cien

The Central Science Management Bureau confirmed on the 9th that TSMC's Central Science Phase II 1.4 nanometer factory expansion is fully accelerating. The first P1 factory has completed its steel structure and is expected to begin trial production in April next year, with mass production anticipated in the second half of next year, ahead of the originally scheduled mass production in 2028. TSMC has applied to the Central Science Management Bureau to set up two temporary offices at the site, which are expected to be completed in April next year, with the first batch of over 5,400 operational and outsourced personnel moving in.The advanced process new factory for TSMC's Central Science Phase II park broke ground last October, planning to build four 1.4 nanometer factories, with nearly 2,000 workers working day and night. The P1 factory is currently undergoing floor and exterior wall construction, with the factory building expected to be completed early next year. The P2 factory has begun basic construction and is scheduled to be completed in October next year, with both factories expected to start mass production successively next year. The P3 factory has obtained a construction permit, while the P4 factory is in the process of applying for a construction permit, planning to be built with a six-month gap. P3 is expected to be completed in the second quarter of 2028, and P4 is scheduled for completion in the fourth quarter of the same year. After the P2 factory is completed in the second half of next year, an additional 1,000 operational personnel will be added, with the total number of employees expected to be between 9,000 and 10,000 when all four new factories in Phase II are completed and put into production.

The U.S. Treasury Department announced that it will purchase at least $4 billion and up to $6 billion of long-term government debt in its expanded repurchase program. U.S. Treasury bonds continued to decline after the announcement, indicating that the scale is smaller than some investors had expected.Economist Guha stated that it remains to be seen whether the impact of the intervention can be sustained.

Dwarkesh Patel and Jerry Han published research on September 8, 2026, breaking down the contributions of data and model improvements in pre-training progress from 2019 to 2025. The study was conducted on a smaller scale, focusing on pre-training, corresponding to the publicly released model recipes and publicly available data corpora for each year, and was trained at a computational scale of up to 1e19 FLOPs.The results show that under the 1e19 FLOPs computational budget, the computational efficiency gain from data improvements is 12.0 times, while model improvements yield 3.7 times, with the data-side gain being approximately 3.24 times that of the model side. The gains from data and model improvements are largely independent and interact very little, with the additive effect of both explaining 88% of the variance in OLMES scores under a linear model.On the model side, the evolution from GPT-2 to OLMo-2 encompasses optimizers, positional encoding, normalization, activation functions, and initialization, among others. On the data side, the corpus evolved from approximately 9 billion tokens of OpenWebText in 2019 to larger and more finely filtered corpora like UltraFineWeb by 2025.

Gate Europe CEO Dr. Giovanni Cunti stated in a LinkedIn post that financial markets are becoming increasingly interconnected, with digital assets, stablecoins, and tokenized assets merging with traditional finance, reshaping the pathways for cross-border capital flow. The core of liquidity is no longer just depth and trading volume, but also the efficiency of capital circulation and the ability to access it instantly.He emphasized that future liquidity not only means more capital but also aims to make capital flow more easily and be accessed more efficiently, achieving better connections between different markets. Broad market access can break the fragmentation of liquidity and enhance the efficiency of capital operations. As stablecoins accelerate payment settlements and tokenization introduces new assets, the focus is on promoting seamless collaboration across all links.Relying on the compliance foundation of the MiCA and payment institution PI dual licensing system, Gate Europe has established highly regulated operational standards to enhance the confidence of institutions, enterprises, and investors with a clear regulatory framework. He pointed out that Europe is advancing digital asset innovation alongside regulatory compliance. Gate Europe stated that it will continue to empower this transformation and work with industry partners to shape the future together.

According to the Financial Times, citing informed sources, Chinese memory chip manufacturers Changxin Storage and Yangtze Memory Technologies have purchased enough deep ultraviolet lithography machines (DUV) from ASML to meet their expansion plans for the next three years.

