Pogbagram

Apple officially announced its new CEO, John Ternus, who will take over in September.

Early this morning, Apple officially announced that Tim Cook will step down as CEO, and John Ternus will take over. Simultaneously, two internal letters were leaked, one from current CEO Tim Cook and the other from John Ternus.

In his letter, Cook stated that he would step down as CEO and become chairman later this year. He had several in-depth conversations with Tenus and found him to be visionary and possessing an exceptional character.

John is the right leader to help us move into the future and explore bold new directions.

Cook confirmed he will continue as CEO this summer and hand over the reins to Tenus in September; Tenus also confirmed in a letter that he will take over as CEO this September.

This September, I will become Apple's new CEO. It has been a tremendous honor to lead the hardware engineering team, participate in such outstanding work, and witness each and every one of you dedicate yourselves to doing the best for users. I look forward to working closely with you in this new role, and of course, I will continue to be personally involved.

Starting today, Tenus will step down as head of hardware engineering, with Tom Marieb taking over and reporting to Johny Srouji. Apple also announced that Srouji will assume the newly created position of Chief Hardware Officer.

Related reading: Apple just announced its new CEO; Cook's successor is more like Jobs.

Huawei releases the Pura X Max and Pura 90 series, featuring a large foldable smartphone. Yu Chengdong, CEO of Huawei, stated that prices may increase if the market cannot withstand the pressure.

Yesterday, Huawei officially launched the Pura 90 series and its first large foldable screen phone, the Pura X Max, at its Pura series and all-scenario new product launch event. The entire Pura 90 series features a flat screen design and focuses on imaging upgrades; the Pura X Max, on the other hand, boasts a large foldable screen and is positioned as a flagship foldable form factor phone. Let's look at the pricing:

The Pura 90 series starts at 4,699 yuan, the same as the previous generation Pura 80 series. The Pura 90 Pro starts at 1,000 yuan less than the Pura 80 Pro. The Pura X Max starts at 10,999 yuan.

At the press conference, Yu Chengdong, Huawei’s executive director and chairman of the terminal BG, said that due to the continuous rise in the price of memory chips, the average cost of a single Pura 90 series unit has increased by about 1,500 yuan, putting significant pressure on pricing.

We're under a lot of pressure, and we've really been working incredibly hard. The pricing pressure for the new phones is immense right now, and if we can't keep up with the costs, we might have to raise prices later.

• The Pura 90 standard edition is equipped with the Kirin 9010S, featuring a rear camera setup of a 50MP f/1.8 main camera + a 50MP f/2.2 periscope telephoto lens + a 12.5MP ultra-wide-angle lens + a red maple lens, and a 50MP front camera; it has a built-in 6500mAh battery and supports 100W wired fast charging and 50W wireless fast charging;
• The Pura X Max is positioned as the industry's first large foldable screen phone, and all models are equipped with the Kirin 9030 Pro. Its design is inspired by the golden ratio of A4 paper, and it focuses on a "wide inside and out, immersive in both horizontal and vertical orientations" audio-visual, gaming and office experience.
• The Pura 90 Pro and Pro Max will both debut with the Kirin 9030S, featuring a 50MP main camera with a 10-stop variable aperture (f/1.4-f/4.0), a built-in 6000mAh battery, and support for 66W wired fast charging and 50W wireless fast charging.
• The Pro Max version further enhances imaging specifications, upgrading the telephoto lens to the industry's first ultra-large sensor 200-megapixel RYYB lens based on the Pro version, supporting 20x optical HD video recording and CIPA 7.0 image stabilization; the screen is upgraded to 6.9 inches, featuring the first anti-reflective and scratch-resistant Kunlun glass, reducing reflections by 70%; and it supports 100W wired + 80W wireless fast charging.

Also launched at the event were the Huawei AI Glasses, starting at 2499 yuan; the MateBook 14 HarmonyOS Edition, starting at 6599 yuan; the Huawei Watch FIT 5 Pro, starting at 2099 yuan; the Huawei Watch Buds 2, starting at 3488 yuan; and the Huawei Watch Master, priced at 29999 yuan.

ByteDance heavily invested in AI, but its net profit shrank by more than 70% year-on-year last year. Executives say the decline wasn't that significant.

According to reports from Jiemian News and 21st Century Business Herald, sources familiar with the matter revealed that ByteDance's net profit in 2025 declined by more than 70% year-on-year, with a significant contraction in net profit margin.

The core reason is that the company significantly increased its resource investment in the field of artificial intelligence in the third and fourth quarters of last year, including the procurement of high-end AI chips and the development of underlying models, among other core aspects. The huge expenditure directly dragged down the company's profit performance for the whole year.

Despite a significant decline in net profit, ByteDance's revenue still maintained strong growth. According to sources familiar with the matter, ByteDance's overall revenue continued its growth trend last year, with domestic revenue increasing by about 20% year-on-year and overseas revenue increasing by nearly 50% year-on-year. The proportion of overseas business revenue increased from 25% in 2024 to more than 30%, setting a new record.

In its communication with shareholders, ByteDance revealed that it will further expand its investment in technology resources this year. Coupled with supporting investments related to compliance in overseas markets, net profit margin will continue to be under pressure in the short term.

In response, ByteDance executive Li Liang issued a clarification, stating that the aforementioned net profit decline of over 70% was based on international accounting standards and included factors that "do not reflect the actual operation," such as changes in preferred stock and option costs, in addition to increased investment in emerging businesses.

He stated that the actual operating profit margin only declined slightly in the second half of the year, "far less than the significant decline reported." Excluding the aforementioned non-operating factors, overall revenue and profit still achieved growth.

Tesla China: There are currently no specific plans to mass-produce robots at the Shanghai Gigafactory.

According to the Securities Times, Tesla China responded yesterday to online rumors that "Tesla is about to mass-produce robots in Shanghai," clarifying that the claim was a misinterpretation.

Tesla currently has no concrete plans to mass-produce robots at its Shanghai Gigafactory. Please do not describe in reports that Tesla is about to mass-produce robots in Shanghai; this is incorrect information.

According to reports, on April 14, when Tesla Vice President Wang Hao was asked about robots at a media event, he only discussed that the Shanghai Gigafactory has good large-scale mass production capabilities and great potential in the future, without making any specific commitments regarding the mass production of robots.

It is reported that Tesla's third-generation humanoid robot is planned to be released at the end of 2026. The Fremont factory is transforming the Model S/X production line into a production line for the humanoid robot Optimus, with a final planned annual production capacity of one million units.

In pursuit of Yu Hao: The mobile phone market should be divided into three parts with Huawei and Xiaomi, but I can't tell you how to do it right now.

According to LatePost, Dreame Technology founder Yu Hao gave an in-depth interview yesterday lasting six and a half hours. Regarding his previous public statement that Dreame phones would compete with Huawei and Xiaomi in the future, Yu Hao stated that Dreame originally did not intend to disclose its mobile phone business, but "it couldn't be kept secret any longer."

We didn't originally intend to talk about it, but we couldn't keep it a secret any longer. When the car team reached 800 people and the mobile phone team had several hundred, there were many rumors circulating outside, and these rumors led to many misunderstandings.

As for how to achieve a three-way division of power, he did not provide a clear path, but rather incorporated it into his consistent methodological framework:

Getting from point A to point B, constantly testing and providing feedback to adjust the route—this approach might be better in today's increasingly internationalized and diverse technological landscape. You want me to tell you how to do it from day one? I can't do that.

Beyond smartphones, Yu Hao also revealed plans to manufacture cars, also targeting the high-end market, aiming to create the "fastest pure electric car" with a price higher than Tesla's.

