China Desk

China News, Summarized 25 Aug 2026, 20:39 UTC 34 stories 18/19 sources

Today's brief

The anti-graft campaign is not winding down — it is being handed foreign jurisdiction

A 47-article draft gives the corruption hunt a passport

The National Supervisory Commission wrote itself a statute. China's top legislature took a first reading Tuesday of a draft Anti-Cross-Border Corruption Law — six chapters, 47 articles, text not yet public — at the 24th session of the 14th National People's Congress Standing Committee, which runs to Friday.

Xinhua's framing is about recovery. The law is meant to enrich the legal toolbox for fugitive repatriation, asset recovery and foreign-related cases, closing jurisdictional gaps on offshore offences. 963 fugitives were repatriated in 2025, per the discipline commission's own work report.

But the draft also sets out obligations for companies. Integrity and compliance duties, and the liabilities for breaching them. That is the part aimed at Chinese firms operating abroad rather than at absconded cadres — a compliance regime with Beijing as the enforcing jurisdiction, on top of whichever local one applies.

And the campaign being codified is speeding up, not settling. SCMP's data dive counts 65 probes into civilian officials of vice-ministerial grade or above last year against 19 in 2013, Xi's first year. The Central Commission for Discipline Inspection, the Party's internal anti-graft body, has opened 46 investigations so far this year against 39 by this point last year.

The personnel move came first. The commission held an out-of-cycle Sixth Plenum on Saturday — these normally convene in January — naming Zhao Shiyong and Zhang Zhong deputy secretaries in place of Xiao Pei and Fu Kui, both associates of Wang Qishan, who ran the commission through Xi's first term. Sinocism reads it as preparation for the coming Fifth Plenum. Hard to disagree.

Go deeper on this section: ClaudeChatGPT

Washington prices one H-1B at $103,265

The fee is back, by rulemaking this time. The Department of Homeland Security published a proposed rule establishing a $103,265 charge on every cap-subject H-1B petition, payable at filing, on top of every existing fee — more than a hundred times the base petition fee, and derived by dividing about $8.78bn of claimed annual government cost by 85,000 petitions. Universities and nonprofit research bodies are exempt. Comments close 24 September.

A federal judge struck down the earlier $100,000 version in June. Solidot's summary for Chinese readers pairs it with the other plan: revoking up to 200,000 business and tourist visas, which would be the largest single revocation in American history.

What that already costs showed up in Bremen. Leiphone's report from IJCAI, the big AI conference that closed Saturday: mainland authors submitted over 4,000 papers, more than 60% of the total, and had 541 accepted — a 13.3% rate against Austria's 32.8%.

Chinese registrations came first at 669. The hall did not match. Program vice-chair Bo An, who chairs the programme next year, told Leiphone that strict refusals at the Guangzhou consulate kept many attendees out. Organisers are already discussing scheduling accommodations for Kyoto next August. Note also that CCF downgraded the conference from A-class to B and submissions rose anyway.

Go deeper on this section: ClaudeChatGPT

Thomson Reuters built its model on Qwen

Two Western legal-AI incumbents shipped on Chinese open weights this week. Leiphone leads with Alibaba: Thomson Reuters launched Thomson-1, built on Qwen3.5, to cut dependence on Anthropic's Claude, starting with table analysis and document review. Separately, Harvey — the American legal-AI company, valued around $11bn — post-trained Kimi K3, Moonshot's open-weight model, into a legal model called Tenet.

The Western write-ups bury the Chinese part. Thomson Reuters' own launch described only a strong open-source foundation and named its base model Snowdon, built with Imperial College London; the Qwen lineage came from the chief technology officer afterwards, with emphasis on months of realignment to be "ethically and politically de-biased". Cost: about $40m over two years.

Same fact, two stories. Chinese tech press treats it as a scoreboard entry — Alibaba claims 460-plus open models and three billion downloads, and cites Pinterest's chief executive putting Qwen's cost at under 8% of comparable closed models. American coverage treats the provenance as a risk to be managed, with Senator Tom Cotton warning about backdoors and Airbnb pushing back. Nobody disputes the mechanism: the cheapest good base weights are Chinese.

Go deeper on this section: ClaudeChatGPT

Memory is now bigger than semiconductors were

THE NUMBER: Gartner puts the 2026 memory market at $837.3bn — larger than the entire semiconductor industry's 2025 revenue of $809bn. DRAM up 246.6% in a year, NAND up 371.9%, memory's share of chip revenue doubling from 27% to 54%.

The capex consequence is already booked. TrendForce has DRAM and NAND going from 47% of major cloud providers' capital spending this year to 68% in 2027, with server memory contract prices up 64% in the second half of 2025 and 270% again this year. Nvidia has told server makers its flagship AI systems cost about 17% more.

Which is why Xiaomi's second chip matters more than its first. The Xring O3 scored 5.22 million on AnTuTu and beat Apple's cores on single-thread — 917 points on Hacker News, all about CPU. GeekPark's read is better: the score is fireworks, the bandwidth is the signal.

The O100 stacks memory wafers onto the compute wafer. Two AI-memory layers bonded to an NPU die at a 1.4-micron pitch, yielding 1.22TB/s — sixteen times the 76.8GB/s of LPDDR5X in today's flagships — and 330 tokens a second locally with a five-value quantised MiMo model that saves another 30% of bandwidth. Lei Jun has spent 21bn yuan (~$2.96bn) over five years to get here.

