Morning Brief: Google AI Updates / Huawei Leads China Phone Market / Xiaomi YU7 GT Breaks Nürburgring Record

💻

Google Releases Seedance 2.0's Strongest Competitor, Plus a Wave of AI Updates

📱

iOS 27 Shortcuts May Get Major Overhaul, Create with Just One Sentence

🤖

DeepSeek Responds to Model Anomalous Responses

🍎

Apple Releases AI-Powered Accessibility Features

🖥️

AMD AI Developer Conference Debuts in China, Lisa Su Introduces New "AI Agent Host" Category

🚗

7 Minutes 34.931 Seconds, Xiaomi YU7 GT Breaks Nurburgring Record

📱

Huawei Leads China's Smartphone Market During May Day Holiday, Android Vendors Raise Average Prices of Older Models by 13%

🚙

Xpeng Robotaxi Mass Production Begins, Pilot Operations Expected in Second Half of Year

🔧

Hesai Becomes Mercedes-Benz L3 Lidar Strategic Partner, Production to Begin at Thailand Factory

🌐

Tesla Abandons India Factory Plan

🧠

Meitu Visual Agent Strategy Takes Shape, Chance AI Completes Multi-Million Dollar Seed Round

️

Xiaomi Imaging Algorithm Wins Three Awards at CVPR 2026 NTIRE

💰

Yangtze Memory Starts IPO Coaching

⚡

BYD Battery CTO Discusses Flash Charging, Lifespan, and Safety Boundaries

📱

Reinforcement Learning Pioneer Sutton Reiterates "Bitter Lesson" Sparking Controversy

📱

vivo S60 Series Set for May 29 Launch

🎥

Starting at 698 Yuan, Insta360 Mic Pro Officially Released

💻

Starting at $1,299.99, Microsoft Unveils New Commercial Surface

Google Releases Seedance 2.0's Strongest Competitor, Plus a Wave of AI Updates

Today, Google unveiled a series of new products at the just-concluded I/O conference, including two new models, a new AI agent product, and a preview of audio smart glasses. Gemini's capabilities are further extending into search, productivity, creation, and hardware.

AI:

• Gemini Omni is officially launched, positioned as a "universal input-output" multimodal generative model, supporting video generation, style editing, and element replacement, with conversational modifications during generation;
• Gemini 3.5 Flash is simultaneously released, showing improvements over 3.1 Pro in programming and real-world AI agent task benchmarks, with output token speed 4x faster than other frontier models, available to all users starting yesterday; Gemini 3.5 Pro is expected to be available next month;
• Personal AI agent Gemini Spark debuts, running on Google Cloud dedicated virtual machines, supporting 24/7 background task execution, and can connect with third-party applications via the MCP protocol. Google also added a new $100/month AI Ultra tier, while the previous $250 top tier was reduced to $200;
• Gemini App gets a major overhaul, introducing a new design language called Neural Expressive, with fluid animations, vibrant colors, new fonts, and haptic feedback.

Hardware:

• Audio smart glasses preview debuts, with hardware built by Samsung, designed by Warry Parker and Gentle Monster, equipped with camera, supporting AI vision and multimodal input, Gemini responses played through earphones, compatible with Android and iOS;
• XREAL and Google brought the first public hands-on experience of Project Aura XR glasses at the event, first to run on Android XR system, equipped with XREAL's self-developed X1S chip, with 70° field of view, confirmed for global release this year.

🔗 Related Reading: Just Now, Seedance 2.0's Strongest Competitor Launches! This is AI Video's "Nano Banana Moment"

iOS 27 Shortcuts May Get Major Overhaul, Create with Just One Sentence

According to Bloomberg, Apple is preparing to introduce several new AI features in the upcoming iOS 27 and iPadOS 27, including system-wide grammar checking, natural language Shortcuts creation, and AI-generated custom wallpapers.

According to sources, the core of Apple's upgrade is expanding system-wide AI writing tools, adding a Grammarly-like grammar checker.

The feature appears as a semi-transparent menu from the bottom of the screen, displaying original text alongside modification suggestions, allowing users to accept individually, apply all, or dismiss with one tap. The system also integrates "Write with Siri" and "Assistive Writing" quick options at the top of the keyboard.

