Global Soul Limited Company News

YINCAE Company (YINCAE), a leading innovator in advanced material solutions, has announced the launch of its ground-breaking underfill material UF 158UL. Designed to meet the growing demand for large chips, this cutting-edge product demonstrates unmatched performance in room-temperature fluidity, fast curing, and high reliability. The UF 158UL has excellent fluidity and can easily fill gaps as small as 10 microns, even in large chips up to 100×100 mm in size. Its unique formulation ensures rapid curing at room temperature, significantly reducing production times and energy costs. In addition, the reliability of the UF 158UL provides the chip with strong protection against thermal stress, moisture and mechanical shock, ensuring long-term stable operation of the equipment. YINCAE's Chief Technology Officer said: "We are very excited to launch the UF 158UL, a transformative product in the field of underfill technology. With its superior fluidity, fast curing and high reliability, the UF 158UL enables manufacturers to achieve unprecedented levels of efficiency and quality in the production of large chips." UF 158UL is suitable for a wide range of applications, including: · High performance Computing · Artificial Intelligence · Automotive Electronics · Aerospace systems Key features and benefits: · Room temperature fluidity for easy application · Fast curing for improved production efficiency · High reliability for long-term performance · Excellent filling of 10 micron gaps · Compatible with large chips up to 100 x 100 mm.

Inherits the high speed and high precision characteristics of YRP10 high-end models, while focusing on ease of use Yamaha Motors Europe's Robotics division Surface Mount Technology (SMT) announced that it will launch the new YRP10e solder paste press (1) on April 1 this year. The YRP10e is an entry-level model that inherits the basic performance of the high-end YRP10 press while emphasizing ease of use. The new generation YR series platform features excellent rigidity, Yamaha's proprietary 3S nozzle (2) and steel mesh adsorption for high-speed printing performance with a core beat time of 6 seconds and printing accuracy of up to ±6σ : ±30μm, Process Capability Index (Cpk) ≥2.0 (measured with CeTaQ measuring equipment under optimal Yamaha Engine conditions). Compared to the previous generation entry-level model (YCP10), the YRP10e is approximately 25% faster and has approximately 40% better printing accuracy (3). In addition, the YRP10e comes standard with features such as printed pressure feedback control for precise pressure regulation and universal steel mesh bracket for one-click replacement of steel mesh for more efficient setup. It also retains some of the popular features of high-end models, including comprehensive print inspection with a dedicated high-resolution, wide-field inspection camera, and residual solder volume inspection, which monitors solder volume in steel mesh to ensure consistent print quality. Yamaha Engine, with its strength as a manufacturer of a full range of mounting equipment including printing presses, surface mounters, dispensing machines and inspection systems, brings its ideal "one-stop smart solution" concept to life. The company is also promoting smart factory systems that enable more efficient mounting processes across the board through smooth and precise collaboration between devices without the use of black boxes. In the mounting process of electronic components, the use of ultra-small chip components and narrow-pitch electrode components is increasing, and the trend of products towards smaller size, higher density, higher functionality and diversification is accelerating. Accordingly, the process of applying solder paste to a printed circuit board (PCB) becomes quite challenging. The difficulty of producing basic circuit boards using entry-level machines is also increasing year by year. In response to this, Yamaha Engine has developed the YRP10e entry-level solder paste press. This model combines the high-speed, high-precision printing performance of high-end models with carefully selected, user-friendly basic features, while ensuring an extremely high price/performance ratio. Despite being an entry-level model, it comes standard with features that are prized in high-end machines, including print pressure feedback control for precise adjustment of print pressure, a universal steel mesh bracket for one-click switching between various steel mesh sizes, and a buffer conveyor that reduces transfer time and speeds up beat times. About Yamaha Robotics SMT Division Yamaha Surface Mount Technologies (SMT), a division of Yamaha Robotics, Yamaha Engine Corporation, produces a full line of equipment for high-speed online electronic assembly. This "one-stop smart solution" includes solder paste presses, component placement machines, 3D solder paste testers, 3D printed circuit board testers, dispensers and management software. These systems bring Yamaha's philosophy to electronics manufacturing with a focus on intuitive operator interaction, efficient coordination between all online processes, and modular design that enables users to meet the latest manufacturing needs. Yamaha Group's capabilities in servo motor control and image recognition of vision (camera) systems ensure extremely high accuracy at high speeds. The current product line includes the latest generation of YR devices with advanced automation capabilities for programming, setup, and conversion, as well as the new YSUP management software with a state-of-the-art graphical interface and built-in data analysis capabilities. Yamaha's SMT division combines design, engineering, manufacturing, sales and service capabilities to ensure operational efficiency and provide easy access to support for customers and partners. With regional offices in Japan, China, Southeast Asia, Europe and North America, the company has a truly global presence. www.yamaha-motor-robotics.eu (1) A machine that uses a shovel-like tool called a "scraper" to coat a paste solder made of fine solder particles, viscous fluid flux, and binder onto a printed circuit board (PCB). By heating in a reflow furnace, the solder melts and bonds the surface-mount electronics to the printed circuit board. (2) Swing single scraper nozzle. This proprietary printhead uses a spatula to control the Angle of attack, speed, and force applied to the print. (3) Under the best conditions of Yamaha Engine Company.

