Is AI buildout a once-in-a-lifetime investment opportunity? BlackRock proposes a new 50/30/20 asset allocation framework
According to a CNBC report, BlackRock, the world’s largest asset manager, describes the AI buildout boom as a “once-in-a-lifetime” investment opportunity. The traditional 60/40 stock-bond allocation is no longer sufficient to diversify risk. It recommends a new structure of 50% equities, 30% fixed income, and 20% private assets, with an expectation that the size of alternative assets could reach $3 trillion by 2030. Why discuss allocation now? Inflation has surged to 3.3%, and the rate path is hard to read. The backdrop for BlackRock’s argument is that global bond holdings are facing dual pressures from both inflation and interest rates. In the eurozone, August inflation rose to a nearly three-year high of 3.3%, and the ECB increased rates again by 1 basis point on the 10th. In the U.S., the rate path is similarly difficult to gauge.
22-Year-Old Singaporean Pleads Guilty in the U.S.! Led a Cross-Border Group to Steal Bitcoin, Laundered Over $245 Million
A major development has emerged in a large social-engineering theft case that shook the U.S. cryptocurrency community. On September 8, the U.S. Department of Justice announced that Malone Lam, a 22-year-old Singaporean citizen, pleaded guilty in federal court in Washington, D.C., admitting that he led a multinational network that stole and transferred cryptocurrency worth more than $245 million using methods such as social engineering, break-ins to access information, and money laundering. Lam granted approval in court for a guilty plea to a RICO (Racketeer Influenced and Corrupt Organizations Act) conspiracy charge, which carries a maximum sentence of 20 years in prison. The case is being heard by U.S. District Judge Colleen Kollar-Kotelly, and the court has scheduled a status hearing for December 8. A sentencing date has not yet been formally set.
The first advanced packaging cluster in all of Taiwan! Kaohsiung’s Phase 1 investment solicitation is instantly sold out—TSMC secures 25 hectares
The first advanced packaging industrial cluster in all of Taiwan takes root in Kaohsiung’s Bai-Pu! After the Nanzih advanced manufacturing facility area, Kaohsiung’s semiconductor map has officially expanded toward advanced packaging equipment, materials, and R&D verification. TSMC (2330) and its supply chain were the first to reserve 25 hectares for the move-in. Currently, the park’s Phase 1 investment solicitation has been fully subscribed, with construction expected to begin this September. The first advanced packaging cluster in all of Taiwan has arrived! TSMC has secured more than half of the industrial-only zone According to a report by 財訊, the Kaohsiung Bai-Pu Park is located on the Taiwan Sugar Bai-Pu Farm, with a total area of 88.73 hectares. Of this, about 53 hectares are planned as an industrial-only zone, including 25 hectares to be allocated for use by TSMC and its key supply chain. On the south side, TSMC is expected to establish two centers, while on the north side the Kaohsiung City Government will hold an investment solicitation briefing in mid-September to attract international companies.
Has GPT-6 Astra made prompt engineering obsolete? Codex: More isn’t always better for Skills, AGENTS.md
As Coding Agent capabilities advance rapidly, the enormous amount of prompts, rules, and operational workflows that were previously stacked to prevent the AI from making mistakes may, in the GPT-6 Astra era, instead become a burden that limits the model’s capabilities. OpenAI engineer Eric Provencher (@pvncher) recently posted "Rethinking skills and prompts for GPT-6 Astra," saying that if developers have continued using AI agents over the past year, they likely have accumulated a large number of instructions meant to "guide the model along," including Skills, AGENTS.md, and various task prompts. But as GPT-6 Astra improves its understanding of ambiguous contexts, decision-making, and autonomous execution, the rules originally designed for older models may not only be unnecessary anymore, but could also reduce the performance of the new model.
Millennium Prize breakthrough sparks ethical controversy! Mathematicians use Codex for research—then does OpenAI move in first to solve it?