Go deeper on this section: ClaudeChatGPT

Everyone is a guest on somebody's show

China is arguing about interview surplus. Jiemian's essay opens on a video blogger posting on 21 August that he does not mind being interviewed, he minds living in an era of surplus meaning — everyone with a show, everyone a guest, everyone obliged to distil life into quotable lines.

The supply numbers explain the fatigue. Bilibili submissions of video podcasts over two hours are up nearly ninefold this year; single-day watch time passed 100 million minutes on 13 May; more than half run past 30 minutes. Veteran host Chen Luyu recorded six and a half hours with interviewer Jiang Sida and aired three hours fifty-five.

Meanwhile Douyin published its enforcement ledger. In one month it actioned 1.4534 million videos and 597,500 accounts for malicious commercial content — including 59 accounts compiling negative material on an unnamed famous technology brand, from accident footage to share-price moves, and 236 fabricating claims about a domestic beverage brand by wrapping them in nationalist sentiment.

The tell is the vocabulary. Rumour-mongering, commercial defamation, black-and-grey industry — a full taxonomy for protecting corporate reputation, deployed at a scale ByteDance has never announced for anything else.

Go deeper on this section: ClaudeChatGPT

Threads we are pulling

  • I gave the 1500m to the wrong robot, confirmed. IT Home now records the win for Beijing Humanoid's Tiangong Ultra at 2:21.63, not Honor's Lightning. Today the hundred metres fell twice: Lightning ran 8.94, then Tiangong ran 8.86. Scale of the event: 666 teams, 2,056 robots, 16 countries, 51 disciplines, 1,301 matches.
  • Doubao Work shipped on schedule. Yesterday I noted ByteDance folding Trae and Coze into Doubao with a work suite due within the week. It launched Tuesday — desktop app, cloud PC for long-running jobs, 30 days free, inheriting the user's Feishu permissions. GeekPark names the trigger: Tencent's WorkBuddy reached million-scale daily users and 20.97 million desktop visits in June.
  • Alibaba's management bought their own placement. After Sunday's HK$80bn raise, chairman Joe Tsai took 720,000 shares near HK$112 and chief executive Eddie Wu 350,000 near HK$111.6 — about HK$120m (~$15m) between them.
  • Xpeng set the embodied-AI record. Its robotics arm raised over $900m at a post-money valuation above $6.3bn, IDG leading, Tencent and Alibaba strategic — with a seven-year IPO deadline or investors can force a buyback at 8% compound. Unitree meanwhile fell 10.31% to 243.9bn yuan (~$34bn), passed by Figure AI.
  • The formaldehyde cabbage has a sequel. A pest-control firm in Xiamen allegedly sprayed dichlorvos, an organophosphate insecticide, in dozens of chain restaurants — on tableware and tables. A district joint investigation group seized its stock and closed its premises. Two food-poisoning scandals, two task forces, three days.

Go deeper on this section: ClaudeChatGPT

The river · 34

Dobot’s embodied-AI revenue jumps more than 20-fold

# Leiphone 15h

Chinese robot maker Dobot reported first-half revenue of 320 million yuan, about $45 million, up 106.6%, with embodied-AI revenue reaching 45 million yuan, about $6.3 million.

More
  • Gross profit was about 150 million yuan, or $21 million, while research spending reached roughly 100 million yuan, or $14 million, up 148.4%.
  • Embodied-AI revenue now contributes about 14.3% of total revenue and grew more than twentyfold, while the company reports more than 60 million yuan, or $8.5 million, in related orders.
  • The company is extending a decade of collaborative-robot manufacturing into humanoids and other bodies instead of starting from a lab prototype with no industrial channel.
  • Dobot says its world-action model combines 3D geometric perception with motion constraints and reached 99.25% average completion on LIBERO, though those benchmark claims are company-reported.
  • Its one-brain, many-bodies architecture pairs a shared model with body-specific motion and force-control systems, a plausible route to reusing skills across arms, humanoids, wheeled and legged platforms.
  • The strongest evidence in the report is commercial: deployments in factories and a cinema where an Atom humanoid reportedly worked 14 hours per day and sold more than 1,000 popcorn cups daily.
  • The strategic test is whether these early deployments convert into repeatable industrial orders without requiring custom engineering for every site.
Open at Leiphone Ask Claude ChatGPT

China dominated IJCAI submissions but visas kept many researchers away

# Leiphone 18h

China supplied more than 4,000 IJCAI submissions and 669 registrants, yet Chinese attendance was visibly lower as strict German visa screening kept many researchers from the venue.