Yu Hao also revealed that Dreame has maintained an average revenue growth rate of 100% for eight consecutive years since its establishment, and its global net profit margin has exceeded 20% in the most recent quarter. Its main business continues to be profitable, which he believes is the fundamental strength supporting its multi-line expansion.

The next generation of Leica cameras will be equipped with a domestically made custom CMOS sensor.

Gpixel, a domestic CMOS image sensor supplier, officially announced a strategic partnership with Leica yesterday. The two companies will jointly develop a high-performance CMOS image sensor chip for the next generation of Leica cameras.

It is understood that this collaboration will integrate Leica's years of professional experience in the field of high-end imaging with Changguang Chenxin's leading CMOS image sensor design capabilities to create a customized chip that precisely matches the stringent standards of Leica cameras in terms of color reproduction, noise optimization, dynamic range, and detail rendering.

The two parties will also collaborate closely on chip verification, image calibration, and mass production preparation, covering the entire chain from R&D to mass production. Engineering and technical teams from Wetzlar, Germany; Antwerp, Belgium; and Changchun, China will participate in the project.

Ford CEO: The reason for choosing to test drive the Xiaomi SU7 was because Tesla "didn't have an updated model".

Recently, Ford CEO Jim Farley explained on the podcast "Rapid Response" why he chose to drive a Xiaomi SU7 instead of a Tesla in 2024 to understand market competition.

Farley responded that if the goal is to compete head-on with Chinese automakers in the automotive sector, the focus doesn't necessarily have to be on Tesla.

This isn't directed at Tesla. They're doing very well; it's just that they don't have any updated models right now.

Farley believes that Chinese automakers, represented by BYD, possess significant advantages in cost control, supply chain management, and manufacturing capabilities, making them arguably the "strongest in the industry." He stated that if Ford can learn from BYD's cost competitiveness and apply it to its more familiar market segments, it could gain an advantage in the next round of competition.

He also pointed out that the next phase of electric vehicle consumers in the United States still favor models such as pickup trucks and SUVs, but price expectations have dropped from the early $50,000 to around $30,000, with "affordability" becoming the core demand.

Earlier, he revealed on a podcast that he drove a Xiaomi SU7 for six months straight in 2024 and refused to change vehicles. In early April of this year, he went even further on the Fox & Friends show, warning that if Chinese cars were to enter the US market on a large scale, it would have a "devastating" impact on American manufacturing.

China accounts for over 80% of global shipments of humanoid robots.

According to CCTV.com, China's shipments of humanoid robots have accounted for more than 80% of the global market share, and robots are becoming a new label for China's foreign trade.

While China's domestic robotics industry continues its explosive growth, Chinese-made robots are accelerating their overseas expansion, with exports continuing to grow rapidly. Data from the General Administration of Customs shows that in the first quarter of this year, China's exports of machinery and electronic products reached 4.34 trillion yuan, accounting for more than 60% of total exports.

Among them, as a representative of new quality productivity and an important carrier of future industries, various types of "robots" produced in China have been exported to more than 140 countries and regions around the world.

Lingguang App invests 100 million yuan to launch a creator incentive program

Ant Light App officially announced yesterday the launch of the "Light Flash App Creator Incentive Program", investing a special fund of 100 million yuan to support high-quality flash apps and outstanding creators on the platform.

According to the program rules, up to 1 million yuan will be awarded daily to 10,000 high-quality flash apps, and up to 1 million yuan will be awarded weekly to 100 outstanding creators. Users can participate in the selection by creating a flash app on the Lingguang App and publishing it to the Lingguang Circle.

Alongside the incentive program, a brand-new feature called "Aura Circle" is also being launched, a content platform designed specifically for the Flash App ecosystem. Within Aura Circle, personally created Flash Apps can be shared, liked, and used for secondary creation.

Related reading: What if apps could spread like social media?

The company scheduled the interview for Sunday, which the job seeker found incredibly convenient.

According to Business Insider, as artificial intelligence makes it easier to inflate resumes, more and more employers are introducing "work trials," requiring job seekers to complete real projects or work on-the-job before being formally hired, in order to verify their actual abilities.

However, this trend has also put new pressure on employed job seekers—they often have to take time off work to participate in these more time-consuming recruitment processes.

Faced with this dilemma, the startup legal services firm Crosby offered its own solution: scheduling interviews on Sundays. Crosby CEO Ryan Daniels stated that this arrangement initially worried him, as it might provoke resistance from job seekers, but many candidates actually felt relieved.

Many people would even say, "This is so convenient."

Sunday interviews allow employed job seekers to avoid using their annual leave, while also aligning with the work habits of company executives, who often handle business on Sundays and have relatively free calendars on those days, giving them more time to engage in in-depth interviews.

In terms of specific format, Crosby adopts differentiated assessment methods for different positions. Software engineer candidates are directly placed in real projects, where their coding skills are assessed not only but also how they use AI-assisted tools in actual work; business position candidates participate in group interviews with the senior management team.

After the interview, candidates will have lunch or dinner with the team to assess their cultural fit. Daniels believes this step is crucial for attracting top candidates: "If they fit in with the team, we certainly won't miss them."

Kimi K2.6 Released and Open Source

Yesterday, Dark Side of the Moon officially released and open-sourced its new model, Kimi K2.6, with key upgrades to the code, AI agent, and office capabilities.

• Enhanced coding capabilities: Internal benchmark scores improved by approximately 20%, capable of coding continuously for 13 hours and processing over 4000 lines of code, supporting multiple languages ​​including Rust, Go, and Python;
• AI agent cluster expansion: Supports up to 300 sub-AI agents to collaborate in parallel, completing 4,000 steps, significantly improving task completion quality;
• Long-term autonomous operation: Supports continuous autonomous operation for 5 days without human intervention within frameworks such as OpenClaw and Hermes Agent;
• Upgraded design capabilities: It can generate web pages and web applications with better visual effects, and is significantly better than Gemini 3 in front-end design evaluation;
• New Skills Feature: Supports "Office Document to Skill" conversion, and includes hundreds of officially recommended skills built into the system.

In industry benchmark tests, K2.6 performs on par with or better than closed-source models such as GPT-5.4, Claude Opus 4.6, and Gemini 3.1 Pro in tests such as SWE-Bench Pro and Humanity's Last Exam.

This release also previewed a small-scale beta test of the "Claw Group" feature, which allows humans and multiple AI agents to collaborate within the same group, with the OpenClaw framework being among the first to be supported.

Hugging Face: huggingface.co/moonshotai

ModelScope: modelscope.ai/organization/moonshotai

Inspiron Luna Ultra Hands-on Review: Leica Dual Camera, 1-inch Sensor, 14-stop Dynamic Range

Recently, blogger David Manning got his hands on InSight's upcoming dual-camera handheld gimbal camera, the Luna Ultra, and revealed five key specifications.

• The lens was jointly developed by Inkstone and Leica;
• The phone features a dual-camera design and supports switching between multiple focal lengths, but the specific focal length values ​​have not yet been announced.
• It features a 1-inch sensor, an F1.8 aperture, a dynamic range of 14-15 stops, and supports 10-bit video recording;
• In the video, Manning also hinted at "another innovation" on the front of the fuselage, but said it could not be disclosed at the moment.

Luna Ultra has been confirmed to be showcased at NAB 2026, held from April 19th to 22nd this year. Inspiron CEO Jingkang Liu publicly stated at the company's annual meeting in February that Luna would be officially released in the first half of this year, emphasizing that it is "definitely not another Pocket."

Starting at 428,000 yuan, the all-new Tank 700 is officially launched.

Yesterday, the all-new Tank 700 was officially launched, with a starting price of 428,000 yuan.

