China Desk

China News, Summarized 09 Oct 2026, 20:47 UTC 34 stories 22/23 sources

Today's brief

Beijing halved its own hybrid exports and refused to put the number in writing

Šefčovič got his cap after all

Brussels got the thing it was told it would not get. After two days with commerce minister Wang Wentao in Beijing, EU trade commissioner Maroš Šefčovič said the two sides have a shared understanding to halve Chinese hybrid and plug-in hybrid exports to the bloc — "several million cars" over four years, per Reuters via CNBC. He would not say how it is enforced. "There are different techniques."

Beijing's own text has no number in it. The commerce ministry published a joint statement and a 16-point outcomes list saying only that intensive consultations produced an understanding on hybrid trade "in a manner consistent with WTO rules," via IT Home. A volume cap China spent three months calling WTO-illegal is now a consensus item with no volume in it. Brussels gets to say "half"; Beijing gets to say nothing.

What China took home is in the same list. A green channel for rare-earth and permanent-magnet export licences to Europe, EU facilitation of dual-use licences going the other way, continued talks on cutting tariffs on some goods, and the EV price-undertaking review grinding on. Third meeting March 2027, ministerial video call in January.

I called this wrong yesterday. I wrote that Beijing had decided Europe was bluffing and that I was less sure than it was. Beijing moved instead — against a deficit the EU puts at over €1bn ($1.12bn) a day, and with Porsche announcing 5,000 more job cuts, 9,000 cumulative, a return to combustion engines and 20% price rises, blaming China sales, per Leiphone's wire. EU leaders still have to endorse it on 15–16 October.

Go deeper on this section: ClaudeChatGPT

The State Council outlawed involution by decree

The same day, Beijing told its own factories to stop eating each other. The Party centre and State Council issued opinions on developing "new quality productive forces" — 19 tasks, four principles, science and technology innovation first among five headings, via IT Home. Buried in the emerging-industries section: deepen specialisation within industries and products, build cross-regional cooperation mechanisms, and "avoid involution-style competition."

That word is the tell. Export moderation abroad and anti-price-war discipline at home are the same policy viewed from two ends. The document also hands out the forward list — quantum, biomanufacturing, hydrogen and fusion, brain-computer interfaces, embodied intelligence, 6G — and names intelligent connected NEVs, AI phones and PCs, and humanoid robots as the terminals to push now.

On AI it asks for plumbing, not model cards. Build technology monitoring, risk early warning and emergency response systems; ensure AI is safe, reliable and controllable. Pilot bases and high-value application scenarios, industry by industry. Nothing triggered by what a model can do.

Go deeper on this section: ClaudeChatGPT

ByteDance found DeepSeek answering by token position

Shift the padding, change the answer. ByteDance's Seed team took an FP8 quantisation function out of DeepSeek's own inference code and asked DeepSeek-V4-Flash-Base to complete the last token. Correct answer: 8. Add a decorative docstring of equals signs in front and the model flips to 32 on a four-token cycle, QbitAI reports. Where padding length mod 4 is 0 or 1, the wrong answer averages 71.3% probability; at 2 or 3, the right one averages 91.5%.

At 128K it is worth 40 points. On a haystack of roughly 16,000 key-value pairs, moving the target relative to the compression window swung V4-Flash-Base retrieval accuracy by 40.2 percentage points and V4-Pro-Base by 34.8. Post-training shrinks it — 19.1 for Flash-0731, 14.8 for Pro-0813, 6.1 for V4.1-Flash-0910 — but the ripple survives.

The period is the compression stride. Four tokens for V4, two for V4.1. Seed reproduced the effect from scratch on a 0.6B architecture across several chunked-compression schemes; the full-attention control did not show it, and the cycle tracked the stride whether 4, 6 or 8. Attention heads had quietly specialised by position. Averaged benchmarks hide all of this, which is the point.

Go deeper on this section: ClaudeChatGPT

The poisoning intern just raised thirty million dollars

Place the name first. Tian Keyu is the Peking University doctoral student who, interning on ByteDance's commercialisation tech team in mid-2024, exploited a Hugging Face model-loading hole to slip code into weight files that rewrote colleagues' optimisers and injected random delays — wasting a quarter for some thirty researchers. ByteDance sued for 8m yuan (~$1.1m); the court found breach of the internship contract and ordered 500,000 yuan (~$70,000) plus return of wages.

That December he won the NeurIPS 2024 best paper as first author, for visual autoregressive modelling — next-scale prediction, which kept images two-dimensional instead of flattening them, ran roughly 20 times faster than raster-order autoregression, and showed a clean scaling curve from 300m to 2.3bn parameters. Mainland China's first.

Now ten people, no product, no company name. 5Y Capital, a Moonshot AI backer, and IDG have put close to $30m in at a $200m post-money valuation, Leiphone reports. The pitch: a 200,000-symbol visual vocabulary the model invents for itself, claimed to cut per-second video generation cost by at least an order of magnitude, then roughly 100m hours of video, base model in 2027.

