Tuesday, July 28, 2026

Well, among today's bunch of signals, the most interesting things are actually two things: the carnival of semiconductor capital and the open source closure of large models. The rest of the summer holidays, the death of Keigo Higashino, and the explosion of Youth Travel Services are all to fill the traffic pool and have little to do with the technical circle.

Let's start with the listing of Changxin Technology. This is the most violent capital story in 2026. Country Garden sold Changxin back then, but now it is a symbol of no pattern, and 47 billion yuan has been wasted. But what deserves more attention is that Li Bin went to eat food and Wei Lai fought to make a profit of 700 million yuan. These new car-building forces now do not rely on selling cars to make money, but rely entirely on investing in chip companies as cash machines. For a company like Changxin, once it was listed on the Science and Technology Innovation Board, its market value directly reached 3.6 trillion yuan, surpassing ICBC. What does this mean? It shows that the will of the country is renewing the life of semiconductors. Behind this wave of market is the perfect resonance of industrial logic and capital logic. But don't be happy too early. Changxin's profitability and technical moat are far behind those of TSMC. When this valuation bubble breaks, it will depend on the next earnings season. For ordinary developers, this has nothing to do with you. Don't think about stud stocks, you should write code.

Then there is the open source and closed source battle for large models. Today, Anthropic issued a statement supporting the open source model on Hacker News, and OpenAI also signed an open letter. But look at these two companies, one makes money from closed sources, and the other uses open source as a cover. Both of them are actually engaging in commercial monopolies. Large models such as the Kimi-K3 and Solar-Open2- 250B are on Hugging Face today, but how many can really run? The cost of building a GPT-4 local model for ordinary developers is still ridiculously high. It's also very interesting that Reddit is challenging Google. Reddit wants to re-price community content. To put it bluntly, it doesn't want to work for Google for free anymore. But the question is, the quality of Reddit's user-generated content is uneven. Can it really sell for a high price? I'm skeptical. This is essentially a redistribution of data ownership in the AI search era, but as a developer, you should pay more attention to tools that can be used directly, such as the Agent operated by TikTok and the Agent that physically operates the iPhone. This is a practical thing that can help you save money and make money. Don't stare at big models all day long.

Finally, TouchGrass on Product Hunt today is quite interesting. It is a software that reminds you to rest, but it requires no intrusion rights. This shows that these people finally realized that no matter how powerful AI is, people still have to eat and sleep. Stop rolling, boys.

The ultra-high-speed listing of Changxin Technology on the Science and Technology Innovation Board marks an accelerating node in the transformation of China's semiconductor industry chain from "assembly" to "self-research". Capital's enthusiasm for memory chips coincides with U.S. technological restrictions on China, forming a "reverse supply" drive: domestic capital is willing to bet on localization breakthroughs with high valuations, and the government quickly gathers R & D resources through policy support and tax incentives. In the short term, Changxin's market value expansion will attract more financing into upstream materials, lithography machines and packaging and testing companies, forming a spiral upward trend of capital penetration into the entire chain; but the rapid expansion of valuation also hides the hidden danger of technological maturity and market demand. Once the technology node fails to fulfill expectations or is subject to export control upgrades, the stock price retracement may exceed the fluctuations of traditional IPOs.

At the same time, AI is evolving from "perception-generation" to "perception-prediction-action". The Physical AI Alliance, co-sponsored by NVIDIA and other top companies, embeds world models, spatio-temporal reasoning and action planning into the large model training pipeline, aiming to allow models to directly participate in closed-loop decision-making of robots, autonomous driving and intelligent manufacturing. Breakthroughs in geometry learning in the mathematical community have been rapidly transferred to explanatory research on large models, indicating that high-order mathematics is becoming a key tool to solve the "black box" bottleneck of large models. If this path matures, AI will no longer be a search engine that only provides information, but a "digital craftsman" with execution power, deeply reshaping the efficiency of traditional manufacturing, logistics and urban operations. The risk lies in safety and controllability: Strengthening the model's ability to act means that the potential harm of losing control is amplified, requiring the industry to jointly develop safety benchmarks and audit mechanisms.