OpenAI Chief Scientist Jakub Pachocki published an article titled "An Alien Mind" on September 6, 2026. The article reviews the results of the RLSlow research project, which emerged in mid-2023, demonstrating the first scalable training of reasoning models. It states that three years later, reasoning language models have become part of rapid economic growth, beginning to push scientific boundaries, capable of operating computers and graphical interfaces, collaborating with humans and other AIs on research projects, while also changing the landscape of computer security and introducing new dangers.The author anticipates that the current pace of progress may continue towards recursive self-improvement based on internal results. If AI development continues along the current path, systems may experience equivalent or greater leaps in capability in the coming years, increasingly driving their own development. The author calls for extreme caution, believing that no one is prepared to deal with the consequences of the rapid rise of machine intelligence. OpenAI will continue to seek technical solutions for alignment and monitoring, build defensive systems, and unilaterally halt further scaling when necessary, but believes broader intervention is needed.The article states that machine intelligence is primarily driven by increased computational power, with AI growing more than being designed. In terms of alignment, it distinguishes between goal alignment and value alignment, with the core challenge being generalization. GPT-6 Astra shows significantly better alignment than GPT-5.6 Sol, but more progress is still needed. In monitoring, the focus is mainly on chain-of-thought monitoring, with assessments showing that reliance on capabilities is gradually weakening. In terms of scalable defense, models are becoming superhuman in breaking into computer systems, currently in a narrow window where using the best available models significantly enhances the security of critical systems.

OpenAI Chief Scientist Jakub Pachocki stated that he hopes to avoid a race towards "unmonitorability" of models due to misleading reports. He mentioned that current cutting-edge models, including Astra, have a computational graph depth that differs from GPT-4 by no more than a factor of two. Since the launch of the first inference models, OpenAI has been committed to preserving and utilizing chain-of-thought monitoring, believing that this technology helps observe how the model's alignment capabilities generalize from the training distribution.Pachocki also pointed out that chain-of-thought monitoring is relatively fragile, and the trend is not optimistic, but it can still be strengthened through research and has become a core objective of OpenAI's current research agenda. Previously, The Information reported that Astra uses recurrent depth, allowing the same set of Transformer layers to compute repeatedly. As a result, some inferences can occur internally within the model, without needing to be entirely expressed as a chain of thought, raising concerns about whether the model will become increasingly difficult to monitor.

Fidelity Global Macro Research Director Jurrien Timmer stated that Bitcoin has maintained its power law curve bottom and has adjusted long enough to meet the time factor of the four-year cycle.

Tencent Group Senior Executive Vice President, CEO of Cloud and Smart Industries Group, Tang Daosheng, published an article in Tencent's internal publication, responding to external doubts about "Tencent being slow in AI." He stated that it is impossible to be completely untroubled; the Hunyuan large model has undergone multiple reconstructions, and Hunyuan Hy3 has shown outstanding performance among large models with equivalent parameters. However, due to the company's overall lack of computing power, the training of models and product development has been significantly slowed down, resulting in a considerable impact.Tang Daosheng mentioned that waking up early may not be the most critical factor; the second half of AI has just begun, and enduring longer provides the opportunity to seize the continuously emerging new opportunities.

Research institution Digitimes stated that with the surge in AI-driven infrastructure spending, major cloud service providers are expected to place greater emphasis on operational efficiency and financial discipline in the second half of 2026. Amazon, Alphabet, and Microsoft reported a combined cloud business revenue of $106.3 billion in the second quarter, with an annual growth rate accelerating to 43%. These companies are expected to increase their total capital expenditures to $74.5 billion.

The research director of Galaxy posted on the X platform stating that even with a high probability of witch nodes, the peak proportion of BIP-110 nodes is only about 20% of the entire network.He mentioned that it is quite ridiculous to see Mechanic and others so shocked by the failure of BIP-110; even accounting for this approximately 20%, its hash power proportion is still far from sufficient, the scale of hash power is very small, and the node proportion is only about one-fifth, still a long way from the target.