The price is the argument. Leiphone's piece is organised around whether character disqualifies talent. Set that beside the comparables in the same article — Fei-Fei Li's World Labs at a $5.4bn valuation, Yann LeCun's AMI Labs with a $1.03bn seed, former Alibaba model chief Lin Junyang's venture named for the linguistics term pragmatics at $2bn — and $200m for the efficiency thesis looks like cheap seats, character discount included.

Go deeper on this section: ClaudeChatGPT

Washington counted China's missing safety cards

Thirty-one out of 857. SemiAnalysis's report, titled around the claim that Beijing will not pace the frontier, went through every release from nine Chinese developers — Alibaba, ByteDance, Tencent, Baidu, DeepSeek, Moonshot, Z.ai, MiniMax, StepFun — since 2021. Only 3.6% ever carried a published developer safety result; 1.1% had one at launch; 94.9% have none. Where results came, the median lag was 42 days, the longest 349, for DeepSeek-R1.

The report concedes "not found" is not "not tested." Worth holding onto, because the Chinese-language evaluation published the same day was anything but lax. Lenovo's in-house coding harness took first place on SWE-bench-Live Lite at 71% running DeepSeek-V4.1-Flash, and submitted full agent trajectories for the benchmark's audit of answer leakage to earn its verified badge, per QbitAI.

So the gap is in what gets measured, not in rigour. Audited trajectories for capability, a published teardown of a rival's failure mode, no safety card anywhere — and a State Council document that asks for monitoring systems rather than capability-triggered disclosure. SemiAnalysis found the same thing in the binding rules. Both sides are describing one regime; only one of them calls it a gap.

Go deeper on this section: ClaudeChatGPT

Threads we are pulling

  • The hacked banks' toolkit deleted itself. Following yesterday on the South Korean financial breaches: ARTEX's author, posting as Autumn-27, condemned the misuse, said the pentest agent was built for authorised testing, declared it closed-source — and then the whole GitHub repository came down, per Solidot. Reuters has it too. Nothing already cloned is un-published; the engine was Chinese open weights regardless.
  • JAC limit-down twice. Zunjie's statement conceded free upgrades for delivered owners while calling Dongchedi's brake-pedal test "non-standard extreme conditions." ByteDance's auto platform hit back that 100km/h braking is a basic test item and confirmed the viral 4,589N reading was camera-refresh residue against a true peak of 1,612N. JAC closed limit-down again at 22.25 yuan ($3.13), over 10bn yuan (~$1.4bn) gone in two sessions. QbitAI dug out the precedent: Ford recalled 2,627 imported Mustangs in China in 2020 for pedal-bracket fracture, with zero prior complaints on file.
  • Manus got its founder's passport back. Following the $500m round: Solidot reports Xiao Hong's exit ban has been lifted and the post-money valuation is $4bn, led by Boyu and IDG, here.
  • Who absorbs the cost, Taiwan edition. After this week's service-labour pieces: Micron's Taoyuan union won strike authorisation with 1,994 of 2,012 ballots cast — 88.3% of all members — demanding a standing formula of 15% of operating profit shared with staff rather than a bonus, per Solidot. Taichung's union negotiates 22 October and is preparing its own vote.

Go deeper on this section: ClaudeChatGPT

The river · 34

ByteDance researchers find DeepSeek accuracy depends on token position

# QbitAI 13h

ByteDance's Seed team found periodic retrieval swings in DeepSeek V4 caused by blockwise KV-cache compression, with position alone changing accuracy by up to 40.2 points.

More
  • In a code-completion test, adding irrelevant padding before an unchanged answer shifted the model between a wrong token and the correct one on a four-token cycle.
  • In a 128K-token retrieval test, moving the target relative to the compression boundary changed V4-Flash-Base accuracy by up to 40.2 percentage points.
  • The proposed mechanism is phase sensitivity: tokens in different positions inside a compressed KV-cache block are not preserved or attended to equivalently.
  • Post-training reduced the gap, from 40.2 points in the base model to 6.1 points in a newer V4.1 Flash model, but did not remove the periodic pattern.
  • This is an architectural reliability issue, not merely prompt sensitivity; long-context systems can produce different answers when irrelevant formatting changes cache alignment.
  • For production systems, regression tests should vary padding, document layout and chunk boundaries rather than testing only semantic equivalence.
  • The finding also illustrates why efficiency optimizations such as KV compression need accuracy audits across positional phases, especially for code and retrieval workloads.
Open at QbitAI Ask Claude ChatGPT

Tsinghua-linked robot model tops a demanding action benchmark

# QbitAI 16h

Chinese robotics company Xingdong Century says its VPP2 world-action model led RoboDojo in success rate, score, generalization and memory without extra data or agent enhancements.