With the rapid iteration of technology, delivery service robots and stir-frying robots have begun to appear more and more in our lives, and these humanoid robots with appearance and action closer to humans have frequently refreshed people's cognition of science and technology. As a highly integrated industrial product involving multiple cutting-edge technologies, what can humanoid robots do now? How long does it take to get into ordinary people's lives? What bottlenecks need to be broken through in the future? With the help of a motion tracking system, a silver and black humanoid robot slowly raised its right hand as workers raised their arms, caught a bottle of mineral water with an open palm and clenched it, then handed it to a worker next to it before letting go. The whole process is fluid and natural, like a human interacting with a partner. This is the test scene seen by the reporter in Leju (Shenzhen) Robot Technology Co., LTD. The "Kua Fu" high dynamic humanoid robot developed by the company, weighing about 45 kilograms, height of about 1.6 meters, walking speed of up to 5 kilometers per hour, rapid continuous jump height of more than 20 centimeters, can be realized through the open source Hongmeng operating system intelligent link application, has been gradually applied in scientific research, education, commercial services and other fields. Humanoid robots, also known as humanoid robots and bionic robots, usually have features similar to human appearance such as head, torso, arms and legs, and have certain movement and perception abilities. The rapid development of China's humanoid robot field has attracted the attention of the global scientific and technological community. Brett Adcock, founder and CEO of Figure AI, an American humanoid robot star company, has praised Leju Robot on social media. In recent years, technology giants around the world have continued to increase artificial intelligence and intelligent interconnection, and breakthroughs and innovations have been made in the field of humanoid robots. At the International Consumer Electronics Show 2025, technology giants such as Nvidia demonstrated the latest progress of their humanoid robots, which became one of the core highlights of the exhibition. The humanoid robot "Optimus Prime" developed by Tesla is already able to walk like a human, can achieve natural turns and other abilities, and is able to accurately pass eggs between left and right hands. Tesla CEO Elon Musk has said that Tesla will have thousands of humanoid robots working by 2025, and formal commercial mass production will start in 2026. According to incomplete statistics from the Research Institute of Shenzhen New Strategic Media Co., LTD., as of June 2024, there are more than 160 humanoid robot manufacturing enterprises in the world, of which more than 60 Chinese enterprises, accounting for 37%, are the largest number of humanoid robot manufacturing enterprises in the world. The US and Japan came in second and third, with 19% and 11% respectively. At present, the mainstream humanoid robots in the market are expensive, and the high price even exceeds 500,000 yuan. In order to let more customers can feel the charm of humanoid robots, Yao Qiyuan, founder and marketing director of Shenzhen Zhongqing Robot Technology Co., LTD., said that the company adheres to the cost-effective strategy, launched the SE01 humanoid robot price of about $20,000 to $30,000, to universities and scientific research institutes as the main target customers, has sold hundreds of units. In January 2024, the "Implementation Opinions on Promoting Future Industrial Innovation and Development" issued by the Ministry of Industry and Information Technology and other seven departments proposed that it is necessary to strengthen future high-end equipment, break through high-end equipment products such as humanoid robots, and drive the industrialization of new technologies by the whole machine to create the world's leading high-end equipment system. Leng Xiaokun, the chairman of Leju Robotics, said that when humanoid robot technology is stable, "it can do everything people do, and there is a lot of room for imagination." According to the International Robotics Association, from 2021 to 2030, the compound annual growth rate of the global humanoid robot market will be as high as 71%. Data released by the Chinese Society of Electronics also shows that by 2030, China's humanoid robot market is expected to reach about 870 billion yuan. Tan Min, chief brand officer of Youbi, said that industrial manufacturing, commercial services and family companionship are the three major application scenarios for humanoid robots at present. Humanoid robots have been initially applied in commercial services and industrial manufacturing scenarios, "We are very optimistic that humanoid robots will continue to be applied in intelligent manufacturing, commercial services and home fields." It is expected that humanoid robots will enter thousands of homes in the future and become a necessity in every family." A special report on the robot industry by Zhongtai Securities pointed out that humanoid robots have been experimentally applied in many fields such as home services, shopping mall reception, and flexible manufacturing. In the future, the replacement of artificial robots will start from the flexible production link of intelligent manufacturing. The greatest significance of the application of humanoid robots in the industrial field is that enterprises do not need to transform production lines for mechanization. "We found that there are still a large number of links in the factory scene that rely on manual labor and cannot be solved with industrial robots. Humanoid robots can be 1:1 adapted to existing production lines and can be deployed without modification." Leng Xiaokun said. Similar to the human body and kinetic energy design, meaning that humanoid robots are more likely to become participants in human social activities. Under the ability of artificial intelligence technology, humanoid robots will have the ability to learn autonomously, be able to think like people, and to achieve communication and emotional companionship to a certain extent. Humanoid robots have significant potential to alleviate the aging population crisis, especially in providing daily assistance, health monitoring and emotional companionship, Tan said. "Humanoid robots can not only help the elderly with some simple daily tasks, but also monitor their physical condition in real time." At present, humanoid robots are still far from large-scale mass production and application. "Hiring" humanoid robots to work in factories and supermarkets is still a "fashionable" rather than an absolutely "affordable" choice. For example, in the factory sector, humanoid robots are still in the small trial stage, and only a small number of advanced factories have begun to explore the use of certain workflows. Industry insiders predict that it may take three to five years for humanoid robots to enter the production line more, and it will take longer for them to truly enter people's lives, requiring higher accuracy and safety. In addition, how to overcome the technical bottleneck and reasonably control the cost has become the key to the mass production of humanoid robots. Set machinery, electronics, materials, computers, sensors, control technology and other disciplines in one of the humanoid robots, the industrial chain is highly complex, and there are many core technologies to be improved. At present, each company's humanoid robot products show a high degree of customization, lack of truly universal parts, resulting in high manufacturing costs. When it comes to the future development of the humanoid robot industry, Tan Min believes that "patient capital" is crucial. He said that as a cutting-edge technology, the research and development and commercialization of humanoid robots require long-term investment and continuous innovation. The industry generally expects that 2025 to 2030 will be a critical period for humanoid robots to shift from "proof of concept" to "scale commercial". Yao Qiyuan suggested accelerating the landing of humanoid robot related standards, constantly opening up application scenarios, and letting humanoid robots "apply first." Leng Xiaokun believes: "Compared with foreign countries, China has a richer application scenario, which helps humanoid robots to be fully trained. It is expected that policy, technology and demand will jointly promote the development of the humanoid robot industry.

