⚠️ Editorial note: The open source ecosystem in China operates under a distinct institutional framework — characterized by state-led initiatives, intranet-like boundaries, and top-down governance. Readers should be aware that this context differs from the community-driven open source model common in other regions. The term “open source” as used in Chinese media may refer to practices that diverge from the conventional definition.
China Open Source Daily — 2026-11-15
🏛️ Policy & Regulation — The Committee Arrives Before the Community: MIIT’s Industrial Operating System Standards Working Group
1. The event, and the institution it sits under. On 12–13 November 2026 the 2026 Industrial Operating System Conference (2026工业操作系统大会) was held at the Zhangjiang Science Hall in Shanghai under the theme “Intelligent control’s new foundation, manufacturing’s new paradigm” (智控新底座 制造新范式). It was hosted by the China Center for Information Industry Development (CCID / 赛迪研究院, 中国电子信息产业发展研究院). Attendance, as reported by the Shanghai Economic and Information Technology Commission: more than 800 representatives, including Wang Jiangping (王江平), member of the National Committee of the CPPCC; Wang Guodong (王国栋) and Qian Feng (钱锋), both academicians of the Chinese Academy of Engineering; Liu Wenqiang (刘文强), CCID party secretary; and Qiu Wei (裘薇), deputy director of the Shanghai Economic and Information Technology Commission — plus more than thirty central state-owned enterprises including China Baowu, CNPC, Sinopec, Ansteel, China Changan, FAW and China Huadian.
The institution that matters was created five weeks earlier. On 21 October 2026 MIIT held the founding meeting of its Industrial Software Standardization Technical Committee (工业软件标准化技术委员会) in Hangzhou, addressed by vice-minister Xiong Jijun (熊继军), with its secretariat housed at MIIT’s Fifth Electronics Research Institute (电子第五研究所). Its stated remit: standards for key technologies, data interoperability, products and services, integrated applications, and ecosystem building.
2. What was actually convened, and on what directions. During the conference, that committee “formally launched the establishment (正式启动筹建) of an Industrial Operating System Standards Working Group”, organised around four directions stated in the source: reference architecture (参考架构), interconnection (互联互通), compatibility and adaptation (兼容适配), and test evaluation (测试评价) — with the stated purpose of solving “interfaces that do not connect, data that is incompatible, and evaluation bases that are not unified.” Simultaneously, a Pilot Validation Formation for Industrial Operating Systems in Key Industries (重点行业工业操作系统中试验证方阵) was launched, assembling industry users, software and hardware firms and testing bodies to run technical adaptation, system integration and application validation in real production settings across steel, non-ferrous mining, rail transit and nuclear power — explicitly to solve the problem that innovative products “cannot get into the field and onto the line.”
3. The report, and the theory of value in it. CCID released 《工业操作系统技术产业探索与实践研究(2026)》, which is said to systematise the technical content, architecture, deployment forms and industrial ecology of the industrial operating system and to trace a path “from control support toward the coordination of data, models and agents.” Zhai Yanfen (翟艳芬), deputy director of the China Software Testing Center (中国软件评测中心), gave the division of labour that is the report’s real institutional content: construction is moving “from building around a single project to the coordinated development of a general foundation and industry capabilities” — basic-software vendors supply common software capability; industrial-technology firms convert industrial mechanisms, process rules and professional tools into reusable components; and industry users extend from being requirement-raisers to being co-builders and validators.
4. The surrounding numbers, and the policy that pays for them. These are the claims the sources make rather than figures this briefing verified: industry operating systems are already deployed across petrochemicals, chemicals, steel and non-ferrous metals; in the first three quarters of 2026, China’s software business revenue reached ¥11,803.1 billion, up 8.4% year on year, with industrial-software product revenue up 7.3%; IDC projects Chinese industrial enterprises’ AI spending to reach ¥280 billion by 2030 at a 52% compound annual growth rate for 2025–2030. The policy frame is the “AI + Software” special action implementation plan (《“人工智能+软件"专项行动实施方案》), whose targets are that by 2028 promotion reaches 20,000 industrial enterprises above designated size, one hundred intelligent technical-renovation projects land, and more than one hundred smart-manufacturing demonstration factories are built.