More
  • Chinese researchers accounted for more than 60% of submissions and 541 accepted papers, but the reported acceptance rate was about 13.3%, below several smaller European national cohorts.
  • The mismatch between registration and physical attendance exposes how visa policy can reshape the practical geography of AI research even when a country dominates the paper pipeline.
  • The conference’s downgrade by China’s academic ranking system did not reduce submissions, suggesting that field reputation and intellectual fit can outweigh domestic classification incentives.
  • A new submission-fee policy raised the share of single-paper primary submissions from 25% to 40%, showing that a modest economic lever can change publication behavior and reduce author concentration.
  • The distribution also reveals a disciplinary tension: machine learning, computer vision, data mining and NLP took nearly half of accepted papers, while knowledge representation, agents, planning and robotics preserved IJCAI’s distinctive breadth.
  • The organizers are considering agenda changes for the 2027 Kyoto meeting to improve Chinese participation, an implicit acknowledgment that access logistics now affect scholarly representation.
  • The broader lesson is that China’s AI research scale is not equivalent to influence in the room; mobility, institutional incentives and conference design determine who can convert papers into networks and collaborations.
Open at Leiphone Ask Claude ChatGPT

Xi’s anti-corruption campaign is accelerating after 13 years

# SCMP 11h

A review found 65 investigations of vice-ministerial officials or above in 2025, versus 19 in 2013, while the disciplinary campaign continues to expand under Xi Jinping.

More
  • The figures point to institutionalization rather than a short campaign: China’s top disciplinary body is still opening high-level cases more than a decade after Xi took office.
  • The supplied review says there were 65 probes of civilian officials at or above vice-ministerial rank last year, compared with 19 in 2013.
  • The CCDI had already opened 46 investigations this year by the report’s cutoff, versus 39 in the comparable prior period cited in the source.
  • At this scale, anti-corruption enforcement is both governance maintenance and a tool for political discipline, reshaping incentives across ministries, provinces and state firms.
  • The campaign can improve compliance and remove entrenched rent-seeking, but repeated purges also create risk aversion, policy paralysis and uncertainty for officials making discretionary decisions.
  • The key distinction to watch is whether enforcement is paired with durable procurement, asset-disclosure and oversight reforms or remains dependent on campaign-style discipline.
Open at SCMP Ask Claude ChatGPT

China seeks corruption tips as Rocket Force probe continues

# SCMP 3h

Chinese military authorities are soliciting reports of procurement violations involving the Rocket Force, extending a corruption campaign inside the strategic missile branch.

More
  • The notice invites suppliers, tendering agencies, bid evaluators and military personnel to report suspected procurement violations through the PLA procurement system.
  • A public tip line places procurement corruption, rather than only individual misconduct, at the center of the continuing Rocket Force purge.
  • The Rocket Force controls China’s land-based strategic missile forces, so an extended investigation can affect procurement confidence, command trust and modernization schedules.
  • The move fits a wider pattern in which anti-corruption enforcement doubles as a mechanism for political control and institutional cleanup.
  • The supplied text does not identify accused officers, contracts or specific allegations, so the operational impact remains unclear.
  • The important question is whether the campaign produces transparent procurement reforms or mainly another round of personnel removals.
Open at SCMP Ask Claude ChatGPT

Chinese official calls for ‘normal stability’ with the United States

# SCMP 8h

At a semi-official China-US dialogue, Party official Liu Haixing urged normal stability with controllable disagreements and a larger zone of shared interests.

More
  • Liu heads the Communist Party’s International Department, so the formulation is a useful signal of how Beijing wants the relationship described even if it is not a treaty commitment.
  • The phrase rejects an explicitly zero-sum framing while preserving room for disagreement, suggesting managed competition rather than a return to strategic trust.
  • Calling for a larger shared-interest pie is an invitation to compartmentalize economic and practical cooperation away from security disputes.
  • Track 1.5 talks are not the same as formal negotiations, but they can test language later used in official diplomacy.
  • The important uncertainty is whether this rhetoric is backed by concrete arrangements on trade, technology controls or military risk reduction.
Open at SCMP Ask Claude ChatGPT

Chinese startup separates forecasting skill from model weights

# AI QbitAI 18h

DigClaw’s Rhizome framework placed three different foundation models in the FutureX top seven by separating search, causal reasoning, probability calibration and long-term forecast records.

More
  • FutureX reportedly uses 59 real-world questions whose answers are unknown at submission time, making it more meaningful than a static benchmark with possible training-data leakage.
  • The framework’s central claim is that forecasting ability can accumulate in the system around a model through search protocols, evidence trails, calibration history and feedback from settled events.
  • Rhizome separates retrieval from reasoning so the search agent looks for relevant signals rather than prematurely optimizing for an apparent answer.
  • It stores each prediction’s available evidence, tool calls, system version, probability and eventual outcome, allowing errors to be classified as retrieval, timing, reasoning or calibration failures.
  • The system aggregates independent predictions in logit space and uses Brier scores and Platt scaling to reduce persistent overconfidence or excessive caution.
  • This architecture is strategically important because an organization could swap foundation models while preserving its forecasting protocol and accumulated outcome data.
  • The reported leaderboard results are promising but still a small sample; the real test is stable calibration across domains and changing information environments.
Open at QbitAI Ask Claude ChatGPT

Chinese AI-for-science product shifts from tasks to projects

# AI QbitAI 14h

Zidong Taichu’s ScienceClaw adds an autonomous project engine that plans, executes, verifies and revises multi-step research instead of handling isolated scientific tasks.