The Slow Mist security team disclosed that the cross-chain bridge project Allbridge was attacked on August 19, 2026, resulting in a loss of approximately $190,000. Notably, this attack was not executed instantly; the attacker had begun laying the groundwork nearly a month prior and bypassed the verification mechanism by forging cross-chain messages. According to Slow Mist's analysis, on July 26, the attacker directly called Circle's MessageTransmitterV2.sendMessage function on the Polygon chain, constructing a cross-chain message disguised as a CCTP style message, claiming that a transfer of 1 million USDC existed, but in reality, no USDC destruction operation took place. Subsequently, Circle generated a valid verification proof (attestation) for this complete message according to normal procedures.About 24 days later, on August 19, the attacker waited for the Base Router to receive a real CCTP deposit, increasing the balance to approximately 191,000 USDC, and initiated the attack just 6 seconds later. The attacker utilized the previously forged message and verification proof to call Allbridge's receiveCctpMessage function. Due to the project's lack of critical verification, the system mistakenly recognized the false cross-chain message as a real deposit and recorded a limit of 1 million USDC. The attacker then temporarily borrowed approximately 809,000 USDC through an Aave flash loan, matching the Router balance with the forged amount, and used the internal credit record to call the transfer function, ultimately transferring out approximately 999,000 USDC (after a 0.1% fee). After repaying the flash loan and fees, the attacker netted a profit of about $189,800. The root cause of this vulnerability lies in Allbridge's failure to verify the identities of the sender and receiver of the cross-chain message, as well as not confirming whether USDC was genuinely minted and whether the balance actually increased, instead directly trusting the amounts and message hash data constructed by the attacker. Slow Mist emphasizes that on-chain message verification does not equate to the actual arrival of real assets. Cross-chain protocols not only need to verify the authenticity of messages but must also ensure that the message source is trustworthy, that the receiver is Circle's official TokenMessengerV2, and that asset accounting can only proceed after confirming the actual minting of assets and changes in balance. This incident once again highlights the security risks of cross-chain bridges in the message verification and asset settlement processes.

According to market news, Strive's SATA fundraising amount is expected to be sufficient to purchase more than 160 bitcoins.

Former NVIDIA research scientist Sanja Fidler announced the co-founding of VeedaAI with long-time collaborators ZGojcic and HuanLing6. The company is dedicated to building simulated realities for Physical AI, serving as a key infrastructure layer for robotic interactive learning and evaluation, allowing robotic intelligence to learn at scale through trial and error in virtual worlds.Fidler stated that real-world trial and error training for robots is unsafe, costly, and time-consuming, necessitating the use of simulated realities to achieve scalable interactive learning. The team believes that world models are the core technology, capable of learning from vast amounts of sensor and physical world data to generate simulated environments with quality, diversity, and physical realism. VeedaAI has received support from Khosla Ventures and Radical Ventures, and team members have years of experience in simulation, generative AI, 3D, robotics, and embodied intelligence.

According to a report by Fortune, OpenAI Chief Financial Officer Sarah Friar stated that AI can help finance teams identify issues more quickly, optimize forecasts, and enhance decision support capabilities, but it cannot replace human judgment, accountability, and financial rigor.She pointed out that OpenAI is building an "AI-native" financial system, with goals including achieving "zero-day closing" and continuously updating forecasts, and suggested that companies restructure processes around decision-making to enhance efficiency while strengthening control and accountability mechanisms.

According to on-chain analysts, a certain ancient whale that acquired 40,000 ETH during the 2015 ICO extracted 7,020.84 MKR at an average price of $828.92 between September 2018 and May 2019, worth $5.81 million.After holding for over seven years, the whale transferred 3,510.42 MKR to a new address for the first time 4 hours ago, worth approximately $4.41 million, with an unrealized profit of $1.506 million, and has not yet transferred or sold.

Flashbots strategic leader Hasu stated that charging fees to passive holders, allowing them to pay for projects investing in your ecosystem (such as Lido, Aave, and DAT, which collectively fund Ethereum client teams, multiple EF derivative projects, overall DeFi construction, etc.), is actually very good. The problem with Ethereum has never been excessive ETH issuance, but rather insufficient ecosystem investment.Reducing inflation will inevitably suppress investment reduction. Unfortunately, the Ethereum research community has never been good at predicting these second or third-order effects.

According to Jinshi reports, Google's chief scientist has left after 27 years in office and plans to start his own artificial intelligence company.

According to on-chain analyst Yu Jin's monitoring, an ancient Bitcoin miner address transferred 500 BTC that had not moved for a long time about half an hour ago, worth approximately 31.27 million USD.Data shows that these 500 BTC have not been transferred for over 13 years since they were consolidated from mining addresses in 2013. At that time, the price of Bitcoin was around 1,020 USD, making the 500 BTC worth about 510,000 USD.It is reported that 1,020 USD was one of the price peaks of Bitcoin in the years around 2013, and it wasn't until 2017 that the BTC price returned to above 1,000 USD, indicating that this address experienced a "holding" period of about three to four years. Currently, the value of this batch of BTC has increased by approximately 60 times compared to that year.