Institutional Analysis & Commentary:
This series has spent six weeks documenting an ecosystem that builds certifiers, registries, vouchers and procurement loops — and has not yet found the institution that receives a stranger’s patch. Yesterday’s briefing found two curricula pointed in opposite directions. Today the same asymmetry arrives in the most consequential domain it could: the factory floor, written in the language of a standards committee.
Start with what kind of instrument a standards working group is, because the distinction decides everything downstream. A standard is not a certificate and not a subsidy. It is a coordination device: a written answer to the question what must be true of two things for them to work together, published in advance so that parties who will never meet can make compatible investments. Williamson’s whole account of hierarchy and market begins from the fact that this device is expensive to produce and impossible to produce privately, which is precisely why standardisation bodies exist. On that test, a national reference architecture for industrial operating systems is a genuine public good, and MIIT building one is not a scandal — it is the correct institution for the problem named. The problems named are also real: interfaces that do not connect, data that is incompatible, evaluation bases that are not unified. The series has argued for six weeks that a commons is a set of rules that strangers can rely on; a published interface standard is exactly such a rule.
Now the second thing, and it is where the analysis must be sharp rather than grateful. Read the four directions end to end — reference architecture, interconnection, compatibility and adaptation, test evaluation — and one notices that all four are buyer-side instruments. They tell a purchaser what to require, and they tell a vendor what will be accepted. Not one of them governs how the software that meets the standard is maintained: who reviews a patch, who holds commit rights, who owes a fix for a defect found in a steel mill at three in the morning, who is allowed to say no to a change. Those are not questions a standards committee is designed to answer, and the omission is not negligence. It is the same structural fact this series recorded on 2026-11-11 about municipal procurement and on 2026-11-13 about certification: the instrument that fits the transaction is the one that gets built, and standardisation fits procurement perfectly.
Third, and this is the entry’s real subject: the ordering. Note the sequence. October 2026: a MIIT technical committee is founded. November 2026: it launches a working group for industrial operating systems and opens its first round of standard project solicitations. What has not happened — in any published account available to this briefing — is the formation of an industrial-operating-system community: an open repository, a contribution guide, a maintainer roster, a licence discussion, an issue tracker with outside participants. The ecosystem has built the committee before the community, and it has built it with the participation of thirty-plus central SOEs and four technical directions but no stated contributor list.
One comparison is unavoidable and should be made honestly rather than rhetorically. The openEuler project, covered on 2026-11-14, reached self-described community self-governance only after Huawei open-sourced it in 2019 and donated it to OpenAtom in November 2021 — code first, governance second, in a sequence that is the normal open-source order. Industry operating systems are being stood up in the reverse order: governance and standards first, product validation second, and whatever code commons eventually exists, third. That does not make the standardisation wrong; a control system has safety and liability properties that a server OS does not. But it does determine what kind of artefact will result. A reference architecture written by a committee and validated in a pilot formation of state-owned enterprises and testing labs produces, at the end, certified products — and the evidence for that prediction is not speculation, it is the pattern this series has now recorded across every instrument it has examined.
Fourth, the strongest counter-argument, and it is stronger than it first looks. The report’s division of labour is unusually well-specified, and Zhai Yanfen’s sentence about users “extending from requirement-raisers to co-builders and validators” is the only line in the entire Chinese industrial-software corpus this briefing has read that describes an inbound role for the user. Read at its most generous, the pilot formation is a formal channel through which a steel plant that has found a bug in a plant-floor operating system can put that finding into a validation process that the vendor must respond to. If the four technical directions are published as open specifications and the validation findings are published as issues in a public tracker, this is a commons institution wearing industrial clothing. That is a real possibility and it costs almost nothing to test. If they are published as committee documents and the validation happens behind enterprise confidentiality, it is a procurement-grade conformance programme. The instrument is identical in both cases; only publication distinguishes them.