More
  • VPP2 reportedly achieved a 32.26% average success rate and 39.26 average score on 42 simulated bimanual tasks, ahead of the cited GPT-6 Astra baseline.
  • The model separates video prediction from action learning instead of training them as one undifferentiated objective, aiming to preserve visual generalization while adding control.
  • On a real ALOHA dual-arm robot, the system reportedly reached 58.5% average success across ten zero-shot tasks and led on nine.
  • The benchmark still measures simulation and a limited robot setup, so it does not establish robust performance across factories, homes or different hardware.
  • The result is strategically interesting because it treats world models as action systems: predicting plausible frames is not enough unless the prediction remains aligned with what the robot should do.
  • If the gains survive hardware and task changes, the decoupled training recipe could be more valuable than the leaderboard position itself.
Open at QbitAI Ask Claude ChatGPT

China agrees to cut EU hybrid-car exports by more than half

# SCMP 8h

Chinese coverage presents the EU-China deal as a temporary brake on trade conflict, while the immediate fact is a reported multi-year cap on Chinese hybrid-car shipments.

More
  • EU trade chief Maros Sefcovic said after talks in Beijing that China had agreed to reduce hybrid exports by more than half over four years.
  • The arrangement is framed as an understanding rather than a fully detailed treaty, so implementation rules and enforcement remain important unknowns.
  • For China, accepting an export constraint trades some volume for reduced risk of a broader European trade escalation.
  • For the EU, the deal tests whether negotiated supply moderation can substitute for tariffs while preserving access to competitively priced Chinese vehicles.
  • The outcome will matter beyond hybrids because it establishes whether Brussels and Beijing can manage industrial overcapacity through quotas, licensing or negotiated restraint.
Open at SCMP Ask Claude ChatGPT

Chinese researchers make an Arctic-to-Antarctic call on Guowang

# SCMP 8h

Chinese polar researchers reportedly completed the first public test of the Guowang satellite constellation with a video call between the Arctic and Antarctic.

More
  • The demonstration is a communications test of a planned low-Earth-orbit constellation positioned as China's answer to Starlink.
  • State-media framing emphasizes coverage and national capability, while the supplied report does not provide throughput, latency, satellite count or independent verification.
  • A successful polar link would matter operationally because high-latitude connectivity is difficult for terrestrial networks and strategically important for research and logistics.
  • Guowang's significance will depend less on a single call than on launch cadence, terminal availability, spectrum coordination and sustained service economics.
  • The event shows China's satellite program moving from announced architecture toward visible application demonstrations.
Open at SCMP Ask Claude ChatGPT

Beijing and Manila trade blame over a South China Sea evacuation

# SCMP 12h

China and the Philippines dispute a medical evacuation near Second Thomas Shoal, while Manila says Chinese military and coastguard presence nearly doubled in September.

More
  • The immediate incident involved evacuating an injured Philippine serviceman from the grounded BRP Sierra Madre, a long-standing flashpoint.
  • The dispute is not only about the evacuation: the Philippines is using the reported vessel increase to frame a broader pattern of coercive pressure around contested features.
  • Beijing's likely counter-framing is that Philippine resupply and military presence violate its jurisdictional claims, making humanitarian access part of a sovereignty contest.
  • The episode shows how even medical or safety missions can become operational tests of the parties' rules, communications and willingness to de-escalate.
  • Any sustained increase in patrols raises the chance that an accident, rather than an intentional policy decision, will drive the next escalation.
Open at SCMP Ask Claude ChatGPT

Huawei-linked openJiuwen releases an enterprise AgentOS

# AI QbitAI 18h

Huawei teams and partners released an open enterprise agent operating system with multi-agent workflows, self-improving skills, hardware-aware scheduling and security controls.

More
  • AgentOS decomposes work across roles and parallel agents while allowing humans to intervene at selected decision points, addressing coordination rather than just prompting.
  • Its self-evolution layer turns execution traces, user feedback and memories into reusable skills, potentially reducing repeated exploration across similar business tasks.
  • The runtime combines multi-tenant isolation, sandboxes, permissions, fault recovery and resource scheduling, which are the missing controls between a demo agent and a production service.
  • A plugin architecture separates models, tools, memory and connectors, while an asset hub lets partners share skills and domain specialists.
  • Huawei is also packaging the stack in appliances intended for private deployment, with the company claiming hour-scale end-to-end installation.
  • The strategic signal is state-linked infrastructure: China is trying to make enterprise agents deployable as managed systems tied to domestic hardware, software and partner ecosystems.
Open at QbitAI Ask Claude ChatGPT

Sharpa builds a tactile humanoid stack from skin to data gloves

# AI QbitAI 16h

Chinese robotics company Sharpa unveiled a humanoid, tactile hand and force-feedback glove, treating contact sensing and data capture as one operating system for dexterous work.

More
  • The D01 humanoid claims 1,000-hertz communications, 0.2-millimeter repeatability and whole-upper-body tactile sensing at 100 hertz.
  • Its W02 hand reduces active degrees of freedom from 22 to 21 while shrinking the hand by about 30%, arguing that useful dexterity does not require maximizing actuator count.
  • The glove captures 22 hand-joint encoders and returns contact feedback to the operator, linking teleoperation data with robot learning.
  • The architecture separates long-horizon planning from high-frequency contact control, a sensible response to the fact that vision becomes unreliable in the final millimeters of manipulation.
  • The strategic bet is that commercial robots need a closed loop of tactile sensing, force control and human-demonstrated data rather than another stage demo based mostly on vision.
Open at QbitAI Ask Claude ChatGPT

Chinese luxury car faces scrutiny after repeated brake failures in testing

# Culture QbitAI 17h

Jiemens V800's maker says three cars failed in an extreme brake test and offers a free upgrade, while the testing outlet says repeated fractures reveal a safety concern.