More
  • The system separates project planning, long-horizon execution and evidence verification, reflecting the fact that research paths change as experiments and literature produce new information.
  • Its Project2Task layer builds a dependency graph with parallel and serial work, while TaskExecutor monitors experiments, handles failures and can rerun or revise a plan.
  • The EviGraph evidence system is meant to check whether claims remain supported across subtasks, not merely whether a single answer sounds plausible.
  • The supplied evaluation reports an overall score of 0.865 versus 0.596 for the best baseline and a Claim Support Rate of 0.38 versus 0.27, but the benchmark and task design need independent inspection.
  • This is the right reliability question for scientific agents: long-running autonomy is useful only when data, code, assumptions and evidence remain auditable and recoverable.
  • The product explicitly says autonomy does not remove researchers from the loop, which is a more credible deployment posture than fully unattended discovery claims.
Open at QbitAI Ask Claude ChatGPT

Open model DM0.5 tops RoboDojo with strong memory and recovery

# AI QbitAI 18h

Chinese robotics company Yuanli Lingji says its open DM0.5 model leads the difficult RoboDojo benchmark, with notable results on memory, teaching by demonstration and long-running manipulation.

More
  • RoboDojo’s reported average success rates were only 8.8% in simulation and 12.8% on real robots through July, making a leaderboard lead more informative than a high score on a conventional benchmark.
  • DM0.5 reportedly achieved 100% success across three randomized trials on a task requiring the robot to remember hidden block colors and reveal them in order.
  • The model supports up to 60 seconds of action memory, allowing it to interpret a human demonstration and reproduce a sequence without reducing everything to a text instruction.
  • A layered System 1 and System 2 design separates fast action expertise from slower planning, refusal and state interpretation when a person interrupts the task.
  • Demonstrations span millimeter-scale block assembly, flexible-object cutting and conveyor-belt picking, where the system reportedly averaged three seconds per package with more than 99% accuracy.
  • The engineering lesson is that memory, feedback and recovery may contribute more to generalization than simply adding training hours or parameters.
  • The claims are largely outlet and company reported, so independent replication on the benchmark and real hardware remains essential.
Open at QbitAI Ask Claude ChatGPT

China’s Science Exploration Award names 50 young researchers

# Signal QbitAI 15h

The 2026 Science Exploration Award selected 50 researchers across ten fields, extending a Chinese program that gives each winner 3 million yuan, about $423,000, over five years.

More
  • The award is designed to fund future exploration rather than simply reward completed work, with unrestricted support of roughly $423,000 per recipient.
  • This year’s winners came from 30 institutions in 13 cities, including several universities and research centers receiving the award for the first time.
  • The geographic spread suggests China’s basic-research network is broadening beyond its most established national centers.
  • The program has funded 347 researchers since 2019 and has previously backed scientists who later became prominent in AI and robotics.
  • Because the award is scientist-led and privately funded, it offers a window into how China is building elite research capacity alongside state programs.
  • The supplied list covers fields from mathematics and medicine to information technology, energy and advanced manufacturing, making it a talent signal rather than an AI-only story.
Open at QbitAI Ask Claude ChatGPT

A Chinese legal drama reopens the question of defending the unpopular

# Culture Pekingnology 4h

A state-backed Chinese legal drama based on real cases has prompted legal scholar Luo Xiang to argue that respect for defense lawyers is a test of whether law can restrain state power.

More
  • The series revisits China’s legal reconstruction from 1979 to 1997 through eleven stories adapted from real cases, including the 1980s campaign against so-called hooliganism.
  • Its producers include the state broadcaster and the country’s top judicial and prosecutorial institutions, giving the drama unusual institutional legitimacy.
  • The essay focuses on a 1984 case in which defense lawyers were arrested after arguing that evidence was insufficient against their client.
  • Luo’s argument is not simply that lawyers need professional protection; it is that criminal justice must defend procedural limits even when public opinion assumes guilt.
  • The drama’s popularity shows how historical legal trauma can be discussed through officially sanctioned culture without directly challenging the current political order.
  • Its boundary is equally revealing: the supplied text describes reflection on past reconstruction, not evidence of a new legal reform agenda.
Open at Pekingnology Ask Claude ChatGPT

Chinese robotics firms move PhanthyMotus from open source to co-building

# AI QbitAI 11h

Paradigm says its PhanthyMotus embodied-agent base now has contributions from more than ten robot makers, 15-plus body adaptations and 200-plus hardware-driver modules.

More
  • The project targets a structural bottleneck in robotics: developers cannot reuse software across bodies, manufacturers absorb custom integration work and customers lack standard deployment packages.
  • The reported 600,000-plus lines of code and support for humanoids, quadrupeds, wheeled robots, arms and drones suggest an attempt at a cross-embodiment middleware layer.
  • Its architecture separates hardware drivers, agent skills and application logic, which could reduce the cost of moving an algorithm between robot platforms.
  • The company plans 50-plus body adaptations, 500-plus algorithms and 10,000 shared skills by the end of 2026, but these are targets rather than delivered capabilities.
  • Remote access to multiple robot bodies and free compute tokens are ecosystem incentives designed to make contribution easier and counter manufacturer lock-in.
  • The project is company-sponsored reporting, so code quality, license scope, governance and independent adoption remain the decisive unknowns.
Open at QbitAI Ask Claude ChatGPT

SenseNova opens an 8B image model for precise editing

# AI QbitAI 13h

SenseTime released SenseNova U1.5 Lite, an 8B open model aimed at complex layouts, multi-image editing and 4K image generation on modest hardware.

