services that harm social relationships; must not harm users’ physical or mental health; and must not induce users to make unreasonable decisions through algorithmic manipulation, misleading information, or emotional traps. Service providers are required to have safety capabilities such as mental health protection, emotional boundary guidance, and dependency risk warnings, and must not take replacing social interaction, controlling users’ psychology, or inducing addiction and dependence as design goals.

    The draft also proposes introducing a number of intervention mechanisms. For example, while protecting users’ personal privacy, providers should assess users’ emotional states and their level of dependence on products and services, and take necessary measures to intervene when extreme emotions or addiction are identified. When users explicitly express intentions to commit suicide, self-harm, or other extreme actions, the conversation must be taken over by human operators, and timely measures should be taken to contact users’ guardians or emergency contacts (required to be provided during registration for minors and older adults). In addition, if a user has been using the service continuously for more than two hours, dynamic reminders—such as pop-up notifications—should be displayed to prompt a pause; providers must not obstruct users from actively exiting the service. Providers should also prominently inform users that they are interacting with AI rather than a natural person.

    The draft further brings training data under regulatory oversight. Providers should strengthen the management of training data, ensure that data sources are lawful and traceable, take necessary measures to safeguard data security, and prevent data leakage risks. Unless otherwise stipulated by laws or administrative regulations, or with separate user consent, providers must not use user interaction data or users’ sensitive personal information for model training.

    Since the beginning of this year, international media have repeatedly reported on incidents linked to AI chat services. The U.S. Federal Trade Commission is investigating chatbot products from major AI companies, and some states have begun to legislate oversight. For example, in October, the Governor of California signed a series of restrictive bills targeting AI applications, aimed at strengthening protections for minors. These measures require AI products to verify users’ ages, remind users every three hours that they are interacting with AI rather than a human, and promptly intervene in cases involving suicide or self-harm.

    Xiaohongshu responds to a public “call-out” from the Lijiang Old Town Culture and Tourism Bureau

    According to a report by Observer Network, the Culture and Tourism Bureau of Lijiang’s Old Town District in Yunnan recently issued a public letter accusing the Xiaohongshu platform of failing to adequately fulfill its responsibilities in information oversight. The bureau claimed this had led to a large number of consumers using unsubstantiated “avoid-at-all-costs” posts to threaten wedding photography businesses into issuing refunds.

    In response, on December 24 Xiaohongshu customer service stated that the platform provides public reporting channels and will handle confirmed violations in a timely manner. People familiar with the platform said that preliminary reviews have not found evidence of consumers fabricating experiences or engaging in other violations. The Old Town District Culture and Tourism Bureau later deleted the letter. However, the Lijiang Old Town Travel Photography Industry Association said it has refused to continue negotiations with Xiaohongshu and is currently commissioning lawyers to collect evidence and prepare a collective public-interest lawsuit.

    Several local travel photography business owners told the media that some consumers, after signing off on services and receiving their final photos, subsequently post negative reviews on Xiaohongshu to extort full refunds. In some cases, social groups have even emerged that coach members on how to obtain free services through complaints. Merchants said that even when they submit signed satisfaction agreements or certificates showing no complaints filed with regulators, their appeals often fail because the platform’s system deems_toggle “insufficient evidence.” As a result, order volumes at some studios have dropped sharply, and some sales teams have even resigned en masse. According to the industry association, as of December 2025, direct economic losses caused by false “avoid-at-all-costs” posts have exceeded RMB 1 million, with indirect order losses reaching more than RMB 5 million.

    Public opinion online has been divided. Some users believe that certain “avoid-at-all-costs” posts have deteriorated in nature and have expressed concern about Xiaohongshu’s overall content environment. Another perspective shifts the focus to the travel photography businesses and the industry itself, arguing that the key to avoiding such posts lies in improving service quality.

    Industry experts note that social platforms such as Xiaohongshu have effectively become part of the infrastructure for tourism services. Meanwhile, structural issues within the travel photography industry—such as cutthroat low-price competition, hidden fees, and non-standardized services—objectively increase the difficulty for platforms to verify the authenticity of content. Experts suggest that platforms should consider establishing collective appeal and communication mechanisms tailored to regional industries, while industry associations should help optimize the environment for public opinion oversight through standardization efforts.

    Cyberspace Administration issues regulations to standardize the behavior of online influencer accounts

    On December 26, the Cyberspace Administration of China (CAC) released the Notice on Regulating the Management of Online Influencer Account Conduct, establishing a negative list for online influencer account behavior and defining clear boundaries for such conduct.