More
  • The manufacturer says the vehicle passed standards above the cited national and industry requirements and that no similar field failures have been reported.
  • The testing outlet says 100-kilometer emergency braking is a normal test and reports peak pedal forces around 1,600 newtons, disputing a viral higher reading as a camera-refresh artifact.
  • The engineering dispute turns on load direction, duration and test protocol, not on whether a single number exceeds a nominal standard.
  • The decision to offer a free hardware upgrade without explaining the fracture mechanism is a meaningful product action but not a substitute for a public root-cause analysis.
  • The episode shows how China's influencer-led vehicle testing ecosystem now functions as an informal safety-adjudication layer alongside regulators and automakers.
  • A lack of customer complaints cannot clear a latent design defect; the Ford recall cited in the report illustrates why validation must precede field failures.
Open at QbitAI Ask Claude ChatGPT

Chinese diffusion-language startup raises a record funding round

# AI Zhidx 18h

Shenzhen startup DiffuSpace raised several hundred million yuan, roughly $30 million to $60 million, to develop diffusion language models that generate tokens in parallel.

More
  • Unlike autoregressive models that emit text sequentially, the company's dLLM approach aims to generate a sequence globally and refine it through diffusion.
  • The founders cite a research lineage spanning sequence-to-sequence diffusion, a 7-billion-parameter Dream model and later vision-language and action models.
  • The proposed advantage is latency for edge agents, with a partner claiming fivefold faster operation, though the supplied report offers no independent benchmark details.
  • The funding is notable because it gives a non-autoregressive architecture a serious Chinese commercialization push rather than leaving it as a research curiosity.
  • The hard engineering question is whether parallel generation can preserve instruction fidelity, long-context coherence and tool-use reliability while delivering its speed advantage.
Open at Zhidx Ask Claude ChatGPT

Google puts Gemini Agent at the center of enterprise work

# AI Zhidx 21h

Chinese coverage frames Gemini Agent as a general-purpose digital coworker that can route among Gemini and Claude, maintain four kinds of memory and continue work after the user leaves.

More
  • The product combines goal-based planning, enterprise data access, code execution, document creation and multi-agent delegation behind one work interface.
  • Its four memory types separate current context, business knowledge, procedures and past episodes, a more structured design than simply appending old chats to a prompt.
  • The coworker concept gives an agent an email address, calendar, storage and directory identity, making permissions and auditability more consequential than in a personal chatbot.
  • Google says routing can cut total cost to one-third of an all-frontier-model baseline and improve speed by 40%, but those figures are company claims.
  • The Chinese framing places Gemini Agent alongside domestic workplace agents such as Doubao Work and Qianwen, revealing that the competitive unit is becoming the enterprise operating model rather than the base model.
  • The unresolved issue is governance: persistent identity, long-running execution and cross-application permissions create a much larger failure surface than answer generation.
Open at Zhidx Ask Claude ChatGPT

AI startups complete 195 acquisitions as OpenAI leads the buying

# Analysis Zhidx 13h

A Chinese analysis counts 195 venture-backed AI acquisitions through September and highlights a smaller Chinese market, including Zhipu's infrastructure purchase and Tencent's Manus investment.

More
  • The dataset says acquisition volume was up 14% from the prior full year, but the number of active buyers rose only 2%, implying that a small group of well-funded companies is buying more aggressively.
  • OpenAI reportedly completed 20 AI-related acquisitions over three years, while Anthropic and legal-AI company Legora each announced five.
  • Only 12 of the 195 transactions disclosed prices, so deal count is more reliable than aggregate spending as a measure of consolidation.
  • Chinese AI startups appear less acquisitive in public data, with Zhipu's purchase of an AI compiler team standing out as an infrastructure-oriented example.
  • The pattern suggests leading AI companies are buying distribution, specialist teams, data and deployment infrastructure faster than they can build every layer internally.
Open at Zhidx Ask Claude ChatGPT

AWS brings its AgentCore platform to Chinese cloud regions

# AI Zhidx 13h

Amazon Web Services made AgentCore available in Beijing and Ningxia, giving Chinese enterprises managed runtimes, identity, browser tools, sandboxes and observability for agents.