Additionally, the Shortcuts app will also receive a significant overhaul. The beta version allows users to directly input natural language requests in the "What would you like your Shortcut to do?" text box, with the system automatically building and configuring automation workflows, significantly lowering the barrier to use.

The new system will also integrate a generative AI wallpaper tool based on Image Playground. The operating system is expected to be announced at the Worldwide Developers Conference (WWDC) in June this year and released to the public in September.

WWDC26 will be held from June 9 to 13 (Beijing time). Ifanr will also be at Apple Park to bring you firsthand on-site coverage.

DeepSeek Responds to Model Anomalous Responses

Yesterday, DeepSeek's official Xiaohongshu account released a latest statement, responding to the issue of model anomalous responses triggered by the <think > characters.

DeepSeek stated that they have recently noticed user feedback where after entering special characters like <think > in conversations with DeepSeek models, the model occasionally returns unexpected content, causing some users to worry about "conversation leakage."

After a comprehensive investigation, DeepSeek's technical team found that:

The anomalous content triggered by inputting characters like <think > is a model hallucination caused by special characters, involving no security issues or privacy leaks.

DeepSeek also stated that they will enhance the model's ability to recognize and handle special characters through targeted training, fix known issues, and optimize the model's performance in such scenarios.

Apple Releases AI-Powered Accessibility Features

Apple yesterday released a series of accessibility feature updates, announcing that Apple Intelligence will be brought to core accessibility features like VoiceOver, Magnifier, Voice Control, and Accessibility Reader, with official release planned for later this year.

For visual assistance, VoiceOver introduces the "Image Exploration" feature, which can generate more detailed descriptions for photos, bills, and other image content; Magnifier simultaneously introduces AI capabilities, providing visual descriptions under high-contrast interfaces and supporting voice command control.

Voice Control introduces natural language input, allowing users to describe interface elements in everyday language without memorizing button labels, initially supporting English for users in the US, Canada, UK, and Australia.

Additionally, Apple announced the following new features:

• Video captions: Using on-device voice recognition to automatically generate captions for videos without subtitles, available on iPhone, iPad, Mac, Apple TV, and Apple Vision Pro, initially supporting English for US and Canadian users;
• Electric wheelchair control: Apple Vision Pro provides alternative input for users who cannot use joysticks through its eye-tracking system, initially supporting Tolt and LUCI two alternative drive systems, compatible with Bluetooth and wired connections;
• Other updates: visionOS adds vehicle motion cues; tvOS supports larger fonts; "Personal Voice" expands to over 50 languages globally; Sony Access controller now works with iOS, iPadOS, and macOS.

AMD AI Developer Conference Debuts in China, Lisa Su Introduces New "AI Agent Host" Category

The AMD AI Developer Conference was held for the first time in Shanghai yesterday, attracting over 2,000 developers. AMD CEO Lisa Su delivered the opening speech and held a dialogue on AI agents with Kai-Fu Lee, CEO of 01.AI.

The conference highlighted the new "AI Agent Host" category concept, centered around the Ryzen AI Max+ series processor, supporting up to 96GB of dedicated GPU memory, capable of running models with up to 200B parameters locally. HP, ASUS, Lenovo, Acer, and other manufacturers have already launched over 35 related products.

On the software side, ROCm 7.2 adds expanded support for Windows and Linux, and adds compatibility with the Ryzen AI 400 series processors. AMD also announced collaboration with Alibaba Cloud to provide free developer cloud based on Radeon GPUs, and launched exclusive programs for Chinese developers.

🔗 Related Reading: Jensen Huang Just Left, Lisa Su Arrived | Full Speech Text

7 Minutes 34.931 Seconds, Xiaomi YU7 GT Breaks Nurburgring Record

Xiaomi Auto announced yesterday that the Xiaomi YU7 GT broke the Nurburgring SUV lap record with a time of 7 minutes 34.931 seconds. This result was set by Xiaomi Auto's chief test driver Ren Zhoucan, making him the first Chinese driver to receive official Nurburgring lap time certification.

Lei Jun previously stated that the Xiaomi YU7 GT is a "pure-blooded GT" with pure driving fun, while also providing comfortable long-distance driving experience.

In terms of specs, the YU7 GT has a maximum power of 1003 horsepower, a top speed of 300km/h, and a CLTC pure electric range of 705km.