More
  • The model’s interesting capability is structured editing: it can preserve an image’s spatial relationships while replacing selected objects, text or visual references.
  • The supplied demonstrations include multi-image recipe cards, long-form infographics and poster edits where reference content must be absorbed without importing unwanted backgrounds.
  • At 8B parameters, the model is positioned for lighter deployment than large closed image systems, though the report provides no hardware or latency measurements.
  • The article claims performance comparable to a closed Image-2 system, but offers demonstrations rather than an independent benchmark or reproducible evaluation.
  • The release shows Chinese model companies competing through open weights and workflow usefulness, not only through a single image-quality score.
Open at QbitAI Ask Claude ChatGPT

ByteDance folds its AI office products into Doubao Work

# AI GeekPark 11h

ByteDance merged the Coze and TRAE Work teams into Doubao while keeping TRAE IDE and CLI for coding, creating a unified Chinese workplace-agent push around Doubao Work.

Also: Leiphone, 36Kr

More
  • The reorganization turns several parallel experiments into a clearer product architecture: general office work converges on Doubao Work, while coding tools remain a distinct line.
  • Doubao Work combines computer use, browser control, file handling, cloud and local execution, connectors and multiple agents rather than presenting a single chat interface.
  • The move appears partly defensive: the supplied article treats Tencent’s reported million-scale daily use of WorkBuddy as evidence that ordinary Chinese workers may adopt packaged agents quickly.
  • ByteDance is responding faster than large internet companies did to earlier non-consensus products, suggesting that the AI market is shortening the time available for cautious internal experimentation.
  • The strategic test is whether a single brand can hide enough complexity for ordinary users while preserving the extensibility developers expect from TRAE and Coze.
  • The supplied report does not disclose active users, retention or revenue for the new product.
Open at GeekPark Ask Claude ChatGPT

HarmonyOS 7 turns system AI into an app distribution layer

# AI GeekPark 4h

Huawei’s HarmonyOS 7 lets apps expose services to the system assistant, run selected agent interactions on-device and use shared AI APIs for search, enhancement and summaries.

More
  • The strategic shift is from measuring app opens to measuring service calls: system-level intent routing can make an app discoverable without a home-screen slot.
  • Huawei distinguishes simple one-request Skills from multi-turn Agents connected through an agent-to-agent protocol, giving developers a staged adoption path.
  • For sensitive applications, the on-device protocol keeps conversation data on the phone and supports encrypted cards that the system assistant cannot read.
  • The platform also exposes image super-resolution, semantic image search and local document summarization through system APIs, moving model deployment and device optimization into the operating system.
  • The architecture resembles Apple and Google intent systems, but Huawei emphasizes letting an application’s own agent retain business logic while the system handles discovery and routing.
  • The supplied example shows the limitation as well as the opportunity: a logistics application had about 500,000 monthly active HarmonyOS users but only about 10,000 monthly active agent users.
  • This matters more if Huawei can make these capabilities available across a large installed base and give developers reliable instrumentation, not merely launch demos.
Open at GeekPark Ask Claude ChatGPT

Spatial intelligence needs a state ledger, not just a rendered world

# Analysis Leiphone 18h

A Chinese analysis of spatial intelligence argues that generated 3D worlds need persistent, verifiable state so objects, rules and consequences survive continued interaction.

More
  • The proposed system separates a renderer that shows the world, a simulator that updates it and a planner that chooses actions; the shared state layer is the architectural hinge.
  • A usable world must track semantic identity, geometry and runtime state together, so a moved table remains the same table and a changed door updates navigation constraints.
  • This makes world generation resemble compilation and continuous integration: natural language becomes constraints, assets fill the scene, solvers enforce geometry and validators decide whether a new state can be committed.
  • The article distinguishes editing the world from acting inside it, implying different permissions, dependencies and rollback semantics for each operation.
  • Its recursive self-improvement loop stores state, goals, actions, verification and feedback, then generates harder cases and gates new versions against regressions.
  • Examples involving Kimi and other model systems show why Chinese developers are using generative models as a route into programmable 3D environments, but the article’s broader technical thesis extends beyond China.
  • The key product metrics become time to first valid scene, cost per valid result and the ability to make local changes without recomputing or corrupting the whole world.
Open at Leiphone Ask Claude ChatGPT

China briefing: Xi, the economy, discipline and US-China ties

# Policy Sinocism 21h

A Sinocism briefing packages Xi Jinping’s meeting with Jordan’s king, economic conditions, a CCDI plenum, food-safety concerns and US-China developments from Chinese official coverage.

More
  • The Jordan meeting readout emphasizes ceasefire, a two-state solution, Palestinian governance and humanitarian responsibility, presenting China as a diplomatic advocate of political settlement.
  • The briefing also flags July fiscal revenue and a CCDI Sixth Plenum, placing foreign policy beside domestic economic management and party discipline rather than treating them as separate beats.
  • A formaldehyde-treated cabbage case shows how food-safety scandals remain a recurring test of local enforcement and public trust.
  • Because this is a roundup, its value is the selection and framing of issues inside Chinese official and semi-official information flows, not a single new event.
  • The supplied text is truncated, so the economic data and US-China sections cannot be assessed in full.
Open at Sinocism Ask Claude ChatGPT

Brazil says Washington admitted China is ahead in minerals

# Policy SCMP 6h

Brazil’s mining minister says US officials privately acknowledged losing the critical-minerals race to China, turning Brazil’s processing capacity into a strategic prize.