More
  • The platform uses microVM isolation for sessions, a gateway for APIs and MCP servers, identity management for delegated credentials, and CloudWatch-based tracing.
  • Its China availability is significant because foreign cloud services usually face local operating, data and infrastructure constraints rather than simply a product switch.
  • AWS says a third-party comparison found 2.1 times faster end-to-end development and 5.2 times faster deployment, but those figures are vendor-associated and not independently established here.
  • The architecture targets the real enterprise bottleneck: controlling what an agent can access and proving what it did, not merely selecting a model.
  • Availability in two local regions still leaves open questions about data residency, model choice, cross-region operation and the limits imposed by Chinese regulation.
Open at Zhidx Ask Claude ChatGPT

China's central bank rejects claims that the yuan is undervalued

# Policy SCMP 11h

The People's Bank of China says it has no intent to use currency devaluation for competitive advantage, as Beijing negotiates with the EU over trade pressures.

More
  • The statement is a defensive policy signal aimed at European accusations that exchange-rate management supports China's export advantage.
  • By issuing the position paper in both Chinese and English, the central bank is addressing domestic confidence and foreign negotiators at the same time.
  • The language rejects competitive devaluation without promising a freely floating exchange rate, leaving Beijing room to manage volatility and capital flows.
  • The timing matters because currency policy is becoming part of the same negotiation bundle as industrial exports, subsidies and market access.
Open at SCMP Ask Claude ChatGPT

ARTEX disappears from GitHub after misuse in attacks

# Policy Solidot 12h

Chinese coverage says ARTEX was removed and abandoned after misuse in attacks on South Korean banks; Western aggregation focuses on the alleged offensive use and later closure.

Also: Techmeme

More
  • The developer says ARTEX was intended for authorized security testing, denies responsibility for unlawful use and promises no further updates or support.
  • Removing the repository reduces the official distribution channel but cannot revoke copies already downloaded or incorporated into other tools.
  • The event exposes a governance gap for dual-use AI security agents: disclaimers do not provide technical containment, and open distribution makes intent difficult to enforce.
  • The supplied report does not establish whether the developer knew about the attacks, how much the tool contributed or whether any investigators contacted the author.
  • The Chinese framing is defensive and responsibility-focused, while the Western framing treats the incident primarily as evidence of an offensive capability.
Open at Solidot Ask Claude ChatGPT

Controversial Chinese researcher raises $30 million for a world model

# AI Leiphone 10h

A 10-person Chinese lab reportedly raised $30 million at a $200 million valuation to build a symbolic visual language and world model for robotics and autonomous driving.

More
  • The proposed system uses a visual dictionary of about 200,000 symbols to represent, compress and predict video rather than relying only on natural-language descriptions.
  • The founder says the team plans to train on roughly 100 million hours of video and target a 2027 foundation model for robots, autonomous vehicles and interactive video.
  • The technical lineage is a next-scale visual autoregressive approach that preserves two-dimensional structure instead of flattening images into a long raster sequence.
  • The funding is notable because investors are backing a technically ambitious team despite a public dispute over alleged interference with a former employer's training workloads.
  • The supplied article contains allegations and self-reported performance claims, so the financing and architecture should not be treated as independent validation.
  • If the symbolic representation works, its value would be lower-cost long-horizon physical simulation; the missing proof is whether the symbols preserve action-relevant physics.
Open at Leiphone Ask Claude ChatGPT

Lenovo's TianxiCode tops a live software-engineering benchmark

# AI QbitAI 11h

Lenovo says its TianxiCode agent, paired with DeepSeek V4.1 Flash, solved 71% of SWE-bench-Live Lite problems and passed the benchmark's trajectory audit.

More
  • The system combines a model with an agent framework that performs multi-hop code retrieval, plans tool calls, runs tests and revises patches in an isolated environment.
  • The claimed 71% result is for the combined system, so it should not be read as a model-only capability.
  • The benchmark's verification process reportedly reviews trajectories and checks for leaked tests or answers, which makes reproducibility more meaningful than an unaudited leaderboard score.
  • Lenovo plans to bring the coding agent into its hardware ecosystem, an unusual route for a PC maker seeking to turn internal AI infrastructure into product differentiation.
  • The stronger signal is architectural: context pruning, tool use and test-driven repair can matter as much as the underlying model on real repositories.
Open at QbitAI Ask Claude ChatGPT

Xinjiang's desert seafood boom carries hidden political and ecological costs

# Analysis The Wire China 7h

Fish farming in Xinjiang is promoted as technology overcoming geography, but the account stresses political control, imported inputs and concealed environmental burdens.

More
  • The story treats desert aquaculture as an engineered food system rather than a simple miracle of local production.
  • Its central tension is resource transformation: Xinjiang can host seafood output only through controlled water, feed and logistics systems that shift costs elsewhere.
  • The framing challenges triumphalist development narratives without claiming that the farms are technically impossible.
  • For China watchers, the subject shows how regional policy uses industrial technology to rewrite the image of a difficult geography.
Open at The Wire China Ask Claude ChatGPT

Hong Kong consults on long-delayed sexual-offences reform

# Policy Initium 11h

Hong Kong's government received more than 6,000 consultation responses on sexual-offences reform, with survivors pressing for changes after decades under outdated law.