In terms of exterior design, the new Tank 700 continues the narrative of Eastern aesthetics, introducing the world's first mechanical rotating and opening laser headlight "Rainbow Eyes" and the industry's first exclusive rear wing for hardcore off-road vehicles "Auspicious Flame Tail". The rear adopts a 6-light layout; the new color "Dunhuang Green" is inspired by the original color of lapis lazuli in Dunhuang murals.

• In terms of power, the standard version is equipped with the Hi4-Z super hybrid architecture for off-road use, and adopts the world's first power split technology; the peak version is equipped with a self-developed 3.0T twin-turbocharged hybrid engine.
• All models come standard with ventilated/heated/massaging fiber cloud seats, a 21-speaker 7.1.4 Dolby Atmos system, a 5.4L intelligent refrigerator, and a 17.3-inch 3K rear ceiling screen;
• It features VLA large model and Coffee Pilot 4.0, supporting spoken commands.

Also launched was "Qilin Club," positioned as a high-end user community. It initially opened up five interest groups: equestrian, sea fishing, photography, and business, and launched the "Top 100 Managers" program to provide members with niche hidden routes and exclusive tour leader services.

Qwen 3.6-Max-Preview is now online.

Yesterday, the Qianwen team released an early preview version of its flagship model, Qwen3.6-Max-Preview. As the next-generation successor to Qwen3.6-Plus, the new model has achieved significant improvements in three dimensions: agent programming, world knowledge, and instruction compliance.

• SkillsBench improved by 9.9 points, SciCode by 10.8 points, NL2Repo by 5.0 points, and Terminal-Bench 2.0 by 3.8 points;
• In terms of knowledge and skills, SuperGPQA improved by 2.3 points and QwenChineseBench improved by 5.3 points;
• In terms of instruction compliance, ToolcallFormatIFBench improved by 2.8 points.

According to the official statement, the model achieved the highest scores on six major programming benchmarks: SWE-bench Pro, Terminal-Bench 2.0, SkillsBench, QwenClawBench, QwenWebBench, and SciCode.

According to CBN, Alibaba's ATH announced yesterday that its AI video generation project HappyHorse-1.0 will gradually open its API for testing through the Alibaba Cloud Bailian platform on April 27, with enterprise customers being the first batch of invited testers.

Related reading: Just now, the strongest model from Qianwen was released, topping the list of best domestically produced models.

Alibaba releases Fun-ASR 1.5, a large-scale speech recognition model.

Yesterday, Alibaba released its new generation end-to-end speech recognition model, Fun-ASR 1.5, which covers 30 languages ​​including Chinese, English, Japanese, Korean, French, and German with a single model. The underlying architecture adopts MoE (Hybrid Expert) architecture and has achieved state-of-the-art (SOTA) results in multiple open-source multilingual test sets.

In terms of Chinese dialects, the model is trained on hundreds of thousands of hours of real speech data, and the average word error rate (CER) has decreased by 56.2% compared with the previous version. It covers seven major dialect systems including Cantonese, Minnan, Hakka, and Shanghainese, and adds a special recognition of ancient poetry. The internal evaluation shows that the character-level accuracy rate reaches 97%.

In addition, the model can automatically insert punctuation and normalize spoken expressions into standardized formats, such as converting "fifty-eight thousand yuan" into "58,000 yuan", which greatly reduces the manual proofreading costs in scenarios such as meeting minutes and legal transcripts.

Mengshi Motors, in collaboration with Huawei Qiankun, released a new intelligent off-road architecture.

Mengshi Motors held a spring refresh launch event yesterday, announcing that its cooperation with Huawei Qiankun has been upgraded from "product co-creation" to "deep empowerment", and jointly released a number of new technologies for intelligent off-road scenarios.

The two parties launched a brand-new intelligent off-road architecture, equipped with the industry's first 896-line lidar, and adopt a "dual optical path" design. The official claim is that it can stably detect obstacles 14 centimeters high and 120 meters away, and improve the road perception accuracy to over 95%.

In terms of chassis, the Rock Chassis 2.0 incorporates Huawei's Qiankun XMC digital chassis engine; in terms of power, the Extreme Power 2.0 boasts a combined range of 1450km, a 0-100km/h acceleration time of under 4 seconds, and supports 6C fast charging. The "Whale Fin" communication system supports dual SIM dual standby, with the official claim of signal strength exceeding industry standards by 60%.

Mengshi Motors stated that it will implement the above technologies in more than four new models in the next two years, and plans to expand its off-road ecosystem base to 100 locations next year.

Gree speaks out on "real copper materials": Discontinuing production of engineering models does not mean that home models will be downgraded.

According to Sina Tech, Gree Electric Appliances issued an industry publicity and standards initiative statement yesterday, and made a public response to the online dispute with Hisense Air Conditioner over "real copper materials".

In its statement, Gree pointed out that "genuine copper materials" is a clear standard and original proposition put forward by Gree based on its commitment to product quality. Gree believes that Hisense's use of this expression in its advertising and its slogan "Good air is made by Hisense" are highly similar to Gree's classic slogan "Good air conditioners are made by Gree".

In response to online allegations that "some Gree models use aluminum wire motors," Gree Electric Appliances clarified that the models in question are 5-horsepower commercial ducted air conditioning units and 5-horsepower commercial ceiling-mounted air conditioning units. These are specially customized small-batch engineering units, not household retail models for ordinary consumers, and the aforementioned products have long been discontinued.

Gree further emphasized that the company implements unified and stringent standards for all domestic retail air conditioners: all household models use genuine copper materials for both the heat exchanger and the motor, and the motor uses 100% copper wire. "We will never substitute aluminum for copper, and we will never reduce the quality or cut corners."

Gree advocates that the industry return to the essence of products, promote transparency in the materials used in end products in the consumer market, and disclose the materials of the four core components: condenser, evaporator, compressor and motor.

Jia Guolong's braised noodles opened three stores in Shanghai, and the affiliated company completed its name change.

According to Red Star Capital Bureau, "Tianbian Casserole Noodles," a new brand under Xibei founder Jia Guolong, opened three stores simultaneously in Shanghai yesterday, located in Jing'an, Pudong, and Jiading districts. The Pudong Dongfang Road store is the largest in Shanghai, with a building area of ​​over 2,400 square meters and 350 seats.

The brand's main product is the clay pot braised noodles, a specialty of Bayannur, Inner Mongolia, the hometown of Jia Guolong. The core categories include clay pot braised noodles and clay pot soup noodles, accompanied by side dishes such as braised dishes, clay pot dishes, and cold dishes. The average cost per person is about 50 yuan.

In addition, business registration information shows that Beijing Jia Guolong Air Bun Catering Management Co., Ltd. has recently completed its business registration change, changing its name to Tianbian Brand Management (Beijing) Co., Ltd., and its legal representative has changed from Fu Guo to Li Zhihong.

Inner Mongolia Xiyue Fendou Catering Co., Ltd. was recently established, with Qin Yongjun as its legal representative. It is jointly owned by Tianbian Brand Management (Beijing) Co., Ltd. and Qin Yongjun, and its business scope covers catering services, food business management and liquor business.

Jia Guolong had previously revealed that one of the core purposes of launching the new brand was to take over some of Xibei's idle stores and employees.

Multiple platforms respond to 12306's refusal to issue tickets.

According to China News Service, during the peak period of ticket purchases for the May Day holiday, the 12306 platform launched risk control upgrades, and the train ticket issuance functions of multiple third-party business travel platforms experienced abnormal fluctuations.

Customer service representatives from Tongcheng Business Travel, Ctrip Business Travel, and Didi Enterprise Edition have all confirmed that the train ticket issuance system is experiencing abnormal fluctuations or delays due to the 12306 risk control upgrade and the May Day booking peak. All platforms stated that they are working on fixing the issues.

