Monday, July 13, 2026

Today's technological signals actually reflect two major trends: one is the real conflict between the implementation of AI and the implementation of ideals, and the other is the capitalization of domestic hard technology, industrial upgrading and the acceleration of going to sea.

Let's talk about AI first. You see, the market value of Smart Music has risen to one trillion Hong Kong dollars, which can overtake Baidu and Xiaomi three streets. Of course, there is the AI model craze. Both at home and abroad are talking about the dream of "AI emerging". But calm down, today's 36 krypton hot article "Why more and more companies find that AI cannot replace manual labor" is simply a hammer. In the early days of large-scale factories, they imagined that AI could do everything, but found that everything was inseparable from people-the process was too complex, the data was dirty, the AI was unreliable, and the chicken feathers were everywhere. Now I turn around and regret having laid off my staff too early. This is completely contrary to last year's global public opinion that "AI is about to swallow all white-collar jobs." From an engineering perspective, the vast majority of large learning/language/picture/video models nowadays have only the "semi-automatic + human flesh-taking" hybrid process that is reliable. Fully automatic either has unstable output or a major accident occurs when it goes wrong. Ordinary developers don't be impulsive, don't believe in sales painting pie, really want All in AI, the project has no budget, no one tuning, the last drop pit must be their own.

Capital did not stop in this round of AI ebb tide. You see Tencent pile up Shenzhen Yunbao intelligence, DPU domestic first stock brand stood up. DPU this kind of computing infrastructure, ten years ago we still think far away, now AI landing reasoning depends on it to top traffic, reduce costs. Companies such as Yunbao and Yuexin are essentially building the foundation for AI, cloud computing, and even new energy. The IPO of DPU and domestic semiconductors shows that these hard technologies in China have finally ushered in a window of confidence in the secondary market. As long as cash flow is sustained and mass production continues, valuations can rise. But making hardware is a hard job, with high technical barriers, quick burning of money, and tight mass production and ecological constraints. It is not uncommon for a piece to fall. Ordinary developers and entrepreneurial teams don't just envy financing and IPOs. There is really no aura of a big manufacturer when making hardware. Once the financing window is closed, wages cannot be paid for half a year.

Another noteworthy trend is the evolution of AI development tools and automated productivity. Hacker News and GitHub hot lists are all about "AI helps you do your work" gadgets such as Claude Code, OpenWiki, OfficeCLI, and AI Job Search. Everyone is not advocating that AI should replace you, but how to make AI your right-hand man. For example, AI automatically writes Cover Letters, code documents, and automatically finds bugs. These things can really save duplication of labor. The most interesting thing is that users are getting smarter: discovering that a bigger Claude model does not mean that it is smart, and changing the model is not as good as changing the "effort". This shows that the "usage method" of AI products is far from certain, and parameter tuning and scenario adaptation have become the decisive point instead. Ordinary developers, even small teams, really want to improve efficiency. Don't imagine that the big model can dominate the world. It's better to think about how to use AI to segment and combine processes to create "small but beautiful" business scenarios.

Overall, the popularity of capital and media has not changed: AI tells stories, hard technology talks about listing, and implementation talks about efficiency. For developers, don't follow suit and hype, let alone be deceived into the trap. The ideals of AI are full, the reality is very bony, and the hardware capitalization window is still there, but not everyone can wait until the spring of the IPO. If you really want to seize the dividends, you have to work hard, trial and error, and review the offer in your own scene, and don't be the last one to take over the offer.

The most significant signal of today's data flow points to the interweaving of technology implementation and business model reshaping under the wave of artificial intelligence. First of all, the internal letter from Zhipu AI founder Tang Jie and the discussion that "more and more companies are discovering that AI cannot replace manual labor" jointly depict a picture of AI moving from "conceptual carnival" to "pragmatic implementation." The market value of Intelligent Spectrum has soared to a trillion-dollar club, indicating that the capital market is still full of confidence in the long-term potential of AI. However, the questioning of "what is more important" in the internal letter suggests that AI companies are moving from pure technology stacking to more commercial value scenarios mining and productization. This shift means that the next stage of AI will no longer be a model-sized arms race, but a competition for who can truly solve practical problems and bring considerable commercial returns. Companies that rely too much on the concept of "PPT" AI, and those that try to replace all labor costs with AI across the board, may face serious challenges.

Echoing this, the application of AI in the field of code development has also shown a trend of evolution from "tools" to "partners". Tencent WorkBuddy is positioned as a "Codex that is more suitable for the physique of Chinese people", emphasizing that AI serves as a collaborator to help humans disassemble problems and advance work rather than replace them. The AI models and Spaces on Hugging Face, as well as the AI application frameworks on GitHub, such as AI-driven job search frameworks, automated document generators, and AI assistants that can read and write Office files, are proving that AI is penetrating into all aspects of software development, improving development efficiency and even reshaping the development process. However, the token spending discussion on Claude Code and OpenCode on Hacker News, as well as the article "I Love LLMs, I Hate hype", also warn us that the application of AI in the code field still needs to balance efficiency, cost and practical results to avoid falling into the quagmire of excessive marketing. In the next six months to a year, we will see the emergence of more AI tools that focus on specific development scenarios and provide precise value, rather than all-encompassing "all-round AI."