Fifth, the falsifiable prediction. If, within twelve months, the industrial-operating-system work produces a public code repository with a contribution guide and merge rights held by someone outside the founding vendors, the committee-first ordering will have been a shortcut rather than a substitute. If it produces instead a conformance suite, a certification mark and a list of certified products — with the reference architecture available only as a purchased standard — then the ecosystem will have extended the pattern of the previous six weeks to the last large domain it had not yet covered. The observable marker is the cheapest one available: whether the reference architecture is on a public URL under an OSI-approved documentation licence, or behind a standards-purchase page.
Checkable questions.
(i) Is the working group’s membership list public, and does it contain any unit that is not a state-owned enterprise, a testing institute or a vendor selling an industrial operating system? A standards body whose members are exclusively suppliers and buyers will write a specification; a standards body with maintainers in it may write something else.
(ii) Will the reference architecture and the test-evaluation rules be published in full, in a freely accessible form, and under what terms? This is the whole difference between a coordination device and a gate, and it is decided at publication rather than at drafting.
(iii) Do the pilot validation findings become bug reports in a public tracker, or acceptance documents in a confidential file? The first is how a standard becomes a project; the second is how a standard becomes a procurement annex.
Sources: 上海市经济和信息化委员会 via 搜狐 — 智控新底座 制造新范式!2026工业操作系统大会在沪召开 (2026-11-13; 12–13 November at Zhangjiang Science Hall; CCID as host; 800-plus representatives; Wang Jiangping, Wang Guodong, Qian Feng, Liu Wenqiang, Qiu Wei; 30-plus central SOEs including Baowu, CNPC, Sinopec, Ansteel, Changan, FAW, Huadian; the formal launch of the Industrial Operating System Standards Working Group under MIIT’s Industrial Software Standardization Technical Committee; the four directions — reference architecture, interconnection, compatibility, test evaluation; the CCID report; the pilot validation formation across steel, non-ferrous mining, rail transit and nuclear power) · 证券时报 e公司 via 新浪财经 — 2026工业操作系统大会在上海举办 (2026-11-13; confirmation that the working group was formally launched in establishment, the report released, and the pilot formation started) · IT之家 — 工信部:将筹备成立工业操作系统标准工作组 (2026-11-14; CCTV News as source; MIIT statement on accelerating the working group; the report’s finding that industrial operating systems are moving from single-point exploration to systematic product supply; Zhai Yanfen of the China Software Testing Center on common foundation plus industry capability, firms converting industrial mechanisms into reusable components, users as co-builders and validators; the “AI + Software” plan targets for 2028; first-three-quarter software revenue ¥11,803.1bn, +8.4%, industrial software +7.3%; IDC’s ¥280bn and 52% CAGR to 2030) · 新浪财经 — 工业和信息化部,新成立一个重要委员会 (2026-10-22; the Industrial Software Standardization Technical Committee founded 21 October 2026 in Hangzhou, addressed by vice-minister Xiong Jijun, secretariat at the Fifth Electronics Research Institute) · Xinhua — “2026工业操作系统大会"即将召开 (the conference framing, the “AI + Manufacturing” action plan, and the 12–13 November programme) · This series’ 2026-08-10 briefing — OpenAtom’s Control Systems Open Source Community launch, the industrial-control layer
🏛️ Institutional Change — A Conference That Is a City, and a Foundation That Is Not: COSCon'26 in Hangzhou
1. The event, and the eleven-year fact in it. On 14–15 November 2026 the 11th China Open Source Conference (COSCon'26, 第十一届中国开源年会) opened at the Yun Valley Center (云谷中心) in Hangzhou’s West Lake district, on the theme “Open Source. The Next Generation” (开源下一代). It is the eleventh edition and the first held in Hangzhou; COSCon was founded by KAIYUANSHE (开源社) in 2015, and KAIYUANSHE’s own description of itself is the institutional datum of the day: “a volunteer-run, vendor-neutral, non-profit open source community founded in 2014, run on the principles of contribution, consensus, and co-governance” — the first OSI member in China, and, as the Europe–China Digital Commons Observatory’s November research note records after checking, an organisation for which no registered legal entity or annual inspection record was found.