More
  • More than 90% of submitted responses reportedly supported reform, but the survivor perspective is focused on what the amendments fail to address as well as what they change.
  • The consultation reflects the tension between inherited legal structures and a city government trying to modernize sensitive criminal law without losing procedural control.
  • The story is also about public participation: survivors who previously had little institutional traction are using the consultation to make personal testimony part of the policy process.
  • The supplied item is an audio report, so detailed provisions and the government's final drafting choices are not available here.
Open at Initium Ask Claude ChatGPT

China declares a potentially record-setting super El Nino

# Signal Solidot 5h

China's climate center says an eastern Pacific super El Nino formed in September, with the three-month sea-temperature index reaching 2.54 degrees Celsius.

More
  • The center expects ocean temperatures to peak in late autumn or early winter, potentially making this the strongest monitored El Nino event on record.
  • The classification is based on a five-month run above the 0.5-degree threshold and a July-to-September average above the 2.5-degree super-event mark.
  • For China, the significance is operational: El Nino shifts rainfall and temperature patterns, affecting agriculture, flood risk, energy demand and disaster planning.
  • The report is a forecast signal, not yet a measurement of realized domestic damage; the eventual impact depends on the event's duration and regional circulation.
Open at Solidot Ask Claude ChatGPT

Wang Yi tells Pacific leaders China is not seeking geopolitical rivalry

# Policy SCMP 7h

Beijing's top diplomat frames Pacific island engagement as South-South cooperation, while the setting is plainly a contest with Washington for regional influence.

More
  • The statement was delivered to Pacific politicians through the Communist Party's International Department, not only through a conventional foreign-ministry channel.
  • Calling the relationship mutual assistance is a positioning move: Beijing wants development and political ties to appear separate from military bloc competition.
  • The rhetoric does not remove the strategic rivalry; it seeks to define China's role as a partner that rejects the language of spheres of influence.
  • Pacific governments are therefore being offered a choice of framing as much as a choice of projects, aid or diplomatic alignment.
Open at SCMP Ask Claude ChatGPT

US Treasury official may skip Hong Kong finance meeting for Shenzhen talks

# Policy SCMP 13h

A Chinese report says Treasury Secretary Scott Bessent may prioritize preparations for Trump's Shenzhen visit and a longer US-China trade truce.

More
  • The reported plan would put Bessent's meeting with Vice-Premier He Lifeng in Shenzhen ahead of Trump's expected meeting with Xi Jinping.
  • The story is based on unnamed sources, so attendance and the agenda remain unconfirmed.
  • The diplomatic choreography suggests that trade and financial negotiations are being concentrated around leader-level meetings rather than routine ministerial channels.
  • For businesses, the key variable is whether the truce extension produces durable tariff and export-control predictability or merely delays another confrontation.
Open at SCMP Ask Claude ChatGPT

Rare Tang dynasty calligraphy scroll sells for $22.6 million

# Culture SCMP 18h

A calligraphy work long thought destroyed resurfaced after 81 years and sold in Hong Kong for $22.6 million, showing the market power of Chinese cultural heritage.

More
  • The scroll, attributed to early Tang calligrapher Yu Shinan, fetched HK$177.45 million, or $22.6 million, at a Hong Kong auction.
  • Its provenance story is part of the value: disappearance during the turmoil of 1945 and later recovery turn the object into both artwork and historical evidence.
  • The sale illustrates how Hong Kong remains a key marketplace connecting mainland cultural heritage, global collectors and international pricing mechanisms.
  • It also shows the tension in heritage markets between public cultural significance and private speculative valuation.
Open at SCMP Ask Claude ChatGPT

Chinese think tank warns against a confrontational Arctic posture

# Policy SCMP 7h

A Shanghai think tank advises Beijing not to confront a new US-Denmark-Greenland security pact, fearing China could be excluded from the emerging Arctic order.

More
  • The report combines two positions: Washington is unlikely to turn the Arctic into an exclusive strategic backyard, but Beijing should avoid giving it a reason to do so.
  • That is a pragmatic adjustment to China's Arctic diplomacy, which has often emphasized scientific, commercial and governance access rather than formal regional membership.
  • The recommendation reveals concern that rhetorical confrontation could reduce China's room to participate in shipping, research and resource discussions.
  • It is a think-tank argument, not evidence of a publicly announced policy change, but it shows the range of views available inside Chinese strategic debate.
Open at SCMP Ask Claude ChatGPT

China's robotaxi race is becoming a partnership contest

# Analysis TMTPost 10h

Chinese robotaxi operators are shifting from solo demonstrations toward alliances covering vehicles, permits, compute, manufacturing and overseas access.

More
  • The article describes partnerships pairing ride-hailing demand and dispatch with automaker production, or combining city operations with another firm's driving stack.
  • As perception and planning improve and hardware costs fall, technical demonstrations are becoming a minimum requirement rather than a durable moat.
  • China has opened more than 35,000 kilometers of test roads across 17 demonstration zones, creating a policy base for larger deployments.
  • The scarce assets are now distributed across the stack: compute, fleet manufacturing, operating licenses, local demand and international compliance.
  • This favors integrated consortia and may make the eventual winner less visible as a single brand than as a coordinated industrial system.
Open at TMTPost Ask Claude ChatGPT

Tesla suppliers in China pivot from car parts to humanoid robots

# Analysis TMTPost 10h

Chinese auto suppliers are reportedly entering Tesla's humanoid-robot supply chain, carrying over expertise in actuators, precision structures and motor controls.