Xiaomi Auto announced on the 18th that the YU7 GT will be officially launched on May 21 (Thursday of this week) at 19:00.

Huawei Leads China's Smartphone Market During May Day Holiday, Android Vendors Raise Average Prices of Older Models by 13%

According to Counterpoint Research's "China Smartphone Weekly Sales Tracking Report" released yesterday, China's smartphone sales during the two weeks around the May Day holiday this year decreased by 16% year-over-year. Rising storage prices driving up device prices and weaker promotional efforts compared to previous years are the main reasons for the sales decline.

• Huawei continues to lead the market, with weekly sales share consistently above 25% in April. The Enjoy 90 Pro Max performed exceptionally well, and the Pura 90 Pro series and the new foldable Pura X Max, launched at the end of April, saw strong sales;
• Apple saw 3% year-over-year sales growth during the May Day holiday, but experienced the usual sequential slowdown. iPhone 17 series demand is gradually returning to normal, and the collective price increases by domestic brands have objectively enhanced iPhone's relative competitiveness;
• The average prices of older models from other Android brands have already increased by an average of 13%, which may further suppress replacement demand in the short term.

Counterpoint Research expects the 618 promotional activities to drive a modest market recovery in early June; China's smartphone shipments for the full year are expected to decline by 9% year-over-year, but still outperform the global average.

Xpeng Robotaxi Mass Production Begins, Pilot Operations Expected in Second Half of Year

Yesterday, He Xiaopeng, Chairman of Xpeng Group, announced that Xpeng's first mass-produced Robotaxi has officially rolled off the production line at the Guangzhou factory.

He Xiaopeng stated that this Robotaxi is based on the Xpeng GX prototype vehicle, equipped with Xpeng's second-generation VLA large model, 4 self-developed Turing AI chips, with local computing power reaching 3000TOPS—the highest globally—achieving L4 autonomous driving capability.

He also emphasized that this is China's first mass-produced pure vision Robotaxi + steer-by-wire chassis + full redundancy pre-installed.

It is reported that Xpeng will begin Robotaxi pilot operations in Guangzhou in the coming months.

Hesai Becomes Mercedes-Benz L3 Lidar Strategic Partner, Production to Begin at Thailand Factory

On May 19, global smart technology company Hesai Technology announced that according to the company's Q1 financial report, Hesai has become Mercedes-Benz's lidar strategic partner and has been confirmed as the supplier for their L3 autonomous driving vehicles.

It is reported that the newly signed supply agreement will fully support Mercedes-Benz's vehicle projects in European and Chinese markets, with the required production capacity to be solidly guaranteed by Hesai's newly completed Thailand "Galileo" manufacturing center.

Tesla Abandons India Factory Plan

According to Drive Safe and Fast, India's Minister of Heavy Industries Kumaraswamy officially confirmed yesterday that Tesla has abandoned its plan to build a factory in India.

Analysts point out that the decision to abandon the factory involves multiple obstacles. India's local supply chain and industrial ecosystem have obvious gaps that cannot support Tesla's manufacturing needs. Additionally, the two parties never reached a consensus on the core issue of whether "tariff reductions should come before the factory" or "the factory should come before tariff reductions."

India's insufficient infrastructure, the gap between local car market purchasing power and Tesla's premium product positioning, as well as policy uncertainty and regulatory risks in India's business environment, were also cited as main reasons.

Meitu Invests in Visual Agent, Chance AI Completes Multi-Million Dollar Angel Round

Chance AI announced yesterday the completion of a multi-million dollar angel round of financing, led by Meitu, with participation from NYX Ventures, Alibaba-affiliated investors, and others. The funds will be used for model iteration, North American growth, and commercialization exploration.

Founded in 2025, Chance AI's core product is a Visual Agent application with camera as the main entry point, adopting the interaction logic of "see—understand intent—invoke Agent—complete action." It can determine scenarios and intents based on user-uploaded images and dispatch corresponding Agents to execute tasks.

Xiaomi Imaging Algorithm Wins Three Awards at CVPR 2026 NTIRE

Xiaomi Technology announced yesterday that it won three awards at this year's CVPR 2026 NTIRE competition: champion in the Efficient Super-Resolution track, champion in the Portrait Restoration track, and runner-up in the Reflection Removal track, all secured by Xiaomi's Xuanjie Multimedia Algorithm Team and Large Model Application Team.