    According to the notice, the negative list of prohibited behaviors for online influencer accounts includes: (1) spreading vulgar content; (2) promoting unhealthy value orientations; (3) creating “ugliness-as-spectacle” or otherwise harmful personas; (4) disseminating false information; (5) distorting or misinterpreting policies and public events; (6) inciting group antagonism; (7) doxxing or publicly exposing individuals at will; (8) organizing offline fights or online flame wars; (9) soliciting negative leads or accusations; (10) calling on fans to gather or mobilize; (11) engaging in coercive or extortionary practices for personal gain; (12) conducting operations without required qualifications; and (13) covertly engaging in illegal or gray-market activities.

    The notice requires all websites and platforms to conscientiously implement relevant laws, regulations, and policies, strictly fulfill their primary responsibilities in accordance with the negative list, and further improve community rules and user agreements related to the management of online influencer accounts. Platforms are also instructed to strengthen, in accordance with the law and platform rules, the management of content publishing, live-stream interactions, topic creation, and other behaviors by online influencer accounts, guiding operators to use their influence responsibly and to standardize online speech and conduct.

    Honor releases the WIN series gaming smartphones

    On December 26, Honor launched the WIN series smartphones in Chengdu, introducing two models—Honor WIN and Honor WIN RT—with starting prices of RMB 3,999 and RMB 2,699 respectively. The WIN series is a brand upgrade of the previous GT lineup and is positioned squarely in the professional gaming phone market.

    In terms of performance, the entire Honor WIN lineup is powered by Qualcomm’s flagship Snapdragon platforms. The Honor WIN features the fifth-generation Snapdragon 8 Elite chipset, paired with LPDDR5X memory and UFS 4.1 storage. To improve endurance, the WIN series raises battery capacity to 10,000mAh for the first time, supporting 100W wired fast charging and 80W wireless fast charging (standard model only). The new phones also support bypass charging, allowing the charger to power the motherboard directly to reduce heat buildup while gaming and charging simultaneously.

    The Honor WIN series introduces the “Dongfeng Turbo” active cooling system, which uses an ultra-compact fan and a direct-drive high-efficiency cooling air duct. Official figures claim a 30% improvement in cooling efficiency, with fan noise as low as 25 dB. On the display front, the phones are equipped with a 185Hz ultra-high-refresh-rate esports screen, support 5,920Hz high-frequency PWM dimming, and are optimized for ultra-frame-rate modes in multiple mainstream games.

    The Honor WIN series also features dedicated esports antennas to ensure low-latency performance in weak network conditions. Additional features include AI-assisted control enhancements and anti-mis-touch functions. At the launch event, Honor announced a strategic partnership with Delta Force, with the WIN series designated as the official device for its professional league. Honor stated that it will further build out a gaming ecosystem around the WIN series, including future products such as tablets.

    Plaintiff presents forged evidence bearing Doubao AI watermarks, receives a judicial reprimand

    Recently, the Civil Division II of the Dawu County Court in Xiaogan, Hubei Province, identified a plaintiff’s attempt to fabricate evidence using AI technology during the trial of a housing lease contract dispute, and issued a formal reprimand in accordance with the law.

    In May 2024, Li signed a housing lease contract with Xiong, agreeing on the lease term and rent, and specifying that the rent for the latter half of the term must be paid by March 2025. After the lease expired, Li sued Xiong for failing to pay the remaining rent as well as water and electricity fees, appointing his daughter Dong as his litigation agent.

    During the hearing, Li was unable to provide valid proof that Xiong owed water and electricity charges. Dong stated in court that Xiong was the first tenant of the property and that all utility costs were incurred by him, and promised to submit photos of the water and electricity meters from before and after the lease period as supporting evidence after the hearing. However, the photos Dong submitted clearly bore a “Generated by Doubao AI” watermark, which immediately raised the judge’s suspicions. Further examination revealed inconsistencies in Dong’s statements: she initially claimed that the meters were used independently, but later admitted under questioning that two households shared the same meters. Faced with the judge’s inquiries, Dong was unable to justify her claims and ultimately admitted to fabricating the evidence, subsequently submitting the genuine photos to the court.

