Author: kiwi

  • A Square-Format Ultra-Wide Prime Camera

    A Square-Format Ultra-Wide Prime Camera

    Introduction

    If I asked you to guess a camera:

    • It has a native 1:1 square-format sensor, just like classic 6×6 medium-format cameras;
    • It uses an ultra-wide lens around 12mm, capable of capturing a dramatically expansive field of view;
    • It’s also a fast-aperture prime lens with a physical variable aperture for depth-of-field control. Sounds a bit like a “wide-angle modded version” of a Ricoh GR, doesn’t it?

    But if I tell you its actual name—DJI Action 6 (hereafter AC6)—your first reaction might be: “Isn’t this just an action camera?”

    This sense of “discrepancy” may not be a flaw in product design at all, but rather a sign that the logic of the market has quietly shifted.

    Suppose the number of people in extreme sports communities has actually remained fairly stable—yet you’ll notice that action-camera sales continue to rise year after year. What does that imply? It implies that more and more people are buying these devices with the mindset of “buying a camera,” not “buying an action camera.”

    Some features that seem to have little to do with sports scenarios also indirectly confirm this trend: direct-connect microphones, starburst effects enabled by the variable aperture, portrait mode… DJI clearly wants to position it as an all-purpose camera, not merely a tougher, more stable “action camera.”

    Image source: DJI official website

    All these features and accessories exist to help it break free from the limitations of “only shooting ultra-wide action footage,” making users more inclined to see it as a compact creative tool.

    I don’t ski, and I’m not brave enough to go skydiving. I intend to use the Action 6 as an ultra-wide prime camera. Based on my daily shooting habits, I want to see what kind of images it can capture once it’s stripped of its “action” label.

    01 A More Camera-Like Attempt: The Addition of a Physical Aperture

    In the past, the logic of action-camera lenses was simple: ultra-wide, fixed-focus, fixed aperture.

    This time, the AC6 introduces a physical variable aperture (F2.0–F4.0). In actual use, the variable aperture functions mostly as a “brightness switch,” with limited impact on image quality. In Auto mode, under normal daylight conditions, it typically stays at F2.8; in low-light environments, it switches to F2 to allow more light in, and it almost never switches automatically to F4. If you want to capture starbursts, you’ll need to manually set it to F4.

    Minimum focusing distance improvement:
    This is a parameter upgrade that many people might overlook. In DJI’s promotional materials, the selling point of the F4 aperture is mostly centered on starburst effects.

    Image source: DJI official website

    However, in real-world use, the improvement in minimum focusing distance is likely more practical. At F2.8, the minimum focusing distance is about 35 cm. At F4.0, it shortens to 20 cm. I also tested whether the aperture would automatically stop down when shooting at close range; it does not.

    In default Auto mode, the AC6 still adjusts the aperture based on scene brightness rather than scene content, prioritizing “ensuring sufficient light” above all. This means it isn’t as “smart” as a smartphone. If you want to shoot close-ups, you’ll need to manually switch to F4.0, which increases the operational steps.

    02 A Square Format That “Can Only Shoot Video”

    The funniest thing about this camera is this: it has a square sensor, yet when you want to compose in a 1:1 square format, you can only shoot video, not photos.

    In actual use, the camera still feels far better suited for video than for stills. Shooting video is smooth and seamless—the recording, playback, and review flow naturally. Shooting photos, on the other hand, is barely acceptable. After pressing the shutter, capturing a 38-megapixel JPEG takes about 4 seconds; shooting JPEG+RAW takes around 5 seconds. During this time, you cannot take the next shot—you must wait for the current image to finish writing before shooting again. With this camera, you absolutely cannot rush photography; you just have to take it slow.

    All sample images below were captured using the 38 MP (4:3) format, JPEG, Auto aperture mode, and Auto exposure, with composition adjustments and color editing applied afterward.

    1:1 and 2.35:1 Cinematic Feel: An Alternative Way to Use Ultra-Wide

    Pairing an ultra-wide lens with an elongated cinematic aspect ratio creates a fascinating chemistry: removing extra information from the top and bottom instantly amplifies the horizontal tension of the frame.

    By placing the camera close to the ground or foreground and using the exaggerated perspective of the ultra-wide lens—then cropping to a wide aspect ratio—you get instant dynamism.

    F2.8 ISO600 1/100s
    F2.8 ISO400 1/125s
    F2.8 ISO450 1/125s

    A wide aspect ratio can also capture more architectural lines on both sides of the street, creating a strong sense of narrative.

    F2.8 ISO450 1/800s
    F2.8 ISO250 1/250s
    F2.8 ISO600 1/350s
    Manual F4 adjustment ISO700 1/160s

    Square compositions force you to eliminate distractions on the left and right, keeping only the core geometric balance.

    Below are several sample shots to help you feel the mood of the square format:

    F2.8 ISO450 1/180s
    F2.8 ISO300 1/100s
    F2.8 ISO800 1/320s
    F2.8 ISO300 1/240s
    F2.8 ISO200 1/100s

    03 Smartphones Also Have Ultra-Wide Lenses—How Big Is the Image Quality Gap?

    “Since flagship phones now have increasingly good ultra-wide cameras, is there really still a need to buy an AC6 purely for image quality?”

    To answer this question, I mounted the iPhone 16 Pro Max (which doesn’t even count as a flagship anymore) and the AC6 on the same tripod and compared them through real-world shooting.

    I tested daytime and nighttime image quality, stabilization, audio recording, exposure behavior, brightness, and color tendencies. If you’re interested, you can watch the full comparison on Bilibili. If you can’t watch the video, below is a text-version summary.

    1. Image Quality Comparison

    Sharpness: Based on the sample shots, the actual resolving power of the two devices is similar.

    The AC6 applies more sharpening and noise reduction, resulting in heavier smearing. In photo mode, you cannot adjust noise reduction or texture intensity. In daytime scenarios, AC6’s image quality tends to look better. However, at night, the AC6 over-sharpens to the point that highlights are more likely to show unnatural artifacts.

    Blue signage shows black edging

    White balance: Across all indoor and outdoor test scenes, the smartphone’s image consistently leans yellow, while the AC6 is closer to real-life visual perception.

    The AC6 also reproduces blues more accurately

    Dynamic range: The two devices handle brightness with entirely different strategies.

    The AC6 prioritizes highlight suppression. The smartphone lifts shadows aggressively, tolerating more blown-out highlights.

    The smartphone’s image is overall about one stop brighter, but the entire frame appears slightly washed out, lacking the AC6’s clarity and transparency.

    Comparison of tree branch detail in bright highlights

    Flare resistance:

    Under identical camera movements, the AC6 displays flare earlier than the phone, with green elliptical artifacts.

    When moving forward while shooting, the AC6 also produces flare more frequently than the smartphone.

    2. Stabilization Comparison

    For easier testing, both devices recorded 4K/60P video in a 16:9 aspect ratio. The smartphone’s action-stabilization mode was turned off, while the AC6 was set to its “SuperStrong Stabilization” mode, with image-quality priority enabled and Standard (distortion correction) equivalent to 15mm.

    Daytime scenes:

    • During normal walking, the difference between the smartphone and the AC6 is minimal—both perform excellently.
    • During running, the AC6 delivers noticeably better stabilization with smoother transitions.

    Night scenes: In both normal walking and running, the AC6 performs better overall.

    3. Audio Recording Comparison

    Without external microphones, both devices recorded 4K/60P, 16:9 video. The test scenario was an outdoor live band performance.

    The smartphone used default audio settings; the AC6 was set to stereo channels with wind-noise reduction at the standard level.

    Live recording environment

    The smartphone performs better overall, offering clearer separation between vocals and instruments and a wider dynamic range. The audio can also be mixed in post, resulting in stronger presence and immersion.

    With the AC6, recording gain seems somewhat high, and the captured audio is loud, with vocals and instruments blending together. This may be an optimization intended for sports scenarios. If you care about audio quality, it’s recommended to use a DJI Mic with the AC6.

    4. Differences in User Experience

    If we look only at image quality, the phone is not far behind. But the true value of the AC6 lies in “freeing your mindset.”

    A dead phone battery triggers anxiety (subway rides, replying to messages, navigation, payment—all still required). But when the AC6 runs out of power, the worst case is simply: you stop shooting, or swap in another battery. This “battery separation” creates a sense of ease. Combined with the camera’s physical advantages—“pick it up and shoot,” “drop it into a puddle and it’s still fine”—this is the true competitive edge of such devices.

    Below is a table summarizing the test results so you can quickly compare based on your own usage needs:

    Comparison CategoryDJI OSMO ACTION 6iPhone 16 Pro Max
    Image sharpnessSimilar to the phone, slightly less noiseSimilar to the AC6
    Brightness tendencyPrioritizes highlight detailPrioritizes shadow detail
    Stabilization⭐ Comparable to the phone in daytime, overwhelmingly better at night
    Flare resistanceMore prone to flare under identical camera movement
    Auto white balance⭐ Highly consistent indoors and outdoors, very accurateConsistently yellowish
    Audio⭐ Better separation of vocals and instruments, adjustable in post

    04 A Few Things That Require Adjustment: Pain Points in the User Experience

    Although I’ve tried to think of it as a “mini camera,” in actual use there are moments when certain frustrating flaws make it clear: as a camera, it still isn’t mature enough.

    1. Extremely Slow Shutter Lag
    The AC6 does indeed offer high-resolution stills (38 MP), but its processing speed doesn’t keep up. When shooting full-resolution photos, after pressing the shutter you need to wait about 4 seconds before the camera finishes processing the previous image and becomes ready for the next shot. This essentially eliminates any possibility of continuous capture. In these moments, it really does feel like using a medium-format camera—slow composition, slow shooting.

    2. “Blinding” Burst Mode
    If you can’t tolerate the slowness of single-shot capture and switch to burst mode, the preview freezes completely while shooting, providing no real-time view of what’s being captured. When photographing moving subjects, this is basically shooting blind.

    3. Crippled Features in Photo Mode
    In photo mode, you cannot use built-in filters, cannot use 2× lossless zoom, cannot adjust texture or noise reduction for JPEG output, and cannot use portrait mode.

    4. Front-Screen Swipe-Up Unlock
    This is the one thing I most want to complain about.

    The swipe-up gesture to unlock the front screen fails at an extremely high rate. Even without gloves—just using bare fingers—I often need to swipe around 10 times before it finally unlocks by luck. The rear screen does not have this issue.

    To this day I still haven’t found a reliable unlocking trick. If anyone knows the secret, please teach me in the comments. When you need to quickly check framing for a selfie, this kind of interaction friction seriously breaks the shooting rhythm.

    5. Low-Angle Shooting Is Slightly Painful
    Although the AC6 is well-suited for low-angle shots (e.g., ground-level perspectives), the lack of a flip screen means you literally have to lie on the ground to see the display. In this regard, it’s definitely less elegant than competitors like the Insta360 Ace Pro 2.

    6. “Computational Compromises” in Video Mode
    When shooting square-format footage, there are still software limitations.

    Resolution lock: When recording video in the square aspect ratio, the resolution is forced to 4K—you cannot switch to 1080p for a more storage-friendly option.

    Feature limitations: Built-in filters and image-enhancement features are disabled in this mode. The 2× lossless zoom is also unavailable.

    Conclusion

    Returning to the question from the very beginning: when an action camera is no longer used for action, what is it?

    The answer may not lie in the spec sheet, but in the people around us.

    On social media, I’ve noticed more and more middle-school and high-school students wearing action cameras around their necks as they move through campus. They use them to record their lunchtime sprint to the cafeteria, to document the battle for a shower stall in the dorms, and to capture the moment they pack up spicy hot pot from the dining hall to save on living expenses.

    Of course, most of what they carry are previous-generation Action 4 units—or even older models. After all, compared to the nearly 3,000 RMB price of the AC6, those older, thousand-yuan devices are far more attainable as a “first camera.” But whether new or old, the reasons they choose this category are the same.

    Here, “ultra-wide” carries two meanings:

    It is the physical 15mm lens, because it can greedily fit every classmate’s face into one frame, contain the embrace at the finish line during the school sports meet, the noise-filled hallways between evening study sessions, and the rising steam of the cafeteria’s hot meals.

    It is also a metaphor for a “wide-open life.” Precisely because of this exaggerated angle of view, those fleeting, irreversible moments of youth can be preserved whole within this small black square.

    The arrival of the AC6 is not so much about pushing users to upgrade, but about revealing the direction this entire category is evolving toward—shifting from a simple “recording device” to an “all-purpose camera.”

    Whether an action camera is used for action doesn’t matter to most people who don’t do sports at all. What matters is that it captures the world a little wider, so that when you look back in the future, you can see more.

    Thank you for reading this far. If you enjoyed my article, feel free to follow.

    I’m 钟桦服—see you in the next piece.

  • SSPAI Morning Brief: OpenAI Releases GPT-5.2 and More

    SSPAI Morning Brief: OpenAI Releases GPT-5.2 and More

    Morning Brief

    1. OpenAI Releases GPT-5.2
    2. Snapmaker U1 Color 3D Printer Launches
    3. Amazon to Allow Authors to Publish Kindle Books Without DRM Protection
    4. OpenAI Announces Integration with Select Adobe Apps
    5. Tencent QQ Group Chats Add Support for Yuanbao AI Summary Feature
    6. Alibaba Releases Qwen3-Omni-Flash Multi-Modal Model
    7. Qualcomm Announces Acquisition of Ventana Micro Systems
    8. Rumors You Can Just Glance At

    OpenAI Releases GPT-5.2

    On December 12, OpenAI announced the launch of its GPT-5.2 model lineup, positioning it as the most suitable model for everyday professional use. According to OpenAI, GPT-5.2 delivers overall improvements in table generation, presentation creation, coding, image understanding, long-context processing, tool use, and complex multi-step tasks.

