Thursday, May 14, 2026

The core of today's news is just two words: work. AI must not only be able to speak well, but also be able to work hard, from e-commerce to writing code, to moving bricks and errands. Don't just stare at the "IQ" of the big model. Now everyone is more concerned about whether it can handle the work for you.

You see that Ali stuffed Qianwen directly into Taobao and competed with Byte for the mentality of "AI e-commerce". This is quite interesting. In the past, AI helped you recommend, but you had to skip over and click on links to place an order. Now it provides one-stop service directly in AI. What's behind this? It is AI agentization. It is not just a chat robot. It is a "digital employee" who can help you run processes, make decisions, and complete transactions. Didn't Cursor CEO also say that AI can now handle 75% of code, and engineers have become "ghost colleagues" managers? The several "agent skills" libraries on GitHub prove that everyone is thinking about how to make AI, a "working person", more reliable. For ordinary developers, this means that you can't just know CRUD. You have to start learning how to "manage" AI, how to break down complex tasks into instructions that AI can understand, and how to verify what AI generates. It's not here to replace you, it's here to change the way you work. Commercially, who doesn't want to be able to directly transform AI e-commerce? But there are many pitfalls. Who is responsible for problems with the products recommended by AI? What about the return process? These are real engineering and business challenges.

Another general direction of "working" is embodied intelligence. CVPR 2026 has been overshadowed by a specific intelligence, such as robotic arm grabbing and robot navigation. The Lei Feng website article said well that machines not only need to recognize images, but also need to "intervene in reality." This is not as simple as running a simulation on the screen. It is to let the machine understand the real three-dimensional space and be able to walk, grasp and interact. Yushu Technology's 3.9 million manned mecha may sound like science fiction, but it is an extreme manifestation of embodied intelligence. FF (Jia Yueting's company) also announced that it had signed up for a "data factory", saying it was the fuel for the EAI brain. What does this mean? To implement embodied intelligence, it requires massive amounts of real-world data, and it must also be data that can be feedback in a closed-loop manner. In engineering, the difficulty of this thing is hellish. Sensor fusion, real-time decision-making, physical interaction, and security redundancy are no joke. For developers, if you are still engaged in traditional CV image classification, you may really want to think about it. Nowadays, to develop specific intelligence, what you need is cross-border capabilities such as software and hardware combination, control algorithms, and simulation. This wave is really going to pull AI from the cloud to the ground and let it deal with the physical world with real weapons.

So, to sum up, whether it is AI's process automation in the digital world or its specific actions in the physical world, the core is to upgrade AI from "answering questions" to "solving problems." This is not only a technological breakthrough, but also a change in the entire industrial model. Don't always focus on the parameters of the big model, think about how to make these AIs really do the work. This is the future.

Today's technology signals reveal several important trends that are intertwined to jointly outline a future full of change and opportunities.

First of all, the in-depth application of artificial intelligence in the field of e-commerce has become an irreversible trend. In the battle for AI e-commerce, Ali and ByteDance are advancing technological integration at an unprecedented speed. The comprehensive connection between Qianwen and Taobao and the launch of "AI Shopping Assistant" mark a key step in the application of AI in the e-commerce field from assisting decision-making to full-process integration. This change will not only change consumers 'shopping habits, but will also promote the transformation of traditional e-commerce models into intelligence and personalization. However, this also means that the e-commerce industry will face more intense competition and higher technical thresholds.

Secondly, the application of AI in various fields is gradually changing from a "guest" to a "main battlefield". The appearance of embodied intelligence at CVPR 2026 demonstrates the ability of machines to move from recognizing images to intervening in reality, which will bring disruptive changes to the field of computer vision. At the same time, Cursor CEO's speech on AI coding revealed the huge potential of AI in the field of software development. When 75% of the code is generated by AI, the role of engineers will undergo a fundamental transformation, and the lowering of software development thresholds will drive the craze of entrepreneurial innovation. However, this may also lead to changes in the talent structure and a redesign of the education system.

Finally, data has become a key element of corporate competition. FF's data factory business completed its first sales order, marking the successful implementation of the data-closed-loop business model. In the digital economy era, data has become the most valuable asset of enterprises. For enterprises, how to effectively utilize data and realize the value-added of data assets will be the core of future competition.