2. Who convenes it, stated as an organisational chart. The official announcement of the conference lists, in this order: guided by (指导) the Hangzhou AI Open Source Ecosystem College (杭州市人工智能开源生态学院); hosted by (主办) KAIYUANSHE and Datawhale; jointly hosted by (联合主办) Trend News (潮新闻), Chuanbo Danao (传播大脑) and Yun Valley Center; supported by (支持) the Hangzhou Open Source AI Foundation (杭州市开源人工智能基金会), the West Lake District Party Committee Talent Office, the Hangzhou Zijingang Science and Technology City Management Committee, and the Zhejiang AI Academy. A volunteer community and a company are the host line; a municipal college is the guiding unit; a city district’s talent office and two foundations sit in the support line.
3. What was published. Three things, in the order the reports carry them. (a) The first batch of “open source mentors” (首批"开源导师”) was credentialed at a ceremony. (b) 《中国人工智能开源生态蓝皮书》— the China AI Open Source Ecosystem Blue Book — was released, jointly by Li Pan (李攀), dean of the Zhejiang AI Academy, Qian Weigang (钱伟刚), deputy head of Zhejiang Daily Press Group, and Han Bin (韩斌), standing committee member of the West Lake district party committee. The Blue Book’s stated finding is that China’s AI open-source ecosystem shows a distinct pattern of “application traction plus multi-party coordination” (应用牵引+多元协同) in which leading enterprises, local governments, universities and communities all play important roles, while several key bottlenecks remain; it is organised around five questions — what open source is, why it matters, what the current state is, what the problems are, and how to advance — and closes with coordinated development paths and policy recommendations addressed to government, enterprises, universities and social organisations. It also names six “typical cities” for AI open-source ecosystem development: Beijing, Shanghai, Shenzhen, Hangzhou, Suzhou and Zhuzhou (株洲) — Zhuzhou selected, per the Hunan report, for an “industrial-scenario-driven” (工业场景驱动) model. (c) The conference’s stated intent, in the organisers’ own words: “the next era of open source will be defined jointly by next-generation technology, talent, community and business.”
4. The policy stack behind the venue. Hangzhou’s claim, stated by Fan Jingjing (范晶晶), founder of Datawhale and a COSCon'26 co-chair: “Today no Chinese city is pushing AI and the open-source ecosystem as a core strategy the way Hangzhou is.” The instruments named: the national 15th Five-Year Plan listing the open-source ecosystem as a key path to an AI innovation highland; Hangzhou’s goal of being the “first city for AI innovation and development”, with the open-source ecosystem as one of three special actions, and the West Lake district as the main front; the city’s 《杭州市打造全国"AI+OPC"创业新高地若干政策措施》; the Hangzhou AI Open Source Ecosystem College, opened in the West Lake district in April 2026 as the country’s first such college, with a talent-integration practice base; and the college’s OPC nine-level talent capability standard (OPC九级人才能力标准(试行)), published 8 June 2026 as a graded evaluation framework for open-source talent. Hangzhou’s 15th Five-Year Plan, per a Hangzhou Daily round-table published 2026-10-08, sets the goal of building “the city with the strongest open-source ecosystem” (开源生态最优势城市) and positions a Hangzhou Open Source AI Foundation as the vehicle that closes the loop from model foundation → developer co-building → industry deployment.