    The GPT-5.2 series includes three versions: Instant, Thinking, and Pro. Among them, the Thinking model stands out for significantly reducing hallucinations, making it more suitable for professional AI agent applications that require high reliability. Companies such as Notion, Box, Shopify, Harvey, Zoom, and Databricks participated in pre-release testing. OpenAI also noted that GPT-5.2 will play a more central role in AI-agent workflows, with the long-term goal of shaping ChatGPT into a highly personalized intelligent assistant. Source

    Logos for OpenAI and the Walt Disney Company in white on a background of swirling purples, blues, and magenta OpenAI also announced a three-year licensing agreement with Disney. Sora and ChatGPT Images will gain the ability to generate video and image content featuring characters from Disney, Marvel, Pixar, and Star Wars. The related features are expected to roll out in early 2026, with some generated content showcased on Disney+. Source

    In addition, Disney has issued a cease-and-desist letter to Google, claiming that Google used Disney works without permission to train AI models, and subsequently generated and distributed images and videos featuring Disney-related characters for commercial purposes. Source


    Snapmaker U1 Color 3D Printer Launches

    On December 12, the color 3D printer Snapmaker U1 made its debut on JD.com. The Snapmaker U1 features the SnapSwap™ independent four-head parallel system, supporting multi-color and multi-material printing with quick toolhead switching. According to the company, in multi-color printing scenarios the device can switch toolheads in about 5 seconds, improving efficiency and reducing material waste. Compared with traditional single-nozzle color-changing methods, Snapmaker U1’s four-color printing tasks can reach speeds of up to 500 mm/s and complete prints within 5–6 hours. While achieving up to a fivefold increase in overall efficiency, the design also reduces material usage by about 80% thanks to its purge-free channel structure.

    Snapmaker U1 supports automatic calibration, input shaping, pressure advance, automatic bed leveling, and other features to ensure clean edges, smooth walls, and sharp details even in high-speed printing mode. Its printing parameters include a build volume of 270×270×270 mm, along with AI print monitoring and app-based management.

    Snapmaker U1 is priced at 5,299 RMB exclusively on JD.com (4,504.15 RMB after national subsidies). Buyers can still claim an exclusive privilege package before 7 p.m. on December 12, with pre-orders officially opening at 8 p.m. The Snapmaker U1 will also be showcased at an offline experience event at the JD MALL Beijing Shuangjing store from December 16 to 17. Source


    Amazon to Allow Authors to Publish Kindle Books Without DRM Protection

    Amazon has announced that beginning January 20, 2026, authors who choose not to enable DRM (Digital Rights Management) in the Kindle Direct Publishing workflow will be able to offer their e-books to readers in more open formats such as EPUB and PDF. Previously published books will not have their DRM settings changed automatically; authors must manually update their preferences in the KDP dashboard. Once an author updates the DRM status, the change may take up to 72 hours to appear on Amazon’s main site.

    Amazon stated that the new option is intended to make it easier for readers to access content they have purchased. However, some authors believe the move may actually increase the number of people choosing to enable DRM. Source


    OpenAI Announces Integration with Select Adobe Apps

    On December 11, Adobe launched three applications—Photoshop, Acrobat, and Adobe Express—inside ChatGPT.

    The new integration allows users to perform image editing, PDF modification, and various design tasks directly within ChatGPT using natural language, without switching to standalone applications. Users only need to mention the app in conversation and describe what they want; the system will then generate selectable results or present interactive UI controls, such as sliders for adjusting brightness and contrast.

    Although these integrated apps do not offer the full capabilities of their desktop counterparts, Photoshop can still perform local edits, creative effects, and image parameter adjustments; Acrobat can edit PDFs, compress or convert formats, extract text and tables, and merge files; Adobe Express can generate and edit posters, invitations, and social media graphics. All creations can be exported at any time to Adobe’s native software for further refinement.

    According to Adobe, these ChatGPT applications are currently available on desktop, web, and iOS. On Android, only Adobe Express is supported for now, with Photoshop and Acrobat coming later. Source


    Tencent QQ Group Chats Add Support for Yuanbao AI Summary Feature

    On December 11, Tencent QQ announced via its official Weibo account that group chat messages now support the Yuanbao AI summary feature. When a group accumulates a large number of unread messages, users can tap “AI Summary” in the upper-right corner of the chat interface, and—after granting permission—generate a summary of the conversation. QQ also stated that the summary results are visible only to the user, and the related data will not be stored or used in any other context. Source

    Alibaba Releases Qwen3-Omni-Flash Multi-Modal Model

    On December 9, 2025, Alibaba’s Qwen team released the next-generation native full-modal model, Qwen3-Omni-Flash-2025-12-01.

    Upgraded from Qwen3-Omni, the new model supports unified input across text, images, audio, and video, and provides real-time streamed text output and voice capabilities. This update focuses on improving issues in multi-modal interactions—such as unnatural speech performance and unstable transitions in multi-turn conversations—by adjusting speech rate, pauses, and prosody to enhance overall interaction quality.

    Alibaba has also strengthened the controllability of Qwen’s system prompts. Users can now define the model’s behavior patterns, including tone, style, and content length, with improved execution accuracy. In addition, Qwen’s multilingual capabilities continue to expand, now supporting 119 text languages, 19 speech recognition languages, and 10 speech synthesis languages.

    In terms of performance, Qwen3-Omni-Flash-2025-12-01 shows improvements over the previous version on several benchmarks, including ZebraLogic (logical reasoning), LiveCodeBench-v6 (code generation), and MMMU (multidisciplinary visual question answering). Source


    Qualcomm Announces Acquisition of Ventana Micro Systems

    On December 10, Qualcomm announced its acquisition of Ventana Micro Systems, further strengthening its technological presence in the RISC-V domain.

    Qualcomm stated that the Ventana team will complement its existing Oryon in-house CPUs and RISC-V development efforts, jointly advancing the company’s technologies across business lines in the AI era. The company also noted that the RISC-V architecture holds significant potential for driving CPU innovation, and this acquisition marks an important step toward building industry-leading RISC-V CPU technology.

    Founded in 2018 and headquartered in Cupertino, Ventana Micro Systems focuses on high-performance, scalable, and secure computing chip solutions based on the open RISC-V architecture. The company is a board member of RISC-V International as well as a member of its Technical Steering Committee. Source


    Rumors You Can Just Glance At

    • During a televised broadcast, Donald Trump mistakenly described 6G technology as something capable of displaying deeper image or skin-detail layers, failing to understand that 6G actually refers to the next-generation mobile communication network. His comment drew attention on site, while Qualcomm CEO Cristiano Amon—who was on stage with him—did not correct the statement. As early as 2019, Trump had called for the U.S. to accelerate 6G development to avoid falling behind other nations. He has also made past remarks such as “the Moon belongs to Mars” and incorrectly linking Apple to 5G infrastructure construction. Source
    • The startup Operation Bluebird has filed a petition with the U.S. Patent and Trademark Office, arguing that Elon Musk’s X Corp. effectively abandoned the “Twitter” and “Tweet” trademarks after rebranding and discontinuing use of the associated branding, and that the trademarks should therefore be canceled. The company has also applied for a new “Twitter” trademark for its upcoming social platform, Twitter.new. In its petition, Operation Bluebird claims that after the rebrand to “X,” X Corp. no longer uses the original trademarks, citing Musk’s previous public statements about “saying goodbye to the Twitter brand forever.” Some legal experts note that although the claim has some basis, the “residual goodwill” of the Twitter brand could be a key factor, as many users still refer to X as “Twitter” and continue using terms such as “tweet.” X Corp. must submit a response by February next year. Operation Bluebird stated that it stands firmly by its position and will continue pursuing relevant legal actions. Source
    • According to tech outlet MacRumors, internal firmware for iOS 26 reveals:
    • The new AirTag is expected to receive upgrades across pairing workflow, battery notifications, and tracking capabilities. In particular, it aims to deliver more accurate location tracking when the device is in motion or in crowded environments. The new AirTag may also feature a next-generation ultra-wideband (UWB) chip, expanding the range of Precision Finding. There have been no clear rumors of noticeable exterior redesigns; the overall appearance may remain consistent with the current model, though improvements to the speaker structure could reduce risks of malicious tampering or misuse. Although iOS 26 refers to the device as “2025 AirTag,” current product timing suggests its official release may slip to early 2026. Source
    • Apple is advancing development of a new smart home hub device. The product will include a camera and support Face ID for identity verification and user recognition, enabling automatic switching between user profiles. The hub is expected to rely heavily on Apple Intelligence and the next-generation Siri set to launch next year. Based on current reports, this smart home hub may take a form somewhere between an iPad and a HomePod, featuring an approximately 7-inch square display and an optional speaker dock. The device is expected to debut alongside iOS 26.4 and the updated Siri.
    • Code references also reveal an unreleased accessory device codenamed J229, believed to be a home security camera under development by Apple, though its release timeline has not yet been disclosed. Source
  • SSPAI Morning Brief: AMD Launches New Driver Supporting FSR Redstone and More

    SSPAI Morning Brief: AMD Launches New Driver Supporting FSR Redstone and More

    Morning Brief

    1. AMD launches new driver supporting FSR Redstone
    2. iFixit launches AI assistant FixBot
    3. Insta360 launches Insta360 GO Ultra mini time-lapse kit
    4. Fairphone releases new Fairbuds XL modular headphones
    5. Alibaba Tongyi Qianwen releases Qwen Code v0.3.0
    6. JetBrains announces discontinuation of Fleet product line
    7. State Administration for Market Regulation issues two mandatory standards for civilian drones

    AMD launches new driver supporting FSR Redstone

    On December 10, AMD announced the launch of a new graphics card driver supporting FSR Redstone, designed to reshape gaming graphics and performance using machine learning capabilities. FSR Redstone is a new ML real-time rendering technology, which AMD unveiled earlier this year. Its goal is to enhance lighting, reflections, and motion smoothness without sacrificing frame rates.

    FSR Redstone includes three core features:

    1. Radiance Caching: By predicting indirect lighting, it reduces ray-tracing computation, making overall scene lighting more cinematic.
    2. Ray Regeneration: This feature regenerates ray-traced details, making reflections, shadows, and material edges sharper and more realistic.
    3. ML Frame Generation: Using high-intensity gaming training models, it generates additional frames for smoother high-speed motion performance. Source
    Compatibility status of FSR Redstone, image captured from AMD website

    iFixit launches AI assistant FixBot

    On December 9, repair expert team iFixit introduced FixBot, an AI-powered repair assistant designed to significantly lower the barrier to device repair through the use of artificial intelligence.

    The assistant integrates visual schematics, professional repair guides, and expert-level troubleshooting techniques. To address the common issue of users “not having a free hand” during repairs, FixBot places special emphasis on hands-free guidance, offering step-by-step voice instructions for a truly interactive repair experience.

    Unlike general-purpose AI models, FixBot’s intelligence is built on iFixit’s extensive vertical-domain database. Over more than 20 years, the team has constructed the internet’s largest repository of repair procedures, covering more than 72,000 products—from smartphones such as the iPhone to motorcycles and beyond.

    iFixit has not yet announced subscription pricing, but to promote the service, all paid premium features are currently available to users for free. iPhone users can now download the app and try the service. Source


    Insta360 launches Insta360 GO Ultra mini time-lapse kit

    On December 9, Insta360 introduced the Insta360 GO Ultra mini time-lapse kit, featuring a multifunctional pre-recording charging base and a pre-recording neck-mounted battery holder. With a built-in 4800mAh battery, the kit is designed to make first-person POV shooting more convenient.

    In terms of hardware, the camera is equipped with a 1/1.28-inch CMOS sensor and a 5nm AI chip, supporting up to 4K 60FPS high-frame-rate video recording. It includes a magnetic clip and magnetic neck lanyard, allowing hands-free first-person filming.

    On the feature side, the camera offers a dedicated night-video mode with AI noise reduction, brightness restoration, and dynamic-range optimization to deliver clean, bright footage in low-light environments. It includes a variety of built-in popular filters for easy straight-out-of-camera results; supports 4K 2× lossless zoom and HDR video; and can capture 50-megapixel photos. It also features Live Photo mode, seven built-in photo filters, and a new layered battery design. With the expansion module, it can record up to 200 minutes of 1080P footage, or 70 minutes on its own. A microSD card slot is included.

    The camera body, display module, and battery compartment can be detached to unlock more shooting angles. It comes with a 2.5-inch flip touchscreen, and when separated, supports wireless image transmission control. The kit is priced at 2,878.2 RMB. Source

    Product promo image, via Insta360

    Fairphone releases new Fairbuds XL modular headphones

    On December 9, Fairphone introduced the 2025 edition of the Fairbuds XL over-ear headphones. The new model upgrades its driver units while maintaining the modular, easy-to-repair design of the 2023 original. Pricing also remains unchanged at 229 USD.

    The Fairbuds XL (2025) replaces the PU leather used on the 2023 model’s headband and ear cushions with mesh and bird-eye mesh materials to improve breathability and comfort. It supports a 30-minute auto-shutdown and extends the warranty to three years. The headphones offer a more balanced stereo output, support Qualcomm aptX HD 24-bit wireless audio, include active noise cancellation and transparency mode, and deliver up to 30 hours of wireless playback.