One code is difficult to find, "the world's first universal Agent", built by the Chinese team, 12 application scenarios. March 6, Manus screen circle of friends, some of the media used "GPT moment", "burst", "Silicon Valley sleepless night" and other titles, has become another stimulant after DeepSeek AI circle. In the mode of "seeking a Manus invitation code", on a second-hand platform, the price of its invitation code was once stir-fried to a sky-high 100,000 yuan. Shells have their uses, and Manus crosses the first Agent gap On X, MIT PhD @ZengyiQin shares his thoughts. "Manus is a good product, but it's not a technological breakthrough," he says. In some domestic AI communities, there are also mixed reviews. Some technicians believe that Manus may be a good Agent, but it is not enough to pursue so much, let alone the so-called "XX moment"; Some investors also resent the satellite-based reporting; Others admire the Manus team for its successful marketing disruption. After Manus triggered the attention of the technology industry, Zhang Tao, a partner of Manus AI, responded in a post on the social platform. He said that "I totally underestimated everyone's enthusiasm," and pointed out that Manus is still a baby in a baby, and implored everyone to be more tolerant and understanding of this company of dozens of people, "the team is working hard to let everyone experience a better product as soon as possible." Shells have their uses, and Manus may be a good Agent, but it's not up to the hype Is Manus a revolutionary innovation? Maybe it's too early to tell. Let the bullets fly for a while. 01 What is the technical principle of Manus? Where is Qiang? What is Manus? Manus officials only gave a four-minute demo video, and the rest of it was "communicating with everyone with the product." According to reports, Manus is the world's first universal Agent product, with independent thinking and system planning capabilities. Most importantly, unlike traditional AI, which only provides suggestions or answers, Manus can autonomously invoke tools in a virtual environment, such as writing code, intelligently browsing the web, manipulating various applications, and directly delivering complete task results. Perform complex tasks ranging from travel planning, stock analysis, educational content creation, to insurance policy comparison, supplier procurement, financial report analysis, and more. Shells have their uses, and Manus crosses the first Agent gap In the GAIA benchmark, a benchmark that evaluates how general-purpose AI assistants solve real-world problems, Manus achieved new state-of-the-art (SOTA) performance across all three difficulty levels. Shells have their uses, and Manus crosses the first Agent gap Official information shows that the core of Manus uses Multiple Agent architecture (multi-agent collaboration architecture) to build closed-loop task flows in independent virtual machines. Specifically, it deals with complex tasks through the division of labor and cooperation mechanism of planning agent, executing agent and verifying agent. In this process, Manus can create a new task environment in the virtual machine, and call various tools, including writing and executing code, browsing the web, operating applications, and so on, and ultimately generate high-quality replies or solutions for users. Take "Customize a trip to Japan in April" as an example, according to the official demo, after entering the prompt word, Manus will take apart the task and determine the idea according to your instructions. Shells have their uses, and Manus crosses the first Agent gap Then, it will follow its plan, step by step, retrieve the data, analyze the content and so on. You can see the corresponding generated code in real time through the right side of the interface. Manus then consolidates the code and feeds it back to the LLM, updating memory and doing reasoning and content generation. Finally, a complete and personalized travel plan is created. Shells have their uses, and Manus crosses the first Agent gap On X, MIT PhD @ZengyiQin shares his thoughts. "Manus is a good product, but it's not a technological breakthrough," he says. Manus, he says, mainly follows predefined flows to perform tasks. However, even if it can optimize system functionality in the short term, the predefined processes will greatly limit the maximum performance of the system; At the same time, Manus can only work in a relatively limited environment, that is, in some way between the browser level and the operating system level, and it is not effective outside of predefined software tools. For example, it can't control the slides to make slides for you. And much of that software is actually outside of Manus' toolbox. Shells have their uses, and Manus crosses the first Agent gap After the exchange of NetEase technology and a number of technical personnel, the relatively consistent view is that Manus may be a good Agent, but it is not enough to pursue so many, let alone the so-called "XX moment". 02 What is the background of the Manus team? Xiao Hong, the founder of Manus products, graduated from Huazhong University of Science and Technology in 2015, majoring in software engineering, and is a serial entrepreneur. He founded Nightingale Technology and launched B-end products such as SaaS tools, One Partner Assistant and Micro Partner Assistant, which received hundreds of millions of investment from Tencent, Zhenfund and other companies, and sold the products to a unicorn company in 2020. In the four-minute video, the presenter is Monica's co-founder and chief scientist Ji Yichao, not Xiao Hong. Shells have their uses, and Manus crosses the first Agent gap Peak Ji, co-founder and Chief Scientist of Monica In 2022, Xiao Hong set up "Butterfly Effect" company to develop AI assistant Monica, focusing on overseas markets. Monica integrates various large-scale language models to achieve All-in-One, providing users with diversified functional services such as chat, translation, copy processing, and drawing. But because it did not develop large models, Monica was criticized as a "shell" product from the moment it entered the market. The general view of the market for "shell" products is that it is not long-term, most of its market is phased products, and the end of the product will be eaten by large model companies. In 2024, Monica did not die, and its product form has gradually expanded from browser plug-ins to apps and web pages. The number of users has reached 10 million, and the profits are considerable, belonging to the head of AI Chrome plug-in products. The Manus that brushes the screen this time also continues Monica's product idea. While the team has not announced the Manus dock model, it is clearly an integration of the capabilities of the larger models. 03 Shell has the use of shell, and the interaction between people and AI will be more humane AI Agent is not a new concept, but past products are often difficult to really "use". Many plugins that need to be debugged, as well as complex toolchains, have been the main barriers preventing ordinary users from really using them. Careful analysis of Manus's broken circle may also be because the threshold for ordinary people to understand and really use it has been greatly reduced. According to user feedback, Manus is far more fluent in executing tasks than expected, even a vague task instruction can be disassembled and executed according to the context, rather than requiring the user to give precise input. However, from the perspective of the implementation effect, many people also say that there is still some distance from the real "liberation of hands" symbolized by Manus. Shells have their uses, and Manus crosses the first Agent gap At Superlinear Academy, members of the Manus team say they believe in and practice the philosophy of less structure and more intelligence: When your data is good enough, the model is strong enough, the architecture is flexible enough, and the engineering is solid enough, then concepts such as computer use, deep research, coding agent, etc. go from product features to natural capabilities. Previously, Xiao Hong also shared his views on Agent in a public interview. He believes that looking back at the trend of AI development, from the earliest form input, to natural dialogue, and now you can understand user needs with context, the fundamental logic is that products are becoming more and more in line with the use habits of ordinary users, and the communication between people and AI is becoming more and more smooth and humanized. Two days before the product launch, Xiao Hong posted on social media: In 2025, Chatbot is crossing the second gap, Agent is crossing the first gap, in the next six months, there will be many good fighting opportunities. Finally, you must believe that the shell has its uses! Shells have their uses, and Manus may be a good Agent, but it's not up to the hype Whether Manus is a good product remains to be seen in the market. However, Xiao Hong said that the historic moment of Agent crossing the "first gap" is not far away.   