Ctrip's corporate travel customer service explicitly stated that the train ticket waitlist function is no longer available. "The backlog of business has been quite severe these past two days, and our technical staff are currently working on fixing it." They added that they are still waiting for further notice regarding when the function will be restored.

Meanwhile, customer service representatives from Ctrip, Fliggy, and Qunar all stated that the waitlist ticket-grabbing function is still working normally.

iQiyi released a database of 100 AI-powered artists, and several actors named mentioned had not signed any AI licensing agreements.

According to Jiangnan Metropolis Daily, the 2026 iQiyi World Conference was officially held yesterday. At the conference, iQiyi CEO Gong Yu announced the AI ​​Artist Database plan, revealing that more than 100 artists have already joined the platform's AI Artist Database.

Video footage from the scene shows that Yu Hewei, Zhang Ruoyun, Ma Su, Chen Zheyuan, Zeng Shunxi, Cheng Lei, Wang Churan, Director Fang, Jiang Long, and others were named among those listed.

In his speech, Gong Yu said that AI can help actors have "more time for themselves," and that "actors often work very hard, filming for three or four months at a time, without any personal life. AI can also help actors increase the number of films they can do from four to fourteen a year, giving everyone more time to rest."

However, several of the celebrities named subsequently issued statements denying the allegations. Zhang Ruoyun, Wang Churan, Li Yitong, and Yu Hewei all stated that they had never signed any AI-related licensing agreements. The significant discrepancy between the official announcement and the statements from the individuals involved has plunged the event into a public relations crisis.

In response to the controversy, iQiyi subsequently issued a statement. The statement pointed out that the original intention of establishing the NaDou Pro artist database was to provide AIGC creators with a standardized platform to facilitate their selection of artists during the creation process and to enable them to efficiently communicate cooperation details with artists through iQiyi.

iQiyi emphasizes that being added to the artist database only means that the artist "has the intention to engage in AI film and television projects," but whether they participate in specific projects or play specific roles requires separate negotiation and authorization. "This process is consistent with the traditional cooperation process for live-action film and television projects."

The live-action film adaptation of "Elden Ring" has been officially announced and is scheduled for release in 2028.

Yesterday, Bandai Namco Entertainment announced that the live-action film adaptation of the popular game "Elden Ring" has officially been scheduled for release on March 3, 2028, in North America in IMAX format, with filming to begin this spring.

This film is a co-production of A24 Pictures and Bandai Namco Entertainment, directed and written by Alex Garland—whose notable works include *Annihilation*, *Ex Machina*, and *American Civil War*. George R.R. Martin, the author of the original game, will serve as one of the producers.

Wanda Cinemas will not change its brand name.

Yesterday, after Wanda Cinema officially changed its name to "Ruyi Film Entertainment Co., Ltd.", Chairman and President Chen Zhixi issued an all-staff letter explaining the rationale behind the name change:

"Confucianism" represents the content creation gene inherent in the company's DNA, while "Film and Entertainment" demonstrates the strategic determination to focus on offline scenarios and cultivate diverse entertainment experiences.

It is worth noting that this name change only applies to the listed company entity. Its physical cinemas will continue to use the "Wanda Cinemas" brand, and new brands will be planned in the future based on market changes and user needs.

At the strategic level, Chen Zhixi reviewed the transformation path of Ruyi in the two years since its acquisition in an all-staff letter: the company has evolved from a simple screening terminal to an industry closed loop of "content development – theatrical screening – IP operation", and has introduced diverse content such as the Spring Festival Gala, F1, and the League of Legends World Championship into cinemas.

Chen Zhixi stated that the company will leverage its more than 700 directly operated cinemas nationwide and its Confucian IP reserves to create "super scenes" and "super IPs".

#Welcome to follow iFanr's official WeChat account: iFanr (WeChat ID: ifanr), where more exciting content will be presented to you as soon as possible.

Last week, DJI released the Pocket 4.

This machine can be said to have been born with a silver spoon in its mouth—it perfectly inherits the product form of the Pocket 3, which has had no rivals in the past three years, raking in nearly 20 billion in revenue with just one product, which is even equivalent to half of the total annual shipments of traditional camera manufacturers worldwide.

This time, the Pocket 4 arrives with terrifying "dominance," featuring a 38-megapixel 1-inch sensor, magnetic flash, and 107GB of built-in storage, pushing it further towards the ultimate form of a handheld gimbal camera.

On the other side of the world, in a corner that no one pays attention to, GoPro, the former king of action cameras, has also released a brand new product line – the MISSION 1 series, equipped with ultra-high video specifications and even supporting external M43 lenses, turning action cameras into miniature movie cameras.

The product is decent, but GoPro is really on the verge of collapse.

Action cameras have evolved into "action movie cameras".

The MISSION 1 series is based on the concept of a "compact action cinema projector," and the entire product line is divided into three models according to their form factor and specifications:

MISSION 1

As the foundation of the entire series, the MISSION 1 standard edition features a 50-megapixel 1-inch sensor, powered by the new GP3 processor.

This is also the core dock that is common to all three models in the MISSION 1 series. GoPro has not reduced the most critical hardware performance because of the different product positioning.

In terms of video specifications, the MISSION 1 standard version can shoot ultra-high-definition footage at up to 8K 30fps; for sports scenes such as skiing and surfing, 4K 120fps and 1080P 240fps slow-motion recording are also more than enough.

In line with the current trend of short video creation, the MISSION 1 standard edition retains the 4K 120fps Open Gate full-sensor recording capability. Creators can record the entire frame first and then freely crop the aspect ratio according to the platform's requirements later.

MISSION 1 PRO

With the "PRO" suffix, the MISSION 1 PRO is directly targeted at professional video creators and film crews.

Its underlying hardware remains a 50-megapixel 1-inch sensor paired with a GP3 processor. But within this same shell, GoPro has forged a more powerful soul—

In addition to the specifications of the standard version, the MISSION 1 PRO also supports shooting 8K 60, 4K 240, and 1080P 960 frames per second 16:9 aspect ratio video; Open Gate full sensor acquisition also goes all the way up to 8K 30 frames per second and 4K 120 frames per second.

Numerically speaking, the PRO version's video specifications are almost double those of the standard version.

MISSION 1 PRO ILS

As the name MISSION 1 PRO ILS suggests, this is the most unique model in the series.

The ILS's core configuration and specifications are completely on par with the PRO version, but GoPro made an extremely bold decision with it: completely abandoning the fixed ultra-wide-angle lens and adopting a mirrorless body form, with the 1-inch sensor directly exposed.

The original lens location has been replaced by a large M43 mount, which can be used to connect lenses with the corresponding mount and provides HyperSmooth image stabilization to meet the shooting needs of different scenarios and subjects.

To complete the final piece of the puzzle for professional recording, the camera also features built-in 32-bit floating-point audio recording. For audio recording, this is equivalent to the RAW format in the audio world, meaning it boasts complete audio and video capabilities.

The only regret is that, based on the currently released images, there are no electronic contacts visible on the M43 mount of the MISSION 1 PRO ILS.

In other words, the MISSION 1 PRO ILS will most likely not be able to control the lens aperture through the camera body, and will also completely abandon autofocus—so that's how it got its "movie camera" positioning (laughs).

Judging from the official images, the GoPro MISSION 1 series also comes with a controller, and its design looks very familiar…

Beyond the leap in video specifications and form factor, the foundation upon which the MISSION 1 series truly stands is the GP3 processor.

Long-time users will certainly be familiar with its predecessor, the GP2. This chip has been in service since the Hero 10 in 2021, and it held up well until the Hero 13. It was still quite capable a few years ago, but as time went on, the uncontrollable noise and the serious overheating after recording for a while made the GP2's fatigue impossible to hide.