NTIRE is the largest academic competition in image restoration and enhancement, organized by the CVPR committee. Xiaomi stated that it will continue to push these technologies from research to product implementation.

Yangtze Memory Starts IPO Coaching

Yesterday, according to the China Securities Regulatory Commission website, Yangtze Memory Holdings Co., Ltd. completed coaching registration with the Hubei Bureau of the CSRC for its initial public offering and listing, with CITIC Securities as the coaching securities firm.

BYD Battery CTO Discusses Flash Charging, Lifespan, and Safety Boundaries

According to 36Kr, BYD's Battery Business Group CTO Sun Huajun gave an exclusive interview about the second-generation Blade Battery and flash charging technology, responding in detail to industry controversies regarding flash charging lifespan, battery thermal management mechanisms, and the lithium iron phosphate technology route.

In March this year, BYD launched the second-generation Blade Battery and flash charging technology, supporting 1500kW high-power charging, achieving 10% to 97% charge in just 9 minutes, and requiring only 3 minutes more charging time in -30°C extreme cold environments compared to room temperature.

Regarding the industry criticism that "extreme flash charging damages battery lifespan and crosses safety boundaries," Sun Huajun stated:

The R&D team has completed over 1000 full flash charging cycles as well as extreme temperature difference impact tests; at the technical level, by optimizing cell dimensions and utilizing the natural symmetric heat dissipation characteristics of the blade structure, system internal resistance and temperature difference have been effectively controlled, improving the new generation battery's warranty lifespan by 2.5%.

Regarding industry comments that "using lithium iron phosphate batteries in models priced above 250,000 yuan is essentially cost reduction in disguise," Sun Huajun responded that battery suppliers don't have the right to define premium—it should be determined by the overall vehicle experience and user judgment.

Additionally, in terms of supporting facilities, as of May 6 this year, BYD has cumulatively built 5,924 flash charging stations, with plans to complete 20,000 stations nationwide by year-end to accelerate the full adoption of flash charging technology.

💡 "Bitter Lesson" Reaffirmed by RL Pioneer Sutton Sparks Controversy

Richard Sutton, the father of reinforcement learning, posted on X yesterday, reaffirming his 2019 "Bitter Lesson" in 26 English words:

AI development should not be distracted by human knowledge, but focus on general methods that scale with compute, such as search and learning.

The "Bitter Lesson" was first published in 2019 and is one of the most widely circulated arguments in the AI field. After reviewing decades of AI history, Sutton pointed out that researchers repeatedly tried to encode human expertise into systems, but general methods that scale with compute always won—computer chess, Go, and speech recognition are all examples.

This argument later became an important theoretical basis for the "scaling" approach in large language models.

After this 26-word condensed version was posted, several scholars directly objected. Thomas G. Dietterich, a founding figure in machine learning, replied that researching the abstract structures of human knowledge can identify cognitive blind spots for AI, and such work should not be called a "distraction."

In the comments section, a long article titled "The Sweet Lesson" was referenced multiple times. John Fletcher, Co-founder and Chief Scientist of The Innovation Game, argued in the article that Sutton's argument itself is a form of "encoded knowledge"—it can tell you where the destination is, but not how to get there.

Fletcher believes that with training data facing a written corpus depletion bottleneck, the "tacit knowledge" never written down in expert brains—intuition, experience, and judgment—is the real battlefield for industry competition today.

vivo S60 Series Scheduled for May 29

The vivo S60 series is scheduled for release on May 29 at 19:30, with Tian Ximeng serving as the aesthetics ambassador, featuring the "Star Sea" visual design.

• Gimbal-level stabilized main camera + IMX882 large-base periscope telephoto (3x optical zoom), supporting 4K Live direct output, with new "Starlight Live" dynamic star burst feature;
• 7200mAh ultra-thin Blue Ocean battery + 90W fast charging, supporting "Micro Power精灵" mode;
• IP68 + IP69 dual certification.

Insta360 Mic Pro Officially Released Starting at 698 Yuan

Insta360 officially released its flagship wireless microphone Mic Pro yesterday, featuring personalized customization via a color e-ink screen and the industry's first three-microphone array design. Single transmitter standard retail price is 698 yuan, one transmitter plus one receiver is 1198 yuan, and two transmitters plus one receiver is 1998 yuan.