Li Shu Source: Yu Jiaqi_NBJS28262 One code is difficult to find, "the world's first universal Agent", built by the Chinese team, 12 application scenarios. March 6, Manus screen circle of friends, some of the media used "GPT moment", "burst", "Silicon Valley sleepless night" and other titles, has become another stimulant after DeepSeek AI circle. In the mode of "seeking a Manus invitation code", on a second-hand platform, the price of its invitation code was once stir-fried to a sky-high 100,000 yuan. Shells have their uses, and Manus crosses the first Agent gap On X, MIT PhD @ZengyiQin shares his thoughts. "Manus is a good product, but it's not a technological breakthrough," he says. In some domestic AI communities, there are also mixed reviews. Some technicians believe that Manus may be a good Agent, but it is not enough to pursue so much, let alone the so-called "XX moment"; Some investors also resent the satellite-based reporting; Others admire the Manus team for its successful marketing disruption. After Manus triggered the attention of the technology industry, Zhang Tao, a partner of Manus AI, responded in a post on the social platform. He said that "I totally underestimated everyone's enthusiasm," and pointed out that Manus is still a baby in a baby, and implored everyone to be more tolerant and understanding of this company of dozens of people, "the team is working hard to let everyone experience a better product as soon as possible." Shells have their uses, and Manus may be a good Agent, but it's not up to the hype Is Manus a revolutionary innovation? Maybe it's too early to tell. Let the bullets fly for a while. 01 What is the technical principle of Manus? Where is Qiang? What is Manus? Manus officials only gave a four-minute demo video, and the rest of it was "communicating with everyone with the product." According to reports, Manus is the world's first universal Agent product, with independent thinking and system planning capabilities. Most importantly, unlike traditional AI, which only provides suggestions or answers, Manus can autonomously invoke tools in a virtual environment, such as writing code, intelligently browsing the web, manipulating various applications, and directly delivering complete task results. Perform complex tasks ranging from travel planning, stock analysis, educational content creation, to insurance policy comparison, supplier procurement, financial report analysis, and more. Shells have their uses, and Manus crosses the first Agent gap In the GAIA benchmark, a benchmark that evaluates how general-purpose AI assistants solve real-world problems, Manus achieved new state-of-the-art (SOTA) performance across all three difficulty levels. Shells have their uses, and Manus crosses the first Agent gap Official information shows that the core of Manus uses Multiple Agent architecture (multi-agent collaboration architecture) to build closed-loop task flows in independent virtual machines. Specifically, it deals with complex tasks through the division of labor and cooperation mechanism of planning agent, executing agent and verifying agent. In this process, Manus can create a new task environment in the virtual machine, and call various tools, including writing and executing code, browsing the web, operating applications, and so on, and ultimately generate high-quality replies or solutions for users. Take "Customize a trip to Japan in April" as an example, according to the official demo, after entering the prompt word, Manus will take apart the task and determine the idea according to your instructions. Shells have their uses, and Manus crosses the first Agent gap Then, it will follow its plan, step by step, retrieve the data, analyze the content and so on. You can see the corresponding generated code in real time through the right side of the interface. Manus then consolidates the code and feeds it back to the LLM, updating memory and doing reasoning and content generation. Finally, a complete and personalized travel plan is created. Shells have their uses, and Manus crosses the first Agent gap On X, MIT PhD @ZengyiQin shares his thoughts. "Manus is a good product, but it's not a technological breakthrough," he says. Manus, he says, mainly follows predefined flows to perform tasks. However, even if it can optimize system functionality in the short term, the predefined processes will greatly limit the maximum performance of the system; At the same time, Manus can only work in a relatively limited environment, that is, in some way between the browser level and the operating system level, and it is not effective outside of predefined software tools. For example, it can't control the slides to make slides for you. And much of that software is actually outside of Manus' toolbox. Shells have their uses, and Manus crosses the first Agent gap After the exchange of NetEase technology and a number of technical personnel, the relatively consistent view is that Manus may be a good Agent, but it is not enough to pursue so many, let alone the so-called "XX moment". 02 What is the background of the Manus team? Xiao Hong, the founder of Manus products, graduated from Huazhong University of Science and Technology in 2015, majoring in software engineering, and is a serial entrepreneur. He founded Nightingale Technology and launched B-end products such as SaaS tools, One Partner Assistant and Micro Partner Assistant, which received hundreds of millions of investment from Tencent, Zhenfund and other companies, and sold the products to a unicorn company in 2020. In the four-minute video, the presenter is Monica's co-founder and chief scientist Ji Yichao, not Xiao Hong. Shells have their uses, and Manus crosses the first Agent gap Peak Ji, co-founder and Chief Scientist of Monica In 2022, Xiao Hong set up "Butterfly Effect" company to develop AI assistant Monica, focusing on overseas markets. By integrating various large-scale language models, Monica

    On Thursday evening, March 7, the United States time, SpaceX conducted the eighth test flight of the Starship comprehensive test, designed to test the spacecraft to the limit and complete the goal that was not achieved in the seventh test in January. During this test flight, the superheavy booster used by Starship successfully returned to the ground and was perfectly pinched by the chopstick-like robotic arm on the Mechazilla launch pad, marking the third time SpaceX has successfully accomplished this feat. Starship's eighth test flight booster recovery success (Source: NetEase Technology report) However, the engine on the top of the starship failed. At that point, the spacecraft fired its own engines in preparation for a suborbital flight of about an hour. However, SpaceX lost contact with the spacecraft as it was about to complete its ascent burn phase. Live footage showed several engines apparently stalling. The eighth test flight of the starship was detonated (Source: NetEase Technology Report) Later, it was witnessed exploding over the Bahamas. "Once enough of the center engine was lost, the spacecraft lost attitude control," SpaceX spokesman Dan Huot said. "We saw the spacecraft begin to spin, at which point we lost contact with the spacecraft. We can no longer get data from the spacecraft, and it is not clear whether its auto-abort system caused it to self-destruct for safety reasons." "We have implemented a number of measures, such as a debris response area, and are working closely with air traffic control," Hult said. "We have taken a number of steps to ensure public safety prior to the launch. These measures worked the last time they were launched and are currently in continuous operation." In January, a Starship on its seventh test flight exploded over the Caribbean, causing a series of flight delays and debris hitting a car on South Caicos Island. It is unclear exactly where the spacecraft exploded during today's mission. But the Federal Aviation Administration temporarily suspended inbound flights to airports in Miami, Fort Lauderdale, Palm Beach and Orlando because of "falling space debris." The FAA notified that the ground grounding order will remain in effect until 20:00 Eastern Time. Following the explosion, SpaceX posted an update on the X platform saying: "During the Starship ascent burn phase, the spacecraft experienced an unplanned and rapid breakup and lost contact. Our team immediately coordinated with security authorities to activate emergency plans. We will review today's flight test data to determine the root cause. As always, success comes from the lessons we learn, and this flight will provide more lessons for improving starship reliability." Breakthrough Max Q The launch of Starship broke through the Max Q, a critical moment in any rocket launch. Max Q is the moment when the aerodynamic Pressure on a rocket reaches its