Wednesday, July 29, 2026

The substantial breakthrough in domestic deep ultraviolet lithography machines is the most noteworthy thing today. U.S. export restrictions on lithography equipment to China have forced China chip manufacturers to be self-sufficient in a "black box." Mass production of DUV means that domestic wafer mills can no longer rely on ASML at the most core process nodes. At least there is a "second leg" on the road of 7 nm and 5 nm. In terms of technical realization, the light source power, photomask accuracy and optical system stability of the immersion DUV are all long-standing difficulties. The domestic team has achieved results in these key links, indicating that they have shifted their R & D investment from paper talk to production line verification. The risk is not the technology itself, but the supporting equipment of the supply chain-photoresists, photomasks, and optical components are still purchased abroad. If the United States further blockade, this chain will still be stuck. For ordinary developers, there are no direct benefits in the short term, but the long-term decline in chip costs and the ecological acceleration of domestic CPUs and GPUs will eventually make the computing power in our hands cheaper and more controllable, which deserves attention now and be prepared to access domestic hardware platforms.

Along with hardware is the wave of open source for large models. When Kimi K3 came out, the parameter scale surged to 2.8 trillion yuan, and thousands of likes were received on HuggingFace within 30 minutes, directly pushing Anthropic and OpenAI's "model monopoly" to the front door. The advantages of open source are that the community can iterate quickly, audit security, and lower thresholds; the disadvantage is the lack of a unified commercialization path, and model maintenance and computing power costs are often borne by a few large companies. The pain points in the industry now are computing power costs and security compliance. The emergence of K3 allows domestic AI entrepreneurs to deploy locally and not rely on external APIs in privacy-sensitive scenarios, reducing compliance risks. However, training data, copyright disputes, and alignment security of open source models are still potential bombs; without a mature governance framework, companies will be caught by regulations when commercializing. Ordinary developers can play prompt and fine-tune first, but if they want to implement models into products, they still need to prepare computing power budgets and security reviews.

Returning to the software level, WorkBuddy rushed to 12 million daily activities in three months, which shows that the "productization" of internal tools is no longer a special case, but an inevitable trend of enterprise-level innovation. The core of its success lies in turning the phrase "everyone is a product manager" into an actual collaboration platform, rather than cramming ordinary users into a pile of complex features. The difficulty of implementation lies in how to find a balance between low code, scalability and security compliance; the pitfalls are that excessive pursuit of growth will lead to functional expansion and increased risks of data leakage. For ordinary developers, the rapid iteration of such internal tools provides a lot of design patterns and operational experience to learn from, which is worth using as a learning case or even trying to replicate in their own team.

Overall, the rapid iteration of domestic lithography machines, open source large models and internal tools are all weakening the boundaries of foreign technology monopoly. The risk lies in supporting supply chains, regulatory compliance and sustainable business models. As long as we can overcome both the hard-core difficulties in technological realization and the soft-core obstacles to commercial implementation, this wave of localization can truly transform into long-term dividends of cost reduction, accelerated innovation, and safety and controllability. Either get started now and seize the opportunity, or wait until the ecology matures before chasing after it. Hesitation will only leave you behind foreign technology.

The signals of today's scientific and technological news, such as scholar Wang Hong's insistence on chalk and blackboard, the progress of domestic deep ultraviolet lithography machines, and Dunhuang's hiring of foreign anchors, all point to a more diversified and humane technological future.

Scholar Wang Hong's persistence reflects that even in the digital era, traditional educational tools still have their irreplaceable value. It reminds us that technological progress should not only pursue efficiency, but should also consider the nature of education and human needs. This kind of reflection is a necessary balance for the development of science and technology.

The progress of domestic deep ultraviolet lithography machines reveals the importance of independent technological innovation to national strategic security. As ASML shares plummet, we can foresee that China chip makers will have more choices, which will intensify competition in the global lithography equipment market. This is both an opportunity and a challenge for China. It requires us to be prepared in many aspects such as technology, industrial policies and international cooperation.

The incident of Dunhuang employing foreign anchors reflects the tension between cultural inheritance and internationalization. In the context of globalization, how to inherit and promote local culture while attracting international talents is a question that requires careful consideration. It is not only related to the spread of local culture, but also affects international cultural exchanges and understanding.

