And to be strong。
"01"
The world congress on artificial intelligence (waic) is about to open in shanghai。
This was an opportunity for insiders of the country's ministries, scientific institutions and large internet factories to talk and to say the first words。
In response, it is a phenomenon that has emerged in the us ai circle:
Many ai industry-related practitioners came to china on a field visit, returned after the trip, opened the computer and wrote a long article saying, "i just returned from china"
They have documented in detail their research in cities such as beijing, shanghai and hangzhou, replaying the real state of operations of the local ai laboratory and the science enterprise。
And what runs through the whole text is a strong cognitive shock: their multi-year industry perception is being subverted by chinese ai。
In the past, they used to judge a country's ai development path with a few simple variables: open or closed sources, government-driven or market-driven, and sufficient computing or limited resources。
In china, however, ai development is not a single-dimensional technological upgrade, but covers the full range of algorithms, data flows, technology overlaps, talent connections, landscapes, governance rules, etc。
This different development model promotes a shift in the perspective of overseas observers and points to a more fundamental issue:
What kind of development logic is chinese ai developing
In the course of the conversation, tan's judgment became clear:
Systematic co-sharing is becoming the most prominent feature of china's ai。
Open source is the most visible label in china's ai. At the technical level, models require more user use to sustain evolution; at the commercial level, open sources can nurture markets。
The cooperative boundary extends to “full factor sharing”. In communicating with insiders, tan feels that, in their perception, calculus, algorithms and data remain at the heart of, but not limited to, artificial intelligence development, and that applications, know-how, industry governance experiences, etc. Are part of a cooperative vision that can be shared。
The form of cooperation was upgraded to “ai eco-construction”. It is characterized by the mutual empowerment of industries and ai, and the creation of a mutually inclusive and mutually reinforcing ecology in the process of integrating ai into the needs of the industry and its application in the evolution of ai。
This explains why overseas observers feel that “the original framework is inadequate”。
They are no longer confronted with traditional models of scientific and technological cooperation in terms of technology export and capacity provision, but rather with an entirely new development path of capacity-building, complete factor sharing and ecological symbiotic coexistence。
"02"
And china's path to development is based on its intrinsic endowment, which is naturally growing。
The following three key points are summarized in light of what has been seen in recent times。
It is more important for china to provide an effective technology base than building a technology high ground
In order to truly understand china's ai development logic, the first step is to get away from the inherent perception that “only ai enterprises do models”。
When it comes to large models, the first reaction of many is often an artificial intelligence enterprise such as openai and anthropic。
In china, however, resources are being invested in modelling, not only in artificial intelligence enterprises, but also in consumer electronics, communications operators, office software and logistics。
For example, in june 2026, a science and technology retail company released a trillion-billion-parameter model to complete full-process training and reasoning on a 50,000-carat calculus cluster。
Behind this is that china does not follow the “model to application” path, but rather the “application to model” that anchors the product from its first day of construction on real business needs。
Engineers from a major internet plant told tan that on one of their lines of business, there are intelligent development teams that do one direction at a time。
How can the intra-enterprise workflow platform be seamlessly connected
How can operational know-how be sequestered and reusable skills modules developed
How can industry expert experience be translated into accessible knowledge base content
These extremely detailed problems of landing are the real problems they face every day。
So the chinese ai team tends to have extreme “engineer thinking”。
Many teams do not make final products, but are closer to infrastructure such as cloud computing and grids. Through this layout, the ai technology base can be embedded in more industries and business processes in the future to support industry-wide upgrades。
Building ecology is itself a way of developing technology
Self-study of self-used and adapted industries is only the first step in china's ai development. The next step is to create an open shared ai technology ecology。
In one detail, the chinese ai team, after issuing the open source model, often publishes the fine-tuning tools, reasoning frameworks, development documents, etc. That accompany it。
As can be seen, the chinese team is delivering more than just technical results. They are lowering the developers ' threshold to make the technology truly usable and useful。
There is a clear timescale。
During the last three years of the world congress on artificial intelligence, tan found a clear trend in the collection of cases published by the national development and reform commission (ndrc) called " ji hui world, china " :
The cases in 2024 were mostly the output of individual models, individual technologies and, starting in 2025, many cases were no longer just the delivery of an ai product, but rather the provision of a complete set of coverage scenarios, tool chains and iterative capabilities。
In this year's case, this feature will become more prominent. In-house sources associated with the general assembly have told lord tan that this year we attach particular importance to the high level of ai open cooperation。
From the output product, to the output program, to the output capacity, this is the construction of ai ecology through open cooperation。
For example, a large model of domestic production has generated more than 170,000 derivative models. When the uganda research team developed a large model covering 31 languages, it was selected as the base。
The technical director of a cloud computing company has indicated that feedback from developers and ecological support from open-source communities are important contributors to their large-scale model technological advances。
It is a positive cycle: openness attracts global participants, and the scene feedback reverses bottom-up models and technology systems。
At this point in time, ecological construction is no longer a “by-product” of technological leadership, but is precisely the path that leads itself。
The logic of “full factor sharing” is thus clear: using the model as the bottom seat and thus willing to help others build the bottom seat; pursuing maximum coverage and the fastest overlap with a problem-oriented “engineer's mind”; and providing a full chain of low-threshold capacity packages to build ecology。
Not everything is shared, but it's organized and controlled
One interesting detail is that, in communicating with a number of practitioners, they say, in the end, they always add the following sentence: china's ai's co-sharing is not the complete liberalization of the bottom line and borders。
When and how many elements are shared, they need to be designed, based on a set of institutional norms that cut across the entire process of industrial development。
First, technology has borders. The china export-restricted technology directory has always retained export management requirements for artificial intelligence-related technologies。
The second is that there are rules for data. The provisions on the promotion and regulation of cross-border movement of data improve the institutional framework for cross-border movement of data and regulate the flow of important data and personal information to the required scale。
In short, the opening of china's ai is an orderly, controlled and high-quality opening above the security floor — all international cooperation is being promoted within an institutional and normative framework。
As models, algorithms, talent and ecological synergies mature, ai international cooperation is moving towards institutional and regulatory dimensions. Since the beginning of this year, china has called for the advancement of artificial intelligence global governance in multilateral forums and has advocated the establishment of a world body of artificial intelligence cooperation, which is to make rule design, governance experience a shared part。
When china started talking about “governance sharing”, the last piece of the puzzle of “full sharing” was in place。
After that, we'll look back at the confusion of the overseas ai practitioner's trip to china。
Their well-known ai development logic is based on an old technological order:
The core technology is in the hands of a few, while other participants have access to products and services, but it is difficult to truly build up their technological capabilities. After entering the age of artificial intelligence, models closed down, interface fees, limited computing power and ecological locks, the more technology centralized, the stronger the dependence of the global ai industry。
And what china wants to explore is another order — one that brings together all of humanity and nations to achieve an open-source, all-embracing ai ecological construction。
This is because artificial intelligence is a very different technology that inevitably raises a whole range of security, ethical and other issues, and only through global participation and joint efforts can a safe and effective development path be achieved。
At the 2026 world congress on artificial intelligence, the open logic of china's international cooperation in artificial intelligence, “all-facts sharing”, will provide a clearer and more specific framework for action. The general assembly may also further answer where the global ai order will evolve。
Lord tan will keep watching。