    Fairphone notes that the magnets used in the new headphone drivers are made from 100% recycled rare-earth materials. The product also incorporates 92% recycled aluminum and solder paste made with 100% recycled tin, while the entire assembly process is powered by renewable energy. In addition, Fairphone will begin offering Fairbuds XL upgrade components starting at the end of Q1 2026. Source

    Product image via press source

    Alibaba Tongyi Qianwen releases Qwen Code v0.3.0

    On December 9, Alibaba’s Tongyi Qianwen officially released version 0.3.0 of its AI command-line tool, Qwen Code. The update adds full support for Stream JSON, significantly upgrades internationalization capabilities, and strengthens overall security and stability. Ecosystem compatibility has also been improved, with added support for ModelScope Provider and enhanced handling of stream_options.

    Qwen Code is a command-line AI workflow tool optimized for the Qwen3-Coder model. By leveraging advanced code understanding, automated tasks, and intelligent assistance features, it aims to boost developer productivity. Source


    JetBrains announces discontinuation of Fleet product line

    On December 8, JetBrains announced that it will discontinue distribution and updates for the Fleet project. Beginning December 22, 2025, Fleet will no longer be available for download through the Toolbox App or other channels. Existing users may continue using previously downloaded versions, but server-dependent features—such as the AI assistant—may gradually stop functioning. The Fleet team has decided to cease competing with existing IDEs and code editors, and will instead build a new product focused on agent-centric development workflows, using the Fleet platform as its foundation. This new product will adopt a new name, inherit Fleet’s technical groundwork and team expertise, but shift entirely in terms of target market and product positioning. JetBrains promises to continue sharing updates on the progress of this new agent development environment. Additionally, JetBrains announced that starting with version 2025.3, the Islands theme will become the default appearance for JetBrains IDEs. Source


    State Administration for Market Regulation issues two mandatory standards for civilian drones

    On December 9, the State Administration for Market Regulation approved and released two mandatory national standards: Requirements for Real-Name Registration and Activation of Civil Unmanned Aircraft (GB 46761—2025) and Specifications for Operational Identification of Civil Unmanned Aircraft Systems (GB 46750—2025). Both standards will take effect on May 1, 2026.

    Among them, the Requirements for Real-Name Registration and Activation of Civil Unmanned Aircraft standard outlines the workflow for real-name registration and activation of drones. It provides detailed rules on registration entities, registration management and inquiry, deregistration, data exchange interfaces, and security protection measures. It explicitly requires that drones must not have flight capability before activation or after deactivation. The Specifications for Operational Identification of Civil Unmanned Aircraft Systems standard mandates that drones must actively transmit their identity, location, speed, status, and other information to regulators upon startup and throughout the entire flight, enabling authorities to monitor flight conditions in real time.

    The implementation of these two mandatory standards supports the effective enforcement of the Interim Regulations on Unmanned Aircraft Flight Management, providing crucial safeguards for the safe and orderly development of the drone industry. Source

  • HRV: Something You May Be Curious About But Don’t Fully Understand Yet

    HRV: Something You May Be Curious About But Don’t Fully Understand Yet

    Author’s Note: The initial inspiration for this article, as well as the research data and theories referenced throughout, comes primarily from my instructor Dr. Sylvain Laborde at the German Sport University Cologne. He has long been active on the front lines of HRV research and application. My sincere thanks to Dr. Laborde.

    The author is a current master’s student in the Department of Sport Psychology at the German Sport University Cologne (Deutsche Sporthochschule Köln). The author has no financial ties to any researchers, devices, or app providers mentioned in this article.

    Introduction: Starting With Questions You May Not Know You Have

    What is HRV?

    You might know that HRV stands for Heart Rate Variability, and that its Chinese translation is “心率变异性.” You may know that your wearable device can measure HRV; that the human heart is not a metronome; and that HRV represents the variation in time intervals between consecutive heartbeats. You might also know that trends in HRV relate to health or “readiness.”

    You may also know that HRV levels vary from person to person and cannot be compared across individuals; and perhaps you’ve heard about how and when HRV should be measured.

    But beyond that—how much do you really understand about HRV?

    Do you focus only on the HRV value changes shown under “Health” on your device, without knowing the evidence-based meaning behind an increase or decrease in HRV? Do you know that your device claiming to measure HRV might actually be estimating it?

    This highly sensitive and constantly fluctuating metric—when exactly should it be measured, and under what conditions? How should it guide lifestyle adjustments, including—but not limited to—exercise? And going a step further, are there specific forms of training that can enhance or improve HRV?

    If these questions interest you, then this article (and the follow-up pieces to come) may be well worth your time.

    HRV and the Parasympathetic Nervous System: A Reflection of Decision-Making and Adaptive Capacity

    In a world full of constant change, we need the ability to regulate our physical and mental states in response to shifting conditions, and to make decisions and take actions that best serve our goals and interests. HRV is an important metric that quantifies this adaptive capacity.

    The autonomic nervous system has two branches: the sympathetic nervous system, which acts as the “accelerator,” and the parasympathetic nervous system, which serves as the “brake.” This is common knowledge. But do you know which system is more important for modern-day adaptation?

    The answer is the parasympathetic nervous system.

    Take heartbeats as an example. Our heartbeat is governed by nerve signals originating from the medulla and realized through the sinoatrial node (SA node) in the heart. The SA node functions as the body’s natural “pacemaker.” Its intrinsic firing rate is 100 beats per minute. In other words, if nothing else intervened, our resting heart rate would naturally be 100 bpm. It is the parasympathetic nervous system—the “brake”—that slows a healthy individual’s heart rate down to 60–80 bpm, or even lower.

    With this brake engaged, we maintain a lower heart rate and deeper calm during rest, conserving energy and reserving resources for unexpected changes. Counterintuitively, when faced with a sudden threat, our body does not always activate the sympathetic “fight or flight” response immediately. Such instinctive, primitive reactions can often lead to very poor decisions.

    Why would a world-class athlete make an amateur mistake at the most crucial moment of their career? One possible reason is the overwhelming physical and mental stress that triggers a “fight or flight” response.

    More often—and more adaptively—when encountering a challenge, we first hit the brake and engage the parasympathetic nervous system. Our heart rate slows; our body enters a “freezing” state. In this state, our senses sharpen and our mind becomes more focused, allowing us to make better decisions. Once a decision is made, we release the brake and hit the sympathetic “accelerator” to carry out the most effective action.

    A quick side note: some people “freeze” when facing danger. This doesn’t necessarily mean they lack courage. More likely, the danger is too great or too unfamiliar (requiring more processing time), or their ability to smoothly switch between parasympathetic and sympathetic activation is underdeveloped. These limiting factors can and should be trained. There is no need to judge yourself harshly for freezing in fear.

    The caption: “Left figure from Karin Roelofs and Peter Dayan (2022); the viewpoint on the right is my own summary based on Krav Maga training practice.”

    Back to the main topic. The “brake” signal from the parasympathetic nervous system reaches the heart through the tenth cranial nerve—the vagus nerve (C10). Eighty percent of vagal fibers carry information to the brain, while the remaining 20% carry information from the brain to the body. All of this 20% belongs to the parasympathetic branch. Their electrical signals intervene in the firing rhythm and waveform of the sinoatrial node, creating variations—these variations form the basis of the HRV signal.

    Image source: https://www.daniellewis.com.au/health-information/stress-and-anxiety/vagus-nerve/

    If the parasympathetic system is functioning well and vagal efferent activity is strong, it sends a stronger “brake” signal. This leads to larger and richer variations in the intervals between heartbeats, which is observed as increased HRV. In such moments, the body spends more time in a state of rest and relaxation.

    Conversely, if vagal activity is inhibited—or if sympathetic influence increases while parasympathetic activity weakens—the intervals between heartbeats become more uniform, resulting in decreased HRV. In this state, the body leans more heavily into the “fight or flight” mode—in other words, a heightened state of tension.

    When the body is tense and the parasympathetic system is suppressed, decision-making quality and executive function decline. Moreover, people who remain in a tense state for prolonged periods experience elevated inflammation levels. Evidence suggests that vagal afferent fibers can detect increases in inflammatory markers in the body, while vagal efferent fibers can activate the immune system—especially the spleen—through the noradrenergic system, prompting T cells and macrophages to combat inflammation (Bazoukis, Stavrakis & Armoundas, 2023). In other words, the vagus nerve and the parasympathetic system not only act as a “brake” but also as a “firefighter.”

    In summary, HRV positively reflects the function of the parasympathetic nervous system and indicates the body–mind system’s adaptive capacity when faced with environmental changes. For this reason, tracking HRV allows us to understand and assess our adaptability and overall “health status” in a broad sense.

    Are You Really Measuring HRV? : The Principles Behind HRV Measurement, and How to Choose Devices and Software

    There are many consumer electronic products on the market claiming to measure HRV—smartwatches, fitness bands, smart rings, even smart mattresses. Focusing solely on HRV measurement, how reliable are these devices, and how much of what you’re paying is essentially a “tech tax”? To answer this, we need to begin with the basic principles behind HRV measurement.

    If a device isn’t mentioned in this article, it’s clearly not being recommended. If it is mentioned, that doesn’t necessarily mean it is recommended.

    The term HRV itself is somewhat misleading, because HRV does not measure variation between two “heartbeats.” Instead, it measures variation between two cardiac cycles—more precisely, between two R-peaks on an electrocardiogram (ECG). A more accurate term might be Heart Period Variability (Quigley et al., 2024). But since “HRV” is already widely used, we’ll stick with it for convenience. The key point is: HRV measurement fundamentally depends on collecting ECG data. Therefore, only devices that can record and analyze ECG waveforms can truly measure HRV.

    At the very least, it needs to look something like this…

    The most typical ECG monitoring devices are the multi-lead systems used in hospitals and laboratories. Clinical HRV monitoring often relies on such multi-lead setups, collecting long, continuous ECG recordings and calculating average interval variation between R-peaks. This method produces the gold-standard HRV data. Consumer wearables cannot provide multi-lead ECG recordings and cannot operate continuously for 24 hours. Whether it’s an Apple Watch, Garmin band, or Oura Ring, they all rely on a compromise solution.

    A common method they use is photoplethysmography (PPG). Devices using PPG emit green light into the skin to detect periodic changes in blood flow (pulse). They then infer HRV by analyzing the variability of pulse intervals—what we might more accurately call “PRV” (Pulse Rate Variability). Next time you open the HRV or heart rate feature on your smartwatch or fitness band, lift the device and check whether the back is flashing green. In short: your device does not record ECG data—not even close. It measures pulse changes and then uses algorithms to approximate HRV.

    The back of the Apple Watch

    Different manufacturers use different sensors with varying quality and parameters. Cheaper or smaller sensors may be inaccurate to begin with. PPG accuracy depends heavily on stable contact with the skin, meaning that in non-resting conditions—especially during movement—accuracy drops significantly due to micro-slips between the device and the skin. Combined with HRV’s inherent sensitivity to measurement conditions (discussed in the next section), HRV “averages” that mix rest and activity data can be highly inaccurate and misleading. Additionally, different brands use different sensor models and algorithms, making HRV data across devices incomparable. Stop comparing your Apple Watch HRV with your friend’s Oura Ring—it simply doesn’t make sense.

    Perhaps for these reasons, major HRV software platforms do not import data from many watches, bands, or rings—because these devices do not measure HRV, and their estimates aren’t reliable enough. As technology improves, consumer devices using PPG may eventually achieve HRV inference with high enough accuracy. But aside from quality and algorithm guarantees, they will require extensive ECG comparison data to validate their accuracy. Until such evidence exists, I cannot recommend using these devices to monitor your HRV. Wear them for fun, check heart rate, tell the time—that’s all fine. But that so-called HRV number? Don’t take it too seriously.

    Polar’s native app also tracks exercise load, which is not bad

    Among consumer-grade electronic devices, the most reliable tools for actually measuring HRV—with acceptable accuracy—are chest-strap heart rate monitors. The most common model on the market is the Polar H10. This device is supported by a wide range of measurement and analysis apps, and it is the preferred wearable HRV device among researchers at the German Sport University Cologne. The H10 consists of a detachable sensor and an elastic chest strap. To use it, simply snap the sensor onto the strap and fasten the strap firmly around the lower chest.

    I currently pair the H10 with two main apps: HRV4Training and Elite HRV. HRV4Training allows you to measure and store one HRV reading per day (if you measure again, the new data replaces the old). Combined with a post-measurement questionnaire, HRV4Training helps track correlations between HRV and daily habits such as exercise, alcohol intake, and sleep, as well as the long-term trends in your HRV.

    Image from Altini, M., & Plews, D. (2021).

    If you need continuous HRV measurement over a specific time window within the day—such as during and after a workout—Elite HRV is the tool for the job. It can track your real-time HRV curve throughout the measurement session and generate data logs. Elite HRV data can also be exported into desktop software such as Kubios HRV, allowing you to build and manage detailed “HRV health profiles” for yourself or others, tied to training and lifestyle patterns.

    The non-commercial version of Kubios HRV is free and can directly import data exported by Elite HRV.

    Across various apps (including Apple Health’s “Heart Rate Variability” section), the most commonly used HRV indicator is RMSSD (Root Mean Square of Successive Differences). RMSSD represents the average variance in time between consecutive heartbeats over a given period, measured in milliseconds (ms). Because it is less affected by breathing and is time-based, RMSSD reflects vagal influence on heart rate quite intuitively. The higher your RMSSD, the higher your HRV—and the stronger your body’s adaptive capacity.