5. The room, and the one sentence in it worth keeping. The main forum carried Angie Jones, VP of the Agentic AI Foundation (AAIF) — the Linux Foundation body to which Anthropic donated the Agent Skills standard in October 2026, and which with Linux Foundation Education launched the vendor-neutral MCPA (Model Context Protocol Associate) certification on 29 October — speaking on “the maintainer’s dilemma: ban AI, or prepare for it.” Wang Xu (王旭), vice-chair of Ant Group’s open source technical committee, spoke on “after AI Coding: exploring the next generation of open source technology.” KAIYUANSHE’s conference call, published in English, is the most institutionally literate document this series has read from a Chinese community organiser: “How does governance become more transparent? How does contributor experience become genuinely welcoming? How does a code of conduct actually get enforced rather than published?” and, on the next generation of communities: “not measured by size, but by health; not by how lively they look, but by how long they last.” R. Littauer of CURIOSS and SustainOSS (open source sustainability) and M. Behling of FOSSASIA were on the international round table. And Lin Lüqiang (林旅强), KAIYUANSHE co-founder and chair, put the day’s own framing on the record: “we are all ’the next generation of open source’ — no matter how old we are, no matter where we stand today.”
6. The stated numbers from the previous edition, for scale. COSCon'25 in Beijing: 1,500-plus on site over two days, 86,946 online views, 21 communities co-creating 24 tracks, 54 community and media partners, 230 volunteer applications becoming a crew of 85; and across eleven editions, a keynote track of 190 sessions with 136 speakers.
Institutional Analysis & Commentary:
Two events, one day, and the contrast is not between openness and closure. It is between two institutional technologies that China has now built to a very high standard, and the fact that neither of them is a commons institution.
First, name what COSCon is, precisely, because the temptation is to be nostalgic about it. It is the longest-running independent open-source gathering in China — eleven editions, a volunteer crew, a 2014 founding, OSI membership, an English-language call with a visa paragraph for foreign speakers and an honest warning that the volunteer crew works in Mandarin. That is rarer than it sounds and it deserves credit. But read the masthead. Guiding unit: a municipal college. Hosts: a volunteer community and a company. Support: a city-district talent office, a district-level science-city committee, a provincial academy, and two foundations whose names differ from the national foundation by two characters. The conference is not a community meeting that a city happens to sponsor; it is a city instrument that a community convenes. KAIYUANSHE supplies what a city cannot manufacture — eleven years of standing, a neutral brand, an international address book, a reputation for being the one vendor-neutral room in the country — and the city supplies what a volunteer community structurally cannot: a venue, an agenda inside a five-year plan, a talent policy, and the ability to make a district’s ambitions legible to a national audience. That is a good trade for both sides, and it is still a trade: what the community gains in reach it pays for in the loss of the only thing that made it unusual — that it answered to nobody.
Second, the harder observation, and it is about the Blue Book rather than the conference. The book released in Hangzhou asks five questions “what open source is, why it matters, what the current state is, what the problems are, and how to advance” and answers them for government, enterprises, universities and social organisations. Look carefully at that list of addressees. It contains every party in the ecosystem except the one the ecosystem’s own governance depends on: the individual contributor with no institutional standing. A framework whose every recommendation is addressed to an organisation is a framework that can describe an ecosystem and cannot describe a commons, because a commons is precisely the arrangement in which an individual’s contribution is recognised without being routed through an organisation. This is not a defect invented in Hangzhou. It is the same shape as every instrument this series has examined: the export register addresses firms, the certificate addresses candidates as purchasable credentials, the certification-to-procurement loop addresses suppliers, and the university programme addresses students as users. The Blue Book completes the pattern at the level of the ecosystem’s own self-description: China’s AI open-source ecosystem, described comprehensively, in English and Chinese, by the people best placed to describe it, has no chapter for the person who is not an organisation. The pattern is application traction + multi-party coordination — and multi-party here means multiple parties, i.e. organisations.
