
Tech Reporter @restofworld. Former Beijing-based Correspondent @scmpnews. Alum @jmschku. Tips and story ideas : [email protected]
Joined June 2009
- Tweets1.9K
- Following3.5K
- Followers10.9K
- Likes1.5K
Kinling Lo 盧建靈 retweeted
This piece by someone who helped lead US talks with China on AI under Biden is well worth reading.
It shows the low level of concern that Chinese officials had for AI safety issues at the time.
But now things have changed. AI is far more powerful and difficult to control. China takes the risks much more seriously.
While the Biden-era talks offer valuable lessons, we can only extrapolate so far from that experience.
I believe China is willing to have a more serious discussion this time around, even if it’s out of their own self-interest.
The piece by Seth Center quoted here: nytimes.com/2026/09/16/opini…
Kinling Lo 盧建靈 retweeted
SCOOP: Chinese-founded AI startup Manus is set to double its valuation to $4 billion in its first fundraising since Beijing ordered it to split from Meta bloomberg.com/news/articles/… w/ @Hazeology
Kinling Lo 盧建靈 retweeted
Just in:
TSMC has finally chosen to adopt High-NA EUV. And to develop larger photomasks to go with it.
This is HUGE news for @ASMLcompany
Kinling Lo 盧建靈 retweeted
Shanghai is emerging as a favored destination for patients from America and other rich countries who are seeking innovations in biotechnology. on.wsj.com/4ytK5hf
Kinling Lo 盧建靈 retweeted
A Taiwanese engineer thought he'd landed a dream job at an American AI-chip startup. His coworkers were all in Nanjing. His paychecks came from Beijing.
He never met a single American colleague. Every meeting, every line of code, every chip design went straight to a China-based team. Taiwan raided the company in 2021. It wasn't American at all, a Chinese firm hiding behind a nearly identical name and a fake ownership structure.
That's one case out of 166. Exclusive government data given to Rest of World shows Taiwan has investigated 166 companies over six years for hiding Chinese ties while poaching chip talent, plus 67 separate trade-secret theft cases. Xiaomi. SMIC, China's top chip foundry, registered as a Samoan company to get in. OnePlus's founder has an active arrest warrant for the same scheme.
The playbook repeats: pay Taiwanese engineers up to 10 times local salary, register as a foreign entity to dodge Taiwan's investment law, deny everything if caught.
restofworld.org/2026/taiwan-…
Kinling Lo 盧建靈 retweeted
A lot of robotics researchers/model developers on this site are acting as though the @FCC just ended America’s ability to build the best brains for robots. That’s categorically FALSE.
Yes, the FCC blocked new equipment authorizations for Covered List robots. But FCC rules have an import exemption in 47 CFR § 2.1204(a)(3) that allows up to 4,000 units of a given model to be imported for testing, evaluation, or product development (though the FCC has proposed lowering that to 40 for Covered List equipment).
And before everyone asks: “product development” isn’t some magic loophole. It means bona fide pre-commercial R&D—training robotics foundation models, benchmarking, collecting data, developing software, testing new autonomy stacks, safety systems, integrations, etc. It does not mean importing a bunch of robots and calling your commercial deployment “research.”
So CMU, Berkeley, Stanford, a frontier lab, etc. can still import a limited research fleet and use foreign robot bodies to build better American robot brains. What foreign producers lose is the ability to take new models through the ordinary FCC authorization process and scale them commercially in the United States.
This is a restriction on commercial deployment, not a ban on American robotics-model development. That distinction is a good thing but has been completely lost in the discourse.
NEW
The FCC has now added two new categories of devices to our Covered List, which bans new versions from import or sale in America.
1. Advanced robotic devices, such as humanoids and quadrupeds produced in foreign countries.
2. Power inverters produced in foreign countries.
This action follows determinations by Executive Branch nat sec agencies that the devices pose unacceptable risks to our national security.
It includes exemptions for devices that, based on a DOW or DHS finding, pose no unacceptable threat.
China's fast-moving market has found a solution to the surging demand for material to be used in AI-generated production ranging from ads to films - buying rights to people's faces at $15 to $700.
Read my latest on a fascinating development that emerged from China's AI boom -
A growing number of platforms are paying people to license their likeness for AI-generated dramas and ads, creating a new marketplace for biometric identity restofworld.org/2026/china-a…
Kinling Lo 盧建靈 retweeted
张一鸣目前把 50% 的时间都投入到了 Seed 上。这是徐新在播客里提到的。
大部分公司发展到一定规模后,创始人都会逐渐从业务一线退出来,开始管理战略、资本、组织和文化。这没有问题,也是大公司的必经阶段。
但张一鸣选择了一条不同的路。即使字节已经是一家拥有巨大商业规模的公司,他依然把大量时间投入到 AI 最前沿的技术探索中。这可能也是字节未来能否赢得大模型竞争的关键因素之一。
因为大模型很多时候像炼丹。今天押注什么方向,采用什么训练方法,如何组织最顶尖的人才,这些选择都会直接决定公司的上限。
所以,创始人需要持续阅读论文,和最前沿的研究者交流,理解那些还没有被市场验证的方向。
徐新说,公司的系统、流程、人员都可以逐渐复制和替代,但创始人的大脑,以及他调动资源的能力,没有人能够替代。
Anything with @violazhouyi’s byline is a must read, in case you haven’t figured that out.