Overall, today's technology signals reveal the following impacts, values and risks:

Impact: The rise of AI e-commerce will change consumers 'shopping habits and promote the transformation of traditional e-commerce; the widespread application of AI in various fields will change the talent structure and reshape the education system; data has become the core element of corporate competition, and data asset management will become Enterprise core competitiveness.

Value: AI technology will promote industrial upgrading and improve production efficiency; the application of AI in the field of software development will lower the threshold for entrepreneurship and promote innovation; the value-added of data assets will bring new growth points to enterprises.

Problems: Competition in the AI e-commerce field will intensify and industry thresholds will increase; widespread application of AI technology may lead to unemployment problems; data privacy and security issues need to be resolved urgently.

Risks: Technological changes may lead to increased social inequality; changes in talent structure may lead to changes in social structure; data abuse may cause new social problems.

Today's technology signals reveal three intertwined core narratives: the close combat of AI e-commerce, the dislocation of virtual and real intelligence, and the illusion of productivity caused by AI programming. Behind these seemingly independent dynamics, we jointly outline the systemic risks in the process of technology capitalization.

Ali deeply embeds thousands of questions into Taobao's closed loop, which is essentially the ultimate manifestation of traffic anxiety. The logic of e-commerce platforms using AI to restructure the shopping process is no longer new-from Amazon's Echo order in 2016 to Microsoft's Bing Shopping Assistant in 2023, the actual conversion rate has never exceeded one-tenth of that of traditional search. The problem is not technological maturity, but the underlying logic of consumer decisions: Do users really need to have multiple rounds of conversations with a virtual shopping guide to buy a roll of toilet paper? What is even more noteworthy is the privacy dolls: When AI simultaneously grasps consumption habits, payment data and conversation content, once a data leak occurs (such as the recently named risk of refurbished hard drives), the consequences will be exponentially amplified. The "AI e-commerce mentality" that Byte and Ali compete for is actually a data monopoly competition wrapped in technology.

The grand occasion of the CVPR venue being flooded with robot papers is in ironic contrast to Faraday's future announcement that the "data factory" will sign the first order. The gap between embodied intelligence and commercialization is wider than participants think. The Yushu manned mecha is priced at $3.9 million, but there is still a significant gait lag in the live demonstration video. This high-priced, low-maturity product is destined to be a B-end toy. The real problem lies in cost transfer: FF's "EAI data closed-loop" requires a large amount of real-world scenario training, and its published customer orders are likely to be smart city projects subsidized by local governments (combined with the monthly circulation of 4 trillion yuan of local debt). Such projects tend to fall into a cycle of "financial payment-corporate data-capital speculation valuations-the lessons of the autonomous unicorn are being repeated at the robotic arm track."

Cursor preaches that "75% of code is generated by AI" is more like a dangerous behavioral art. The intensive emergence of agent management tools on GitHub (such as mattpocock/skills, addyosmani/agent-skills) exposes the truth about AI programming: engineers are being downgraded from creators to prompt editors and code correctors. The unemployment and confusion of front-end engineers on V2EX is not an isolated example, but a precursor to the collapse of the talent structure. When companies rely too much on AI to generate code, they will face triple hidden taxes: a surge in debugging time (Stanford 2025 research shows that AI code debugging time is 3.2 times that of manual work), a cumulative technical debt index (the University of Tokyo measured the system crash rate after 5 AI iterations. Up to 47%), and the most critical thing-the exhaustion of innovation capabilities. Lin Junyang's "Shaped Brain" project funded at a valuation of US$2 billion may not escape this trap.

Amid all this noise, signals that are truly time-penetrating were ignored: the Chinese Academy of Sciences 'breakthrough in obtaining the genetic information of Homo erectus for the first time 400,000 years ago had no capital enthusiasm, and the Ministry of National Security's warning about the risk of data retention was only in the developer community. Single-digit discussions. While the technology industry is addicted to rewriting the world with big models, basic research that takes ten years to sit down and the social costs incurred by the popularization of technology are continuing to lose their voice in the public opinion field. Historical experience shows that bubbles are often not burst by competitors, but torn by neglected reefs.