Third, the six city list is the most quotable and least examined line in the document, and it is worth thinking about what a “typical city” is. Beijing, Shanghai, Shenzhen, Hangzhou, Suzhou, Zhuzhou. Five of the six are the obvious candidates; Zhuzhou, a prefecture-level industrial city in Hunan with no national model lab and no major open-source foundation, is selected for “industrial-scenario-driven” development. The category is worth naming: a “typical city” is not a measurement, it is an administrative form. It converts a set of heterogeneous local practices into a ranked and copyable template, which is exactly what a five-year-plan implementation system needs and exactly what the open-source way has historically refused to provide. Compare it with how open source actually spread: not by designating typical cities but by making it cheap for anyone, anywhere, to run the thing on their own machine — the diffusion mechanism is the licence and the repository, not the demonstration zone. A city that becomes typical because a book says so has gained an audience; it has not gained a maintainer. The prediction that follows is testable and slightly uncomfortable: the six cities will be cited in local policy documents within twelve months, and the number of maintainers in them will not measurably change.
Fourth, credit where it is due, and this is the part a critique that is only a critique would miss. The conference’s own call for participation is better written on the subject of governance than anything in the Blue Book: how does a code of conduct actually get enforced rather than published is the correct question, asked in public, in English, by the organiser who would be the one obliged to answer it. Inviting Angie Jones to speak on the maintainer’s dilemma, one month after the AAIF absorbed Anthropic’s Agent Skills standard and launched a vendor-neutral MCP certification, is not decoration — it puts a working model of foundation-governed AI standards in the room on the same day the city announces a talent certification standard of its own. And the presence of Sustainer/CURIOSS people, whose entire subject is the money problem of open-source maintenance, is the closest any Chinese conference in this series has come to programming the unpriced half. A conference is not an institution, but the list of who is on the stage is a statement about what the organisers believe is real.
Fifth, the day’s thesis, and it is the one this series has been building toward since November began. Yesterday’s briefings documented an ecosystem that pays for every outbound transaction and has no budget line for the inbound one. Today shows that the description of the ecosystem has the same asymmetry as the ecosystem. A conference guided by a college, hosted by a community, supported by a district, and a standards working group convened by a ministry with thirty state-owned enterprises — both are superb coordination devices, and neither has written down who answers a stranger. The distance between them is the distance between a country that has learned to organise open source and a country that has learned to be one. The second is not harder to organise; it is smaller, cheaper and requires only that somewhere in an eleven-year conference programme, a ministry reference architecture, a blue book’s five questions and a city’s nine-level talent standard, somebody be named as the person who reads the patch.
Checkable questions.
(i) Who is on the roster of Hangzhou’s first batch of “open source mentors,” and what did they do to be credentialed? A mentor credential is the cheapest possible commons institution — it names individuals and confers standing — and it is the only instrument in today’s material that could, in principle, do the work nothing else does. Its selection criteria are therefore the most important unpublished document in the city.
(ii) Is the Blue Book’s full text published, and does any chapter address individual contributors, maintainers, licences in practice, or the funding of maintenance? The five-question structure suggests not. If the full text does contain such a chapter, this entry’s central claim needs revising, and the revision is a search away.
(iii) Does the Hangzhou Open Source AI Foundation have a published legal name, registration record, governing documents and grant criteria? A foundation that closes the loop from model to deployment is an institution; a name in a press release is an aspiration. The answer is in the civil-affairs registry, and no source this briefing found has looked.