In terms of appearance, Mic Pro features a six-color e-ink screen supporting black, white, red, yellow, blue, and green primary color display, with AG+AR coating on the screen surface and a 0.9mm Corning glass cover plate.

• Transmitter built with three-microphone array, supporting omnidirectional, super-directional, cardioid, and figure-8 pickup patterns;
• 32-bit float internal recording, 48kHz audio quality, 400m transmission distance, automatic gain adjustment;
• Transmitter battery life of 10 hours, up to 30 hours with charging case; built-in 32GB storage, supporting up to 60 hours of internal recording.

Starting at $1299.99, Microsoft Releases New Commercial Surface

Microsoft officially released its new commercial Surface series yesterday, including Surface Pro for Business and Surface Laptop for Business, powered by Intel Core Ultra Series 3 processors, available in select markets starting today.

Surface Laptop for Business comes in 13-inch, 13.8-inch, and 15-inch options, starting at $1299.99 for the 13-inch, and $1949.99 for the 13.8-inch and 15-inch; Surface Pro for Business in 13-inch 2-in-1 form factor starts at $1949.99, with 5G options available.

• Powered by Intel Core Ultra X7, with graphics performance approximately 35% higher than MacBook Air M5, and over 90% improvement over Surface Laptop 5;
• Surface Laptop 13.8-inch and 15-inch offer up to 23 hours of battery life (local video playback test);
• Some 13.8-inch configurations feature an integrated privacy screen for the first time, supporting one-key activation and IT centralized management;
• Supports local AI inference, including meeting transcription, writing assistance, image generation, and real-time translation.

Microsoft stated that versions powered by Qualcomm Snapdragon X2 processors will be launched later this year, with local AI inference performance improved by approximately 80% over the previous generation.

Starting at 2999 Yuan, Lenovo "AI Host" Officially Unveiled: 190 TOPS Computing Power, 300g Body

Lenovo held a new product launch event yesterday at the Three High Furnace in Beijing's Shougang Park, officially launching Tianxi AI 4.0 and multiple new full-scenario terminal products. A total of 10 products have met the highest level requirements of national L3 AI intelligence standards.

Tianxi AI 4.0 is positioned as a system upgrade from "+AI" to "AI+", with the core being AI-native devices actively executing tasks to achieve inter-Agent collaboration. The new version debuts the Tianxi Claw technical architecture, along with the Skills Plaza. Lenovo also launched the "Sky Realm Plan," providing developers with 1 trillion Token computing power and 10 million yuan in incentive funds.

The main new products released include:

• Lenovo AI Host P7: 190 TOPS computing power, 300g total weight, under 30W full-load power consumption, supports 122B local large model
• Lenovo AI Host Mini: 45 TOPS computing power, 100×100×43mm volume, launch discounted price 2999 yuan
• YOGA Air 14 Ultra: 975g total weight, 17.5dB noise, 17-hour battery life, education subsidy national subsidy final price starting at 7650 yuan
• Legion Y7000X: 170W dual full-power computing, under 2kg body, education subsidy national subsidy final price starting at 8499.15 yuan
• Legion Y900 AI Tablet: Supports "AI Eye-Tracking Multi-Window," 13-inch model education subsidy national subsidy final price starting at 2719 yuan
• Legion Phone Y70 New Generation: Supports full-blooded 2K streaming to take over PC games, powered by Snapdragon 8 Gen 5, launch discounted price starting at 2599 yuan
• Lenovo moto razr fold: 4.43mm unfolded thickness, starting at 9999 yuan
• Lenovo moto razr 70 Ultra: 4-inch full-function AI external screen, starting at 7499 yuan

NVIDIA's First Self-Developed CPU Vera Begins Delivery

NVIDIA officially began delivery of its first self-developed CPU "Vera" yesterday, with employee Ian Buck personally delivering the first products to four partners: Anthropic, OpenAI, SpaceXAI, and Oracle Cloud Infrastructure.

Vera is specifically designed for AI agent scenarios, positioned as the scheduling control layer in AI factories: model inference is still handled by GPUs, while control layer tasks such as agent orchestration, tool invocation, and KV cache management are handled by Vera.