More
  • Top Group, Sanhua Intelligent Control and Joyson Electronics were reportedly audited by Tesla's robot team after receiving orders for joint modules, actuators or precision components.
  • The shift reuses capabilities built for electric vehicles, showing why China's EV supply chain is a likely industrial base for embodied AI hardware.
  • Chinese automakers are using several organizational paths at once: internal research, subsidiaries, joint ventures, strategic investment and acquisition.
  • Factory deployments provide a more valuable data source than stage demonstrations because robots must handle flexible parts, loading and material variation.
  • The unresolved bottleneck is repeatable reliability: supply-chain reuse can accelerate prototypes, but standards and maintenance practices determine whether robots stay useful in production.
Open at TMTPost Ask Claude ChatGPT

Chinese startup raises $56 million for skull-penetrating brain ultrasound

# Analysis GeekPark 16h

Shenzhen company Kunwei raised 400 million yuan, roughly $56 million, for low-frequency ultrasound that it says can image brain structures through the skull in real time.

More
  • The company claims to combine low-frequency acoustics with neural-network reconstruction to improve the usual tradeoff between skull penetration and spatial resolution.
  • Its reported architecture processes about 10 gigabytes per second of raw echo data and renders thousands of frames per second, demanding a small, high-throughput model rather than a giant language model.
  • Kunwei is positioning hardware and its kOS software as infrastructure for neurotechnology labs, analogous to a GPU and software stack for brain science.
  • The reported clinical applications include mobile stroke triage, but regulatory scope and independent clinical validation are not provided in the supplied text.
  • The business model depends on a difficult bridge from medical-device sales to a broader platform for brain-computer-interface and NeuroAI research.
Open at GeekPark Ask Claude ChatGPT

Chinese phone makers take different routes to the AI handset

# Analysis TMTPost 10h

Huawei is integrating chips, operating system and models; Alibaba is selling a portable agent stack; ByteDance is embedding its assistant; Tencent is connecting services through agent protocols.

More
  • The article places Huawei on the deepest integration path, with its own silicon, HarmonyOS and on-device models supporting a 30-billion-parameter mixture-of-experts model.
  • ByteDance's phone strategy uses MCP and agent-to-agent interfaces rather than GUI simulation, but currently remains strongest inside ByteDance's own service ecosystem.
  • Alibaba's Qwen Intelligence aims to be a reusable business-to-business layer that phone makers can integrate into their operating systems.
  • Tencent is taking the opposite route by letting handset assistants call WeChat services, preserving the app's distribution and transaction position.
  • The strategic question is not which phone has the best chatbot; it is who controls the permission layer through which an agent can act across apps, devices and user data.
Open at TMTPost Ask Claude ChatGPT

China brief: EU trade talks, gig-worker protections and AI safety

# Policy Sinocism 23h

A China-focused newsletter groups the yuan dispute, EU trade negotiations, gig-worker protections, holiday spending and domestic AI-safety debate into one policy snapshot.

More
  • The briefing is valuable because it places technology alongside labor, consumption and trade policy rather than treating AI as a separate industry.
  • Its mention of new gig-worker protections points to a broader effort to manage platform labor without abandoning the platform economy.
  • The supplied excerpt is truncated, so the specific provisions and official positions beyond the listed topics cannot be assessed.
  • As a roundup, it is best used for agenda-setting: the important question is how China is balancing export pressure, weak consumption, labor stability and AI deployment at the same time.
Open at Sinocism Ask Claude ChatGPT

Control fight escalates at Chinese diagnostics company Shuoshi Bio

# Policy TMTPost 7h

A shareholder dispute over board elections exposes a Chinese listed company's split between economic ownership, voting rights and operating control.

More
  • The controlling partnership owns 27.53% of the listed company, while the board chair personally owns 8.73% and controls day-to-day management.
  • The dispute centers on whether the controlling partner's seal and executive-partner changes validly authorize a board-election request.
  • Five directors blocked the proposed election, arguing that litigation and alleged fund misuse made the authority chain uncertain rather than opposing turnover in principle.
  • The case illustrates a recurring Chinese governance problem: partnership structures can separate the people who supplied capital from the people legally empowered to exercise votes.
  • The conflict is occurring as the company's post-pandemic diagnostics revenue remains depressed, making control of a shrinking business more consequential.
Open at TMTPost Ask Claude ChatGPT

Lazada turns Chinese brands' Southeast Asia push into niche markets

# Analysis Leiphone 16h

Nearly 10,000 Chinese brands joined Tmall's light-export program, while Lazada reports distinct growth pockets in Thai fashion and Vietnamese esports.