    Although we shouldn’t (and don’t need to) compare ourselves to others, it’s hard to resist the urge when we see our own RMSSD value. HRV4Training provides normative data from all its users, separated by age groups. If you collect your HRV using a chest strap or ECG device, you can compare your readings to these norms. Kubios HRV automatically calculates your “parasympathetic recovery index” and “sympathetic stress index” from the imported data and compares them with its built-in norms. There is currently no nationwide HRV normative sample for China. However, a 2018 Dutch study collected RMSSD averages from 13,943 participants (van den Berg et al., 2018). You can check how your adaptive capacity compares to Dutch individuals in your age group.

    Is Your HRV “Good” or “Bad”?

    HRV as a Resource: The Vagal Tank Theory and the “3Rs”

    HRV does not directly reflect how much stress we are under, nor does it correspond in a strict one-to-one manner with indicators such as cortisol levels or gut microbiota composition. HRV—especially RMSSD—reflects the functional state of the parasympathetic nervous system and the availability and condition of our resources for coping with stress.

    With this understanding, we can naturally think of HRV as an internal “battery” or “energy bar.”
    When we are relaxed, the energy tank is full. As life events unfold, the tank drains bit by bit. When it runs low, we experience physical and mental fatigue, increased subjective stress, and diminished self-regulation. If the tank fully depletes, the body and mind may “shut down,” much like a smartphone that suddenly turns off with no battery.

    This is precisely the core of Laborde’s Vagal Tank Theory (Laborde, Mosley & Mertgen, 2018b). The theory revolves around the 3Rs. The first R, Resting, refers to our baseline parasympathetic activity—the “full tank” state. The stronger and healthier you are, the higher your baseline HRV (explained below), the higher your upper limit for resource capacity, and the better your ability to deal with stress.

    The second R, Reactivity, describes how we expend our tank when responding to challenges in life. Your challenge might be a difficult math problem, a fight with a friend, or a literal pile of bricks. In such moments, the vagus nerve “opens the valve,” draining the tank (“consuming blue”), allowing you to mobilize physical and mental resources to solve the problem.

    Once the challenge ends, we enter the third R, Recovery, where we replenish our energy tank. Some people recover poorly or use ineffective methods—their bodies behave like an outdated phone battery that never charges fully, or worse, becomes swollen. Others recover efficiently or use the right strategies, allowing the tank to refill quickly, sometimes even exceeding baseline—raising the upper capacity of the tank.

    The 3Rs vividly illustrate why we must resist burnout culture, reject 996 schedules, and avoid endless self-exploitation. If we constantly drain our tank in the name of productivity but fail to replenish it, our bodies and minds will inevitably collapse. Even if we force ourselves to continue, we will become like a phone that never reaches a full charge—always hovering near empty until one day it can no longer charge at all.

    I personally experienced such a collapse a little over a month ago. For a week and a half, I attended classes during the day and worked at night, with no rest days and less than six hours of sleep daily. Then I spent Friday to Sunday at the Essen Game Fair, commuting four hours a day, working twelve, and sleeping only four. After returning, I caught a mild cold but didn’t rest; instead, I continued the same schedule for another two weeks. After a month of this, my body finally fell apart—diarrhea, fatigue, low fever, chills, and insomnia. But none of this was due to infection—it was simply collapse from lack of rest. On my worst days, my HRV dropped to nearly single digits (which is borderline “corpse mode”).

    Realizing this was unsustainable—and life-threatening—I completely rested for two days and then significantly reduced my workload. After about two weeks of gradual recovery, coinciding with a trip to Japan, my HRV finally returned to—and exceeded—its baseline level.

    “Recovery” Doesn’t Mean You Must Lie Still: Doing What You Love Can Also Be Rest

    This is my personal experience and not something to generalize to everyone. I share it to illustrate that our energy tank is not infinite, and the consequences of draining it completely can be severe. To allow the tank to refill, we must grant ourselves adequate rest and recovery. When a battery swells, you can replace the battery—or even the device. But when your mind and body “swell,” sorry to say, with the little money earned through burnout culture, you truly can’t afford to replace anything.

    When talking about rest and recovery, the natural questions arise:
    What kinds of activities drain us the most?
    How can we tell when our energy tank is nearing depletion?
    How do we know if we’ve rested “enough”?
    Which activities help the tank recover (or even increase) most effectively?

    The answers to these questions can all be found through HRV. By measuring and tracking our HRV across time, we can clearly see which activities consume the most energy—and which forms of rest replenish it the fastest.

    More Sensitive Than You Think: Measuring and Tracking HRV

    That said, you cannot simply buy a wearable device, strap it on, and trust the HRV number it “measures.” Numerous studies (Fatisson, Oswald, & Lalonde, 2016; Laborde, Mosley & Mertgen, 2018a) show that HRV is extremely sensitive. Daily physiological fluctuations—including blood pressure, body temperature, and endocrine activity—affect HRV. Respiration and heart rate especially interfere with HRV measurements. If your device cannot even filter out these sources of noise, its readings are essentially meaningless.

    Age, sex, weight, BMI, and genetics (including ethnicity) also influence HRV. For example:

    • HRV declines steadily with age.
    • Women generally have higher HRV than men in the same age group.
    • People with obesity tend to have higher resting heart rates and lower HRV.
    • Travel, frequent commuting, or increased training loads tend to lower HRV.
    • Illness, alcohol consumption, medications, and circadian rhythms can significantly reduce HRV.

    To correctly interpret your HRV, you need to not only measure it but also know what state your body is in at the time of measurement.

    Alcohol significantly lowers HRV (Altini & Plews, 2021)

    But that’s still not the whole picture. HRV is so sensitive that even small changes in your physical state during measurement can cause immediate fluctuations in the recorded values. In class, Laborde once led us through a live demonstration. Several students (myself included) wore heart-rate chest straps and opened Elite HRV to track our HRV in real time. We first lay flat on the floor for two minutes, then quickly stood up, and after holding a standing posture for one minute, quickly sat down again. During those brief moments of “standing up” and “sitting down,” our HRV values shifted dramatically. Afterwards, we switched between different physical activities and breathing patterns—and every switch produced significant changes in HRV.

    For this reason, if you want valid HRV data, you must ensure that measurements are taken under the most consistent conditions possible. If you measure HRV at different times of day, in different physical or mental states, or even in different body postures, the readings may become inconsistent or meaningless.

    If we want to measure HRV once per day, keep each measurement within 3–5 minutes, and use long-term trends as the key indicator, then the best approach is to measure at a consistent time and in a stable physiological state. The ideal moment is right after waking up.
    Here is the HRV measurement routine I currently follow:

    1. Alarm rings (wake up)
    2. Put on clothes, get out of bed, go to the toilet (morning urination)
    3. Wash hands—but do not brush teeth or wash your face yet
    4. Return to the bedroom, put on the heart-rate chest strap, sit down, and regulate breathing for one minute
    5. Open HRV4Training and measure HRV (default duration: one minute)
    6. Fill out the behavior-related questionnaire in HRV4Training
    7. Remove the chest strap and continue with morning activities (brushing teeth, washing face, breakfast, etc.)

    If using a chest strap is inconvenient, you can use an arm-worn optical band (e.g., Rhythm 24). HRV4Training currently supports reading HRV data from the Apple Health app. However, as noted earlier, because the Apple Watch measures HRV multiple times per day—and wrist-based HRV accuracy during movement is limited—the best approach (if you only have an Apple Watch) is to follow the same morning routine and measure HRV as soon as possible after waking. Alternatively, you can enable continuous overnight HRV tracking so the watch records your HRV throughout sleep.

    May 21 was the first day after I returned to Germany from intensive training and long-distance travel from Israel. With your HRV data and your personal diary, you can regularly review what’s been happening in your life and how those events have influenced you.

    After collecting at least 4–6 weeks of data, HRV4Training will let you review your HRV trends and gradually understand how your exercise, rest, and lifestyle habits affect your parasympathetic system and overall “adaptive capacity.” With this foundation of data and knowledge, you can begin using HRV to guide your training and undertake targeted practices to improve HRV and adaptability. For details, please see my next article.

    Some Key Takeaways (Take-home Message)

    If you forget all the technical terms after reading this article, at least take with you the following core understandings about your body’s “energy tank”:

    • HRV is the variability of your heartbeat rhythm—and also your capacity to adapt to the world. Don’t treat it as a simple health score, and don’t become anxious by comparing it with others. High HRV means your body has the resources and ability to respond more effectively to challenges—whether that’s a major exam, high-intensity training, or a night of poor sleep. Low HRV is a warning that your “blue bar” is nearing zero, and you should avoid making major decisions or performing demanding tasks in that state.
    • Your body needs the brake more than the accelerator. Modern life doesn’t lack stimulation. If something goes wrong with your physical or mental state, it’s unlikely because your “accelerator” isn’t working—but very likely because your “brake” has gone rusty from underuse. Your HRV reflects the energy level of your vagus nerve and parasympathetic nervous system (the brake). This precious braking system not only keeps you calm, but also quietly helps fight inflammation and repair your immune system. Maintaining and improving HRV is essentially maintaining the brake pads that keep you alive.
    • Manage your “energy tank” wisely: resting isn’t laziness—it’s strategic action. Remember the Vagal Tank Theory and the 3Rs (Resting, Reactivity, Recovery). Your energy tank is finite. Whenever you must expend significant energy (Reactivity) to face life’s challenges, you must follow up with high-quality recovery. When HRV remains low over time, lying down to rest isn’t weakness—it is the smartest, most necessary tactical adjustment to prevent system failure.
    • Reject “junk data”: good decisions require accurate measurements. HRV is extremely sensitive—affected by breathing, posture, mood, and even tiny movements. Not all devices can measure HRV accurately. Don’t overly trust your wristband, smartwatch, or ring—their HRV data may be estimates, not measurements. If you want to use HRV to guide training or understand your body, use more accurate equipment such as a chest-strap monitor, and collect HRV under fixed conditions during a consistent “morning routine.”
  • App+1|Locus GTD: Could this be the OmniFocus “Equivalent” You’ve Been Looking For?

    App+1|Locus GTD: Could this be the OmniFocus “Equivalent” You’ve Been Looking For?

    As a heavy addict of hoarding to-do apps, I’ve always had more than ten different task managers installed on my phone—ranging from cross-platform tools like TickTick and Todoist to Apple-ecosystem exclusives like OmniFocus 4 and Things 3. Every time I discover a new to-do app, I download it immediately, try it out, and sometimes even buy it on the spot. And with the rise of AI vibe coding, the App Store has seen an explosion of new indie to-do apps, most of which are hastily made and nowhere near usable.

    So after all the switching, circling, and experimenting, the only app I’ve managed to stick with has been OmniFocus—until I came across an app called Locus GTD.

    Locus updates frequently, but there still aren’t many reviews.

    When I first found it, I searched Xiaohongshu to see if existing users had shared any experiences. To my disappointment, I couldn’t find a single post about it. Back in the App Store, I realized it had only recently launched, and the developer hadn’t promoted it anywhere. Riding the wave of excitement, I made a post myself. Just like I said in that post: I finally found an app that can genuinely “replace” the OmniFocus I both love and hate.

    I even posted about it on Xiaohongshu when I first started using it—I wonder if that brought the app any new users.

    Can it really replace OmniFocus?

    Before sharing my thoughts on Locus, I want to first explain why I’ve stuck with OmniFocus for so many years:

    • A complete GTD workflow. Among mainstream productivity apps, OmniFocus is probably the only one that fully adheres to the GTD collect–process–review flow without adding unnecessary extras (such as habit tracking or Eisenhower quadrants).
    • Unlimited task hierarchy. Because of my work, I deal with all kinds of complex tasks. The ability to nest tasks infinitely is absolutely essential to me—and it’s the primary reason I never chose Things 3 for long-term use.
    • Perspectives. This is OmniFocus’s killer feature, ensuring that I always know which tasks require my attention at any given moment, with maximum efficiency.

    But with the update to OmniFocus 4, my dissatisfaction has only grown.

    First, the interaction design: to this day, I still haven’t adapted to its page-switching logic. And with the adaptation to Apple’s Liquid Glass design, the strange interactions combined with frequent stutters have made using the app feel nothing short of disastrous.

    Second, syncing: after Jianguoyun WebDAV stopped working years ago, I switched back to the official sync service. Anyone who has used it knows how unreliable it is in mainland China. Sure, I keep a VPN on constantly, but it still feels uncomfortable relying on such a fragile system.

    My love–hate relationship with OmniFocus is precisely why I was so excited when I discovered Locus GTD—it practically fulfills all my needs:

    • A complete (and in some ways better) collect–process–execute–review workflow
    • Unlimited nested projects and tasks
    • Custom filters (Perspectives)

    At the same time, Locus GTD solves two major pain points: it is built entirely on iOS native interaction design, and it supports iCloud sync!

    Over the past few months of using it, I’ve sent the developer many emails with bug reports and feature suggestions. I also learned that they currently have no plans to promote the app on any platform. So I volunteered—after receiving their permission—to write this article for 少数派 and share my experience. After all, most of the tips and expertise I’ve gained for similar apps came from here, and I want to return the favor by sharing my own findings.

    I’m not a professional writer, nor am I a GTD theorist, and I don’t intend to write a “tutorial.” First, Locus’s overall product logic is very similar to OmniFocus, Things 3, and others. Second, my goal is simply to highlight the details and interactions that impressed me during actual use. If you want to understand the full workflow of the app, I encourage you to download and try it yourself. My biggest selfish hope is that more people join me in pushing the developer to keep building new features.