On the 9th, OpenAI announced that its next-generation model—demonstrably surpassing GPT-6 Astra—has proposed a solution to a Millennium Prize problem: the Navier-Stokes Millennium Prize problem. This was originally supposed to be a major breakthrough for both AI and the mathematics community. But the announcement raised a sensitive question: if researchers give unpublished research results to AI tools like ChatGPT and Codex for assistance in processing, could this content in turn help OpenAI’s models improve—and even allow OpenAI itself to become a research competitor? Mathematicians use Claude and Codex to assist with research The central figures in this controversy are New York University mathematician Tristan Buckmaster and Levent Alpöge, a mathematician at Anthropic. Over the past year, the two have privately collaborated on related research issues and have also extensively used AI tools such as Claude and Codex to assist in deriving and organizing research content that has not yet been made public.
The bald white-haired stock god is bullish on ESMT: may replicate SanDisk’s performance, with an opportunity for revaluation in niche memory
Serenity, dubbed by the community as the “bald white-haired stock god,” has recently singled out Taiwan-listed ESMT (ESMT, 3006), saying that this memory manufacturer, which has long focused on low-density DRAM, SLC NAND, and NOR Flash, may be able to replicate the strong performance SanDisk (SNDK) previously saw as memory prices rose. Serenity believes that for end products, a few dollars increase in an old-generation memory chip may not matter much; but for ESMT, the ASP increase of those few dollars could translate directly into a very substantial profit increment. If a DDR2 chip rises by $2, consumers may not notice it, but ESMT may feel it a lot.
AI Articles Can Even Be Caught in Japanese! Japanese Author Tests Pangram Tool: 100% Hit on Gemini-Generated Content
Can AI-written articles really be detected? Earlier, when Substack introduced an AI article detection tool called Pangram, it sparked discussion about whether paid content had been heavily ghostwritten by AI. Now Japanese media has also actually tested with “Japanese.” Japan (Gendai Business) author Masakazu Kobayashi recently tested Pangram by inserting both Japanese text generated by Google Gemini and articles he wrote himself before generative AI became widespread. The former was judged as “100% AI-generated,” while the latter was judged as “100% human-written.” Although just two samples are far from enough to prove that Pangram is also officially claimed to be 99.99% accurate for Japanese, it at least shows that: even though Pangram currently primarily offers its service in English, it already has the ability to recognize AI-written Japanese articles.
Can Stablecoins Be Included in M1 or M2? Fed Research Explores the Possibility, Moving Toward Official Money Supply Indicators
The Fed’s research on stablecoin classification suggests they may be included in M1 or M2 in the future. The Federal Reserve (Fed) released its latest FEDS Notes research on September 4, exploring new forms of money such as tokenized deposits, tokenized money market funds, and payment stablecoins, and how they should be incorporated into U.S. official measures of the money supply in the future. The study notes that payment stablecoins have partial functions of money. Depending on the real-world usage scenarios, they may be classified in M1 in the future, or included in M2 as items that are not part of M1. Currently, the Fed publishes monthly data on the monetary base, M1, and M2 via H.6. M1 mainly includes currency in circulation, demand deposits, and other highly liquid deposits, emphasizing assets that can be used immediately for transactions. M2 includes M1 plus additional assets such as small-denomination time deposits and retail-oriented money market funds, which are more geared toward short-term saving purposes.
Korea to launch tokenization of securities in 2027! Three-phase blueprint unveiled: expands to stocks, bonds and funds
The Korean government has released a new blueprint for the digitalization of capital markets. On September 4, the Financial Services Commission (FSC) held the third joint public-private meeting of the tokenized securities task force, announcing that starting in February 2027 it will build the tokenized securities infrastructure in three phases. The planned application scope will expand from existing fragmented investment products to traditional financial products such as stocks, bonds, and funds. Amendments to Korea’s (Electronic Securities Act) and (Capital Markets Act) were passed by the National Assembly on January 15, 2026, and the related regulations are scheduled to take effect on February 4, 2027. The legislative changes formally recognize that issuance and transfer records of securities can be maintained using blockchain distributed ledgers, and they grant legal effect to the relevant ledgers. Korean regulators emphasized that tokenized securities are not a new category of asset, but one method of issuance alongside paper securities and electronic securities; they must still comply with existing requirements for securities registration, information disclosure, and intermediary broker license.