• 88 self-developed Olympus cores, 176 threads;
• Memory bandwidth up to 1.2 TB/s, maximum capacity 1.5 TB;
• Under full load, software environment running speed improved by up to 50% compared to traditional CPUs, with 2x energy efficiency improvement;
• Will serve as the host CPU for the Vera Rubin NVL72 rack system.

Qwen3.7 Preview Version Goes Live on Arena, Ranked First Among Chinese Models on Text Benchmark

Alibaba's Qwen series launched two preview models on Arena yesterday: Qwen3.7-Max-Preview and Qwen3.7-Plus-Preview.

• In the text domain, Qwen3.7-Max-Preview ranks 13th overall, the only Chinese model in the top 15, driving Alibaba Lab's ranking to 6th;
• In the vision domain, Qwen3.7-Plus-Preview ranks 16th, also the only Chinese model on the vision benchmark, with Alibaba Lab's ranking rising to 5th.

Google AI Studio Launches on Play Store, Open for Pre-Registration

Google AI Studio launched its Android version on the Google Play Store yesterday, currently open for pre-registration.

According to the Google Play Store page, the app's main feature is building applications through natural language interaction—users describe their needs via voice or text, and the AI directly generates a runnable application. Generated applications can be shared via links, and recipients can use them in their browsers.

Odyssey Releases Two World Models in Two Days, Supporting Multiplayer Co-op Generation

AI startup Odyssey recently released two world models in succession, completing a technological leap from single-player real-time interaction to multiplayer networked generation within two days.

• Starchild-1 decouples audio-visual context through causal distillation and asynchronous KV caching, allowing users to intervene at any time via voice, text, or actions, with the model responding in real-time and altering subsequent visuals and ambient sounds;
• Agora-1 extends real-time interaction to multiplayer scenarios, supporting up to 4 real or AI players in the same GoldenEye 007 deathmatch game.

As previously reported by TechCrunch, Odyssey was founded by two founders with autonomous driving backgrounds and has raised a total of $27 million in funding.

Horizon Robotics Open-Sources 400M Parameter Humanoid Robot Control Model HoloMotion-1

Horizon Robotics recently released and open-sourced HoloMotion-1, designed for full-body control of humanoid robots. With a parameter scale of 400 million, it achieves approximately 300 FPS real-time inference on edge devices.

HoloMotion-1 is the first phase of the HoloMotion technical roadmap, focusing on learning complex full-body movements from video, MoCap, and teleoperation data. It uses a Mixture of Experts (MoE) Transformer architecture as its foundation.

Horizon stated that open-sourcing aims to lower the R&D barrier for full-body control of humanoid robots and accelerate the scale application of related technologies in real-world scenarios. According to the roadmap, subsequent versions will sequentially address language and task-conditioned motion generation, complex terrain adaptation, and cross-body generalization, all currently in planning or progress.

Anthropic Launches Self-Hosted Sandboxes, Claude Agent Tool Execution Can Remain Within Enterprise Intranet

Anthropic yesterday launched self-hosted sandboxes for Claude Managed Agents, along with the release of MCP tunnels.

The new architecture separates the agent's orchestration layer from the execution layer: task orchestration, context management, and error recovery continue to run on Anthropic's cloud, while tool calls and code execution are transferred to the enterprise's own infrastructure or hosted sandboxes.

Enterprises can configure CPU, memory, and runtime images independently, and compute-intensive tasks can directly leverage local computing power. Currently, four sandbox providers are supported: Cloudflare, Daytona, Modal, and Vercel.

Tencent WorkBuddy Launches One-Person Company Incentive Plan, Distributing 30 Million Credits

Tencent WorkBuddy officially launched a one-person company incentive plan, targeting global entrepreneurs, OPCs (one-person companies), and WorkBuddy power users, distributing 30 million WorkBuddy Credits. Users who receive incentives can directly use AI credits within WorkBuddy for business efficiency and innovation.

The launch of this incentive plan is closely related to the wave of AI agent-driven "super individuals." According to the introduction, relying on AI agents, individual entrepreneurs without offices and large teams are gaining power comparable to traditional enterprises, making the OPC model an emerging trend in the current entrepreneurship market.