maximum during flight. At this point, the rocket was already moving quite fast, while the atmosphere was still relatively dense, resulting in extreme pressure on the surface of the rocket. In simple terms, this is the moment when the atmosphere "tries its best" to crush the rocket. If the rocket structure cannot withstand this pressure, it can lead to accidents such as rupture or loss of control. A series of upgrades During liftoff, the super-heavy booster successfully separated from the spacecraft. SpaceX refers to this process as "thermal separation" because the spacecraft uses its own power to push the super-heavy booster away by firing its engines, rather than using pneumatic thrusters like SpaceX's Falcon rockets, which are separated by violent force shocks. During Starship's eighth test, SpaceX removed a large number of thermal tiles from the ship, the black hexagonal tiles that cover the belly of the ship. The tiles are designed to protect the spacecraft from the extreme heat of re-entry into Earth's atmosphere, which can exceed 2,600 degrees Fahrenheit (about 1,427 degrees Celsius). The aim is to stress test vulnerable areas on the vehicle. In other words, push the vehicle to its limits to determine its point of failure. SpaceX is also testing some new materials for insulating the shingles, saying in a statement: "A variety of metal tile options, including a metal tile with active cooling, will be used to test alternative materials to protect the spacecraft from thermal protection during re-entry." The failed test in January prompted SpaceX to make other upgrades and improvements, including new vents and a "purification system" designed to prevent fires. SpaceX said the mission team's experience led them to adjust the lines that feed fuel to some Starship engines and modify the propellant temperature. Another change adjusts the spacecraft's "operational thrust target," the amount of thrust SpaceX wants the engine to generate during flight. The starship is now 6.5 feet (about 2 meters) taller than before and has about 25 percent more propellant, which could allow the craft to carry out longer missions in the future. The starship's front flaps are now smaller and move up, away from the heat tiles. This design upgrade is designed to reduce the amount of time the flaps are exposed to high temperatures during reentry. At the same time, the super-heavy booster was also upgraded to include a more powerful flight computer. Simulation satellites are to be deployed It is worth noting that in this test flight, SpaceX originally planned to attempt to deploy simulated Starlink satellites using starships. SpaceX had hoped to test how the Starship would accomplish that mission during its seventh flight in January, but the spacecraft suddenly exploded about 10 minutes into the flight. SpaceX said the simulated satellite used in the eighth test flight will be "similar in size" to the company's next-generation Starlink Internet satellite. Since starships are not equipped with payload fairing like most other rockets, these satellites may need to be ejected through a special hatch. However, the simulator deployed this time won't actually stay in space. Instead, they will travel along a suborbital path and, like the spacecraft itself, are expected to crash in the Indian Ocean about an hour after takeoff. The attempt failed due to an explosion in the spacecraft.

A few days ago, according to the American Chinese newspaper "World News Network" reported that TSMC chairman and president Wei Zhe Jia delivered a speech at the Executive MBA of Taiwan University, entitled "If there is no TSMC", in the speech, he responded to the outside world for whether TSMC may become a "US semiconductor" in the future.In his speech, Wei talked about the difficulties he encountered in setting up a factory in Arizona, saying, We were in Arizona, and I was trained with a lot of snot and tears. Mr Wei points to three difficulties in setting up a factory in Arizona. The first is infrastructure. The first is America's inadequate infrastructure capacity. He said it was remarkable to build a house in the United States. Locals don't build factories, and even if there is an Intel plant next door to TSMC's Arizona plant, it was built ten years ago and now the people are gone. "We ended up getting half the construction workers from Texas." The U.S. can supply TSMC with the grade of chemicals it needs, but at five times the price of those imported directly from Taiwan. TSMC can only ship sulfuric acid from the island to the port of Los Angeles, from which it can be trucked to Arizona, and even then it is cheaper than producing the chemical directly in the United States. The TSMC Arizona plant is the rest of the world The second problem is local government, which is reflected in local laws and administrative speed. Wei Zhe Jia said that he has never thought of a local local government various laws and regulations are not the same, and the local government does not know how to amend the regulations, but also requires TSMC to use TSMC money to hire experts to make rules, every meeting has municipal officials, TSMC invited experts, TSMC. It is up to TSMC to speak and jointly formulate the rules before proceeding to the next step. "We ended up with more than 18,000 regulations that cost $35 million." In terms of administrative speed, Wei Zhe jia pointed out that the reason for TSMC's rapid progress is that it can build a production line in two and a half years. Each technological advance changes the design of the entire production line, and TSMC works with manufacturers to immediately modify each change. But in the United States, any change has to be stopped, applied to the local government with the blueprints, and waited for their approval. To that. Wei Zhejia bluntly said: "The latest technology runway to the United States, do you think it is possible?" It takes at least twice as long to apply for everything and then come in." Mr Wei said it would take two and a half years to build a production line again, with land, water, electricity and talent. Each time a new technology is rolled out, TSMC needs to build at least three plants. Taking the 3-nanometer factory as an example, Wei Zhe jia said that if the above conditions are met, it will take about five years to move the entire TSMC to any country and build a factory that meets the needs of all customers. The last problem is the trade union problem, Wei Zhe Jia said: "At the beginning of the establishment of a trade union is a very noble ideal, is to help workers fight for better welfare, but after that, the trade union has become very strange." "World News Network" believes that TSMC has encountered labor and salary problems in the local area. Mr Wei said TSMC would continue to expand its investment in Taiwan. "World News Network" believes that TSMC, the latest technology will only stay in Taiwan, China, it is difficult to move to other countries. Recently, U.S. Commerce Secretary Raymond Mundo said that TSMC has begun production of 4-nanometer chips in Arizona for American customers. "For the first time in U.S. history, cutting-edge 4-nanometer chips are being produced in the United States by American workers, with production and quality comparable to Taiwanese products." She said the production, which had not been previously disclosed, had begun in recent weeks. The New York Times said that about 50% of the employees of TSMC's Arizona wafer fab are still from Taiwan, but with the continued expansion of the fab, this situation may change over time. TSMC had previously expected the first fab in Arizona to be mass-produced in the first half of 2025 and the second fab to produce state-of-the-art 2-nanometer chips in 2028. In addition, TSMC agreed to adopt the most advanced "A16 chip" manufacturing technology at its wafer fab in the state. In December last year, Zhu Fenglian, spokeswoman for the Taiwan Affairs Office of the State Council, said in response to inquiries that TSMC had become a "front name" for the DPP authorities to curry favor with the United States. As public opinion on the island has said, the DPP authorities are determined to "rely on the United States to seek independence", and TSMC will become "the United States Semiconductor" sooner or later. If the DPP authorities are allowed to "sell Taiwan" without a bottom line, the advantages of relevant industries in Taiwan are bound to be weakened, and the interests of enterprises and people on the island are bound to be damaged. When Taiwan's use value is squeezed dry, the "chess pieces" will become "abandoned children", and the island's enterprises and people see this more and more clearly. Zhu Fenglian said that it should be pointed out that many Taiwan enterprises have been able to develop rapidly through participation in Cross-Strait cooperation, which fully proves that the two sides of the Taiwan Strait work together to strengthen the Chinese national economy is the right way.

