xin

The official account of Alibaba Cloud stated that the large model company Step Star has globally released its flagship open-source model Step 5 Preview, positioned as a primary working model for real-world intelligent agent tasks, excelling in AI programming, software engineering, professional knowledge work, and financial scenarios. The model has a total of 600 billion parameters and 27 billion active parameters, utilizing sparse MoE, with a maximum context window of 1 million, capable of processing text, images, and videos. The article mentions that it recently ranked second globally in both weekly and daily call volume on OpenRouter. On the day of the release, Alibaba Cloud announced the completion of DAY0 adaptation, optimized for joint inference based on chips, super node instances, and models, with inference computing power provided by the Lingjun super node.Step Star has developed its own intelligent agent reinforcement learning training platform, with the sandbox platform managed by Alibaba Cloud. The solution is based on Alibaba Cloud's Agent Sandbox and Kubernetes container service, aimed at elastic computing power in training mode, claiming to create tens of thousands of sandboxes in a single minute, with a maximum concurrency of up to one million, compatible with E2B API, and providing native Kubernetes interfaces. Through multi-level image caching, intelligent pre-scheduling, and control link optimization, it achieves millisecond-level application and minute-level creation of 30,000 sandboxes, with the sandbox lifecycle uniformly managed by Alibaba Cloud's backend.In terms of storage, each Pod starts with 15GB of free temporary storage, automatically increasing by 30GB each time when usage exceeds 70%, up to a maximum of 300GB. For security, MicroVM-level isolation is adopted by default, covering computing, networking, storage, and permissions, with zero-trust access to cloud products and large model interfaces. The results provided in the article include: a 90% reduction in image startup time, model evaluation cycles compressed from weekly to daily, and an average savings of about 40% on EBS storage costs; relying on ACK One Fleet, it can support the stable operation of hundreds of thousands of sandboxes simultaneously.

On October 22, at the 2026 World Physics AI Conference and Yushu Partner Ecosystem Conference, Yushu Technology founder Wang Xingxing stated, "I am really very afraid of wasting each day," demonstrating a strong sense of urgency regarding the acceleration of physical AI.Wang Xingxing believes that top AI models have already penetrated the mathematical world, and physical AI has also shown signs of being penetrated, with the breakthrough likely occurring in the next 1 to 2 years or 2 to 3 years. Live demonstrations showed that the Yushu robot equipped with GPT-6 could complete tasks such as painting, playing chess, and recognizing color palettes without additional training, indicating that the model can solve some planning problems. He expects that by the end of this year or next year, the robot is expected to complete tasks like organizing a room as instructed.In light of this time window, Yushu Technology officially launched the "Yuchuang Plan," opening recruitment for embodied intelligent entrepreneurs with core technologies, aiming to incubate over 200 ecological enterprises in the next five years. Wang Xingxing also pointed out that after the breakthrough point of physical AI, robot hardware needs to reach a scale of producing tens of millions of units annually, and emphasized the importance of core resources such as data, computing power, chips, and energy.

The 2026 World Physics AI Conference and Yushu Partner Ecosystem Conference was held today in Hangzhou. Wang Xingxing, Chairman, General Manager, and Chief Technology Officer of the robotics company Yushu Technology, stated in his keynote speech that the most critical challenge for embodied intelligence, surpassing the ChatGPT moment, is the alignment of AI model inputs and outputs with real robots.Wang Xingxing mentioned that for top large model companies, in a sense, they have basically achieved self-evolution of models internally, where models can write programs and collect data on their own, and humans only need to provide appropriate guidance and participation. He also noted that it usually takes 5 to 10 years for technological breakthroughs to diffuse to general users, so the external perception is not that profound, while physical AI has already shown signs of being penetrated.