At the level of information consumption, the game between Reddit and Google reveals the re-pricing of community content in the era of AI search. The real dialogue and segmentation experience of the community provide high-quality training signals for the large model and become a source of first-hand answers to user queries. As AI search portals gradually replace traditional SERPs, the focus of traffic and business monetization will shift from portals to vertical community platforms, prompting content creators and platform operators to accelerate the construction of machine-readable structured discussions. If the platform is unbalanced in terms of privacy compliance and content governance, it may trigger regulatory tightening or a crisis of user trust, which in turn affects the entire ecosystem's business model.

Taken together, these three clues: semiconductor capitalization, technical deepening of Physical AI, and AI repackaging of community content together point to a higher-level trend-"digital materialization." The simultaneous acceleration of technology and capital not only promotes hardware self-sufficiency, but also provides underlying computing power for AI's activism; content platforms provide AI with the blood of semantics and experience. In the next six months to a year, collaborative innovation in the upstream and downstream of the industrial chain will accelerate. The beneficiaries include domestic chip manufacturers, AI entities with cross-modal R & D capabilities, and vertical platforms that can provide high-quality community data; while the technology is immature. Highly valued companies, AI providers that lack security governance, and traditional media that fail to adapt to AI search may face the dual pressure of valuation retractions, regulatory restrictions, or traffic loss.

Among today's technology signals, the most worthy of vigilance and deep investigation are the listing carnival of Changxin Technology, the A-share "technology giant", and the in-depth linkage and reassessment of AI and the content platform ecosystem. On the surface, the two seem incompatible, but in fact they reflect the mood, logic and undercurrent of the technology capital market in 2026: capital and narrative have gradually driven the valuation expansion of technology giants and emerging companies in the past few years. Now, in the new cycle of "China Core" and AI content, old bubble risks and new structural risks are intertwined.

The market value of Changxin Technology has soared fivefold after being listed after half a day, surpassing ICBC to become the number one A-share brother. Behind this is the complex interest game between huge hot money, industrial capital and listed companies. Some people celebrate "keeping the cloud and seeing the moon bright", some people regret "missing the feast", and more institutions are sweeping goods in the secondary market, as if the chip localization dividend and AI accelerated computing power demand have ensured the company's future 10 years of high growth. However, the bubble in China's stock market and even global emerging technology stocks has never been absent in history. When the Internet bubble burst in 2000, and the GEM flash collapse at a high level in 2015, there were similar moments of "only rising but not falling" and "wealth freedom for all". Although Changxin's main income and industry status have strong support, are they against the Nvidia and AMD-style technology bull market in the United States? Can the A-share industrial chain, capital flow, and global market barriers really give it hundreds of billions or even trillions of dollars in space? Don't forget that it is often the early PE/VC and major shareholders that can truly secure their pockets in chips and even technology dividends. After ordinary investors go in, they buy more stories than profits. Are "prophetic" quitters like Country Garden short-sighted or do they see the huge uncertainty behind the valuation? Who is paying the bill and who is dropping the bag has actually given some answers.

At the same time, the reassessment of AI and content platforms has reached a critical point. Reddit, Google, and major content communities are engaged in a game around AI search and data utilization rights, and are no longer satisfied with the old traffic for content exchange logic. On the one hand, AI models are increasingly dependent on real-world data, and purely synthetic content and "automatic canned" information have become difficult to satisfy users. On the other hand, the content community has begun to proactively ask for prices-Reddit is a typical example, from traffic distribution to content charging, to direct negotiations with AI companies. This change may seem to be beneficial to community content, but in fact it has accelerated the rise in information barriers and the prices of "content raw materials". AI companies either pay high prices or crawl in a grayer way. The information obtained by users is increasingly limited by "who can afford the money" rather than "who is right." This is an open and innovative approach to the entire Internet. In fact, it is a negative incentive.

These two themes show that the circling game of capital and technology has entered a new stage. The valuation myth of technology leaders and the "battle for raw materials" for AI data are all mixed with old stories of bubbles, monopoly and scarcity. Ordinary users or investors have to ask: Is what I am involved in value creation or is it a drum to spread flowers? If it is just a short-term profit allocation game, the risk is always at the end of the chain. Looking back at history, the technologies and companies that can truly travel through cycles often do not rely on temporary hype and traffic, but on the few that can still maintain real profits and extensive value creation after the uproar. The current market is far from calm enough to reach this standard.