Anthropic, OpenAI's closest rival, is finalizing a new round of funding, The Wall Street Journal reported on Monday. Anthropic is negotiating a $3.5 billion funding round, much larger than previously expected. Anthropic initially planned to raise $2 billion, but managed to raise the amount during negotiations with investors, highlighting the continued high level of investor interest. Anthropic was previously valued at $18 billion, and the latest funding will roughly triple its valuation to $61.5 billion. The $61.5 billion valuation includes the cash Anthropic raised in this round. The round was led by Lightspeed Ventures, with participation from General Catalyst and Bessemer Venture Partners, according to people familiar with the matter. MGX, an ABU Dhabi-based investment company, is also in talks to participate. Anthropic's annualized revenue (projected for the next 12 months based on recent sales) has reached about $1.2 billion, according to people familiar with the company. Still, the company is losing money. By comparison, OpenAI told investors during a funding round last October that its 2024 revenue is expected to reach $3.7 billion. Details of Anthropic's latest funding round have not been made public. Founded in 2021 by former OpenAI employees, Anthropic is considered one of the few AI startups with enough talent and funding to compete with OpenAI. While Anthropic still lags behind OpenAI in the consumer market, its Claude chatbot is becoming increasingly popular among programmers and enterprise customers. Anthropic plans to use the funding to support the training of more powerful AI models. On the same day, Anthropic released its most intelligent AI model ever, which combines "inference mode" and "traditional mode" : The company claims that Claude 3.7 Sonnet's "hybrid" model is capable of both "inference mode" (stopping to think about complex answers) and traditional mode (generating answers in real time), and says it is the only such "hybrid" model on the market.Analysts believe that this move will help Anthropic gain a much-needed edge in the fierce competition among technology companies.Since the launch of DeepSeek's new model, some Silicon Valley investors have worried about the prospects of AI companies such as Anthropic, because DeepSeek's model can perform as well as the best US AI, but at a fraction of the cost and free to use. Anthropic CEO Dario Amodei has previously blogged that DeepSeek's breakthrough will not change the economics of AI research and development. He believes that DeepSeek's training cost reduction is in line with industry trends and does not represent a breakthrough technical achievement: If the downward trend in AI training costs is four times per year, and if the training cost of DeepSeek-V3 is about eight times lower than the current US model developed a year ago, that is actually completely in line with the normal trend. Even with DeepSeek's training cost data, they are only on the trend line, and may not even be quite there yet.Anthropic's larger-than-expected funding round shows that investors are still willing to bet on companies developing proprietary AI models. As previously reported, OpenAI is also in talks for a new round of funding of up to $40 billion, which would value it at $300 billion. Musk's xAI has also held informal financing talks with potential investors.

On February 24, local time, Marc Raibert, founder of Boston Dynamics, said in an interview with the First Financial reporter during the Dassault Systemes 3DEXPERIENCE World Summit in Houston, USA, His intelligent robotics Institute, the AI Institute, has bought a robot from Yushu Technology, a Chinese robotics startup, to test its capabilities. The founder of Boston Dynamics gives a speech "Our lab bought Uki's humanoid robot, and we also bought several smaller robots, which are relatively cheap." "Even if we are competitors, we can buy it and see what it can do," Raibert told CBN. Raibert said he was impressed by Uki's robot's capabilities. Last summer, Raibert attended the World Conference on Artificial Intelligence (WAIC) in Shanghai. "I learned that they had released 27 human-shaped robots, which was amazing!" "It was my first time in China," he told Yicai. "I talked to the CEO of Yushu and I thought they did a great job." Boston Dynamics developed Spot, a four-legged robot, and Atlas, a humanoid robot. Raibert did not elaborate on the differences between Boston Dynamics' robots and Uki's, but said, "There are a lot of differences between them, and I think competition is inevitable." It will be very interesting to see how they develop." Raibert specifically mentioned the wave of AI innovation in China triggered by DeepSeek. "I'm glad to see innovations like DeepSeek coming from China, but this is just the beginning, the whole field of artificial intelligence is still emerging, we are not at the end of the road, there will be more new ideas that change the development of technology in the future," he told the first financial reporter. Humanoid robots are seen as an important step towards achieving Artificial General Intelligence (AGI). Raibert has been involved in robotics research for over 50 years and has worked at Carnegie Mellon University and the Massachusetts Institute of Technology (MIT). He founded Boston Dynamics in 1992. Raibert believes that to achieve AGI, building hardware is as important as building software. "When AGI will be achieved is a complex issue, and there is some hype involved." "There are many dimensions to intelligence," he said. "I think machines are a long way from being fully intelligent, and it also depends on ethics and regulation, so it's hard to predict when AGI will arrive." He believes that before AGI arrives, AI and robotics researchers, regulators, and businesses still need to address the many issues arising from AI technology and find a balance. "For example, large language models are starting to produce a lot of hallucinations, and some companies are trying to fix that." "Raibert said. When talking about the commercialization of humanoid robots, Raibert said that although the application potential of humanoid robots is huge, there are still many challenges to enter the "ordinary people's homes" : "There are three main aspects: first, the environment of the family is different, and there will be complex scenes; Secondly, there is the issue of safety, whether families can safely use it; And finally, accessibility, the current humanoid robots are still too expensive for families." 'he said. Thirty years after founding Boston Dynamics, Raibert founded the AI Institute Lab in 2022, which focuses on basic science research leading to a new generation of intelligent robots. In his view, humanoid robots have been able to achieve many human capabilities, but in some subtle ways are still very different from human beings. "For example, the construction of small tissues such as muscle fibers, nerves and skin in robots is still very different from that of humans, and these are the problems that bother me now." 'he said.