According to Jiemian News, the autonomous driving technology company Neolix has officially completed a new round of financing of nearly $400 million. So far, Neolix's total financing amount has exceeded 10 billion yuan, making it the leader in the unmanned delivery vehicle industry in terms of financing scale.Insiders in the investment circle revealed that Boyu Capital and the National Artificial Intelligence Industry Investment Fund are the lead investors in this round of financing. Alibaba AEF NextGen Hong Kong Fund, Monolith Lisi Capital, Vertex Growth under Temasek, and GAC Capital participated in the investment, while existing shareholders such as UAE Stone Venture, Xincheng Capital, SoftBank Asia, Yizhuang Guotou, and Gao Rong Venture Capital continued to increase their holdings.The National Artificial Intelligence Industry Investment Fund was established with contributions from the National Big Fund Phase III. Previously, its investment focus was concentrated on AI and computing power chips, high-bandwidth memory chips (HBM), large model algorithms, embodied intelligent robots, AI security, and other fields. It has invested in leading companies in key areas of artificial intelligence such as Kimi, DeepSeek (general large model), Muxi (AI computing power chip), Galaxy General (embodied intelligence), and Changting Technology (AI security).

According to The Defiant, THORChain announced that its v3.21 upgrade will fix the FIN contract vulnerability on Rujira and allow the previously paused application layer to restart. The network has set the upgrade target at block height 28,366,100, with an expected time around October 23. After the restart, trading, lending, borrowing, and automated trading on Rujira will resume. Rujira's products run through smart contracts within THORChain, rather than on an independent blockchain, and the application layer can pause these contracts without affecting the underlying layer of THORChain.According to Rujira, the vulnerability was discovered on September 26, when an attacker transferred a custom concentrated liquidity position without holding a position. The team paused the application layer with the support of node operators, and only one position from the Rujira team itself was affected, with no loss of user funds. Rujira attributed the vulnerability to a flaw in the position transfer function, rather than the trading logic of the liquidity strategy. On September 27, Rujira stated that the fix is located at the underlying layer and has been scheduled for the v3.21 upgrade.Regarding the restart window, Rujira proposed a conditional grace period in its plan on October 15: if the application layer is ready to resume, any credit account position with a loan-to-value ratio above 100% and meeting liquidation conditions will request node operators to pause the credit account contract for an additional 24 hours, allowing borrowers to replenish collateral before liquidation resumes. Borrowers can replenish collateral through Rujira Borrow or by directly transferring assets to the credit account address. Rujira indicated on October 10 that positions still deemed unsafe after the grace period may be liquidated when the liquidator triggers the process.

According to Lookonchain monitoring, a new wallet (0xb15b) has encountered multiple liquidations in the past 10 hours, with a total of 8,163 ETH liquidated, worth 19.78 million USD.

Market research firms Counterpoint Research and TrendForce stated on the 18th that China's DRAM manufacturer Changxin Storage's monthly wafer production capacity is expected to expand from the current approximately 320,000 wafers to 420,000 wafers by 2027, adding about 100,000 wafers in just over a year, an increase of 31.3%.The industry expects that Changxin Storage's monthly DRAM wafer production capacity will reach 500,000 to 600,000 wafers by the end of 2028. If calculated at the upper limit of 600,000 wafers, this would be an increase of 280,000 wafers compared to the current level, an increase of approximately 87.5%.

Changxin Technology announced that it expects its operating revenue for the first three quarters of 2026 to be approximately 265 billion to 270 billion yuan, an increase of 232.916 billion to 237.916 billion yuan compared to the same period last year; it is expected that the net profit attributable to shareholders of the listed company will be approximately 131 billion to 132.5 billion yuan, compared to a loss of 5.28 billion yuan in the same period last year, achieving a turnaround.The company stated that due to the growth in demand for AI computing power and the improvement in the prosperity of the DRAM industry, there is a supply shortage of products and prices have risen sharply; the company is advancing capacity construction, product research and development, and process technology improvements, optimizing operational management and expanding the market.In addition, the company's chairman, Zhu Yiming, proposed to repurchase company shares for 1 billion to 2 billion yuan, with the repurchased shares intended for employee equity incentives, employee stock ownership plans, or reducing registered capital. The relevant plan still needs to go through the review process.