For a long time, the GP2, whose computing power had run out, was the culprit behind GoPro's incremental upgrades.

The old engine couldn't pull the carriage of the new era, so GoPro finally made the tough decision to switch to the GP3 processor in the MISSION 1 series.

The upgrade in underlying technology has led to a dramatic increase in computing power, which is the physical prerequisite for the MISSION 1 series to support massive data streams such as 8K or high frame rates. This also allows for the implementation of 10-bit color, HLG HDR, and GP Log 2 curves and specifications, preserving greater post-processing flexibility. Furthermore, the GP3 introduces AI-based image processing technology, enabling the camera to capture cleaner and brighter images in low-light conditions.

Furthermore, the GP3 is built using a 5nm process, and a smaller process usually means lower power consumption, which theoretically can also improve heat dissipation.

According to the plan, the standard and PRO versions of MISSION 1 will be available for pre-order on May 21, while the main event, the ILS version and various creator sets, will not be available until the third quarter of 2026. The prices of the three versions have not yet been announced.

Overall, the MISSION 1 series boasts strong product capabilities, featuring a large enough sensor size, ample video specifications, and a cleverly positioned product strategy.

Traditional cinema cameras offer excellent image quality, but are bulky and fragile; action cameras are robust and portable, but their image quality is often subpar. The MISSION 1 series attempts to bridge these two gaps.

GoPro Senior Vice President Pablo Lema is clearly also very excited about it:

We expect the MISSION 1 series to expand the creative boundaries of creators worldwide, just as GoPro pioneered the portable camera category.

But will this work?

It's important to understand that GoPro's current situation is far from optimistic.

Those who ride the crest of a wave are destined to fall.

In 2002, Nick Woodman, an avid surfer, went to Australia in search of the perfect wave.

He longed to capture his heroic figure as he swam through the waves, but found that there was no waterproof, cheap, and convenient equipment available on the market. So, Nick Woodman sewed a wristband from shredded old wetsuits and tightly strapped a Kodak film camera to his wrist.

This move unexpectedly created a completely new category called action cameras.

A classic "garage geek" moment.

Over the next decade, GoPro experienced explosive growth, becoming synonymous with extreme sports, adventure, and a passionate lifestyle—skydiving, diving, skiing, surfing—wherever there was adrenaline, there was bound to be a small, square black box.

In 2014, GoPro went public on Nasdaq, and its stock price peaked in September of the same year, with its market capitalization exceeding $10 billion. To put this in perspective, Sina Weibo, which also went public in the US around the same time, had a market capitalization of only $3.4 billion. In the eyes of Wall Street, GoPro, which only sold a small square camera, was equivalent to three Weibo accounts.

The following year, GoPro's revenue reached $1.62 billion, which was its peak period.

However, the good times didn't last. According to Digital Camera World, Nick Woodman delivered a dismal performance in its latest earnings call: the company suffered a net loss of $9.1 million in the fourth quarter and a total loss of $93.5 million for the year.

As a pioneer in this field, how did GoPro end up in this predicament?

The first to be targeted was undoubtedly the smartphone.

With smartphones featuring image stabilization, low-light capabilities, and multiple cameras, the pain points for ordinary people recording their lives have been completely eliminated. While they may not be as hardcore as GoPro, for the average person, who would spend thousands of dollars to carry an action camera just to capture everyday moments?

Interestingly, the timing of the start of the mobile imaging competition in smartphones almost coincided with the collapse of GoPro's stock price; even more counterintuitively, after 2020, while the action camera market grew, GoPro's market share shrank further—DJI and Insta360 entered the market.

InStone's entry point is extremely clever. The biggest pain point of traditional action cameras is the high barrier to entry and the difficulty in composing shots. InStone's panoramic camera, with its "shoot first, then frame" solution and the invisible selfie stick, perfectly solves the pain point of ordinary users not knowing how to compose shots. The subsequent GO series further lowered the barrier to entry for first-person perspective, firmly capturing the attention of casual users.

Once they had established a foothold on the flank, Shadowstone revealed its true colors, pulling out the traditional Ace Pro series and stabbing GoPro in the heart of the market.

DJI's story is even more dramatic.

In its early days, DJI's Phantom 1 drone was equipped with a GoPro mounted underneath. The two companies even enjoyed a brief honeymoon period.

However, after years of experience in gimbals, image transmission, and image stabilization, DJI launched the first Pocket in 2018 using micro-gimbal technology, taking advantage of the rise of social media and vlogs to completely dominate the most lucrative market.

In 2019, DJI suddenly realized that it seemed to have all the necessary technologies for action cameras, and all that was left was assembly—so DJI pulled out the Osmo Action series and sharpened its knives to target GoPro.

Thus, DJI and Inspire, leveraging their respective core strengths, have effectively sandwiched GoPro between them, while also controlling the broader professional market through their diverse product lines, effectively blocking GoPro's escape route.

No matter how many thrilling and perfect waves a person rides on a surfboard, their ultimate fate is to fall into the sea.

But everyone who falls will desperately try to get back on.

GoPro is no exception.

DJI Pocket 4 drives GoPro to the brink of extinction.

The MISSION 1 series is GoPro's breakthrough weapon.

To be fair, from a product perspective, these three MISSION 1 cameras can compete with DJI and Inspire action cameras, and may even sell well—after all, the 1-inch sensor, rich video specifications, and compatibility with a vast array of Micro Four Thirds lenses are all genuine innovations.

Nick Woodman also expressed his confidence:

This is the most powerful product we've ever made; the final visual effects will absolutely be amazing.

So, can GoPro turn its fortunes around with the MISSION 1 series?

I think it's probably not possible.

The configuration of the MISSION 1 is undeniably top-notch, but it still retains the GoPro DNA at its core: you have to be professional enough to be worthy of me; while the Pocket 4's logic is completely the opposite, with its mechanical gimbal, AI tracking, and mindless output… you don't need to learn anything, just turn it on like opening a Zippo, and the Pocket will shoot great pictures.

GoPro's past success stemmed from capitalizing on the public's yearning for "extremes and thrills" during an economic upswing. However, in today's uncertain world, "effortless" experiences have become a basic necessity for the public.

Action cameras have tried to turn things around before – when the previous generation Pocket 3 was sold out everywhere, there was a trend on social media of “action cameras as a cheap alternative to Pocket.”

But this brief experiment ultimately benefited DJI, proving that Pocket was the correct version.

Products are born of the times and also fade with the times.

With the capabilities of the MISSION 1, GoPro may be able to once again take to the surfboard and continue conquering the ocean.

But when it looks back, it will be shocked to find that the vast beach, rich in cash flow and market potential, is already crowded with people holding Pockets, leaving no room for it.

Since there was no way to return to shore, GoPro could only paddle into even more turbulent waves, repeatedly falling and getting up, gradually moving further and further away.

#Welcome to follow iFanr's official WeChat account: iFanr (WeChat ID: ifanr), where more exciting content will be presented to you as soon as possible.

The first conversation between iFanr and John Ternus took place the day after the iPhone 16 launch event in 2024.

He was Apple's vice president of hardware engineering at the time. Today, he is about to take over as Apple's CEO.

That day, the Chief Content Officer of iFanr attended a "breakfast meeting" arranged by Apple, where he sat around with several executives, drinking coffee and chatting. Afterwards, each person had a short formal interview. The equipment was set up, shot with an iPhone 16 Pro, and the footage was handed over to you afterward.

The location is a newly completed building in Apple Park called the Observatory. It's built on a small hill in the park, with a domed entrance at ground level. Inside, there's a circular skylight facing the sky. Sunlight streams in directly from above, giving the entire space a temple-like, ceremonial feel.