In January this year, artificial intelligence startup DeepSeek released two breakthroughs through its new model R1, quietly redefining the economics of artificial intelligence. This model achieves top performance at 1/40 of the cost of the previous model. As of December 2024, DeepSeek's V3 large language model has reduced training costs by more than 90%. Two of DeepSeek's breakthroughs attracted widespread attention: First, DeepSeek revealed that asking AI models to elaborate on their reasoning processes - a research approach known as chain-of-thought prompting - improved accuracy and efficiency. Second, DeepSeek uses artificial intelligence to generate its own data sets, completely independent of manual labeling of the data. While there are arguments that DeepSeek isn't as cheap as it claims, these breakthroughs have certainly ushered in a new era of AI economics. The cost structure of artificial intelligence is changing dramatically. Every dollar of performance surge has had a profound impact on start-ups, enterprise applications, and infrastructure investments. This shift could upend market forces, ultimately helping nimble startups catch up with tech giants in the short term while boosting profit margins. Tech giants have already invested more than $100 billion in AI infrastructure development, and it continues to increase. Now they must consider how to generate a return on these huge investments and maintain an edge over algorithms against more nimble, smaller market competitors. In the face of a rapidly changing environment, both tech giants and start-ups are facing a clear signal: seize the opportunity of technological progress quickly, or be eliminated. The AI market landscape before and after DeepSeek Before DeepSeek's rise, start-ups struggled to compete with infrastructure spending by tech giants, which poured billions of dollars into building huge data centers every quarter and gained huge advantages from advances in artificial intelligence technology. These giants not only have massive data resources, but also gather a large number of doctoral talents, and the progress of algorithms also depends on their strong technical strength. In addition, long-established distribution networks allow them to move products quickly to existing customers and accelerate technological progress through feedback loops. Today, however, startups are big enough to compete with the tech giants. By 2025 alone, the cost of training models will fall by 95%, significantly reducing the infrastructure advantage of the tech giants. Reasoning costs have plummeted nearly a thousand-fold in the past three years and are expected to fall further in the future. The duration of the algorithmic advantage has been reduced to 45 to 100 days and may continue to decrease. When training costs are no longer a key bottleneck, inference performance (that is, how well AI models perform in real-time applications) becomes a new focus. We are entering a new phase: smaller, cheaper models that offer comparable power to larger models and can run on lower performance Gpus, extending the life of older Gpus. If smarter AI products can be delivered at a very low cost, then startups finally have a chance to outperform the tech giants while increasing profits. Efficient manpower allocation further strengthens the challenger's advantage. With no longer needing to hire large numbers of PhD-level talent to assemble a competitive AI team, startups can develop, optimize, and distribute models at a much lower cost than tech giants. And, because they are largely focused at the application level, challengers are able to enjoy higher profit margins in the same way that cloud startups gained an advantage by improving unit economics 15 years ago. This trend is not just good for startups. It also puts companies like Nvidia at greater risk. After DeepSeek's announcement, Nvidia's stock price fell 12%, although it has since rebounded. The risks for chipmakers are heightened because demand is shifting from training-focused hardware to more efficient inference solutions. The rise of consumer-grade neural processing units (Npus) could accelerate this shift, allowing AI models to run natively on devices such as smartphones and laptops. Artificial intelligence spending What's good for challengers is bad for tech giants. Ai giants have almost instinctively tied DeepSeek's dominance to national security implications in an attempt to drum up support for its development of similar technology, ignoring the fact that US researchers, including at Stanford University, have been able to replicate and even surpass DeepSeek's technology. Looking ahead, companies investing huge sums in data infrastructure projects may ask: Was the huge spending on AI model research and development wasted? If cheap technology works just as well as expensive technology, why spend so much money? Historical trends suggest that most AI advances have indeed relied on excessive capital investment in scale. The Transformer architecture was successful because of overtraining, exceeding what was considered algorithmically optimal at the time. New technological advances are proving that we can achieve the same performance at a lower cost. Although efficient solutions like DeepSeek have significantly improved efficiency, even so, the expansion of hyperscale cloud providers still requires larger data centers and must bear ballooning inference costs. However, the tech giants are not sitting still. We're already seeing an arms race for DeepSeek's achievements, with the likes of Google's Gemini model, Microsoft's Azure AI Foundry and Meta's open source LLaMA all vying for dominance. Open source models can play a key role. Mark Zuckerberg, CEO of Meta, stressed the importance of personalized AI - that is, models tailored to the needs, culture and preferences of individual users. This vision aligns with a broader trend in AI development: smaller, more specialized models capable of delivering high performance without the need for a massive cloud infrastructure. Startups win new chips At the same time, open source and closed source giants have different goals, further enhancing the challenger's advantage. Open source models created by companies like Meta will continue to compete and reduce costs across the ecosystem, while closed source models try to charge higher fees through better technology. Startups can leverage the competition between the two camps to achieve the best price/performance ratio for each use, while increasing profit margins. Regardless of the size of the business, the message is clear: Take advantage of the specific advantages available to them - market dynamics, computing power and talent - quickly or face failure. The cycle of technological progress is getting shorter, from the months or even years it used to take to set new performance standards, to DeepSeek's technological breakthrough suggesting it could now take as little as 41 days. Innovation is advancing at an unprecedented rate, and the fault-tolerance space is rapidly shrinking.

