Greetings from a world where…
cicada killer wasps are still back
…As always, the searchable archive of all past issues is here. Please please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay support access for all AND compensation for awesome ChinAI contributors).
Feature Translation: The “Guangdong Faction” in the AI Wave
Context: In Caixian Chen’s latest piece (link to original article on Sinpers.ai WeChat public account), she asks two related questions:
Why is it that the “Guangdong Faction” [广东帮] has produced so many amazing AI talents? This comes in the wake of Kimi K3, for which the two co-lead authors were Jianlin Su and Guangyu Chen, both from Guangdong province. In fact, Moonshot AI co-founder Yang Zhilin and DeepSeek CEO Liang Wenfeng are also from Guangdong province.
And, if it’s the case that Guangdong is such a hub of AI talent, why is there no star AI company rooted in Shenzhen, the province’s tech center? The fact of the matter is that China’s most prominent frontier AI companies are concentrated in Beijing and Shanghai. Recall in ChinAI #340, I pointed out how much local Shanghai papers took pride in MiniMax (based in the city).
Key Takeaways: Before we review Chen’s answers to these two questions, this article stood out to me as a reminder of the intense regional rivalries that shape China’s high-tech ecosystems.
Chinese economist Zhang Jun argues, “These cities’ competition to generate stronger growth than their counterparts – a contest sustained by the political incentives the central government has long provided to local officials – has played a driving role in China’s rapid industrialization and ongoing structural transformation.”
Rumor has it, from Chen’s piece, Shenzhen authorities had talks with two of the “Six Little Tigers” of China’s LLM start-ups (Zhipu AI, Moonshot AI, MiniMax, Baichuan Intelligence, StepFun, and 01.AI) to relocate their headquarters to Shenzhen. The proposed terms including a free office building, but the deal fell through. “In hindsight, this wasn’t entirely a bad thing for Shenzhen: one of those companies later dropped out of the race to build the most intelligence models, while the other was highly unlikely to ever leave Beijing,” Chen muses.
So, why has the large model wave seemed to have abandoned Shenzhen? Chen’s initial answer is that there is at least one company from Shenzhen who is seeking to compete at the AI frontier: Tencent.
Tencent’s WorkBuddy outranks ByteDance’s TraeIDE as the top AI desktop agent in China. Still, Tencent’s technical achievements do not place it in the top tier of Chinese AI labs. The article notes, “Tencent has still failed to live up to the expectations placed upon it by southern China.”
Someone who takes pride in being from Shenzhen might look at DeepSeek’s success through the lens of regional rivalry: “In 2025, the explosive rise of DeepSeek shifted the industry's gaze from Beijing and Shanghai to a new focal point: Hangzhou. Although this history was made by a native of Guangdong, the resulting dividends did not flow back to the province, not to Shenzhen and even less to Guangzhou.”
In the end, Chen is hopeful that Guangdong can produce a local AI champion. “Yet, Guangdong needs a top-tier enterprise, one that can attract AGI resources while operating in a pragmatic, down-to-earth manner. There is a need for a destination to channel the talent across the Guangdong-Hong Kong-Macao-Greater Bay Area for Guangdong natives who studied outside the province, and for local capital,” she writes.
All the ingredients are there — the pragmatic research mindset, the attractive location, the access to global capital and talent. And one more cultural ingredient, according to Chen: “If Beijing embodies a ‘political mindset,’ then Guangdong embodies a ‘business mindset’…The core of the political mindset is control, whereas the core of the business mindset is freedom; one focuses on power, the other on growth.”
If no local champion emerges, could Shenzhen’s competitiveness decline in the AGI era? Chen is confident that either Tencent or an up-and-coming force will “capture the region’s full potential and ultimately secure a proper foothold in the AGI landscape.”
FULL TRANSLATION: The “Guangdong Faction” in the AI Wave
ChinAI Links (Four to Forward)
Must-read: A fatal reaction
If you can only read one article this week, make it this longform essay by Brendan Borrell, co-published by Science and Retraction Watch. It tells a tale of an experimental gene-editing trial that went deathly wrong at a Shanghai hospital, leading to the death of “Mei,” a six year old girl. She was the daughter of “Jason” and “Linda”: the parents “decided to tell her story now because they are angry about what they feel is a lack of accountability by the researchers and the institutions.”