Sources: 潮新闻 via 搜狐 — 第十一届中国开源年会首次在杭州举行:共同探讨"开源下一代” (2026-11-14; COSCon'26 at Yun Valley, Hangzhou, 14–15 November; guided by the Hangzhou AI Open Source Ecosystem College, hosted by KAIYUANSHE and Datawhale, jointly hosted by Trend News, Chuanbo Danao and Yun Valley, supported by the Hangzhou Open Source AI Foundation, the West Lake talent office, the Zijingang Science and Technology City committee and the Zhejiang AI Academy; the first “open source mentor” ceremony; the Blue Book released by Li Pan, Qian Weigang and Han Bin; the “application traction plus multi-party coordination” finding and the five questions; Lin Lüqiang’s remarks; Fan Jingjing on the 15th Five-Year Plan, the three special actions, the district’s AI+OPC policies and the April 2026 college and 8 June 2026 OPC nine-level standard; Angie Jones and Wang Xu keynote topics) · 潮新闻 via 凤凰网大风号 — 中国开源年会首次落地杭州:一座城市的生态共建与长期主义 (2026-11-14; the Qian Weigang speech; the youth forum with 20-plus child speakers and the youngest at 9; Angie Jones on the maintainer’s dilemma; ZHUANG Biaowei on the snowball; the college’s dean Xu Xuchu on “no walls” and “ecosystem first, students second”; the Blue Book’s stated aim of moving China from participant to ecosystem builder and rule co-builder) · 搜狐 — 共话开源下一代!COSCon'26 第十一届中国开源年会主论坛议程与嘉宾阵容重磅官宣! (2026-11-12; the full speaker roster: Lin Lüqiang, Fan Jingjing, Li Mingkang, Wang Xu, Xu Shiwei, Angie Jones, Kevin Xu of GitHub, Silvano Luciani of Kong, Mario Behling of FOSSASIA, Richard Littauer of CURIOSS/SustainOSS, Jiang Tao of CSDN and AtomGit, Michael Yuan of WasmEdge, Xu Xuchu, Wang Wei of ECNU with the preview China Open Source Annual Report 2026, Li Meng of ISCAS Open Summer, Zhuang Biaowei of Tiangong Kaiwu as moderator; high-school and middle-school speakers; the four sub-themes) · 开源社 KAIYUANSHE — COSCon'26 theme and call for co-produced events (2026-10-12; the four sub-themes in English; “How does a code of conduct actually get enforced rather than published?”; “not measured by size, but by health”; the volunteer crew works in Mandarin; visa and invitation-letter guidance for foreign speakers; COSCon'25’s 1,500-plus attendance, 86,946 online views, 21 communities, 24 tracks, 54 partners and 85 volunteers; 190 keynote sessions and 136 speakers across eleven editions; KAIYUANSHE as first Chinese OSI member, founded 2014, volunteer-run and vendor-neutral) · 华声在线 via 腾讯新闻 — 株洲入选中国AI开源生态发展典型城市 (2026-11-14; the Blue Book’s “typical cities”: Beijing, Shanghai, Shenzhen, Hangzhou, Suzhou and Zhuzhou, with Zhuzhou noted for an industrial-scenario-driven model) · 每日商报 via 杭州网 — 何以开源?何以杭州? (2026-10-08; Hangzhou’s 15th Five-Year Plan goal of being “the city with the strongest open-source ecosystem”; the ¥5 million GOAI prize; the Hangzhou Open Source AI Foundation’s stated role closing the model–developer–deployment loop; HDU’s open-source college executive dean Lin Fei on the three levels of learning, practice and co-building, and on open source entering the credit system) · Europe–China Digital Commons Observatory — Chinese open-source foundations and OpenAtom finances (research note 2026-11-03, updated 2026-11-08; KAIYUANSHE “founded 2014, volunteer community with an elected board… we found no registered legal entity or 年检 record”; FY2021 income CNY 604,062, of which public donations CNY 43,000 and COSCon tickets CNY 5,062; OpenAtom’s CNY 600m-plus cumulative figure and the absence of an audited breakdown) · LFOSSA — Just Launched: Model Context Protocol Associate (MCPA) Certification (2026-10-29, updated 2026-11-11; AAIF and Linux Foundation Education; Angie Jones quoted as AAIF VP of Developer Experience) · This series’ 2026-08-10 briefing — KAIYUANSHE, COSCon and the community form versus the state-chartered codebase form
🔍 Commentary — Who Pays for the Unpriced Half
First, today’s two events are one finding seen from a ministry rostrum and a city stage, and neither of them is a failure. In Shanghai, an industrial operating system standard working group was launched with four technical directions, a CCID report, a pilot validation formation across steel, mining, rail and nuclear power, and 800 people including thirty state-owned enterprises. In Hangzhou, an eleven-year-old volunteer conference opened under a municipal college’s guidance, released a blue book addressed to government, enterprises, universities and social organisations, credentialed a first batch of open-source mentors, and named six typical cities. Both are competent, both are funded, both were built by people who understand exactly what they are doing. That is the uncomfortable part: the asymmetry this series has documented for six weeks is not the result of anyone’s incompetence, and it is not going to be fixed by better people doing the existing jobs.