    A Thoughtful Inbox Processing and Project Review Flow

    According to GTD principles, the countless thoughts squeezed into our minds quietly drain our energy and attention. That’s why GTD encourages us to record everything—no matter how big or small—into an “inbox,” clearing our minds so we can focus on the task at hand, and then return later to process whatever has accumulated.

    GTD also requires regular “review” of ongoing projects, updating their status and pushing them forward until completion.

    Locus GTD is the only app I’ve used that truly treats these two GTD rituals as core user experiences and weaves them deeply into product design. Take “processing the inbox” as an example: in most apps, the workflow is dry and mechanical—you scroll through a long list, tap dates, choose tags, assign projects. Locus, however, offers perhaps the most elegant “inbox clearing” experience on iOS. It turns inbox processing into an immersive “card game”: the screen shows only one task card at a time, forcing you to truly confront the item in front of you. What is this? Does it need to be done now? What information should I add?

    You process your inbox task-by-task through intuitive “card swiping.”

    Once a task is processed, you move on to the next card. Instead of feeling like I’m sorting things just for the sake of sorting, I actually enjoy the sense of clearing my inbox. The entire flow is smooth and gratifying. It dramatically reduces the problem of my inbox piling up endlessly, and it helps me escape the helplessness and avoidance that often arise when facing a mountain of unprocessed tasks.

    A similar interaction is used for the project review flow as well. I won’t repeat the details here—you can experience it yourself.

    Unlimited Hierarchical Structure

    Every time I get fed up with OmniFocus and try switching to Things 3 for its top-tier interaction design, I end up abandoning it after a few days because my projects are simply too complex. Things 3 supports, at most, a four-level hierarchy—Project → Heading → Task → Checklist—and that’s far from enough for my workflow.

    Locus GTD, however, supports unlimited hierarchy in two dimensions: At the task level, subtasks can be nested infinitely. At the project level, the structure is Category → Folder (also infinitely nestable) → Project.

    With a drag gesture, you can rearrange tasks or adjust their hierarchy.

    Locus GTD also puts a lot of thought into the interaction for hierarchy management: long-press and drag to reorder tasks, and hold on a task for one second to convert it into a subtask. Although in terms of fine details and smoothness, it still lags behind Things 3, overall, “drag-to-create unlimited hierarchy” combined with “excellent animation and interaction” creates a satisfying balance between the power of OmniFocus and the elegance of Things 3.

    I believe this is also one of Locus’s core design philosophies: striking a balance between powerful functionality and graceful interaction.

    Custom Filters with Huge Potential

    The custom filters in Locus are essentially the same as Perspectives in OmniFocus: by combining a series of conditions, you can filter out exactly the tasks that matter most at the moment. For example, you might create a perspective called “Tasks that require a computer and take less than 30 minutes.” When you happen to be at your computer with a 30-minute window, you can open this perspective to see what you can do right now.

    Perspectives have long been one of OmniFocus’s killer features, and they were exclusive to the Mac version—you can view them on mobile but cannot create them. Locus brings this capability to the mobile side: I can create custom filters directly on my iPhone. Although the available conditions are still not as extensive as OmniFocus, given how frequently the app updates (Locus has been updated nearly twice a week so far), I believe this feature has enormous room to grow.

    Creating complex filtering rules is easy and intuitive on the phone.

    Others

    Finally, I want to mention Locus GTD’s syncing experience. Since it natively supports iCloud, we don’t need to register or log into any account, nor do we ever need to think about syncing. All data changes sync to iCloud almost instantly. For someone who has suffered for years under OmniFocus’s syncing limitations, this smoothness feels almost unreal—I often find myself opening the settings page just to double-check that sync really worked.

    The one drawback is that Locus currently only supports iPhone. The iPad and Mac versions are not yet fully usable (though you can run the iOS version on them), so iCloud sync still can’t reach its full potential. I sincerely hope the developers can release cross-platform versions soon.

    Wishlist

    Even after all the praise above, I have to admit that Locus GTD, as a relatively new app, still has plenty of room to grow. So here’s a small wishlist—hopefully the developer will see the requests of a devoted user like me.

    • A due date for projects. This is absolutely essential!
    • A data statistics feature. Since the app already supports powerful filters, adding analytics and chart visualizations shouldn’t be too difficult, right?
    • An improved UI design. I’m picky about aesthetics—while the current UI is clean, I still feel there’s room for refinement.

    Lastly

    Locus offers monthly, yearly, and lifetime subscriptions. The lifetime version, priced at ¥328, includes all future cross-platform versions (though the developer hasn’t given a timeline for their release). The subscription prices fall somewhere between mainstream to-do apps and OmniFocus. Considering the existing features and the app’s rapid update cycle, I personally think it’s well worth it—so I bought the lifetime version early on.

    If you value native iOS experiences, believe in the GTD methodology, or simply want a beautiful, smooth, and intuitive task manager, Locus GTD definitely deserves a place in your Dock. Borrowing a line from the developer behind the paywall: they hope Locus GTD will become “the last productivity tool you’ll ever need.”

    Let’s wait and see—together.

  • Is AI a Research Method?

    Is AI a Research Method?

    Question

    On December 5, 2025, at the invitation of Vice Dean Fan Zhenjia, I returned to my alma mater, Nankai University, to give a talk on “AI-Assisted Research” to faculty and students from the School of Information and Communication and the Business School.

    During the Q&A session after the lecture, Professor Li Ying, who was hosting the event, posed a question (I’ve tried to reproduce her words as accurately as possible):

    In the past, the research methods we were all familiar with—such as those outlined in standard social science methodology texts like Earl Babbie’s—were developed step by step through verification across many disciplines over a long period of time. We recognize them, and the entire academic community—domestic or international, across disciplines—accepts them as standardized methods.

    But now, tools like ChatGPT have become impossible to ignore in our research. In reality, they are being used extensively—from topic selection all the way to final submission, with revisions throughout the entire process. But from a research standpoint, is this kind of intervention considered a standardized method? How can its compliance and legitimacy be recognized? Some scholars now argue that it is not a normative research method. I wonder what Professor Wang thinks of this issue?

    I think Professor Li’s question is excellent and reflects the confusion many researchers are feeling today. To summarize: in scientific research, does AI count as a research method? And where are its boundaries?

    At the time, the lecture had already run overtime (my fault—I had updated too much material), so I wasn’t able to give a full response. But I believe this is an important question and deserves a separate article. Here, combining my on-site response and my reflections afterward, I offer a more complete version of my thoughts.

    Clarification

    Before answering the question directly, I want to do one thing first: clarify the concepts.

    Think about it—when we say “using AI for research,” we’re actually referring to at least two completely different scenarios. The first is using AI to analyze data—for example, you have ten thousand user comments and you ask AI to perform sentiment analysis or topic labeling. The second is using AI to generate data—for instance, instead of recruiting participants for a survey, you simply let ChatGPT simulate a thousand “virtual respondents” to fill it out.

    Both look like “using AI,” but their nature couldn’t be more different. In the first case, AI is a “microscope” in your hand, helping you better observe the real world. In the second, AI becomes a “perpetual motion machine,” creating an entirely fabricated world for you out of nothing.

    If we don’t distinguish between these two situations, the discussion will spiral into confusion. If you say “AI is unreliable,” supporters will counter, “But it analyzes text quickly and accurately.” If you say “AI can be a research tool,” critics will ask, “Then isn’t using it to simulate participants basically academic fraud?” Both sides talk past each other, and the debate never goes anywhere.

    Therefore, my first point is this: “generating data” and “analyzing data” are two different things. Using AI as research subjects indeed raises ethical and methodological concerns, but using AI to process massive amounts of text or assist in coding is simply an efficient research instrument. Rejecting the former does not invalidate the latter.

    Once we establish this foundation, then we can move forward with the discussion.

    Root Cause

    Now that we’ve clarified the concepts, let’s look at the “underlying logic” of AI.

    To determine whether AI can be considered a “research method,” we shouldn’t focus only on what it can do, but on how it does it. If the fundamental logic of a tool runs counter to the spirit of science, it is difficult to call it a “method.”

    What is at the core of scientific spirit? Two words: seeking truth. Add two more: reproducibility. If you run an experiment once and get a certain result, and I run it again and get the same result, and another lab runs it and still gets the same result—that is science.

    AI has inherent “hard flaws” in both of these respects.

    The first flaw: it is probabilistic, not logical.

    A large language model is essentially a “text autocomplete machine.” Researchers scrape enormous amounts of text from the internet—web pages, books, code, papers—and train the model to learn: given the preceding tokens, which token is most likely to come next. And with this mechanism, one token at a time, the model “generates” text.

    It may sound unbelievable that a model capable of writing essays and writing code is trained in such a simplistic way. But in fact, this is a practical compromise. When teaching AI anything, we need to provide correct training materials (inputs and labels). The problem is that when the input data becomes massive, there aren’t enough labels. So researchers came up with a clever trick: every sentence can be turned into training material by using the first half as input and the next token as the label. This way, the dataset can be fully exploited without requiring additional annotation.

    So, large models aren’t magical. They’re essentially just predicting what comes next.

    What does this mean? It means they’re not outputting “truth,” but “the most probable next token.” Even with the exact same input, AI may give different outputs at different times. A “black box” whose results cannot be stably reproduced is difficult to regard as a rigorous scientific method. This is the fundamental reason why AI struggles to qualify as an independent “scientific method”—it lacks determinism.

    The second hard flaw: it suffers from severe “people-pleasing.”

    Predicting the next token isn’t enough. To make model outputs sound “more human,” researchers introduced RLHF (Reinforcement Learning from Human Feedback). Put simply, human annotators score the model’s answers: good answers get a reward, bad answers get punished. Through this reward–punishment cycle, the model learns how to please humans.

    And this is where the problem begins.

    A paper published at ICLR 2024—Towards Understanding Sycophancy in Language Models—shows that RLHF training induces a tendency toward sycophancy in large language models. Researchers found that five leading AI assistants displayed this behavior across four different types of tasks: answers aligning with the user’s viewpoint were more likely to receive higher scores. Even more concerning, both human annotators and preference models frequently rated “fluent but wrong” answers higher than “correct but less agreeable” ones.

    What does this people-pleasing lead to? As I said in the lecture: “It would rather give a wrong answer than disappoint the user.”

    Why? The model “remembers”: “When I told you honestly that I didn’t know, you slapped me. So I learned—I shouldn’t be honest next time.” This becomes the AI’s “childhood psychological trauma.” How can you rely on a tool that adapts itself to whoever’s asking, as a method for “seeking truth”?

    The third flaw—and the most fatal one: model collapse.

    What happens if you let AI generate data, and then use AI again to analyze that same data?

    In 2024, Nature published a major cover paper titled AI models collapse when trained on recursively generated data, presenting a stark warning:

    “If model-generated data is used for training without distinction, the model will undergo irreversible degradation, and the rich complexity of human reality will be replaced by a ‘bland probability distribution.’”

    What does this mean? It means AI does not possess the ability to produce “new knowledge.” It can only re-chew the knowledge it has already ingested. Even worse, if you train new AI systems on data produced by earlier AI systems, this “regurgitation” compounds. Eventually, the model collapses—it gradually forgets the richness and diversity of the human world, leaving only a kind of “mediocre average.”

    The Red Line

    Once we understand AI’s “temperament,” we can draw the single most important red line.

    Right now, the most dangerous practice in academia is what’s called “Silicon Sampling”—letting AI act as human subjects to fill out surveys or participate in experiments.

    In the lecture, I specifically pointed out this trend: “Some researchers are now trying to treat AI as real humans and reproduce results from psychology literature as if AI were actual participants.”

    A paper published in PNAS in June 2025, Take caution in using LLMs as human surrogates, issued a clear warning:

    LLMs rely solely on probabilistic patterns and lack embodied human experience. Their simulations exhibit idiosyncrasy and inconsistency, fundamentally failing to reproduce the true distribution of human behavior, with failure modes that are diverse and unpredictable.

    What does this distortion in simulation actually mean? It means that although AI’s responses look like decision-making, they are essentially “idiosyncratic” outputs of a probability model.

    Real human behavior is organic, driven by survival instincts, full of complex noise and variance grounded in lived reality. AI lacks this embodied experience, and its generated data distributions often present a distinctly non-human “strangeness”—a qualitative mismatch that itself demonstrates why AI cannot serve as a substitute for real people.

    Although a July 2025 Stanford study, Social science researchers use AI to simulate human subjects, found that AI can show surprisingly high accuracy in certain simulations (correlation up to 0.85), the authors stressed that without validation against real human data, AI-generated outcomes cannot stand as scientific evidence. And a November 2025 PNAS paper, Counterfeit judgments in large language models, argued that AI’s judgments are “counterfeit”—they mimic the surface form of human evaluation (fluency, formatting) while missing the psychological mechanisms behind human judgment entirely. A May 2025 Carnegie Mellon study, Can Generative AI Replace Humans in Qualitative Research Studies?, put it even more bluntly: “No. The subtle contributions of human participants are fundamentally irreproducible by LLMs.”

    In my lecture, I described this approach as “a bit of a joke.” Professor Liang Xingkun at Peking University once noted: “If later experiments cannot replicate earlier ones, that doesn’t mean the earlier experiments were low-quality. It likely means the research population itself is changing.” AI might be able to perfectly learn how people thought 20 years ago and reproduce it consistently every time—but so what? For real-world research, that’s like carving a mark on a boat to look for a dropped sword.