Zilong Qiu, a neuroscientist at Shanghai Jiao Tong University School of Medicine, enrolled “Mei”, in a sketchy clinical trial for brain-directed gene-editing therapy treatment. The article reports:
Seven experts in fields including genetics, virology, and bioethics who reviewed details of the Nature study and the clinical trial for Science and Retraction Watch expressed concern that Qiu and his team downplayed the trial’s risks in describing them to the parents, overlooked safety signals in animal studies, and proceeded even though success was unlikely. “This shouldn’t have gone to trial,” says Steven Gray of the University of Texas Southwestern Medical Center, who develops viruses for gene therapy.
Ironically, Qiu was one of the Chinese scientists who condemned He Jiankui’s rogue gene editing experiment in an open letter.
Qiu’s behavior gets even more shameless. He goes forward with the publication of a Nature study even after the trial leads to a tragic death: “The family’s genetic data had been removed and a sentence thanking them for their ‘participation and support’ had been eliminated.” The article details so many failings in the peer review process that allowed the paper to get to this stage. After this Nature paper gets an enthusiastic reception, Jason and Linda decide to escalate their complaints.
I’ve only recently started to see Chinese media cover this case after Borrell’s piece came out. Mei died in March 2025, so no reckoning process even started until a year and a half later:
The muted response to Mei’s death sharply contrasts with the Gelsinger tragedy (Jeff’s note: a U.S. gene therapy trial that led to the death of an 18-year-old) decades ago. His death drew widespread headlines and prompted a major self-examination by the gene therapy field. The university where the trial was conducted paid a fine of more than a half-million dollars and had further gene therapies suspended. FDA barred Wilson, the principal investigator, from human research studies for 5 years for violations that included failing to disclose the adverse reactions observed in the primate toxicology studies on Gelsinger’s informed consent form. “Public disclosure … is essential in safely moving these fields forward,” Wilson says.
Should-read: China’s Great City Rivalries
In this Project Syndicate commentary, Chinese economist Zhang Jun distills his research on the positive effects of inter-urban competition on high-tech development. On the flip side, Zhang admits, “Of course, such competition carries risks – in particular, short-sighted efforts to boost growth in ways that exacerbate misallocation of resources, overcapacity, and high financial leverage.”
Should-read: How Russian Economic Policy is Gender
There are many reasons no one is talking about Russia’s AI prowess. Is it possible that one reason is the Russian political elite’s deeply gendered conception of a “real” economy? For The Duck of Minerva, University of Groningen assistant professor Lisa Gaufman explores this question.
Should-read: Silicon Valley’s Other China Problem: It’s Training Their AI
In Forbes, Anna Tong reports on the Silicon Valley startups that are selling data to Chinese AI labs. But wait! I thought China was the Saudi Arabia of data, and that’s why China would dominate everyone in AI, so why would these labs need to buy a bunch of high-quality English data…… Could it be that this analogy was always mistaken and misleading?
Thank you for reading and engaging.
These are Jeff Ding’s (sometimes) weekly translations of Chinese-language musings on AI and related topics. Jeff is an Assistant Professor of Political Science at George Washington University.
Check out the archive of all past issues here & please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay for a subscription will support access for all).
Any suggestions or feedback? Let me know at chinainewsletter@gmail.com or on Twitter at @jjding99
Is part of the answer just that Guangdong is too good at hardware and supply-chain execution? A lot of AI value there may get swallowed into phones, robots, appliances, cars, cameras... which makes it harder to produce one clean “star AI company” story.
the byd navsim paper from this week almost answers the question backwards. shenzhen quietly built an ai champion, just not a chatbot company, a carmaker with its own chip and its own model stack. doesn't fit the llm-lab definition of a guangdong star but it's the same regional bet, just applied to a vertical instead of a frontier lab.