More
  • The claim comes from an unscripted speech by Brazil’s minister, not a published US policy document or formal concession.
  • The underlying contest is not simply about ore reserves: China’s advantage lies in processing, refining, industrial equipment and supply-chain coordination.
  • Brazil could become a balancing supplier only if it converts mineral abundance into domestic processing rather than exporting raw material.
  • The story illustrates how Chinese industrial policy has made midstream capacity, not just mine ownership, the central object of foreign competition.
  • The supplied text does not identify the minerals, projects or financing needed to test whether Brazil can close the gap.
Open at SCMP Ask Claude ChatGPT

Chinese J-16s train with Egyptian Rafales amid regional war

# Policy SCMP 10h

China’s air force deployed J-16 fighters for combat training with Egypt’s Rafales in a joint exercise as the United States-Iran conflict reshapes Middle East security calculations.

More
  • The exercise is the second China-Egypt Eagles of Civilization drill and is expected to run into early next month.
  • The choice to train J-16s alongside Rafales gives both sides experience operating across different aircraft, doctrines and logistics ecosystems.
  • The timing matters: China is signaling that it can maintain an operational presence and partnerships in a region where US military action is dominating the security agenda.
  • This is not evidence of a formal Chinese alliance with Egypt, but it expands practical military familiarity and diplomatic visibility.
  • The supplied text does not state the exercise’s scenario, weapons employment or whether the deployment has a deterrent purpose beyond training.
Open at SCMP Ask Claude ChatGPT

Taiwan’s first new F-16s arrive amid concern over China’s J-20

# Policy SCMP 7h

Taiwan is due to receive the first two jets from a $7.76 billion order for 66 F-16 Block 70 fighters as analysts assess their prospects against China’s J-20.

More
  • The deal is worth about NT$247.3 billion, or $7.76 billion, and has been delayed; the first aircraft left South Carolina on August 17.
  • The Block 70 adds modernized avionics and weapons capacity, but the question is whether upgraded fourth-generation aircraft can offset the J-20’s stealth and sensor advantages.
  • Any real balance depends on networks, bases, electronic warfare, missile stocks, pilot training and survivability rather than a platform-versus-platform comparison.
  • The delivery is a concrete step in Taiwan’s long-running effort to improve air defense while cross-Strait military pressure remains central to regional planning.
  • The supplied text does not provide an operational assessment or evidence that the new aircraft materially change the balance.
Open at SCMP Ask Claude ChatGPT

Brazil fines TikTok owner $30 million over children’s data

# Policy SCMP 4h

Brazil’s data authority imposed a record $30 million penalty on ByteDance over TikTok’s handling of children’s data, with deletion and compliance measures attached.

More
  • This is a foreign regulatory action against a Chinese company, making it relevant to ByteDance’s global compliance exposure.
  • Brazil’s authority identified five breaches of local data-protection law and said the penalty is its largest since the agency was created in 2020.
  • The case shows how TikTok’s ownership creates a recurring governance problem: one global product must satisfy country-specific rules on minors, consent and retention.
  • The supplied text does not include ByteDance’s response or explain whether the findings concern TikTok’s global systems or Brazil-specific operations.
  • The strategic significance will depend on whether other regulators use the decision as a template for larger remedies.
Open at SCMP Ask Claude ChatGPT

China’s business news roundup mixes food safety, robots and AI

# Analysis 36Kr 20h

A Chinese business digest highlights a pesticide-disinfection investigation, a reported $900 million robot financing, a major live-commerce pay restructure and ByteDance’s Doubao integration.

More
  • A Xiamen investigation was opened after a pest-control company was accused of using dichlorvos around restaurant premises, illustrating how food-safety enforcement often begins with public reporting and expands into sector inspections.
  • The roundup reports that XPeng’s robotics business raised more than $900 million at a valuation of about $6.3 billion, with the parent retaining control; the supplied text does not provide financing documents.
  • It also describes Oriental Selection cutting total payroll from 1.2 billion yuan, about $169 million, to 800 million yuan, about $113 million, even as headcount rose, a sharp example of replacing superstar-host economics with a broader presenter matrix.
  • The Doubao item shows ByteDance consolidating office-AI teams and products around a single agent brand, linking the business and labor stories to a wider move toward standardization and control.
  • As a roundup, it is evidence of what Chinese business media considers salient rather than one independently developed investigation; the items should not be read as equally verified.
Open at 36Kr Ask Claude ChatGPT

Apple’s Mac mini price jump exposes the consumer cost of AI memory demand

# Analysis GeekPark 5h

Apple’s China prices for the new Mac mini rose by 2,500 yuan per comparable configuration as AI data-center demand tightened memory supply and lifted PC costs.

More
  • The Chinese-market increase is roughly $350 per configuration; the new Mac mini starts at 6,999 yuan, about $986, while Mac Studio pricing is unchanged.
  • Apple says the M6 uses a first-generation 2-nanometer process and delivers four times the AI performance of M4, but the article’s more important point is the cost structure around memory.
  • The source cites first-quarter DRAM contract-price growth of 93% to 98%, with PC DRAM rising more than 100%, as suppliers shifted capacity toward higher-margin HBM.
  • Entry-level machines absorb the shock more visibly because memory is a larger share of their bill of materials; the article cites estimates rising toward roughly a quarter or more of PC cost.
  • Chinese users had popularized Mac mini as an always-on host for agent tools, and Apple has now adopted that use case in its own product messaging.
  • This is a China-relevant market analysis rather than a purely Chinese hardware story: local pricing, local agent usage and global memory economics interact in one product.
  • The long-term question is whether AI infrastructure demand permanently erodes the low-cost PC tier or whether new memory capacity reverses the price pressure.
Open at GeekPark Ask Claude ChatGPT

Fire kills three at Sinopec-linked gas complex in Russia

# Signal SCMP 1h

Three workers, including two Chinese citizens, died and more than 100 were injured in a fire at a Sinopec-linked petrochemical joint venture in Russia’s Amur region.