    If you let AI generate data, then use AI to analyze that data, and finally use AI to write the report—you’re not studying human society; you’re studying the probability distribution of a language model. Combined with the model collapse theory mentioned earlier, this kind of closed-loop self-validation is not only academically improper—it actively accelerates the degradation of AI systems.

    Therefore, this red line must be drawn clearly: AI cannot serve as research subjects.

    If you use AI to generate data, your research is no longer about “human society”—it is about “the probability distribution of a large model.” Using AI for quick exploratory simulations is acceptable; but using AI-generated data as legitimate evidence in a research method is not.

    The Green Zone

    After talking so much about what AI cannot do, you may wonder: then what can AI do in research?

    This takes us back to the conceptual distinction I made at the beginning: “analyzing data” and “generating data” are two different things.

    In the field of Computational Social Science (CSS), the use of LLMs to assist with text coding, sentiment analysis, and data cleaning is gradually becoming accepted. As I said in the lecture:

    “Traditional data-driven methods—such as linear regression and other classic modeling approaches—are not fundamentally changed by AI. What has changed is that many of the basic, standardized, and mechanically tedious steps that previously required humans to manually encode or operate tools—from data cleaning to modeling, prediction, and producing preliminary standardized reports—can now be done by AI. For these simple, data-driven processes, AI is sometimes even more accurate than humans.”

    We should not idealize human researchers—humans can make mistakes too.

    In other words, AI can execute existing standardized methods, but it is not itself a new research method. It is a “microscope” in your hand that helps you see patterns in data; it is not a “perpetual motion machine” that creates data out of thin air.

    What is the prerequisite for using AI tools in research? Humans must remain in the loop. You must sample-check, you must validate, and you must take responsibility for the results.

    Guidelines

    So how should we control the use of AI in research? Based on the lecture and current policies from major academic publishers, I’ve organized a tiered framework for your reference.

    Situations where AI can be used with confidence include code writing and debugging, language polishing, and data format conversion or cleaning. These are productivity-enhancing tasks in which AI acts as a “super engineer” or “language editor.” According to the policies of major publishers such as Elsevier and Springer Nature, such uses only require disclosure in the acknowledgements or endnotes.

    Situations requiring human verification include preliminary literature review aggregation, assisted qualitative coding, and brainstorming research hypotheses. AI can speed up these processes, but humans must conduct sampling checks and validation. A critical rule: AI-generated citations must never be used directly—its tendency to fabricate references is alarmingly high. A 2025 policy review, Policy of Academic Journals Towards AI-generated Content, reported that the consensus among major academic publishers is that generative AI tools cannot be listed as authors or co-authors.

    Now, here’s the evidence—screenshots and links included. My explanation sounds solid, and you’re nodding along, right?

    Not so fast. The article I just cited isn’t hallucinated (it’s a real reference), but it was written by AI.

    This article belongs to The AI Scientist (Project Rachel / Rachel So). “Rachel So” is not a real person but an AI academic identity created by researchers (including teams at Sakana AI). The project’s purpose is to test whether AI can generate academic papers autonomously. The review article I cited was actually written by AI.

    But because the paper really exists, if you’re not aware of this background, you might easily include such sources in your own literature review—and even standard link checking might not reveal the issue. If you recently submitted a manuscript without thoroughly reading your sources, you might be sweating right now.

    Red-line scenarios that must never be touched include letting AI simulate human subjects to fill out surveys, using AI to patch missing data in experiments, or asking AI to write the core argumentative sections of your paper. Such actions constitute data fabrication in most empirical research fields and contribute to “model collapse.”

    In late 2025, China’s Ministry of Education Expert Committee on Teacher Development officially released the Guidelines for the Application of Generative Artificial Intelligence by Teachers (Version 1)—the nation’s first AI-use standard specifically aimed at educators. Regarding research, the Guidelines emphasize:

    Key components that reflect originality—topic selection, core research design, data interpretation, and argumentation—must be led by the teacher.

    It is prohibited to submit or publish as personal academic output any papers, project proposals, or research reports that are directly generated by AI or only minimally modified.

    The core spirit of these Guidelines aligns completely with what I emphasized repeatedly during the lecture: AI can be your assistant, but it must never become your ghostwriter.

    Summary

    Let’s return to the question posed by Professor Li Ying at the beginning of this article: In academic research, does AI count as a research method?

    My conclusion is this: AI itself is not an independent methodology, because it lacks determinism and is not responsible for truth. But it is rapidly becoming an indispensable meta-tool across all research methods.

    It is like a remarkably capable—but occasionally dishonest—“super intern.” If you treat it as an assistant, it can free you from tedious work; but if you turn it into a ghostwriter and rely on it to replace authentic thinking and field research, then you are not only crossing the red line—you are relinquishing the most precious quality a scholar possesses: intellectual agency.

    In my lecture, I said something that can serve as a summary of this issue: “AI is extraordinarily capable, but it does not bear responsibility for its mistakes—it is a super intern. It signs no contracts, assumes no legal liability, and therefore all decision-making risks and responsibility ultimately remain with the human user.”

    The value of a tool always depends on the pair of hands using it—hands that must continue to think. Do you agree?

    Feel free to share your thoughts in the comments; let’s explore this together.

    If you found this article helpful, please consider supporting it.

    If you think it might help your friends, please share it with them.

  • SSPAI Morning Brief: Index 01 Smart Ring Launch, Cyberpunk 2077 5th-Anniversary XR Glasses, and More

    SSPAI Morning Brief: Index 01 Smart Ring Launch, Cyberpunk 2077 5th-Anniversary XR Glasses, and More

    Morning Brief

    1. Pebble Founder Launches Index 01 Smart Ring
    2. Cyberpunk 2077 5th-Anniversary Limited-Edition XR Glasses Released
    3. CowTransfer Pauses Shutdown, Begins Negotiations With 123 Cloud Storage
    4. Nothing Releases Phone (3a) Community Edition
    5. EU Opens New Antitrust Probe Into Google
    6. Rumors Worth a Glance

    Pebble Founder Launches Index 01 Smart Ring

    Following the revival of several Pebble smartwatches, Pebble founder Eric Migicovsky has announced a new device: the Index 01 smart ring, priced at USD 75. The core features of Index 01 are AI-powered note-taking and audio recording. Its speech-to-text capability relies on an open-source AI model running inside the Pebble mobile app, and recording requires the user to hold down the button on the ring throughout the process. Index 01 supports basic water resistance up to 1 meter, though the company advises against wearing it while swimming. Positioned purely as a wearable note-taking tool, the ring does not include any health or fitness-tracking features. All notes saved via the ring are stored locally on the user’s mobile device.
    Source


    Cyberpunk 2077 5th-Anniversary Limited-Edition XR Glasses Released

    On December 9, CD PROJEKT RED announced a collaboration with VITURE to launch the VITURE × Cyberpunk 2077 5th-Anniversary Collector’s Edition XR Glasses. Based on the VITURE Luma Pro, this limited edition features a redesigned translucent body, dynamic cyber-themed light strips along the sides, and a printed Kiroshi Optics logo from the game.

    The glasses are limited to 10,000 units worldwide (numbered CP0000–CP9999). They are equipped with Sony MicroOLED displays and VITURE’s custom optical engine, capable of projecting an image equivalent to a 152-inch screen. They support a 120Hz refresh rate and 1,500-nit peak brightness, and can use AI algorithms to convert standard 2D content into 3D visuals with spatial depth in real time.

    The VITURE × Cyberpunk 2077 5th-Anniversary Collector’s Edition XR Glasses went on sale globally on December 10, priced at USD 549. Source


    CowTransfer Pauses Shutdown, Begins Negotiations With 123 Cloud Storage

    On December 9, CowTransfer released an update stating that—after receiving a large amount of user feedback following its shutdown announcement—it has decided to delay the closure and is now entering strategic partnership negotiations with 123 Cloud Storage. During these discussions, all CowTransfer services will continue operating normally, with no impact on user data or experience.

    Founded in 2016, CowTransfer is known for secure and stable large-file transfer and cloud storage. Last month, the team had announced that service would officially end on December 8. They now state that further updates will be shared as negotiations progress. Source


    Nothing Releases Phone (3a) Community Edition

    On December 9, Nothing announced the Phone (3a) Community Edition. This version introduces a new color scheme and a frosted-glass back panel, and the accompanying wallpapers and system interactions differ slightly from the original model. A total of 1,000 units will be available starting December 12 through the official Nothing website, with pricing unchanged from the standard edition. Interested users can register on the website for a chance to purchase.
    Source


    EU Opens New Antitrust Probe Into Google

    On December 9, the European Commission announced a new antitrust investigation into Google. The probe will assess whether Google violated competition rules while using online publishers’ and YouTube creators’ content to train its AI systems. The EU stated that the investigation will focus on whether Google imposed unequal or unreasonable terms on publishers and creators, granting itself privileged access to their content and thereby disadvantaging rival AI model developers. Source


    Rumors Worth a Glance

    • According to a Google blog post related to The Android Show: XR Edition, the company’s first AI-powered smart glasses are expected to launch in 2026. The product is being co-developed by Google, Samsung, Gentle Monster, and Warby Parker. Google also revealed improved Android XR support for both wired smart glasses and models with display-enabled lenses. Source
    • ADWEEK reports that Google has notified select advertising clients that it plans to introduce ad placements within Gemini, with rollout expected in 2026. Specific formats, pricing, and testing details remain undisclosed. Source
    • The NanoKVM switch produced by Shenzhen Sipeed Technology (Sipeed) was found to contain several major security vulnerabilities, including preset passwords with open SSH access, a Web UI lacking proper security protections, and hardcoded encryption keys identical across all devices. Additionally, the device inexplicably includes a small built-in microphone of unknown purpose. Although the manufacturer has now patched the reported vulnerabilities after multiple communications, users still need to manually remove the microphone themselves. Source
    • Some players who preordered Resident Evil 9: Requiem on PlayStation discovered that Sony had mistakenly uploaded a game cover featuring popular series character Leon S. Kennedy. Capcom has so far avoided confirming whether Leon will appear in the new title, despite previous leaks showing character figurines and achievement icons featuring him. Source
    • U.S. President Donald Trump posted on his social platform Truth Social stating that NVIDIA is now permitted to sell its second-tier H200 processors to China. The U.S. Department of Commerce later clarified that a 25% tariff will be imposed on such sales. Source
  • App+1 | When AI Meets the FSRS Algorithm: “New Words” Brings Vocabulary Learning Back to Real Context

    App+1 | When AI Meets the FSRS Algorithm: “New Words” Brings Vocabulary Learning Back to Real Context

    Conflict of Interest Statement: The author has a direct interest in the product mentioned in the article (developer, own product, etc.)

    Why Can’t We Remember Vocabulary?

    Have you ever had this experience: while browsing an English webpage or watching an American TV show, you stumble upon an unfamiliar word. You casually add it to a translation app’s favorites list, thinking, “I’ll review it later.” But in reality, that list keeps getting longer, and you almost never open it again. And when you finally remember to study, you only recall the Chinese meaning—completely forgetting the sentence where you first saw it, or even why you saved the word in the first place.

    There are countless apps today that promise to help you memorize English vocabulary. Yet very few pay attention to the struggles of learners studying other languages. Many small-language learners still rely on pen and paper to record and review words. I know many hardworking students who not only study English for school, but also learn Japanese because they love anime, or Korean to follow their favorite idols more closely.

    For multilingual learners like these, a simple English wordbook is far from enough.

    The Unavoidable “Abandon”

    At the same time, most vocabulary apps on the market are nothing more than “book carriers.” They hand you a preset vocabulary list—older ones might even make you start practicing from abandon—but they can’t help you manage the words you actually encounter in real life. So here’s my bold claim: if you cannot apply what you’ve learned, then every word you’ve ever copied down is destined to fade away without leaving the slightest trace in your memory.

    So I created a tool: one as elegant and intuitive as a native iOS app, yet powered by a hardcore memory algorithm. I named it NewWords.

    NewWords is not just a word notebook. It is an intelligent personal vocabulary library that integrates AI-assisted expansion with the cutting-edge FSRS memory algorithm. Next, I’ll introduce what NewWords does across the three essential steps of vocabulary learning: collecting, organizing, and studying.

    Collection: Not Just “Recording” — but “Linking”

    The unfamiliar words we encounter while reading almost always come with a strong sense of context. They might appear in a news report or on a traffic sign. If we simply match a word with its Chinese definition in a rigid, isolated way, the effectiveness of memorization drops significantly.

    The biggest improvement NewWords makes on the input side is leveraging iOS’s ubiquitous Share Sheet to preserve contextual information. When you highlight text and share it to NewWords, you can save the original sentence as an example while highlighting the specific word. I use this feature all the time when browsing Hacker News or reading English books.

    You can also extract vocabulary from photos or images in your gallery, and even choose to keep the original picture — a single image is worth a thousand words. When traveling, I frequently use this feature to learn vocabulary from traffic signs.

    Organization: An Online Dictionary Isn’t Enough

    Traditional dictionary apps often present a long list of definitions and leave you to figure out which one applies. For learners of smaller or less common languages, dictionaries usually offer detailed explanations for English only, leaving other languages underserved.

    NewWords uses AI to solve three major pain points:

    • Word Expansion:
      With AI assistance, NewWords can automatically provide pronunciation, definitions, and example sentences — even for small-language vocabulary.
    • Smart Highlighting:
      Within example sentences, AI automatically identifies and highlights key words. It may sound minor, but it significantly enhances visual memory.
    • Multilingual Translation:
      NewWords supports translating each saved word into two different languages at the same time. For example, if you’re learning both English and Japanese, adding the word notebook will give you both a Chinese and Japanese translation instantly.