Our question to Ternus was about the camera controls on the iPhone 16: why choose mechanical buttons instead of a pressure-sensitive solution?

Ternus believes that camera controls are very different from Mac trackpads, with different sizes, and that iPhones require IP68 dust and water resistance, thus necessitating a new concept.

The "camera control" actually has a haptic switch at the bottom, and then a very high-precision force sensor on top of it. The entire button is connected to the Taptic Engine and also uses a capacitive sensor, making it possible to press, half-press, and slide.

John Ternus's answer also reflects his qualities as a product manager. He joined Apple in 2001, starting as a product design engineer, and was promoted to VP of Hardware Engineering in 2013. In 2021, he succeeded Dan Riccio as SVP of Hardware Engineering.

Near the end, iFanr added a comment—rather than a question, it's more of a personal observation: Apple Intelligence would make even more sense if combined with Vision Pro or similar AR glasses.

He simply smiled slightly and said thank you. The interview ended.

We didn't include this part in the video—it's not informative. Apple executives share a common trait: they know everything, but they can't keep their mouths shut.

Ternus is now in the leadership position. Before handing over the reins, Jobs told Cook, "Never ask what Jobs would do, do the right thing." Cook has lived up to expectations. In his own way, he has made Apple the most efficient manufacturing company in the world, made services the second engine of hundreds of billions in annual revenue, and incorporated sustainability and inclusivity into the company's values. In fourteen years, he has proven that you don't have to be Jobs to lead Apple forward.

Now it's Ternus's turn. The question he smiled but didn't answer back then, now it's his turn to answer.

Thank you, Tim Cook! Welcome, John Ternus!

Tim Cook, who is about to step down as CEO, also released a thank-you letter to Apple users around the world today, thanking them for sharing their daily experiences of using Apple products during his 15-year tenure.

The full text of the thank-you letter can be found at: https://www.apple.com/community-letter-from-tim/

Over the years, iFanr has had the privilege of exchanging ideas with Mr. Cook on several occasions, experiencing together moments where technology and humanity converge.

#Welcome to follow iFanr's official WeChat account: iFanr (WeChat ID: ifanr), where more exciting content will be presented to you as soon as possible.

"Lobster" OpenClaw is probably the hottest AI keyword after ChatGPT.

But if you ask your friends, they'll most likely never have personally deployed or used a lobster.

There is a clear technological gap between ordinary users and lobsters.

Therefore, a more crucial question begins to emerge: when AI truly demonstrates its potential, what will ultimately determine its widespread adoption?

The answer is slowly shifting from the cloud to the terminal, with device manufacturers that are closer to users beginning to try to fill this "last mile".

Honor, as a terminal manufacturer, offered a straightforward solution: pre-installing the "lobster" feature in the product, allowing users to use it right out of the box without having to worry about anything else. This solves the three major pain points faced by the lobster-based intelligent agent in one go: complex deployment and learning process, high usage costs, and privacy and security risks.

The Honor YOYO Claw technology is easy to use and easy to use.

The newly launched MagicBook Pro 16 by Honor is a rather "new" laptop.

First of all, the configuration is up-to-date: the highest-end third-generation Intel Core Ultra X7 358H processor and 32GB of RAM are standard features for high-end thin and light laptops this year.

More importantly, its positioning is novel—Honor officially defines it as the "chosen shrimp-raising device," which comes with a "Honor YOYO Claw" lobster technology intelligent agent.

No need to deal with annoying command lines; Honor YOYO Claw technology has been packaged into an application, which provides an intuitive and simple graphical interface with a double-click.

This solves the first level of the problem, "how to use it"; the overlooked problem is "what to use it for".

Many people know that lobsters are very strong, but they can't explain why they are strong or how to create value for themselves, so they simply stop at "I've heard they are very strong".

Even if you manage to install OpenClaw on your computer, you still have to deal with the complexities of installing and debugging Skills, as well as embedding it into your workflow. These steps deter most people.

Honor has even eliminated this complexity for users – Honor YOYO Claw technology comes pre-installed with 5 "main shrimp" and 23 "sub shrimp", covering five major scenarios including education, office, academic, content creation, and intelligent assistance.

The hybrid architecture of Honor YOYO Claw's technology platform and cloud also gives it a stronger self-learning potential: every time a user interacts with YOYO Claw, every time a document is uploaded or generated, the platform summarizes the experience and templates for easy reuse in the future.

In other words, it's more than just delivering a single task; it's an assistant that continuously grows, becoming more convenient and understanding of you the more you use it.

Therefore, rather than saying that Honor YOYO Claw technology is "ready to use out of the box," I would prefer to call it "easy to use out of the box."

If OpenClaw is likened to a "racing car"—powerful but too hardcore, only suitable for a select few to show off their skills—then what Honor wants to create is a "family car," where everyone can enjoy the improvements brought by technology.

Just like when computers were first invented, people had heard about their computing power, but the general public didn't understand programming languages ​​or what to do with such high computing power.

It wasn't until the advent of a spreadsheet program called VisiCalc that this powerful computing capability truly became accessible to ordinary people. Without needing to learn technical skills, they could achieve efficiency improvements far exceeding those of manual work, and thus computers truly became part of the masses.

The transformation from capability to application, and the establishment of a foothold from technology to application scenarios, is the right way to "open up" the lobster industry.

Good lobsters should be edible by everyone.

Let's rewind to the end of 2022, back to the time when ChatGPT was just emerging.

Why was ChatGPT able to spread virally, sweeping the globe with lightning speed and quickly taking root in every industry? There are two very important core reasons: ease of use and usability.

With a suitable internet connection, simply enter ChatGPT.com into your browser. The interface is so simple it consists of only a single dialog box. You can type anything you want; even elementary school students can chat with the AI ​​for several rounds. This interactive method of directly obtaining output from natural language input is unprecedented in machines, and it has begun to transform the way humans work overnight.

Like real lobsters, OpenClaw is "delicious" but not easy to eat. Just wanting to taste it is an exponential challenge, comparable to ChatGPT back in the day. Deploying it on Windows PCs, which are used by more people, is even more difficult.

The gap between technology and practical application needs to be bridged by terminal manufacturers. Technologies like Honor YOYO Claw, which are pre-installed on products and can be launched by simply double-clicking the mouse, are likely to be the first "pre-made shrimp" that most users will experience.

The Honor YOYO Claw technology is extremely easy to learn. Users only need to scan a QR code to log in and connect to third-party platforms such as Lark and WeChat. They can then interact with the lobster through the chat box without any additional complex API key configuration.

Keeping the complexity at the technical level, handing the simplicity to the user, and translating the benefits into the user experience—this is more important than anything else.

Solving the "can it be used" problem is just the beginning. In fact, when "Lobster" is placed in front of ordinary users, people will still have concerns. They worry that it will consume too many tokens and that they will run out of money before they even earn any. They also worry that its permissions are too high and that they might accidentally leak their privacy or delete all their files.

Therefore, to truly commercialize lobster products and enable more people to use and dare to use lobster, solving these two issues is equally crucial.

As a terminal manufacturer, Honor has an advantage over AI manufacturers—the local computing power of its devices.

The edge-cloud collaborative architecture and edge-first approach allow Honor YOYO Claw technology to maximize the value of tokens: tasks that can be completed on the edge are completed locally; when cloud computing power is needed, the required skills are precisely matched. The former directly reduces the consumption scenarios, while the latter finely allocates resources.

According to official data from Honor, Honor YOYO Claw technology saves 50% on overall word consumption compared to OpenClaw, and even up to 90% in extreme scenarios.

A typical scenario is local search. Traditional OpenClaw relies heavily on cloud computing power. For tasks with a slightly larger workload, such as local file retrieval, scripts need to be called repeatedly, which is not only extremely inefficient, but also consumes tokens rapidly.