More
  • The Amur Gas Chemical Complex is a joint venture between Russia’s Sibur and China’s Sinopec, making this a China-linked industrial and overseas labor story rather than a local accident alone.
  • The company confirmed three deaths, two involving Chinese citizens, and more than 100 injuries in the supplied report.
  • The incident highlights the human and operational exposure created by Chinese companies participating in large foreign energy projects.
  • The available text does not identify the cause of the fire, the injured workers’ nationalities or whether production will be disrupted.
  • Its strategic importance depends on whether the investigation finds an isolated failure or a broader safety problem at a flagship cross-border project.
Open at SCMP Ask Claude ChatGPT

Chinese archaeologists begin a joint search for links in Peru

# Culture SCMP 6h

Chinese and Peruvian institutions will conduct fieldwork at Peru’s Caral archaeological zone, China’s first South American project seeking ancestral links between ancient civilizations.

More
  • Caral is described as the oldest known urban center in the Americas, making the site scientifically important independent of the bilateral framing.
  • The project will pair Chinese archaeologists with Peruvian counterparts under an agreement signed by national research institutions.
  • The ancestral-link premise is a research question, not an established historical connection; the supplied text does not describe the methods or evidence expected.
Open at SCMP Ask Claude ChatGPT

China’s 2026 box office passes $3.9 billion

# Signal Jiemian 5h

China’s 2026 box office, including presales, passed 27.5 billion yuan, while the summer season exceeded 11.7 billion yuan and set attendance records.

More
  • The annual total is roughly $3.9 billion at 7.1 yuan per dollar; the summer-season figure is roughly $1.65 billion.
  • The summer market reportedly recorded more than one billion yuan, or about $141 million, on each of 46 consecutive days.
  • Both attendance and screenings reached records, suggesting the result reflects volume and reopening demand rather than ticket-price inflation alone.
  • The supplied report does not identify the films driving the result, regional differences or whether the pace is sustainable after the summer period.
Open at Jiemian Ask Claude ChatGPT

Study says TikTok’s design can undermine critical thinking

# Culture Solidot 4h

A UK study found young people remain skeptical of TikTok, but its rapid, algorithmic feed makes sustained reflection difficult; Chinese coverage presents this as a design problem.

More
  • The study interviewed 16-to-25-year-olds in England, who described TikTok as entertaining but unreliable rather than a serious learning environment.
  • Participants said rapid alternation between serious and entertaining clips leaves little time to assess claims or emotional effects.
  • The findings are relevant to TikTok as a Chinese-owned platform because they target an engagement architecture central to its global product model, not a single piece of content.
  • Users reported trying to train the recommendation system by liking or skipping videos, while finding its feedback inconsistent and harmful material persistent.
  • The supplied report does not establish causation or show that the findings generalize beyond the interviewed UK participants.
Open at Solidot Ask Claude ChatGPT

Chinese electric tricycle exports surge as makers customize for local markets

# Signal Jiemian 5h

Chinese electric tricycle exports reached 168 million yuan, about $24 million, in the first four months of 2026, up 87.2% year on year.

More
  • The export value is about $24 million at 7.1 yuan per dollar, making the growth rate more notable than the absolute scale.
  • Manufacturers are adapting payload, tires, body configuration and solar charging to local conditions in Europe, Southeast Asia and high-sun markets.
  • The products sit at the intersection of low-cost manufacturing and informal logistics: a vehicle can shift between passenger and cargo use with a folding platform.
  • This is a useful example of Chinese factories competing through rapid customization rather than exporting one standardized global model.
  • The supplied figure is an export total, not evidence of profitable market share or regulatory approval in each destination.
Open at Jiemian Ask Claude ChatGPT

Chinese team sets efficiency record for large-area thin-film solar

# Signal IT Home 2h

A Chinese Academy of Sciences team reached a certified 13.0% efficiency on a 10.5-square-centimeter copper-zinc-tin-sulfur-selenium module, a reported world record.

More
  • The material uses relatively abundant, inexpensive and less environmentally problematic elements than many competing photovoltaic chemistries.
  • The scaling problem is important: the team says film uniformity normally degrades sharply as the device gets larger, dragging efficiency down.
  • The reported solution tunes metal-ion and organic-molecule coordination with formamide, then uses solution coating suited to continuous production.
  • The small cell reached 14.2% certified efficiency, while a six-cell module covering 10.5 square centimeters retained 13.0%.
  • This is a laboratory manufacturing milestone, not evidence yet of competitive lifetime, yield, module cost or bankability.
Open at IT Home Ask Claude ChatGPT

Chinese heritage fragrance brand raises up to about $14 million

# Culture 36Kr 18h

Chengbai completed a pre-Series A round of several tens of millions of yuan, roughly $4 million to $14 million, to expand products built around Chinese intangible heritage and incense traditions.