    The presiding judge gave Dong a serious lesson in the rule of law, pointing out that using AI-generated images as litigation evidence constitutes the fabrication of key evidence and is suspected of disrupting judicial order. Taking into account that Dong voluntarily confessed during the investigation and promptly corrected her mistake, the court followed a principle of combining punishment with education, opting to handle the matter by way of a reprimand and refusing to admit the fabricated evidence.

    Internationally, the use of AI-generated false materials in litigation has also become increasingly common. A database established by French lawyer Damien Charlotin has collected more than 600 cases worldwide. When such errors are discovered, many lawyers attempt to deflect responsibility with various excuses, such as claiming they did not know they were using AI, did not realize AI could hallucinate, alleging that hackers remotely altered the files, or even saying it was too inconvenient to switch windows on a laptop so they used AI. Some lawyers have even tried to shift the blame to interns or clients.

    Beijing arbitration case clarifies that dismissal due to a role being replaced by AI constitutes unlawful termination

    On December 26, the Beijing Municipal Human Resources and Social Security Bureau released its 2025 annual collection of representative labor and personnel dispute arbitration cases. Among them, an arbitration case arising from AI replacing a job position explicitly clarified that “AI replacing a position does not equal lawful dismissal.”

    According to the case details, Liu had worked for many years at a technology company, responsible for traditional manual map data collection. In early 2024, the company decided to fully shift to AI-driven automated data collection, abolishing the department and corresponding positions where Liu worked. At the end of 2024, the company terminated Liu’s labor contract on the grounds that “the objective circumstances on which the labor contract was based have undergone a material change, making the contract impossible to continue performing.” Liu believed the termination was unlawful and therefore applied for arbitration.

    In its analysis, the arbitration committee pointed out that the company’s introduction of AI technology was a proactive technological upgrade undertaken to adapt to market competition. Although such changes may lead to adjustments in job structures, they do not meet the criteria of “objective circumstances,” which require characteristics of irresistibility and unpredictability. Terminating a labor contract on the grounds that a position has been replaced by AI essentially shifts the normal risks of technological iteration onto the employee. The arbitration committee thus determined that the company’s action constituted unlawful termination.

    The arbitration committee further noted that, in the context of rapid technological development, employers should give priority to properly accommodating affected employees through measures such as negotiating changes to labor contracts, providing skills training, and arranging internal job transfers. If termination of a labor contract is indeed necessary, employers must strictly comply with relevant regulations and avoid mechanically applying “material changes in objective circumstances” as a justification for dismissal.

    Rumors You Can Just Glance At

    • According to the November 2025 Domestic Mobile Phone Market Operation Analysis Report released by the China Academy of Information and Communications Technology (CAICT), domestic brands shipped 23.228 million units in November, down 12.6% year over year, accounting for 77.7% of total shipments. Meanwhile, shipments of foreign-brand phones reached 6.93 million units, up 128.4% year over year. This surge is likely related to adjustments by manufacturers such as Apple in promotions, channel strategies, or new product release cycles.
    • Weibo user “定焦数码” claimed that “a new iteration of the iPhone Air is confirmed to launch and will appear at the fall event. In addition, the iPhone 17e has already entered mass production and will be seen at the spring event.” This differs from earlier reporting by The Information, which stated that due to lukewarm market performance of the first-generation iPhone Air, Apple had informed its supply chain to remove the second-generation product from the 2026 roadmap, possibly delaying it until 2027.
    • X user Kosutami recently shared a set of photos said to be prototypes of the first-generation AirPods, showing two bright color options Apple once experimented with: pink and yellow. The prototypes were described as using an “iPhone 5C color scheme.” Although these colorways never made it into mass production, the level of finish on the prototypes suggests that Apple seriously considered breaking away from its single-white design tradition during the early development stage.
  • SSPAI Morning Brief: Xiaomi Unveils Multiple New Products, Microsoft Announces BitLocker Hardware Acceleration Support, and More

    SSPAI Morning Brief: Xiaomi Unveils Multiple New Products, Microsoft Announces BitLocker Hardware Acceleration Support, and More

    Morning Brief

    1. Xiaomi unveils multiple new products
    2. Microsoft announces BitLocker will support hardware acceleration
    3. ASUS launches RT-BE90U wireless router
    4. Critical security vulnerability discovered in MongoDB
    5. Alibaba open-sources Tongyi Qwen-Image-Edit-2511 image editing AI model
    6. Lenovo Watch GT Pro officially goes on sale
    7. Rumors You Can Just Glance At

    Xiaomi unveils multiple new products

    On December 25, Xiaomi unveiled a range of new products, including the Xiaomi 17 Ultra, 17 Ultra Leica Edition, Xiaomi Buds 6, and Xiaomi Watch 5.