As a terminal manufacturer, Honor can utilize local computing power and models, allowing local search tasks to run entirely on the device side, which is not only fast but also cheaper.

As the generative AI wave sweeps across all industries, the "meta-economy" has also emerged. In March of this year, the daily meta-call volume in my country exceeded 140 trillion, a more than 1,000-fold increase in two years.

When tokens are directly linked to the economy, every token saved directly creates real economic value.

The high risks associated with lobsters' high privileges are addressed by Honor's YOYO Claw technology, which uses "shrimp-to-shrimp" tactics.

Honor YOYO Claw technology is equipped with an "independent security shrimp" that can monitor all operations of other shrimp throughout the process and promptly block high-risk operations such as formatting the hard drive. If it requests sensitive permissions such as payment or camera access, it will also require the user to confirm the request again.

As for the lobster's "memory" ability and the processing of core data, these will also be limited to the edge to avoid privacy issues such as cloud storage or leakage.

From not knowing how to use it to not needing to learn, from not being able to use it to being able to afford it, from not daring to use it to not being afraid to use it, Honor YOYO Claw technology has found a new route for lobsters to be popularized—beyond parameters and performance, what is a "good shrimp" is one that everyone can eat.

Shrimp farming, the new standard for AI PCs

2023 was the year generative AI exploded, and Intel and Microsoft, as rule-makers, were among the first to propose the concept of "AI PC".

Nearly three years have passed, but AI PCs remain just slogans. They are nothing more than having a Copilot key on the keyboard and a built-in AI chatbot—capabilities that can mostly be achieved in a browser.

The crux of the problem is that no one knows what form or path a so-called "AI PC" should take, or how to integrate AI with computers locally to bring value to users. These questions remain unresolved, making it impossible to deliver a satisfactory answer.

Three years have passed since ChatGPT made its name, and working people are no longer satisfied with just playing a question-and-answer game with AI in the chat box. Since AI clearly has great potential, it should not only "answer questions" but should act as an assistant to directly help users solve problems.

Given these expectations, the so-called "AI PC," touted as a productivity tool labeled "AI," has been quite disappointing.

Just then, OpenClaw appeared.

This AI solution, which is more suitable for deployment on computers and features a high degree of hardware and software integration, points to a completely new direction for "AI PCs".

OpenClaw is essentially an experimental project with high barriers to entry and low stability, and was originally a toy for a small group of geeks.

But like the fire of Prometheus, once it was brought to the human world, people could not ignore its value, making more and more people yearn for it.

Ultimately, what we truly yearn for is a system capable of partially or even completely taking over workflows, automatically understanding tasks, automatically breaking down processes, and automatically completing execution. The significance of OpenClaw lies in the fact that it has clearly demonstrated for the first time that this path is feasible in the present.

Even though it is still in its infancy and accompanied by various risks and uncertainties, the industry must take this path because the peak "is right there."

Along this path, the division of labor among different roles is gradually becoming clear: large model manufacturers focus on refining the technical foundation, which is the most "hard" task, but they are the furthest from users; internet companies promote the implementation of AI large model applications, which is flexible and fast, but it is difficult to go "deep".

Terminal manufacturers, who are closer to the user side, play the role of "deepening and solidifying" their expertise. They understand user habits and needs better and have a greater chance of turning AI into a truly usable tool.

Especially in the form of "lobster" which emphasizes the synergy between hardware and software, terminal manufacturers have both natural advantages and bear certain unavoidable responsibilities.

Therefore, pre-installing a packaged "YOYO Claw technology" is not only a natural move for Honor, but also a necessary step, and a sample answer for future AI PCs.

This is not the end for "lobster", but only the beginning. After all, we have only just overcome the hurdle of making lobster "accessible to everyone".

As a terminal manufacturer that understands users better and holds the glory of multiple device ecosystems, Lobster should never be confined to the small space of PC, but should achieve cross-device flow without configuration.

Honor YOYO Claw technology supports multiple family members sharing one PC, with each person having their own dedicated shrimp to meet different needs: children can use it to help with homework, and the elderly can use it to remind them to take their medicine.

iFanr has also learned that Honor is already developing cross-ecosystem capabilities for its YOYO Claw technology, which avoids over-reliance on devices and allows computing power and capabilities to flow freely according to the scenario.

Imagine this scenario: you give a command to YOYO Assistant on your Honor phone to retrieve a data table on your Honor laptop, and then YOYO Claw will refer to previous documents to generate a quarterly report, which can then be sent directly to other colleagues via Lark.

When AI transforms from a single connection and a single tool into a system-level computing power that spans multiple devices, it will truly enter workflows, lifestyles, and the daily lives of every "you and me".

#Welcome to follow iFanr's official WeChat account: iFanr (WeChat ID: ifanr), where more exciting content will be presented to you as soon as possible.

Before the lobsters were even installed, people on social media were already talking about "raising horses." After OpenClaw and Hermes became popular this year, it's safe to say that AI can actually do the work.

However, although friends in the AI ​​community have been working late into the night to "raise lobsters" and have already run through various automated processes, most people are discouraged at the configuration stage.

Even if it runs smoothly, you might only be doing basic tasks like researching information and writing reports. The token consumption is also quite high. If you do a slightly more complex task, the bill might shock you.

Currently, many "lobster-related" and cowork-related products are flooding the domestic market. As an AI media editor, after trying out a bunch of products, I discovered a rather interesting one: DuMate .

This is a desktop agent made by Baidu AI Cloud. The name seems ordinary and easy to overlook, but the Chinese name it gave itself , "搭子" (Dāzi), actually aroused my curiosity .

While large and small companies are all defining AI products in line with current trends, it doesn't call itself Agent, Claw, or Copilot. Instead, it uses the more down-to-earth name "Dazi" (搭子).

After using it for a while, I found that when AI became my work and life partner, it was closer to the ideal agent I envisioned.

It's worth mentioning that DuMate currently offers 1000 free points daily, which is generally enough for normal daily use. Entering the exclusive APPSO invitation code "appso12" in DuMate will also grant you a large amount of additional permanent points.

Let AI clean up my computer—it's easier than a lobster!

The installation process for DuMate is extremely simple: download the client, install it locally with one click, and no development environment configuration is required. After opening it, grant a working folder and tell it what you want to do using natural language.

Compared to the complicated process of "raising lobsters" with OpenClaw, the learning curve for DuMate is basically just that you need to know how to type.

Let's start with the most basic scenario and test our skills a bit.

I have a bad habit of never organizing or categorizing the various files on my computer. As a result, after a while, contract PDFs, project Excels, meeting recordings, papers, and screenshots from six months ago are all piled up together.

Every time I try to organize it, I feel like it's too big a project, so I keep putting it off, creating an endless cycle.

This time, I said to my partner DuMate, "Please categorize and archive the files in my download folder by date and file type."

After saying that, I went to get some water. When I came back, the download folder had been divided into more than a dozen subdirectories: documents were archived by year and month, images were in a separate folder, audio files were grouped into audio files, and I also had it clean up duplicate and invalid files.

Documents that have been gathering dust for years can be processed in minutes.

Here's a side note. If you let an AI access your real work files, your first thought is definitely: Is it safe? This is especially relevant given the numerous incidents OpenClaw has had with accidental file deletions.

DuMate works like this: It runs locally, processing files on your own computer without uploading to the cloud. It can only access folders you've authorized, and the first time you try to modify a new directory, it will pop up a window asking for permission.

Risky operations such as deletion and moving require secondary confirmation, and the entire execution process is displayed on the screen in real time, allowing you to stop it at any time.