QuestMobile: Instant Retail Expands to Lower-Tier Markets, AI Becomes New Variable in Platform Competition

QuestMobile yesterday released the "2026 Local Life Consumption Insight Report." As of March this year, the monthly active user scale for comprehensive e-commerce, local life services, food services, and fresh food e-commerce reached 1.095 billion, 569 million, 189 million, and 137 million respectively. After subsidies receded, platform competition focus has shifted from price to supply chain efficiency and user mindshare.

• Food delivery platforms generally experienced traffic decline, with users migrating to comprehensive consumption platforms; food services industry MAU decreased by 3.6% year-over-year;
• "Store-warehouse integration" has become the mainstream: Meituan launched dual-brand front-line warehouses, JD 7Fresh bet on asset-heavy stores, and Hema accelerated its layout in lower-tier markets;
• In-store group buying has become the new battlefield: Dianping's growth rate dropped to 11.4%, while DouShengSheng's daily active users have approached 16 million, with users in second-tier and below cities accounting for 67.3%;
• Rider capacity extends to lower-tier markets: Meituan's crowdsourcing and Ele.me's crowdsourcing saw year-over-year increases of 6.2% and 12.9% respectively in third-tier and below cities;
• AI entry battle heats up: Alibaba's Qianwen connects consumption and payment chains, while Meituan and JD are respectively deploying AI assistants and unmanned delivery.

Second Domestic Large Cruise Ship Opens Sales Today

According to CCTV News, Adora Cruises, a domestic cruise brand, announced yesterday that the second domestic large cruise ship, "Adora Magic City," will set sail from Guangzhou Nansha International Cruise Port on November 22.

The inaugural voyage season tickets will go on sale on May 20, covering the period from November 22, 2026 to January 2, 2027, including 6 short, medium, and long-route products. The routes target family travel, holiday trips, and cruise experience users.

"Adora Magic City" is the second domestic large cruise ship following the first one. Guangzhou Nansha will serve as its inaugural port and will also accommodate more South China tourism and travel demands.

"The Party of Heroes is Wiped Out." Scheduled for July 22

According to GCORES news, game development team デス組合 and publisher Flashlight Games announced that the dungeon exploration adventure game "The Party of Heroes is Wiped Out." is expected to release its official version on July 22, 2026. The official also released a commemorative illustration for reaching 100,000 wishlists.

The game's premise is that only one person can be rescued from the already-wiped-out party of heroes, requiring players to make trade-offs during dungeon exploration and character selection.

Peanuts Family Joins the Sony Family

Sony announced yesterday early morning that the Peanuts family officially joins the Sony family.

In March this year, Sony Music Entertainment (Japan) and Sony Pictures Entertainment announced the completion of the equity acquisition of Peanuts Holdings LLC, the parent company of Peanuts, with a combined ownership ratio of 80%.

According to the equity structure change, the family of Charles M. Schulz, the creator of Peanuts, will continue to retain the remaining 20% of the IP.

Sony stated that the ownership rights and daily business operations of the "PEANUTS" brand will continue to be managed by Peanuts Worldwide LLC, a wholly-owned subsidiary of Peanuts Holdings LLC.

Hello Kitty's Big Screen Debut Scheduled for 2028, Director of "Moana 2" Leads the Project

According to Deadline, Sanrio's Hello Kitty live-action/animation hybrid film confirmed its director yesterday: David Derrick Jr. and John Aoshima, who previously directed "Moana 2," will co-direct. The film is scheduled for North American release on July 21, 2028.

The film will follow Hello Kitty and her friends on an adventure, with Sanrio characters like Pompompurin, My Melody, and Twin Stars expected to appear. The film is described as a combination of animation and live action, serving as Hello Kitty's official debut in Hollywood theaters.

"The Bureau" Season 2 Scheduled for June 21

According to Deadline, Paramount+ series "The Bureau" has confirmed its release date for season 2, with all episodes dropping on June 21. The official also released a trailer and first batch of stills.

"The Bureau" is adapted from the French series "Le Bureau des Légendes," starring Michael Fassbender, Jeffrey Wright, and Richard Gere. Season 2 continues to focus on internal operations, identity concealment, and character relationships within the intelligence agency.

The Hollywood Reporter also reported on the season 2 trailer information, summarizing the new season's tagline as "Trust No One." This series is one of Paramount+'s major global market projects for this year.