Finally, developments in the consumer goods industry, especially the trademark infringement cases of LV and Jasmine Milk White, and Ali's reduction of the express delivery system, although on the surface, they reflect brand value and supply chain integration. The deep logic of strategic adjustment of giants will become important soil and application scenarios for the implementation of AI technology. The LV case highlights the severity of brand protection and changes in consumers 'perception of brand value, while Alibaba's adjustment to express delivery suggests that the restructuring of traditional business models is accelerating. AI has great potential in brand marketing, supply chain optimization, consumer insight, etc., but its implementation effect will be highly dependent on the understanding of the complex logic of these traditional industries. Those startups that can cleverly combine AI capabilities with specific business pain points, such as Yunbao Intelligence, which focuses on DPU (Data Processing Unit), have heavy investments like Tencent behind them and may become the next breakthrough point. These signals are all pointing in one direction: the future of AI belongs to innovators who can understand and reshape the complexity of the real world.

In today's signal, the most worthy topics are the hidden cracks in the AI bubble and the digital anxiety in traditional industries. Behind these seemingly independent incidents, Tencent's market value has soared tenfold, Tencent's heavy position on the DPU track, and Claude's "becoming stupid" are actually different aspects of the same story: capital's fanatical pursuit in the AI field is encountering real backlash, while traditional enterprises are struggling under the pressure of digital transformation.

The soaring market value of Intelligent Spectrum may seem like a victory for the AI industry, but a closer look at this internal letter shows that Tang Jie emphasizes "more important things after the GLM moment"-which in itself reveals anxiety. The surge in market value is often not because the business is truly mature, but because the capital market is too optimistic about AI. There have been many similar bubbles in history, such as the blockchain boom in 2017, when the market value of countless projects soared, and most of them eventually fell silent. The fact that Zhipu's share price can be "stabilized" after the lifting of the ban is more like a short-term capital game than a reflection of long-term value. More importantly, the narrative of AI replacing artificial labor is being shattered. 36 Krypton's article "Why More and More Companies Find that AI cannot replace manual labor" reveals a cruel reality: AI's performance in reducing costs and increasing efficiency is far less than expected. Many companies find that layoffs are early because AI cannot completely replace the complex judgment and creativity of humans. This is in sharp contrast to the high valuation of Intelligent Spectrum. The enthusiasm of the capital market and the coldness of practical applications are tearing the narrative of the AI industry.

Another clue worthy of attention is the struggle of traditional industries under the wave of digitalization. Nippon spent US$8.6 billion to acquire Akzo Nobel's decorative coatings business, which seems to be an industry integration, but actually exposes the anxiety of traditional manufacturing in digital transformation. The coatings industry itself has limited room for growth. Nippon's acquisition is more like a defensive layout, trying to withstand the impact from emerging materials and smart manufacturing through scale effects. The changes in the express delivery industry more intuitively demonstrate the subversion of digitalization on traditional business models. Ali's reduction of "Tongda" means that e-commerce giants are loosening their control over logistics. Behind this is the rise of new logistics technologies and models, such as unmanned distribution and intelligent warehousing. Traditional express delivery companies may be marginalized if they cannot keep up with the pace of digitalization.

The most ironic thing is that behind these grand narratives, the companies that provide "shovels" may actually make money. Tencent's Cloud Leopard Intelligence focuses on DPU (Data Processor), which is the infrastructure of the AI era. Similarly, there are basic semiconductors, whose price has been raised by 25% due to demand for AI and new energy. These companies are not directly involved in AI hype, but they are the biggest beneficiaries of the bubble. And those startups that shout about the AI revolution may end up just working for these infrastructure companies.

The risk is that when a bubble bursts, it is often the most optimistic participants who are most injured. If the market value of Smart Spectrum falls, it may trigger a chain reaction, causing financing difficulties for AI startups. The failure of traditional industries in digital transformation may lead to a large amount of resources being wasted on ineffective projects. The longer-term problem is that excessive hype about AI and digitalization may cause people to ignore real opportunities for technological innovation and industrial upgrading. For example, behind the chaos of signal jammers is actually the real need for radio management and electromagnetic security, but this need is submerged in the sexier AI story.

After all, today's signals reflect a cliché: there is always a huge gap between the speed of technological progress and the speed of commercialization. Capital markets always try to bridge this gap, but they will eventually be pulled back to the ground by reality. The true value may be hidden in those less sexy corners.