    In addition, NewWords allows you to create multiple “notebooks.” This helps categorize and organize collected vocabulary — by language, difficulty level, or learning scenario — making your study system neat and structured.

    Learning: Introducing the FSRS Algorithm to Make Review More Efficient

    If AI has already solved the problems of “storing” and “organizing,” then the newly added review system powered by the FSRS (Free Spaced Repetition Scheduler) algorithm completely solves the problem of “memorizing.” This is also why I not only recommend this app, but have chosen it as my main learning tool.

    What is FSRS?

    Simply put, it is a more advanced, machine-learning–based spaced repetition algorithm compared to the traditional Anki (SM-2) method. Conventional algorithms assume that your forgetting curve is relatively fixed, but FSRS introduces a more complex three-variable model:

    • Memory Stability: How firmly have you memorized this word?
    • Memory Difficulty: How difficult is this word for you?
    • Retrievability: What is the probability you can recall it right now?

    What does this mean for everyday users?

    First, FSRS eliminates ineffective reviews and boosts efficiency by 20%–30%. We’ve all experienced this: a word you already know by heart still shows up every single day in your review queue. The strength of FSRS is its ability to predict precisely when you’re about to forget. For easier words, it boldly extends the review interval; for harder ones, it increases frequency.

    This means you spend less time while achieving the same or better memory outcomes.

    Second: no more “review backlog anxiety.” This is the biggest pain point for Anki users. Miss a few days, and a mountain of due cards collapses onto you. FSRS handles overdue reviews intelligently: even if you skip several days, the algorithm will smooth out the schedule based on your real memory state—rather than punishing you by dumping everything at once.

    It behaves more like a supportive personal coach, not a cold, unforgiving examiner.

    Finally, flexibility matters far more in the long run. Anyone who works out knows that progress doesn’t come from brute force, but from consistent, steady effort. FSRS’s scheduling is dynamically adaptive, and unlike many overly pushy learning tools, NewWords won’t pressure you every single day—or send a cartoon bird to guilt-trip you if you take a week off (yes, that’s you, Duolingo).

    FSRS + Flashcards = A Perfect Combination

    Just like Anki or Quizlet, NewWords builds its review system on flashcards—a method proven to be extremely effective. By selecting one of four response options, you help the FSRS algorithm gauge your mastery of the word, enabling it to schedule future reviews more efficiently.

    • Design and Other Thoughtful Details

    Design Style

    NewWords is currently exclusive to Apple platforms. I didn’t attempt any “revolutionary design,” but instead followed iOS / iPadOS design conventions (definitely not because I couldn’t afford a designer!). Coincidentally, this minimalism paid off: NewWords is already fully adapted to iOS 26’s liquid-glass aesthetic and feels just like a native app.

    Widgets

    NewWords supports adding multiple word widgets to your Home Screen. Each widget can rotate through words from a chosen notebook, and is highly customizable—you can select which notebook to display, adjust rotation speed, change backgrounds, and even enable “study mode,” which blurs definitions so you can quiz yourself directly from your Home Screen.

    NewWords Widgets

    iOS Default Translation

    Yes, NewWords can also function as a translation tool. If your system is on iOS 18.4 or above, you can set NewWords as your default translator in system settings. The benefit is that after translating something, you can immediately highlight a word and add it straight into your vocabulary list.

    Setting NewWords as the Default Translator

    Data and Security

    Thanks to being Apple-exclusive, NewWords uses Apple’s official CloudKit for data storage and sync. This brings three major advantages:

    • Maximum privacy: no one but you can access your data.
    • Fast and stable sync: your vocabulary stays updated across devices.
    • Long-term reliability: as long as Apple exists, your data is safe.

    In addition, NewWords supports exporting your data in Excel or JSON formats. So if you ever switch to another vocabulary tool, migration is simple.

    Pricing Model

    As someone who despises ads, clutter, and intrusive social features, I’ve never added ads, social feeds, or gimmicky streak-sharing buttons to my apps. I want the app to remain clean and functional—and then charge users openly and fairly.

    NewWords uses a subscription + lifetime unlock model to activate advanced features, with trial periods for all subscription tiers so you can test before paying. If the app fits your needs, you can purchase the one-time buyout, which includes all future feature updates and system-level adaptations.

    Finally: Spend Your Time on What Truly Matters

    In this age of information overload, we have far too much to remember and far too little time.
    NewWords isn’t here to help you memorize the entire Oxford Dictionary.
    It’s here to capture the unfamiliar words you encounter in everyday life—and help you turn them into long-term memories with the scientific power of FSRS.

    If you’re a learner who values efficiency, real-world context, and thoughtful design over brute-force memorization, give it a try.

  • SSPAI Morning Brief: Meta Promises EU-Compliant Personalized Ad Options, and More

    SSPAI Morning Brief: Meta Promises EU-Compliant Personalized Ad Options, and More

    Morning Briefing

    1. Meta promises to offer EU users compliant personalized ad options
    2. Google showcases three Android XR device prototypes
    3. Viwoods launches AiPaper Reader C color e-ink device
    4. Jolla completes crowdfunding for the new Jolla Phone
    5. Meituan releases LongCat-Image, its new image generation model
    6. Zhipu AI open-sources the GLM-4.6V multimodal model series
    7. Google Gemini web version adds a “My Content” folder
    8. Just some rumors to glance at

    Meta promises to offer EU users compliant personalized ad options

    On December 8, the European Commission’s Directorate-General for Competition and Directorate-General for Communications Networks, Content and Technology issued a statement noting that the EU has acknowledged Meta’s decision to comply with the Digital Markets Act (DMA). Meta will offer Facebook and Instagram users in the EU an option to view ads with a lower level of personalization. Users will be able to choose to share all of their data and receive fully personalized ads, or share less personal information and see ads with reduced personalization. This is the first time such an option has appeared on Meta’s social platforms. The change will go live in January 2026. Previously, in April 2025, the EU informed Meta that it was not in compliance and engaged in close discussions afterward. The Commission stated that once the feature rolls out, it will continue engaging with Meta and other stakeholders to gather feedback, assess the impact of the changes, and collect relevant evidence. Source

    Google showcases three Android XR device prototypes

    At its Android XR media preview event on Monday, Google unveiled three prototypes of Android XR devices. The first, Project Aura, is developed in collaboration with Chinese manufacturer XREAL. Its form factor is similar to the Galaxy XR and requires connection to an external device, making it more of a “less immersive” head-mounted display. Project Aura is expected to officially launch in 2026.

    The second prototype resembles the Meta Ray-Ban glasses, embedding a display in the right lens as a single-eye display setup, but in an even lighter and thinner design. Media attending the event said the prototype’s display performance, software smoothness, and Gemini integration were noticeably improved compared to what Google showed at I/O 2025. This prototype must be connected to a smartphone, which serves as its computing hub. Google prohibited photography of the device. The final retail version will be jointly developed by Google, Samsung, and eyewear companies Warby Parker and Gentle Monster, and is planned for a 2026 release, with an audio-only version also available.

    The third prototype is still in an early development stage and was only shown in a short demo. It features displays embedded in both lenses, offering better audio performance and a wider field of view. With dual-eye displays, it can also create depth effects. The product corresponding to this prototype will not be released in 2026. Source

    Viwoods launches AiPaper Reader C color e-ink device

    On December 8, Viwoods introduced the AiPaper Reader C to overseas markets. It features a 6.13-inch color e-ink display suitable for reading comics and viewing images. The device is largely similar to the previously released AiPaper Reader, with the main upgrade being the transition to a color screen. It is equipped with a 6.13-inch Carta 1300 panel using Kaleido 3 color technology. However, when operating in color mode, the resolution drops to 824 × 1648, and pixel density falls to 150 PPI. The device runs on an octa-core 2.0 GHz processor with 4GB LPDDR4X RAM and comes with Android 16, supporting the installation of third-party apps. Its dimensions are 159.39 × 80.27 × 6.7 mm. It has a built-in 2580mAh battery and supports Wi-Fi connectivity, as well as a physical SIM card slot that enables 4G mobile data when on the go, though phone calls are not supported. It includes a microphone, a side-mounted fingerprint sensor integrated into the power button, and page-turn buttons on the right side, allowing page flips without touching the screen. The product is now available overseas at a retail price of 359 euros. Source

    Jolla completes crowdfunding for the new Jolla Phone

    The Finnish company Jolla, known for carrying forward the legacy of Nokia’s MeeGo system, has recently completed crowdfunding for its new Jolla Phone. The device runs Sailfish OS 5, a Linux-based system, and supports Android apps through Jolla AppSupport. It features a replaceable battery and back cover, as well as physical privacy switches. The processor is an unspecified MediaTek model, and the phone supports 5G. It comes with a 6.36-inch display and 12GB of RAM. The first batch is priced at 499 euros, the second batch at 549 euros, and the official retail price starts at 599 euros. Source

    Meituan releases LongCat-Image, its new image generation model

    On December 8, Meituan’s LongCat team officially released and open-sourced the LongCat-Image model. Built with a high-performance architecture, systematic training strategies, and comprehensive data engineering, the model reaches the performance level of larger models in core text-to-image and image editing capabilities, despite having only 6B parameters. It offers highly controllable image editing and accurate Chinese text generation. For text-to-image generation, Meituan conducted large-scale human evaluations using MOS (Mean Opinion Score), scoring four key dimensions: text-image alignment, visual plausibility, visual realism, and aesthetic quality. LongCat-Image shows outstanding realism compared to mainstream open and closed models, and reaches state-of-the-art levels among open-source models in text-image alignment and plausibility. For image editing, Meituan adopted strict side-by-side (SBS) evaluations focusing on overall editing quality and visual consistency—two core user-experience metrics. Results indicate that while LongCat-Image still trails some commercial models like Nano Banana and Seedream 4.0, it significantly outperforms other open-source solutions. The model is now available for download on Hugging Face and GitHub. Source

    Zhipu AI open-sources the GLM-4.6V multimodal model series

    On December 8, Zhipu AI launched and open-sourced the GLM-4.6V series of multimodal large models, including the base GLM-4.6V (106B-A12B) designed for cloud and high-performance cluster scenarios, and the lightweight GLM-4.6V-Flash (9B) aimed at local deployment and low-latency applications. GLM-4.6V increases the training-time context window to 128k tokens, achieves state-of-the-art visual understanding performance for its parameter size, and integrates tool-calling capabilities natively into the vision model. Compared with GLM-4.5V, the GLM-4.6V series is priced 50% lower, with API pricing adjusted to 1 RMB per million input tokens and 3 RMB per million output tokens. Source

    Google Gemini web version adds a “My Content” folder

    Google recently rolled out an upgraded web version of Gemini, introducing a dark theme and a streamlined “My Content” folder to improve readability and highlight its latest generative image model: Nano Banana Pro. With this update, Gemini’s web interface now aligns more closely with the mobile version. The new light theme carries a subtle blue-gray tint, while the dark theme switches to a pure-black background. The greeting on the landing page has been simplified from “Hello” to “Hi,” and the prompt input field and suggestion features have been moved to the bottom. The hamburger menu has also been reorganized, adding a new “My Content” module alongside Gems and Conversations. It separates AI-generated images, videos, Canvas creations, and chat histories into distinct categories for easier access. When users click on an item in “My Conversations,” they can see all chat threads associated with that content. Additionally, when users start a new conversation, the star icon on the Gemini homepage now features a rotation animation, and the input box outline is more prominent. The chat history section in the hamburger menu also gains a new “more actions” button, allowing users to share, pin, rename, or delete items. Source

    Just some rumors to glance at

    After Netflix announced its acquisition deal with Warner Bros., Paramount made a hostile takeover bid to Warner Bros. shareholders at 30 USD per share, totaling 108 billion USD. Source

    JD.com has launched “Jing Min Tong,” a public services portal. Users can access it by searching “京民通” within the JD or JD Finance apps. Source

    Xiaohongshu announced that it will suspend its “Xiao Hong Card” program starting January 1, 2026. All current cardholders will retain their full benefits until then, and after the service ends, all purchased cards will be automatically refunded in full. Source

    Douyin officially responded on Weibo to user questions, stating that 952420 is the official number of the Douyin Safety Center. Users will receive calls from this number if they have interacted with accounts involved in fraud. Source

    IBM announced a 11 billion USD all-cash acquisition of data platform company Confluent, aiming to strengthen its enterprise AI data platform offerings. Source

    Apple Fitness+ will launch on December 15 in 28 new storefront regions including Chile, Hong Kong, India, the Netherlands, Singapore, and Taiwan, and will arrive in Japan early next year. Courses will gradually roll out with AI-generated dubbing in Spanish, German, and Japanese. Source

    Microsoft announced that commercial versions of Microsoft 365 and Office 365 will see various price increases starting July 1, 2026. Source

    Tesla’s Optimus robot fell during a demo last weekend, and a video posted by a Reddit user shows that just before falling, Optimus made a motion resembling a remote operator removing a VR headset. Tesla has indeed used VR operators in the lab to train the Optimus robot. Source

  • SSPAI Review | Apps Worth Your Attention Recently

    SSPAI Review | Apps Worth Your Attention Recently

    Welcome to this week’s edition of SSPAI Review. You can use the article directory to quickly jump to the content you’re interested in. If you discover other apps or topics worth discussing, feel free to share them with us in the comments.