    First up is the Xiaomi 17 Ultra, Xiaomi’s latest imaging flagship. In terms of design, the Xiaomi 17 Ultra series features a 2D flat display paired with a completely flat, straight-edged frame. With a thickness of 8.29 mm, it is the slimmest Ultra model in the series to date. Color options include black, white, purple, and green, and the device supports IP66, IP68, and IP69 ratings for dust and water resistance.

    On the imaging front, the Xiaomi 17 Ultra is equipped with a Leica 1-inch Light Fusion Master main camera and a Leica 200-megapixel optical zoom telephoto lens. The main camera uses the Light Fusion 1050L sensor, featuring a 1-inch large sensor, 3.2 μm pixels, an f/1.67 aperture, and third-generation LOFIC ultra-high dynamic range technology. Its single-pixel full well capacity has increased to 6.3× that of the previous generation, supporting high dynamic range scenarios such as Fireworks mode. The telephoto camera adopts a new-generation HPE sensor, reducing power consumption by 40% compared to its predecessor. It measures 1/1.4 inches and delivers 200 megapixels. This lens supports continuous optical zoom, covering 75 mm to 100 mm mid-telephoto full-sensor optical output, with telephoto relay options at 200 mm and 400 mm. It is also the first Leica APO optical lens on a mobile device, supporting 4K Live Photos and multiple camera movement modes.

    For the display, the Xiaomi 17 Ultra features a 6.9-inch low-power “sunlight” display using M10 emissive materials and a new red primary emissive material, protected by Dragon Crystal Glass 3.0. Powering the device is Qualcomm’s fifth-generation Snapdragon 8 Supreme Edition processor, paired with a 6800 mAh battery that supports 90 W wired fast charging, 50 W wireless fast charging, and the PPS universal fast charging protocol. In terms of connectivity, it supports Xiaomi Star Communication and UWB ultra-wideband interconnection technology. Pricing starts at RMB 6,999 for the 12 GB + 512 GB configuration.

    Following the launch of the Xiaomi 17 Ultra, Xiaomi also introduced the 17 Ultra Leica Edition. This version continues Leica’s design language in its exterior details, including the iconic Leica red dot, a black-and-white two-tone body, finely knurled frame, and exclusive Leica engravings. The rear camera area features a Master Zoom Ring for focal length adjustment, with customizable parameters such as exposure compensation and white balance. In terms of imaging styles, the 17 Ultra Leica Edition offers one-tap Leica modes, including Leica M9 image rendering and Leica M3 + MONOPAN 50 black-and-white film aesthetics, as well as exclusive imaging options like a 3:2 full-frame aspect ratio. This edition also adds Beidou + Tiantong dual-satellite communication and a secure privacy encryption chip. The Xiaomi 17 Ultra Leica Edition starts at RMB 7,999 for the 16 GB + 512 GB version.

    At the same event, Xiaomi also unveiled the Xiaomi Buds 6. Xiaomi states that these earbuds feature a bionic curved structure, weighing just 4.2 g per earbud, and are available in four color options. They are equipped with triple-magnet gold-plated drivers, co-tuned with Harman, and support end-to-end true lossless audio, spatial audio, intelligent translation recording, and real-time interpretation. Battery life can reach up to 35 hours, with a price of RMB 699.

    In addition, the Xiaomi Watch 5 was announced. It features a 47 mm, 1.54-inch AMOLED display and a dual-sided synthetic sapphire glass body, powered by the Snapdragon W5 processor and a BES2800 low-power chip. It supports standalone eSIM, an on-wrist control center, EMG micro-gesture control, ECG electrocardiogram analysis, and over 150 sports modes. Battery life reaches up to 6 days in performance mode and up to 18 days in power-saving mode. Pricing starts at RMB 1,999. Source

    Microsoft announces BitLocker will support hardware acceleration

    Microsoft has announced that BitLocker in Windows 11 (25H2) and Windows Server 2025 now supports CPU hardware encryption. After enabling BitLocker, the encryption workload is shifted from the software layer to a dedicated encryption engine, reducing the performance overhead typically caused by software-based encryption. According to Microsoft, the new approach can significantly improve storage read and write performance: sequential read speeds increase from 1,632 MB/s to 3,746 MB/s, write speeds rise from 1,510 MB/s to 3,530 MB/s, and CPU usage is reduced by approximately 70%.