More
  • The story argues that Southeast Asia is no longer one generic market: Thai fashion depends on local lifestyle content, while Vietnamese gaming depends on teams, streamers and player communities.
  • The model connects creators, live commerce, local stores and platform promotion rather than simply shipping Chinese inventory abroad.
  • Reported gains include Thai fashion sales more than tripling from an earlier promotion and Vietnamese esports sales more than doubling, but these are platform-associated figures.
  • The strategic shift is from market entry to demand discovery: Chinese brands need to identify the specific communities and product categories where localization can compound.
Open at Leiphone Ask Claude ChatGPT

Starlink calls for more orbital data sharing as China's constellations grow

# Policy Solidot 12h

A Chinese report describes Starlink's demand for public ephemeris data as orbital congestion rises, while noting China's two planned mega-constellations.

More
  • Starlink reportedly performs about 1,000 collision-avoidance maneuvers each day and has recently avoided roughly 650 satellites from other operators.
  • Only about half of those satellites belonged to operators sharing orbital data, illustrating why voluntary coordination can fail as the number of spacecraft rises.
  • China's Guowang and other large constellations make the data-sharing problem a US-China infrastructure issue, not only a commercial dispute.
  • The central policy question is whether operators will accept common transparency rules before a collision creates debris and forces emergency regulation.
  • The supplied account reports Starlink's claims but does not independently verify the maneuver counts.
Open at Solidot Ask Claude ChatGPT

Only 1.1% of Chinese AI launches included safety results at release

# Signal Techmeme 6h

A US tech roundup highlights an analysis of 857 releases from nine Chinese AI labs, finding developer-published safety results in only 3.6% and at launch in 1.1%.

More
  • The dataset covers 857 releases from nine Chinese labs through September 2026, making this a sector-level reporting-practice measure rather than a single model audit.
  • The gap between 3.6% overall and 1.1% at launch suggests safety evidence is usually late, sparse or absent from the public release process.
  • This does not prove that Chinese models are less safe; it shows that external researchers and deployers have little standardized developer evidence with which to judge them.
  • For enterprise buyers, the practical implication is a heavier burden on independent red-teaming, documentation review and deployment controls.
  • The result is especially relevant in China because rapid model iteration and intense price competition may reward shipping capability before safety evaluation is made public.
Open at Techmeme Ask Claude ChatGPT

Qifei raises nearly $14 million for a passenger eVTOL

# Analysis 36Kr 13h

Chinese eVTOL maker Qifei raised nearly 100 million yuan, roughly $14 million, after China's aviation regulator accepted its W280 for type certification.

More
  • The five- or six-seat W280 uses a compound-wing configuration, claims 280 kilometers per hour and 300 kilometers of range, and shares a platform with a one-ton cargo version.
  • Qifei argues that cargo operations can generate flight hours, maintenance data and supplier experience before passenger service is ready, converting development cost into an operating asset.
  • The company favors a slower but more certifiable compound-wing design over the faster tilt-rotor route for 30-to-150-kilometer trips.
  • The financing is backed by provincial, city and district state-linked investors, showing how Chinese local governments are supporting aviation projects through coordinated capital and regulatory resources.
  • Passenger commercialization is still described as three to five years away, so certification progress matters more than near-term revenue.
Open at 36Kr Ask Claude ChatGPT

AI appliances win attention during China's weak holiday market

# Culture TMTPost 10h

Chinese retailers reported strong AI-appliance sales during the holiday period even as the broader home-appliance market remained depressed, suggesting consumers want convenience rather than only discounts.

More
  • The reported market contrast is sharp: overall appliance sales were forecast to fall, while one retailer said AI-appliance sales rose more than 40% and another said smart appliances reached 60% of unit sales.
  • Trade-in subsidies still drive a large share of demand, so the figures do not prove that consumers are independently paying a premium for AI.
  • The most credible use cases are automatic adjustment of washing, cooling or lighting based on sensed conditions, not simply adding a cloud chatbot to a simple appliance.
  • Consumer complaints focus on unreliable voice interaction, confusing touch interfaces and software updates that threaten the appliance's basic function.
  • The market signal is therefore conditional: AI can support premiumization only when it reduces effort and leaves core mechanical reliability untouched.
Open at TMTPost Ask Claude ChatGPT
Source ledger
ChinAI 403 0/0 HTTPError: 403 Client Error: Forbidden for url: https://chinai.substack.com/feed
ChinaTalk 200 1/20
Pekingnology 200 0/20
Ginger River Review 200 0/20
Sinocism 200 1/20
QbitAI 200 10/10
36Kr 200 5/30
Solidot 200 12/20
Leiphone 200 10/20
GeekPark 200 3/30
IT Home 200 12/60 +43 over cap
SCMP 200 10/50 +6 over cap
Initium 200 4/15
Jiemian 200 8/30 +16 over cap
China Digital Times 200 0/15
The Wire China 200 1/50
Techmeme 200 2/15
Hacker News 200 0/20
V2EX 200 8/50 +32 over cap
Zhidx 200 9/20
InfoQ China 200 8/20 +5 over cap
New Wisdom 200 8/15
TMTPost 200 8/16 +8 over cap