Second, the mechanism is now visible enough to state in one sentence, and it is a mechanism about instruments rather than intentions. Every institution that appeared today fits a transaction that already has a buyer, a specification, an acceptance test and a unit of delivery: a standardisation committee convenes because a regulator must publish conformity requirements; a city supports a conference because it has a five-year plan and a talent policy; a blue book is written because a foundation and an academy need a deliverable; a mentor credential is issued because a ceremony is scheduled. Not one of these institutions has a buyer for the thing a commons actually needs, which is somebody’s attention on Sunday night to a patch from a person they have never met. The instruments are not wrong. They are simply all instruments for transactions that exist.
Third, and this is the correction this week’s run of briefings has been working toward: the claim should be weaker and more precise than “China’s open source lacks governance.” It has governance in abundance — more committees, standards, certifications, definitions, blue books, city designations and talent ladders per unit of code than any other open-source ecosystem on earth. What it lacks is a governance document in which a named individual has the standing to say no to a change, and a budget line that pays them. The two events today make this visible from opposite ends: the standards working group is a governance instrument with no repository, and the conference is a repository-adjacent gathering with no committee. Neither is the missing piece. The missing piece is the seam between them, and the seam is one paragraph long: a public repository, an OSI-approved licence, a contribution guide, and a maintainer list that includes someone who was not on the founding team.
Fourth, the sentence to keep, and it comes from Hangzhou rather than from the analysts. KAIYUANSHE’s own call for COSCon'26 asks, in public, “how does a code of conduct actually get enforced rather than published?” That question is the whole of this briefing compressed into nine words. A code of conduct that is enforced is the smallest possible commons institution: a written rule, a named body, a process, a sanction, and a community that can observe all four. Every instrument examined this fortnight — the export register, the certification fee, the curriculum, the distribution, the reference architecture, the blue book — is published and not enforced, because enforcement requires an enforcer, and an enforcer requires a budget line. The ecosystem has solved the problem of publishing rules at scale. The next thing it has to publish is a name.
Sources: 上海市经信委 via 搜狐 — 2026工业操作系统大会 (2026-11-13; the standards working group’s four directions and the pilot validation formation) · IT之家 — 工信部:将筹备成立工业操作系统标准工作组 (2026-11-14; Zhai Yanfen on users becoming co-builders and validators; the “AI + Software” 2028 targets) · 潮新闻 via 搜狐 — 第十一届中国开源年会首次在杭州举行 (2026-11-14; the guiding/hosting/support structure, the Blue Book’s five questions and its addressees, the first open-source mentor ceremony) · 开源社 — COSCon'26 call for participation (“How does a code of conduct actually get enforced rather than published?”; “not measured by size, but by health”) · Europe–China Digital Commons Observatory — Chinese open-source foundations (KAIYUANSHE’s absent legal entity; OpenAtom’s unaudited CNY 600m) · This series’ 2026-11-14 briefing — two curricula, one question · This series’ 2026-11-13 briefing — the certification fee and the missing maintainer rung · This series’ 2026-11-11 briefing — the instrument that fits the transaction · This series’ 2026-11-12 briefing — the safe harbor and the donated upstream