Apple on Monday announced plans to invest up to $500 billion, including plans to open a new artificial intelligence server factory in Texas. Apple said it will partner with Foxconn to build a 250,000-square-foot server manufacturing facility in Houston to produce Apple Intelligence servers. The new facility is scheduled to be operational by 2026 and will be part of Apple's major investment plan over the next four years. In addition to the new Texas plant, Apple said it plans to hire about 20,000 new workers nationwide. Apple said most of the new hires will be in research and development (R&D), chip engineering, software development, artificial intelligence and machine learning. In a statement Monday, Cook said, "We are confident in the future of American innovation and are proud to further our long-standing commitment to investing in the United States through this $500 billion investment plan."Apple said its $500bn investment plan would include partnerships with suppliers across the US, production of content for its Apple TV+ streaming service in 20 states, new hires and research and development spending. Apple said it "remains one of the largest tax payers in the United States, having paid more than $75 billion in U.S. taxes over the past five years and $19 billion in 2024 alone." The tech giant also said it would increase its U.S. Advanced Manufacturing Fund to $10 billion from the current $5 billion, create a new manufacturing academy in Michigan, and expand its U.S. research and development investments to support cutting-edge areas such as chip engineering. It is worth noting that the announcement of the investment plan comes after Apple CEO Tim Cook met with US President Trump last week. According to previous articles seen by Wall Street, US President Donald Trump said at a meeting attended by governors on Friday that Cook promised him to move production of Apple products from Mexico to the United States. Cook also pledged that Apple would invest hundreds of millions of dollars in the United States. Trump revealed that Apple is investing in China because it does not want to pay tariffs. Wall Street has previously quoted analysts as pointing out that Apple's move is both a response to the Trump administration's policies and a strategic move to protect its own interests. By increasing domestic investment and employment, Apple hopes to avoid potentially steep tariffs while maintaining its high profit margins and product pricing strategy. Here's the original text of Apple's press release on Monday: Apple today announced its largest investment commitment to date, with plans to spend more than $500 billion in the United States over the next four years. This new commitment builds on Apple's long-term investment in American innovation and high-skill manufacturing, and will support broad initiatives in areas as diverse as artificial intelligence, chip engineering, and skills development for students and workers across the country. "We believe in the future of American innovation and are proud to build on our long-term investment by committing an additional $500 billion to power America's future," said Tim Cook, Apple's chief executive officer. "From doubling the Advanced Manufacturing Fund to building advanced technology in Texas, we are excited to further support American manufacturing." At the same time, we will continue to partner with individuals and companies across the country to write an extraordinary new chapter in the history of American innovation." New manufacturing facilities in the United States As part of the U.S. investment plan, Apple and its partners will begin production of servers in Houston later this year. The 250,000-square-foot server manufacturing facility is scheduled to officially open in 2026 and will create thousands of jobs. Previously, the servers were manufactured outside the United States, and the one to be assembled in Houston will play a key role in Apple Intelligence. These servers are the foundation of Apple's Private Cloud Compute, which combines powerful AI processing power with the largest and most advanced security architecture to date to power AI cloud computing. These servers are the culmination of years of research and development by Apple engineers, bringing the industry-leading security and performance of Apple chips to the data center. The Apple team designed these servers with a special focus on energy efficiency to reduce the energy needs of Apple's data centers, which already run on 100 percent renewable energy. With the expansion of Apple Smart Systems across the United States, Apple also plans to continue expanding its data center capacity in North Carolina, Iowa, Oregon, Arizona and Nevada. Double the American Advanced Manufacturing Fund As part of the new investment, Apple is doubling the size of its U.S. Advanced Manufacturing Fund from $5 billion to $10 billion. The fund was established in 2017 to support world-class innovation and job growth in high-skilled manufacturing in the United States. This additional commitment will focus on promoting advanced manufacturing and skills development across the country. The fund expansion includes Apple's commitment of more than $100 million to TSMC's Fab 21 plant in Arizona. Apple is the largest customer of the advanced manufacturing facility, which employs more than 2,000 workers to produce Apple chips. Mass production of Apple's chips officially began last month. The chips used by Apple are carefully designed to provide users with superior functionality, performance and energy efficiency, playing a key role in Apple devices around the world. Apple's suppliers already produce chips at 24 factories in 12 states, including Arizona, Colorado, Oregon and Utah. Apple's investments in the sector have helped create thousands of high-paying jobs at U.S. companies like Broadcom, Texas Instruments, Skyworks and Qorvo. To date, Apple's U.S. Advanced Manufacturing Fund has supported projects in 13 states - including Kentucky, Pennsylvania, Texas, and Indiana - that help local businesses grow, train workers, and develop a range of innovative manufacturing processes and materials for Apple products. Expand U.S. investment in research and development Apple continues to expand its research and development investment in the United States. Apple has nearly doubled its spending on advanced research and development in the United States over the past five years, and plans to continue to accelerate growth. Recently, Apple announced the latest iPhone 16e. With the A18 chip and the new Apple C1 modem, the phone delivers fast and smooth performance and groundbreaking battery life. The Apple C1 is the first cellular modem designed by Apple, and the most energy-efficient modem on the iPhone to date. This innovation is the result of years of research and development investment, bringing together the efforts of thousands of engineers. The Apple C1 also marks a new chapter in Apple's chip strategy, setting the stage for modem system innovation and optimization for more Apple products in the future. Over the next four years, Apple plans to hire about 20,000 people, the vast majority of whom will be in research and development, chip engineering, software development, and artificial intelligence and machine learning. The expansion commitment also includes significant investments in Apple's research and development centers across the United States, with a focus on custom chips, hardware engineering, software development, artificial intelligence and machine learning. Establish a manufacturing academy in Detroit to support American business To help companies transition to advanced Manufacturing, Apple will establish an "Apple Manufacturing Academy" in Detroit. Apple engineers will work with experts from top universities such as Michigan State University to advise small and medium-sized businesses on implementing AI and smart manufacturing technologies. The school will also offer free offline and online courses covering key skills such as project management and manufacturing process optimization. Through these courses, Apple hopes to drive productivity, efficiency and quality in its supply chain. Apple has a long-standing commitment to investing in the education and skills development of American workers and students. This includes ongoing funding for organizations like 4-H, Boys & Girls Clubs of America, and FIRST, which work closely with Apple to provide free coding skills training to teens across the country. Apple's support for the next generation of innovators also includes the New Silicon Initiative, which aims to train future hardware engineers and chip design talent. Last year, the program expanded to Georgia Tech and now covers eight colleges and universities across the country. This year, Apple will also launch a new collaboration with the Center for Microchip Design Education (CEMiD) at UCLA to further expand the reach of the program.