I wrote about the Chinese math kids turned AI superstars, and what it’s like meeting my much smarter (and richer) classmates in Silicon Valley.
restofworld.org/2026/chinese… via @restofworld
China's leading AI assistant apps according to QuestMobile. Note, just the persistent strength of ByteDance's Doubao.
Ahead of its expected payment plans, will be interesting to see how this affects the usage base. We are likely to see a new Seed model soon to justify the new plans.
The China watcher sphere on X is currently dominated by geopolitics, tech dominance, he green transition, and macro finance. But framing issues exclusively through the lens of "great power competition" masks the intense, complex social changes happening on the ground.
A recent viral controversy in Shenzhen perfectly illustrates what we are missing. (china.caixin.com/2026-04-28/…)
Last week, a man and a woman got into a physical altercation at a bus stop after she tried to stop him from smoking. Both were detained and refused to settle. In response, police subjected the woman to a forced strip search—sparking massive public outrage regarding law enforcement overreach.
But the controversy itself isn't what caught my attention. It’s the woman at the center of it: 29-year-old Wang Ronghao.
It turns out Wang is a meticulous, active chronicler of the grueling service industry. Over the past 9 years, she has worked at Haidilao, Hema Fresh, Lawson, and various other gig economy jobs, documenting the crushing reality of modern labor. Reading her diaries, you realize that the mechanisms of control over labor have evolved right alongside economic development, taking on terrifying new forms. (A report from 2025: news.qq.com/rain/a/20260415A…)
Her writings expose the invisible, often dehumanizing architecture required to maintain "extreme service":
Algorithmic Panic: At Hema grocery, strict "30-minute fresh delivery" timers force kitchen workers to literally reach into 100°C steamers without gloves to grab RFID tags just to avoid algorithmic penalization, leaving them covered in burns.
Panopticon Management: At Haidilao hotpot, a relentless reporting culture and surveillance enforce "16 zero-tolerance rules." If a customer pours their own water before a server can sprint over to do it, the server faces docked pay and public reprimand.
Systemic Overwork: A total deprivation of basic human needs. Convenience store clerks are forced to stand for 10+ hour shifts even when the store is empty, and workers battle for corners on the street just to sleep for 20 minutes.
I don’t want to frame this merely as a "China labor story." It is clear that we are seeing the exact same algorithmic exhaustion in Amazon warehouses and among Uber drivers in the U.S. It could be any country.
The very "substrate" of society is shifting globally. In both countries, and in many others, the driving forces are identical: the relentless optimization of technology and the unchecked leverage of large, powerful companies.
It is not a question of which country is "better" or who is winning a geopolitical rivalry. Beneath the high-level macro narratives and great power posturing, this is plainly an age-old labor struggle, evolving alongside technology, that never actually left.
Kinling Lo 盧建靈 retweeted
Here y'all go!
interconnect.substack.com/p/…
Kinling Lo 盧建靈 retweeted
For the first time ever, the “Fab 5” of China smart EV startups NIO Xpeng Li Auto Xiaomi* and LeapMotor “WeiXiaoLiXiaoLing” were all profitable in the same quarter.
Li Auto and LeapMotor as well as Xiaomi’s Smart EV AI and Other New Initiatives were profitable for FY25. NIO and Xpeng trying to get there for FY26.
Let that sink in a bit.
*Xiaomi with an asterisk as it is already an established brand in smart phones and electronics while others are pure plays.
Kinling Lo 盧建靈 retweeted
Rumor: TSMC Arizona’s Fab 4 is already fully booked despite they haven’t yet broken ground, as Apple, Nvidia, AMD and Qualcomm seek more USA production for supply chain security, media report, adding TSMC projects in the US, Japan and Germany could account for 20% of TSMC’s production by 2028. $TSM $NVDA $AAPL $AMD $QCOM #Semiconductors #Arizona #semicondutor money.udn.com/money/story/56…
Incredible, lol, robot maker Unitree just filed for IPO and not only do they make money, their adjusted net margin is 35%, putting it on par with software companies. Humanoid fever is only going to increase from here I think :) At @TechBuzzChina we have some new special projects in this area we are announcing soon!