The data stays on the device, you have complete control over permissions, and the entire process is transparent . This is the premise that allowed me to later entrust it with my actual work documents and materials.

After confirming that there were no security issues, I tried a task with real business content to increase its difficulty.

Anyone who works in operations knows how tedious it is to manage channel campaign data. Just considering the four platforms—Tmall, JD.com, Douyin, and Xiaohongshu—the backend data formats are all different. Every week, you have to manually download, clean up null and duplicate values, standardize the format, calculate ROI, create pivot tables, and finally compile it into a PPT for your boss.

If this process goes smoothly, it will take half a day; if it doesn't, you'll definitely have to work overtime.

I put the campaign data from the four platforms into a folder on my desktop, and then said to my partner DuMate, "Integrate the campaign data from these four platforms, clean up null values ​​and duplicates, calculate the ROI of each channel, create a pivot table, and finally generate a report PPT."

It began reading Excel files one by one, automatically identifying field differences across different platforms, standardizing column names, removing blank rows and duplicate records, calculating ROI by channel, and generating a summary Excel file with pivot tables. Then, based on this data, it created a PowerPoint presentation: a cover page, a bar chart comparing ROI across channels, a line chart showing campaign trends, highlighted problematic channels, and a summary and recommendations on the last page.

To be honest, the layout looks better than what I did myself. I checked the data accuracy item by item and found no errors. It even corrected some table errors itself, showing more attention to detail than I could.

It read your Excel file, analyzed and organized the data, generated the PPT and tables you needed, and saved them to your specified folder . After entering my requirements, I didn't open Excel once or move the mouse once from start to finish.

It executes and thinks simultaneously, delivering the finished product directly to you, like a smart and highly efficient colleague.

My partner and I have "built" an AI workflow.

The two scenarios above, to put it simply, are still about "you give it a task, and it does the work." What truly made me feel like a "partner" was when it found work to do on its own.

APPSO focuses on AI-related content, an industry characterized by explosive information updates and a daily need to monitor a vast amount of industry and product news. Previously, the approach was to spend one or two hours each morning sifting through various information sources and manually compiling a pool of topics for the day.

I've now set up a trending topic monitoring system on my partner DuMate and set up a scheduled task: every morning at 8 AM, it automatically captures AI-related trending topics from major tech media and social media platforms, compiles them into a structured list of trending topics, and sends it to me.

Every morning when I wake up, the day's top headlines are already neatly laid out there. It categorizes them for me, indicating which are more important and trending, which are breaking news, which are in-depth analyses, and which are suitable for follow-up topics.

The hot topic capture relies on the Baidu search skill built into DuMate, which has been downloaded over 80,000 times globally on ClawHUB , so it's definitely something special.

Speaking of skills, let me add a point. DuMate is compatible with the OpenClaw standard, and third-party skills from ClawHUB can be directly imported. The extended resources accumulated by the OpenClaw community are readily available. Its own skill library also has a bunch of ready-made skills to choose from, including those from Tencent Docs, Fliggy, and Didi.

For me, this means that if I want to add a certain ability, I can just install a skill without waiting for an official update.

My partner, DuMate, has integrated with mainstream IM software such as WeChat, Lark, Ruliu, and DingTalk, with basic one-click binding. Then, what happened next made me realize that AI was truly working seamlessly in my workflow.

Yesterday, I asked it to compile trending news for me, and its partner DuMate gave me the headline: Honor Robot "Lightning" sweeps the top three spots in the 2026 Beijing Yizhuang Humanoid Robot Half Marathon. This news is all over the internet, and I figured it was worth making a short video to capitalize on the trend.

The previous process was: find materials, write scripts, open software, export, and upload, which took two hours if it was fast, and half a day if it was slow.

This time, I didn't switch to any other apps. I directly replied to my friend on WeChat: "I'm an editor at AI media appSO. I'd like to follow up on that robot half-marathon trending topic and make a 10-second short video. I have some video editing solutions and reference images related to the robot half-marathon in a folder on my desktop. Could you help me generate a video?"

It read the editing scheme and reference images on my desktop and automatically generated a 10-second short video with narration and background music.

My partner shared a trending topic at 8 AM, I replied with a message on WeChat at 8:10 AM, and the video was generated by 8:30 AM . From discovering the trending topic to producing the content, the entire process was completed within WeChat. I can even do video production while I'm on the go.

The last scene was probably the most surprising for me.

Jensen Huang recently had a very popular video podcast, and I wanted to turn it into internal sharing material. I told my partner, "Organize this video podcast into a geeky Apple-style PowerPoint presentation, extracting Huang's core viewpoints and key quotes, making the content as detailed as possible. Also, test some suitable topic directions for tech media based on the video content, and give me a detailed report."

It first accessed the video link, extracted the complete audio content and transcribed it, then analyzed it segment by segment, organized Huang Renxun's core arguments into a structured PPT, divided the interview into chapters and compiled Huang Renxun's key quotes.

The style is indeed that minimalist Apple aesthetic. It also generates a separate topic report, extracting five or six angles suitable for tech media to follow up on from the video content, each with its own entry point and reference materials.

Two documents, a PowerPoint presentation and a report, took less than ten minutes in total.

I checked the content, and the key quotes were accurately extracted, the main points were basically summarized, and there were two directions I was already considering in terms of topic selection, as well as another one that I hadn't thought of but was definitely worth doing.

In addition, content layout, writing and publishing on Xiaohongshu and Weibo, and analyzing popular content on various platforms can all be done together. My "partner" and I can build a complete content production pipeline. The way I collaborate with him is really not much different from working with capable colleagues in the team.

Partners, a new kind of human-machine relationship

After using it for a while, my feelings towards my partner DuMate were no longer just "this tool is easy to use", but more like a new kind of "colleague relationship".

When it helped me organize my files, I found myself not nervous. I knew it only accessed folders I authorized, and it would ask me about any risky actions, making the entire process transparent. Trust wasn't about its promise of "I won't mess around," but rather its mechanisms preventing it from doing so. After a few days, a natural division of labor developed between us: it handled repetitive tasks, while I focused on decision-making.

Further on, there's a tacit understanding. If you correct it the first time, it will often serve up exactly the taste you want the second time. This has a lot to do with its memory ability; it learns your aesthetic preferences over time.

Think about it carefully: trust, division of labor, and tacit understanding. These three words are generally not used to describe the relationship between people and tools. You wouldn't say you have a tacit understanding with Excel. They refer to the relationship between people (partners).

Naming products in the AI ​​industry is actually quite similar to self-introductions on a blind date. How you introduce yourself reveals how you understand the relationship.

What's the logic behind a "dinner buddy" relationship? When you invite your buddy to dinner, who looks at the menu first and orders first doesn't matter. In a buddy relationship, there's no captain and co-pilot; it's just "you do your thing, I do mine, and we call it a day when we're done."

The term "partner" truly refers not just to a smarter AI, but to a new human-machine relationship . You are responsible for judgment and decision-making, while it is responsible for execution and delivery. You collaborate as equals, not as master and servant. Once the work is done, you go your separate ways, and come back next time you need something.

AI is no longer a passive, standby role in your life. A work partner is two people working together, only one of them doesn't need to drink coffee .

If you want to experience what a true AI partner is like, you can try contacting DuMate.

APPSO also learned that DuMate will release a series of updates at the Baidu Create conference on May 13-14.

In addition to the AI ​​foundation updates for enterprises, developers, and partners, the main focus of this Create conference was on the latest developments in Agent infrastructure. How to build and implement it was probably the most valuable information to be found at the event.

APPSO will also provide on-site coverage as soon as possible, so stay tuned.

#Welcome to follow iFanr's official WeChat account: iFanr (WeChat ID: ifanr), where more exciting content will be presented to you as soon as possible.