    At present, hardware-accelerated BitLocker is only available on vPro devices equipped with “Panther Lake” processors, with support expected to expand to more devices in the future. Source

    ASUS launches RT-BE90U wireless router

    On December 25, ASUS introduced the RT-BE90U wireless router. The new model supports tri-band Wi-Fi 7 with a BE9400 wireless specification, including advanced Wi-Fi 7 features such as 320 MHz channel bandwidth, 4096-QAM, and MLO, and can expand home networks via AiMesh. In terms of hardware, the RT-BE90U is equipped with four 2.5 GbE RJ45 ports and six external antennas. The rear of the device also includes a USB-A (5 Gbps) port for network sharing with external devices.

    On the software and feature side, the router supports Smart Home Master, allowing users to configure up to three SSIDs for device categorization and parental controls. It also integrates the enterprise-grade AiProtection security solution, providing more comprehensive protection for home networks. Source

    Critical security vulnerability discovered in MongoDB

    A critical security vulnerability has recently been disclosed in the popular non-relational database MongoDB. Assigned CVE-2025-14847, the flaw affects multiple versions of MongoDB and MongoDB Server. The vulnerability stems from inconsistent handling of length parameters, which could allow unauthenticated attackers to execute arbitrary code and gain control of affected systems.

    MongoDB’s security team stated that clients could exploit a flaw in the server-side zlib implementation to trigger the return of uninitialized heap memory without authentication. The company strongly recommends upgrading to patched versions as soon as possible. MongoDB currently advises immediate updates to the following unaffected versions: MongoDB 4.4.30, 5.0.32, 6.0.27, 7.0.28, 8.0.17, and 8.2.3. If an immediate upgrade is not feasible, users can mitigate the risk by explicitly removing zlib via the networkMessageCompressors or net.compression.compressors options when starting mongod or mongos, thereby disabling server-side zlib compression. Source

    Alibaba open-sources Tongyi Qwen-Image-Edit-2511 image editing AI model

    On December 25, Alibaba’s Tongyi Qwen announced the official open-sourcing of its image editing model, Qwen-Image-Edit-2511. This release focuses on strengthening consistency across multiple dimensions and improving overall generation quality.

    Specifically, Qwen-Image-Edit-2511 enhances character consistency, integrates LoRA capabilities for lighting and multi-scene generation, and improves industrial design generation and geometric reasoning abilities. In addition, the new version further boosts consistency in multi-person group photo generation. While the earlier Qwen-Image-Edit-2509 had already improved single-person group photo consistency, this update addresses consistency challenges in multi-person scenarios, enabling the model to generate more coherent group photos based on two individual portraits.

    Currently, Qwen-Image-Edit-2511 has been open-sourced simultaneously on the ModelScope and Hugging Face platforms. Source

    Lenovo Watch GT Pro officially goes on sale

    The Lenovo Watch GT Pro smartwatch officially went on sale on December 25, highlighting up to 27 days of battery life, dual-frequency GPS for precise positioning, 50-meter professional-grade water resistance, as well as outdoor features such as more than 170 sports modes and an altitude, barometer, and compass. The current price is RMB 899.

    In terms of design, the Lenovo Watch GT Pro features a 1.43-inch circular AMOLED display with a resolution of 466 × 466, protected by Corning glass, and supports a wide range of free, customizable watch faces. The watch offers 5ATM professional water resistance and includes built-in swimming algorithms.

    On the functionality side, the watch is compatible with both Android and iOS systems and comes with dual-frequency GPS positioning and 24/7 health monitoring. It supports heart rate, blood oxygen, sleep, and stress tracking, as well as breathing exercises and health reminders. For sports and fitness, it offers more than 170 workout modes. The smartwatch also supports Alipay payments, a compass, Bluetooth calling, message notifications, weather information, remote camera control, phone-finding, and a voice assistant. Source

    Rumors You Can Just Glance At

    According to The Korea Economic Daily, Samsung Electronics is accelerating its push toward GPU in-house development and plans to adopt a self-developed graphics IP in the Exynos 2800, which is expected to launch in 2027. The move aims to reduce reliance on AMD’s RDNA technology and increase autonomy in product iteration. Samsung also intends to extend its in-house GPU beyond smartphones to smart glasses, robotics, and automotive SoCs, while further expanding into the AI ASIC market. Source