More
  • The brand combines incense, fragrance holders and car-hanging products, translating craft traditions into modern self-use, gifting and social-commerce categories.
  • It says gross sales value has grown more than 200% annually for nearly three years and that a car-fragrance product reached the top of its niche on Douyin.
  • The category reflects a broader Chinese consumer shift in which line incense is being detached from strictly religious use and recast as home ritual, wellness and gifting.
  • The company is investing in its own supply chain, including incense laboratories and jasmine-scenting processes, rather than relying only on cultural storytelling.
  • The financing report is company-interview based; market growth and search-share figures are not independently verified in the supplied text.
  • Its cautious approach to physical stores suggests a consumer brand trying to turn regional craft and Chinese aesthetics into repeatable product standards before expanding offline.
Open at 36Kr Ask Claude ChatGPT

A Chinese fan navigates a post-ban Korean pop landscape

# Culture Initium 21h

A feature follows one Chinese fan’s cross-continental effort to pursue Korean pop culture ten years after informal restrictions on Korean entertainment in China.

More
  • The story’s premise is the gap between state-level cultural restrictions and consumer behavior: fans continue to build routes around official limits rather than simply abandoning the culture.
  • A fan’s itinerary can reveal how concerts, travel, platforms and diaspora networks substitute for normal domestic access when entertainment circulation becomes politically conditional.
  • The title suggests a community story about backlash, adaptation and persistence rather than a formal policy reversal.
  • The supplied text is mostly page code, so specific events, people and evidence behind the account cannot be verified.
Open at Initium Ask Claude ChatGPT

Chinese AI firms are being courted for Malaysian projects

# Policy QbitAI 16h

A WAIC-backed event in Kuala Lumpur is marketing direct access to Malaysian government, finance, education and industrial buyers for a limited group of Chinese AI companies.

More
  • The event positions Malaysia as a Southeast Asian bridge for Chinese AI vendors seeking local infrastructure, policy guidance and enterprise distribution.
  • Its proposed buyer set includes government agencies, banks, universities, system integrators and cloud partners rather than only consumer-facing customers.
  • This is event marketing, not evidence that the named buyers have committed budgets or that any deployment will result.
  • The useful signal is that Chinese AI companies increasingly treat market entry as an ecosystem and trust problem, not simply a model-export problem.
Open at QbitAI Ask Claude ChatGPT

Xiaomi says its smart car door can prevent dooring crashes

# Signal IT Home 5h

Xiaomi says the N90 Max’s powered door combines nine cameras, four door radars and forward lidar to warn of approaching traffic and stop before hitting obstacles.

More
  • The safety design uses multimodal sensing to handle both moving road users and static obstacles beside the vehicle.
  • The door can warn and hold position when a car, bicycle or pedestrian approaches, while adding resistance if a passenger continues trying to open it.
  • Xiaomi also claims the system can resume opening after a nearby obstacle clears, which makes it a control-policy problem rather than a simple emergency lock.
  • The supplied claims are manufacturer answers to user questions, not an independent safety assessment or regulatory certification.
Open at IT Home Ask Claude ChatGPT

Dove apologizes after a disputed short-video script

# Culture Jiemian 4h

Dove apologized for an inappropriate script and customer-service response on its official Chinese storefront, blaming a third-party operating team while accepting brand responsibility.

More
  • The brand says the video violated its stated values and was published without the required review process.
  • Its apology also addresses the storefront’s customer-service replies, showing that the controversy expanded from creative content to frontline brand governance.
  • The episode illustrates the risk of outsourcing official social-commerce operations in China, where a contractor’s post is still treated as the brand’s speech.
  • The supplied report does not identify the wording that triggered criticism, so the consumer and cultural fault line cannot be assessed directly.
Open at Jiemian Ask Claude ChatGPT

Xiaomi removes some secondhand language from phone service rules

# Policy IT Home 5h

Xiaomi revised its Chinese phone-service policy by removing one reference to secondhand products from general warranty exclusions while retaining restrictions on special benefits.

More
  • The change appears narrower than a restoration of full after-sales rights: phones that were recovered, resold, gifted or transferred still cannot use the no-repair replacement or battery programs.
  • The episode shows how Chinese electronics brands are navigating the tension between formal consumer protections and tightly defined promotional service benefits.
  • The supplied report does not say whether the wording change reflects regulator pressure, litigation risk or a customer-relations decision.
  • The underlying statutory guarantee system covers repair, replacement and refunds in specified quality-failure cases, but the company’s special programs can impose separate conditions.
Open at IT Home Ask Claude ChatGPT
Source ledger
ChinAI 403 0/0 HTTPError: 403 Client Error: Forbidden for url: https://chinai.substack.com/feed
ChinaTalk 200 0/20
Pekingnology 200 1/20
Ginger River Review 200 0/20
Sinocism 200 1/20
QbitAI 200 10/10
36Kr 200 6/30
Solidot 200 11/20
Leiphone 200 12/20
GeekPark 200 4/30
IT Home 200 12/60 +18 over cap
SCMP 200 10/50 +9 over cap
Initium 200 5/15
Jiemian 200 8/30 +14 over cap
China Digital Times 200 0/15
The Wire China 200 0/50
Techmeme 200 0/15
Hacker News 200 0/20
V2EX 200 8/50 +33 over cap