By connecting these signals together, we can see that a diverse and humane technological future is quietly taking shape. Technological innovation and traditional culture coexist, national strategic security is closely linked to the global market, and cultural inheritance and internationalization are intertwined. In this process, impact and value coexist, and problems and risks coexist.

In terms of impact, breakthroughs in independent technologies will enhance the country's scientific and technological strength and international competitiveness, and at the same time promote changes in the competitive landscape of the global technology market. In terms of value, respecting and inheriting traditional culture can help build a community with a shared future for mankind, while the introduction of international talents can promote cultural exchanges and integration.

However, this process is not without problems and risks. Technological competition may lead to tense international relations, the tension between cultural inheritance and internationalization may lead to cultural conflicts, and talent introduction may also face the challenge of cultural integration.

In general, today's scientific and technological signals indicate the integration of science and technology and humanities, and the harmonious coexistence of technological innovation and cultural inheritance. This is both an opportunity and a challenge, and we need to adopt a more open and inclusive attitude to meet this diversified and humane technological future.

The most noteworthy signal today is actually two seemingly unrelated stories that both point to the same core issue: Kimi K3 open source and domestic DUV lithography machine delivery. One is in the AI field and the other is in the semiconductor field, but both carry strong "breakthrough" narratives, and both have things that make me, a veteran, wary.

Let's start with Kimi K3. An open source model with 2.8 trillion parameters reached the top of Hugging Face in 30 minutes. NVIDIA spent 5 billion yuan on Ilya and signed a guarantee of 250 billion yuan for OpenAI. This picture is too familiar. When BERT came out in 2018, everyone said that the larger the parameters, the better; when ChatGPT broke out in 2022, everyone said that the larger the computing power, the better. Now that the parameter has reached 2.8 trillion yuan, Ilya's SSI needs to be "safe and super intelligent", but the question has never been whether technology can do it, but who will pay for it after it is done. WorkBuddy was launched with 12 million DAUs in three months. Tencent discussed "Mingpin Tang" internally. This is indeed a good product, but behind it is the traffic feed of Tencent's ecosystem, not the 2.8 trillion parameter. It is a good thing for Kimi K3 to be open source, but a model to go from open source to actual commercial value is separated by huge engineering costs, deployment costs and user habits. How many of the most liked models on Hugging Face in history have really changed the industry? Not to mention that Anthropic's Claude has just exposed privacy issues-shared chats are included by Google and contain pornographic content. While everyone is frantically accumulating parameters and computing power, loopholes in security design and data governance are ignored. Ilya wants to build a "secure super intelligence", but can't even manage the most basic privacy of chat records. This is really ironic.

Let's look at the domestic DUV lithography machine. When delivered during the year, ASML's share price plummeted. Changxin Technology's listed market value was 3.66 trillion yuan, and Country Garden lost 47 billion yuan in Changxin's sales. This is indeed a milestone, but look carefully at the risks. First, the domestic lithography machine is DUV, not EUV. The process nodes it can do are limited. It mainly solves the problem of replacing mature processes, not the most advanced ones below 7nm. Second, Reuters reported that "delivery is expected within the year", but the actual production capacity ramp, yield, and equipment stability are all unknown. There have been many "breakthroughs" in China semiconductors in history, but every step from the laboratory to the production line is full of variables. Third, Changxin Technology has a market value of 3.66 trillion yuan, which is larger than many chip companies combined. Country Garden sold Changxin back then, but now it seems that it has lost 47 billion yuan. However, at that time, Country Garden's own capital chain was broken, and it may not have survived until it was sold. What is behind this is the valuation bubble of domestic semiconductors-the capital market's enthusiasm for "domestic substitution" may have overdrawn growth in the next few years. What really needs to be paid attention to is whether so much money pouring in can be transformed into real technological iteration and cost reduction, rather than becoming the next "big fund corruption case."

Finally, when these signals are strung together, the biggest risk is that "narrative trumps reality." Whether it is the 2.8 trillion parameter of AI or the "delivery within the year" of semiconductors, it is easy for people to overlook two questions: Who is making money and who is paying the bill? Nvidia invested in Ilya, which on the surface supports security AI, but in essence it sells more GPUs; domestic lithography machines are delivered, capital is carnival, but in the end, it is downstream wafer fabs who need stable, low-cost, high-yield equipment, not "domestic" label. History tells us that every technology bubble starts with a narrative of "breakthrough" and finally ends with the embarrassment of "landing".