Some stats:
- Unitree’s STAR Market (Shanghai) IPO has been accepted, with a planned raise of RMB 4.2 billion (US$611 million) and an implied initial post-money valuation of at least RMB 42 billion (US$6.1 billion)
- 2025 revenue reached RMB 1.71 billion (US$248 million), - up 335% YoY, while adjusted net profit exceeded RMB 600 million (US$87 million), up 674% YoY.
- In the first 9 months of 25, humanoid robot revenue reached RMB 595 million (US$86 million), surpassing quadruped robot revenue of RMB 488 million (US$71 million) for the first time.
- Unitree shipped over 5,500 units last year, occupying 32.4% of the global humanoid market
- Of the IPO proceeds, the biggest chunk, RMB 2.02 billion (US$294 million), will go toward robot model R&D, followed by RMB 1.11 billion (US$161 million) for robot body R&D. Another RMB 445 million (US$65 million) is earmarked for new product development and RMB 624 million (US$91 million) for a manufacturing base.
The “genius girl” who previously worked at DeepSeek and was recruited by Lei Jun for Xiaomi AI is now on Twitter as well. It feels like more Chinese AI talent is realizing they can come here, speak for themselves, and build influence directly. I’m all for the added interaction and transparency.
MiMo-V2-Pro & Omni & TTS is out. Our first full-stack model family built truly for the Agent era.
I call this a quiet ambush — not because we planned it, but because the shift from Chat to Agent paradigm happened so fast, even we barely believed it. Somewhere in between was a process that was thrilling, painful, and fascinating all at once.
The 1T base model started training months ago. The original goal was long-context reasoning efficiency. Hybrid Attention carries real innovation, without overreaching — and it turns out to be exactly the right foundation for the Agent era. 1M context window. MTP inference for ultra-low latency and cost. These architectural decisions weren't trendy. They were a structural advantage we built before we needed it.
What changed everything was experiencing a complex agentic scaffold — what I'd call orchestrated Context — for the first time. I was shocked on day one. I tried to convince the team to use it. That didn't work. So I gave a hard mandate: anyone on MiMo Team with fewer than 100 conversations tomorrow can quit. It worked. Once the team's imagination was ignited by what agentic systems could do, that imagination converted directly into research velocity.
People ask why we move so fast. I saw it firsthand building DeepSeek R1. My honest summary:
— Backbone and Infra research has long cycles. You need strategic conviction a year before it pays off.
— Posttrain agility is a different muscle: product intuition driving evaluation, iteration cycles compressed, paradigm shifts caught early.
— And the constant: curiosity, sharp technical instinct, decisive execution, full commitment — and something that's easy to underestimate: a genuine love for the world you're building for.
We will open-source — when the models are stable enough to deserve it.
From Beijing, very late, not quite awake.
Kinling Lo 盧建靈 retweeted
No joke, I got this on the way out of the bathroom. Jensen Huang happened to walk by with security and asked if he’d answered my earlier question — I said I had one more. “You said you got China purchase orders, that means you got the green light from both sides?” He said yes.
SCOOP: Nvidia has received purchase orders from China for its H200 chips and is restarting manufacturing, CEO Jensen Huang tells me. Huang says Nvidia now has clearance from BOTH the U.S. and China. cnbc.com/id/108279192 $NVDA
Kinling Lo 盧建靈 retweeted
Photo from Shenzhen: huge crowd of Chinese people (lots of grannies!) lining up to get help installing OpenClaw.
One thing about tech diffusion in China that I feel is underdiscussed and that I’ll admit I don’t fully understand, is how open people of all ages are to jump into new tech. Feels very different from the AI suspicion/resistance you see in the U.S.
Similar with mobile payments and the shift to cashless. Street vendors in the lowest tier cities setting up WeChat Pay and Alipay QR codes almost overnight and Chinese grannies happily using payment apps with no problem at all. And yes that kind of grassroots adoption helped mobile payments scale extremely fast and allowed China to basically skip the credit card phase.
My conjecture is that if something similar happens with AI tools the speed of AI diffusion in China could look very different from what we see in other countries, which obviously would have major implications...
Kinling Lo 盧建靈 retweeted
Interesting - Chinese Premier Li Qiang posing for a pic with Merz + German and Chinese academics - including @m_huotari from Merics, which is still officially sanctioned by Beijing.
Längerfristige Dialogformate mit Klartext sind wichtig für chinesisch-deutschen Beziehungen.
Gut, daß das Deutsch-Chinesisches Dialogforum gestern getagt hat, in der sicheren Hand von @SabineKellerer als Ko-Vorsitzende und mit Klartext-Teilnehmern wie @oertel_janka & @m_huotari.







