Polygon Labs official announcement, Bor v2.1.3 is officially released to the public, previously distributed in private version form to validators and major infrastructure providers. This update fixes two medium-severity denial-of-service vulnerabilities in BlockSTM V2 parallel block execution, both submitted by security researchers through the Immunefi bug bounty program.Unbounded keccak256 cache issue: Attackers can fill hundreds of MiB of cache during the speculative execution phase by constructing specific transactions, leading to node memory exhaustion and delayed block production. The fix is to set a capacity limit on the shared cache. Repeated code hash calculation issue in parallel execution: A single transaction can cause the parallel processor to take about 11 seconds, severely affecting node activity. The fix is to cache code hash results on a per-transaction basis.In addition, this update also includes several internal hardening measures: the transaction pool address filtering adds interception for EIP-7702 SetCode authorization, RPC batch execution pool concurrency limit fixes, consistency fixes for reading balances across self-destruct boundaries in parallel execution, and an automatic cleanup mechanism for pending transactions (default 3 hours). The affected versions are Bor v2.1.1 and v2.1.2. v2.1.3 is a non-hard fork, and nodes can upgrade individually without coordination. No exploitation records of the two vulnerabilities have been found on the mainnet and testnet Amoy.

According to the 21st Century Business Herald, the Chaoyang Branch of the Beijing Public Security Bureau reported that it has recently filed a case for investigation into the suspected illegal activities of related companies of Yixin. Currently, several criminal suspects, including the actual controllers Tang and Zhao, have been subjected to criminal coercive measures in accordance with the law.Yixin Group had previously fully suspended the redemption of quasi-fixed income products on May 22 and issued an apology letter to relevant investors on October 11. Yixin stated that the group has incurred significant losses in multiple businesses, and the market value of asset disposal is relatively low. Preliminary findings indicate that the group may be insolvent and has a significant debt repayment gap.

On October 12, during the in-depth dialogue session at the opening ceremony of the 26th China International Industry Fair, Wang Xingxing, founder of the robotics company Yushu Technology, stated: His faith in AI is growing day by day. The biggest mistake he made in the past decade was not believing enough in AI, and even now, many people may not have fully utilized or embraced AI.Wang Xingxing mentioned that the recent release of GPT-6 has had a significant impact on the cognitive understanding of the embodied intelligence industry. The company team has already deployed it on robots for task testing, and the results have been quite good. He believes that the critical point for embodied intelligence will arrive relatively quickly, possibly within one or two years. However, to achieve large-scale application of robots, issues of generalization and efficiency still need to be resolved, but these problems are expected to be addressed in the coming years.

Tsinghua-based AI infrastructure company Wuwen Xinqiong has completed a C round financing of 2.3 billion yuan, bringing the total financing to 4.3 billion yuan. The company was established in 2023, does not develop chips, nor does it build its own data centers. It integrates the computing power of different chips such as Huawei Ascend, NVIDIA, and AMD through software, providing large model training and inference services to customers.Developers can directly call the models via API without needing to deploy them themselves. Bloomberg reported on October 1, citing informed sources, that Wuwen Xinqiong has secretly submitted a listing application to the Hong Kong Stock Exchange, with the earliest possible listing in the first half of 2027, and the pre-listing valuation is reportedly around 14.3 billion yuan.

Changxin Storage (CXMT), a Chinese memory chip company, announced that it has made key progress on a new storage architecture. Samsung Electronics and SK Hynix are also advancing similar architectures to increase storage density. The company's president, Cao Kanyu, stated at an industry event in Shanghai that Changxin Storage plans to launch new server memory products based on this architecture by the end of 2026, according to a company statement.The architecture is designed as 4F², meaning each memory cell occupies an area of 4F², where F is the minimum line width of the semiconductor. The company claims that this proprietary transistor design can improve storage density and efficiency by accommodating more memory cells on a single chip. Currently, mainstream commercial DRAM uses a 6F² cell structure.

Alibaba Chairman Joe Tsai stated that open-source AI is the best way for Europe to reduce its dependence on American and Chinese technology and to establish technological autonomy. He urged Europe to build its own computing infrastructure for training and running AI models, stating that no country would want to fully trust another country's technology, as a change in government or circumstances could lead to the cutting off of technology supplies.Joe Tsai added that Europe has not fully utilized its AI researchers, as these talents often become targets for recruitment by laboratories in the United States and China. He called on Europe to focus on using proprietary industrial data to train artificial intelligence models for competitive advantage, citing Chinese manufacturing as an example, stating that Chinese factories possess valuable manufacturing data that could be highly valuable for training models. He also expressed optimism about the emergence of new cloud service providers in Europe in recent years, specifically professional cloud computing companies equipped with AI chip data centers.

According to the Financial Times, the European Union is considering implementing a universal tax measure targeting large enterprises to tax major American technology companies. This move aims to increase revenue for the EU while avoiding a strong backlash from the Trump administration.