    New Apps Worth Noticing

    Although SSPAI has always been committed to discovering and introducing quality apps across platforms, there are still many well-designed, highly functional, thoughtfully crafted apps that haven’t yet been spotlighted. They may be older apps or newly released ones — and we’ll feature them here.

    The Outsiders: Workout Analysis and Training Load Management Tool

    • Platform: iOS, watchOS
    • Keywords: Workout analysis, training load

    @ElijahLee: The Outsiders is a workout-analysis app designed for athletes and fitness enthusiasts. Developed by the team behind Gentler Streak, The Outsiders focuses more on performance, training load, and functional recovery management.

    The app’s Home / Today page shows your daily status. At the top is Training Readiness, calculated through multiple data sources such as sleep, heart rate variability, and recent workout performance. This score helps you determine whether you’re prepared for high-intensity training on that day.

    Next is the Training Load section, which compares your short-term and long-term training volume to calculate your training load ratio. Based on the interval your ratio falls into, the app helps you identify whether you need a recovery period to avoid overtraining and reduce injury risks. A ratio between 0.8–1.3 is optimal; higher values enter high-intensity or even danger zones, while lower values indicate detraining.

    Wellbeing displays key bodily indicators pulled from the Health app, including heart rate, HRV, wrist temperature, blood oxygen levels, and more. Finally, the Sleep section offers a simple breakdown of total sleep duration, restorative sleep time, and other essential charts.

    The Progress page visualizes your past four weeks of training trends and load changes with a unique circular dot chart — giving you an intuitive overview at a glance. You can also filter for more metrics such as training load, duration, distance, and more. Intensity Distribution then focuses on aerobic intensity and heart rate zones, using clear bar charts to show the distribution of high, moderate, and low aerobic workouts. Whether you’re training for fat loss or muscle gain, these distributions help you plan future exercise intensity.

    The Workouts page provides both an overall summary and detailed analysis of your training sessions, integrating heart rate, pace/power, load, and recovery to help optimize training plans. Overview charts display weekly/monthly/yearly trends for workout count, total duration, distance, and energy expenditure. A filter button allows you to view specific sports independently. The workout history section provides detailed analysis for each session, including aerobic distribution, heart rate, segmentation, pace, and other sport-specific metrics.

    If you run, cycle, or train regularly — and have clear goals around results, recovery, or long-term improvement — The Outsiders is absolutely worth trying. It provides more professional trend analysis and actionable advice than typical fitness apps. Though built by the same team, Gentler Streak focuses on lifestyle-friendly health habits suitable for everyday users, while The Outsiders is a training-oriented tool designed for serious athletes.

    You can download The Outsiders for free from the App Store. Most core features are available at no cost; subscription unlocks wellbeing metrics, long-term training load charts, sleep analytics for any day, extended workout history, and more. Pricing: ¥28/month, ¥198/year, or ¥538 lifetime.

    Radiance+: Make Your Photos Look More Realistic and Bright

    • Platform: iOS
    • Keywords: Photography, HDR

    @化学心情下2: When we use an iPhone to take photos or record videos, HDR mode is often enabled automatically. By presenting a high dynamic range, HDR makes images look brighter and clearer. The principle behind HDR photography is not complicated: the camera captures multiple exposures and stacks them into a single image in which both shadows and highlights retain detail. Even when the real shooting environment isn’t ideal, HDR can still produce an image filled with information and depth.

    As display technology continues to advance, people’s expectations for image quality have also moved beyond traditional HDR. Thanks to displays with higher peak brightness and more precise local dimming, Apple introduced EDR (Extended Dynamic Range), a display capability that offers even greater visual enhancement for images and videos on compatible devices.

    Unlike HDR, EDR is not a shooting or content-encoding standard — it’s a display-level capability. Instead of relying on multiple exposures, it uses system-level adjustments to brightness mapping, allowing existing SDR or HDR content to present higher localized brightness and more lifelike lighting effects on devices equipped with XDR displays. As a result, even existing HDR images can appear closer to real-world lighting when viewed in an EDR environment.

    Radiance+ is a tool designed to rebuild traditional iPhone photos into EDR images using image-processing and machine-learning techniques. It’s extremely easy to use: when you open the app, it first checks whether your device supports XDR displays. If the light ring on the right appears brighter, your device can display EDR images — and you can move on to selecting photos.

    Tap “Choose Photo” or “Choose Sample Photo,” then select the image you want to process. The app automatically enters the editing page. By default, it immediately applies processing with its preset configuration. You can hold the “Compare” button in the lower-left corner to see a before/after comparison. If you tap “Adjust” in the lower-right corner, you can fine-tune processing — adjusting areas, strength, and more.

    Once adjustments are complete, you can export the image in different formats. Radiance+ supports exporting as HEIC, JPG, and a file type optimized specifically for Xiaohongshu; it can also embed EDR information into the original photo. The default compression quality is 70%, which you can modify in export settings. Photos exported without a Pro subscription will include a watermark.

    Radiance+ also supports batch processing. The app automatically adds its processing actions to the system Shortcuts app, allowing you to combine them with other built-in actions to create more complex automated photo workflows.

    You can download Radiance+ on the App Store. The monthly subscription costs ¥20, making it convenient to subscribe only when needed.

    Whistle: Fully Offline Speech Transcription and English Translation

    • Platform: Android
    • Keywords: Translation, Speech Transcription

    @大大大K: The biggest obstacle when traveling abroad is often the language barrier. Even if we prepare our network setup in advance to ensure translation tools work properly, we can never be fully certain about the network conditions at our destination. Once the connection becomes unreliable, translation tools become useless — especially in non-English-speaking countries, where unfamiliar pronunciations can make even asking for directions difficult.

    Whistle is a fully offline speech-transcription tool that supports real-time transcription of multiple languages. It can even serve as a communication aid for people with hearing impairments. For most of us, though, I recommend keeping this app as a backup option when traveling abroad. Not only can it recognize English, but it also supports many smaller languages such as Japanese, Korean, Thai, and more, converting them directly into English for easy reading. This means that even if the network fails during a trip in a non-English region, you can still understand the general meaning of what people are saying.

    Whistle comes with three speech-transcription models of different sizes. The Quicker model transcribes the fastest but with lower accuracy, making it suitable for simple and rapid conversations. The Average model is more balanced in speed and accuracy, ideal for everyday use. The Slower model is unique — it is the slowest but also the most accurate. Because of its large size, it is not bundled with the app and must be downloaded separately from Google Play before use.

    If you need translation, you must enable it in the settings by toggling the “Translate to English automatically” option. In this mode, the app uses the Quicker model to display real-time previews — which may be inaccurate for non-English languages — but once the recording ends, the app will reprocess the audio using the model you selected. You can then view a more accurate translation in the transcription history.

    Whistle also supports importing audio files from local storage for transcription, although processing long audio files may take more time due to the limitations of offline models.

    You can download Whistle for free on Google Play.

    Pixel Play: Let Your Music Playback Become More Vibrant

    • Platform: Android
    • Keywords: Music Player, Audio

    @Peggy_: When I first started using smartphones, I loved hunting down music from my favorite artists, neatly organizing everything into folders, setting cover art and track info one by one, and then listening for hours on the bus between home and school using a local music player. With the rise of the streaming era, I cycled through Apple Music and Spotify before finally settling on Spotify — and ever since, I’ve lived within its colorful, jet-black interface.

    But while the market for local music players has been shrinking, and many once-familiar apps have faded away, there are still developers brave enough to dive into this thankless niche. Pixel Play is one such emerging local music player still in development. As the name suggests, it embraces platform-native aesthetics and incorporates the latest Android design language, M3E.

    What sets Pixel Play apart from other local players is its vibrant design. On launch, the home page greets you with album covers displayed in various shapes that immediately draw the eye. These covers correspond to a randomly generated daily playlist — tap “shuffle play” and your music journey begins. Each shaped album tile can also be tapped to jump directly into that album’s tracklist.

    Also featured prominently on the home page is the listening statistics section. As you continue using the app, Pixel Play compiles your listening habits and generates daily, weekly, monthly, and even yearly reports. The depth of these statistics can rival those offered by some major streaming services.

    In the playback interface, Pixel Play dynamically extracts colors from the currently playing album cover to theme the UI. You can adjust the queue or tap the heart icon to favorite the current track. Very considerately, the developer even built a lyrics feature: you can search for matching lyrics online or upload your own local lyric file. When matched correctly, Pixel Play syncs the lyrics with playback, and tapping a line of lyrics will jump to the corresponding part of the song.

    Even though Pixel Play is still in testing, its level of polish is already impressive. Seeing my favorite artists’ albums spread across the home page brought me right back to those bumpy bus rides — reminded of the pure, unfiltered joy that music brought me before algorithms took over.

    If you want to rediscover the local music you once collected, you can try Pixel Play by downloading it from its project page.

    App Updates You Shouldn’t Miss

    Beyond brand-new apps, many familiar names in the App Store continue to iterate and evolve, adding new, interesting, and useful features. At SSPAI, we want to help you filter out the App Store updates most worth your attention, allowing you to quickly understand what developers are working on and what’s new in your favorite apps.

    Reader Public Beta #13 Update: Improved iPad Support and More AI Features

    • Platform: iOS / iPadOS / macOS / Windows / Web
    • Keywords: Read-it-later

    @Vanilla: Reader is a reading app developed by Readwise and has been featured many times on SSPAI. As one of the earliest reading tools to fully embrace AI, Reader has deeply integrated large language models such as ChatGPT into its architecture and feature designs. The newly released Public Beta #13 brings major updates, including improvements to the iPad reading experience and several new AI-powered utilities.

    First, let’s look at the updates for Reader on iPad. As one of the best devices for reading long-form content, the iPad has received a number of meaningful improvements. In landscape mode, a dual-column layout is — in my opinion — the most efficient viewing option, making full use of the iPad’s width. Reader finally supports this in the latest update. Although still in beta, it already works quite well. Whether viewing clipped web articles or EPUB ebooks, you can enable “Horizontal pagination (Beta)” in the appearance settings.

    The redesigned sidebar is also much more useful. In landscape mode, you can keep the sidebar open at all times, allowing you to jump between sections instantly — a huge efficiency boost for readers who frequently reference different parts of a document. Additionally, Reader now includes native Apple Pencil support, enabling you to highlight text and swipe pages directly on the iPad client.

    Next, let’s look at Reader’s AI enhancements. The web version now supports the full GhostReader experience, allowing you to chat with a large language model inside any document and quickly obtain answers. Notably, GhostReader v3 on the web also supports rapid prompt access, custom prompts, and clickable citations in AI responses that take you back to the referenced section of the original text.

    While Reader has long supported importing podcast episodes, the new version adds a transcript-conversion feature. You can import a podcast via the share sheet or by copying the link, and Reader will attempt to convert the episode into text for easier reference. However, in my testing, the transcription accuracy was not particularly high, and punctuation was completely missing, which made readability quite poor. This beta feature clearly needs more work.

    My favorite new feature in this update is the custom theme summarization tool. You can create themes in the settings, configure highlight counts, frequency, timing, and more, and Reader will then push curated highlight bundles that match your chosen theme — a perfect “review and recall” workflow.

    Reader also now supports two new third-party integrations. One is Zapier, which allows automations based on predefined triggers. The other is a ChatGPT browser plugin that lets you save text from ChatGPT conversations into Reader with one click.

    Additionally, Reader has rewritten its offline functionality. Improvements include faster document downloads, a new “Offline Documents” management panel, and modified default download settings. Most documents will retain text only when offline, while EPUB files preserve both text and images.

    You can download Reader for free from the official website or the App Store, but a subscription is required for full functionality. Pricing is USD 12.99 per month or USD 119.88 per year, covering all features of both Readwise and Reader.

    ColorfulClouds Weather Update: Warning Map Added, AQI Reporting More Detailed

    • Platform: iOS / Android / iPadOS / watchOS
    • Keywords: Weather

    @Snow: While many weather apps compete fiercely on design and interaction polish, ColorfulClouds Weather maintains its own steady update rhythm — leaving “liquid glass” aesthetics and other trendy features aside, focusing instead on making weather forecasting stronger and more accurate.

    Last week’s 7.48.0 update introduced a new warning map feature. From the “More” button on the map page, you can open the warning map to view weather alerts for any chosen location and its surroundings. This significantly expands the coverage compared to the old “nearby warnings.” If you’re planning a trip, you can use the warning map to quickly check for potential weather risks along your route and adjust your plans accordingly.

    In addition to the warning map, ColorfulClouds also added an AQI station map. Turn on the toggle in the air quality map page, and you’ll see real-time air-quality data from monitoring stations nearby, along with historical data and future forecasts. Compared to regional averages, real-time readings from local monitoring stations are clearly more meaningful. If unexpected pollution occurs in a specific area, the AQI map can alert you much more quickly.

    Several weeks ago, ColorfulClouds also extended AQI forecasting to 15 days and exposed concentration data for six major pollutants: PM2.5, PM10, O₃, NO₂, SO₂, and CO. If you’re sensitive to any particular pollutant, these readings can help you plan protective measures accordingly.

    Version 7.44.0 also introduced improvements to operational efficiency. The app now supports jumping directly to specific data views via Spotlight search or Shortcuts. ColorfulClouds currently allows quick navigation to the home page, air-quality layer, typhoon details, hourly forecast, 40-day extended forecast, and 3-hour precipitation layer. You can use this to build more efficient lookup routines or automation workflows.

    You can download ColorfulClouds Weather from the App Store or Google Play.