According to market news, the U.S. Treasury has withdrawn its previous proposal for cryptocurrency regulation concerning unhosted wallets and convertible virtual currency (CVC) mixing services.It is reported that the proposal aimed to expand the scope of monitoring cryptocurrency trading activities, including requiring financial institutions to collect and report relevant information involving unhosted wallets and cryptocurrency mixing services. This withdrawal means that the related monitoring rules will not be advanced for the time being.

According to on-chain analyst Yu Jin's monitoring, the largest short position established the day before Changxin Storage (Changxin Storage) went public has been completely closed 10 hours ago.This address shorted at a price of $6.5 and closed the position at a price of $8.5 after holding for 2 months, resulting in a total loss of $10.98 million, including a funding fee loss of $5.18 million and a price change loss of $5.8 million.

ZachXBT posted on platform X that UPay may allow the sanctioned illegal market Xinbi to issue its Visa crypto card and requested UPay to clarify the situation.ZachXBT stated that Xinbi is mainly involved in pig butchering scams, fraud, money laundering, and human trafficking.

According to a report by South Korea's "JoongAng Ilbo," a statement obtained from a former head of equipment investment at Changxin Memory Technologies (CXMT), who has a background at Samsung Electronics, reveals that the former CEO of CXMT, Wang Ningguo, summoned the Korean team, including the individual in question, from October to December 2016. He demanded they bring back Samsung Electronics' Process Recipe Plan (PRP) and equipment information, threatening to report them to the South Korean government if they failed to do so, and warned them to verify their connections and not to lie. The individual stated that they felt scared at the time. The report mentioned that CXMT's management referred to Samsung Electronics as "Huangshan."After working at Samsung Electronics for 28 years, the individual joined CXMT and was prosecuted for allegedly leaking and using Samsung's 1.6 trillion won investment in the world’s first 18-nanometer DRAM process without authorization. In April of this year, they were sentenced to seven years in prison in the first trial and are currently in the second trial. The PRP includes the sequence of about 600 steps in DRAM manufacturing, equipment, condition values, and the companies and models of the equipment. Park, a former process design engineer at Samsung, hand-copied the relevant content into a 12-page notebook in September 2016 and brought it out, and is currently under an Interpol Red Notice.During the court hearing, the individual stated that there was a claim made during communications with the Chinese side that investment would not be provided to CXMT if they could not produce the 18-nanometer process. They later learned that the other party had wanted the technology from the very beginning, and CXMT initially had no plans to develop it independently. The pressure mentioned occurred before a meeting to prepare for the investment application. CXMT was established in 2016 with an investment of 14.4 billion yuan from Hefei Industrial Investment, a subsidiary of the Hefei municipal government, and had not been profitable for about 10 years, achieving its first annual profit last year with a net profit of 7.1 billion yuan.

According to Caixin, poker master and seasoned cryptocurrency investor Hu Zheweng made significant bets during the three waves of cryptocurrency frenzy involving ICOs, algorithmic stablecoins, and AI tokens, but faced consecutive failures. Hu Zheweng claimed to have invested in the blockchain project Stratis, achieving a return of "over 1000 times."According to disclosed information, Hu Zheweng invested approximately 80 million USD in algorithmic stablecoin TerraUSD (UST) and its sister token LUNA from May 2021 to May 2022, with the peak market value of his holdings exceeding 800 million USD, but the price of LUNA subsequently nearly dropped to zero.In addition, Hu Zheweng has filed a lawsuit in Chicago, USA against the globally renowned quantitative trading firm Jump Trading and its related companies and executives in the cryptocurrency business, seeking at least 500 million USD in damages. After experiencing Stox and Terra, Hu Zheweng has not left the cryptocurrency market; last January, he bet on a new project by a Chinese computer professor Zhang Yongfeng. Zhang Yongfeng entered the Computer Science Department of Tsinghua University in 2007 and is currently a tenured associate professor in the Computer Science Department at Rutgers University in the USA. He has been sued by Hu Zheweng on multiple charges including "securities fraud." It is reported that the tokens issued by the organization founded by Zhang Yongfeng, in which Hu Zheweng invested millions of dollars, have fallen over 99.6% from their peak in 2025.