SoftBank’s Masayoshi Son: Humans will become super-humans! Business performance indicators should change to "AI return on investment"
The world is moving toward full AI adoption across major industries. At that time, it is expected that 100 trillion AI intelligent agents and 1 billion humanoid robots will come into play. How can companies prepare themselves to become AI-native businesses? In his keynote address at SoftBank World 2026, SoftBank founder Masayoshi Son said that companies need to introduce Return on AI—"Return on Artificial Intelligence Investment"—as an operational metric, and transition systematically into an AI-Native "artificial intelligence-native enterprise." Artificial superintelligence leads global value creation growth Sun Masayoshi explained in his speech that the economic output of Artificial Superintelligence (ASI) is expected to account for 20% of global GDP by 2040, reaching an annual output of 7,000 trillion yen. Over time, the labor market will gradually shift to a hundred trillion AI agents and 1 billion humanoid robots. These systems have the ability to operate independently, communicate, and continuously learn, becoming the primary labor force on production lines. To meet the corresponding computing-power demand, the capacity of global AI data centers (AI Data Center) needs to reach 3 terawatts. The average annual investment is expected to be about $5 trillion, thereby driving computing power to Quetta-level (1030).
Working with Wangdao Bank to Launch the Bito.ONE Pilot Program, BitoGroup Rolls Out Taiwan’s First Enterprise-Grade Stablecoin Settlement Case
[The following content is provided by BitoGroup] Stablecoins are gradually being incorporated into the world’s mainstream financial systems, and Taiwan has also entered a new phase of integrating stablecoins with banking. BitoGroup, a virtual asset firm in Taiwan, announced the official launch of a pilot program for its enterprise-grade stablecoin settlement solution, “Bito.ONE.” Wangdao Bank has selected BitoGroup to build an institutional-level stablecoin settlement service, implementing first a model of collaboration between virtual asset service providers (VASPs) and banks. This covers processes including enterprise payments, transaction verification, exchanges between stablecoins and fiat currencies, and settlement, and will further be rolled out to real enterprise applications to position early for cross-border enterprise stablecoin financial services.
Can fruit flies trade crypto? Coinbase engineer opens source Stonkfly project: gives a fruit fly $100 to trade Bitcoin
Can fruit flies trade crypto? An engineer is giving a "fruit fly" $100 to trade Bitcoin. It sounds strange, but this is indeed the latest open-source project by Coinbase engineer Alex Wormuth called Stonkfly. He also asked curiously: "Can fruit flies become rich?" Stonkfly is a fruit-fly brain-connection simulation system used to operate cryptocurrency trading accounts. It is based on the MaleCNS v1.0 fruit-fly connectome released by Google, retaining 166,700 neurons and 25.6 million connections. The project first converts publicly available Bitcoin prices into RGB charts, then stimulates 3,335 brightness inputs and 811 color inputs. Fixed neural readout mechanisms then generate buy, sell, or hold recommendations. After the Coinbase AgentKit ActionProvider checks the limits, it places orders through Coinbase Advanced.
You can’t just wait for QDay! The U.S. pours $300 million into three quantum companies, and discussions about Bitcoin’s post-quantum migration heat up
The U.S. government is ramping up investment in the quantum computing industry, bringing the “quantum threat” that Bitcoin has long faced back from theoretical risk to a practical engineering issue. On September 8, the U.S. Department of Commerce officially finalized CHIPS R&D incentives for Rigetti, D-Wave, and Quantinuum. Each company can receive up to $100 million, for a combined total of up to $300 million. The funding will be used for technologies such as quantum chips, cryogenic systems, improvements in error rates, and fault-tolerant quantum computing. It is worth noting that the $300 million is not the entirety of Washington’s quantum investment this time. On the same day, photonic quantum computing company PsiQuantum also received the highest possible $100 million in R&D incentives. GlobalFoundries received up to $375 million, including efforts to establish domestic U.S. capabilities for quantum wafer foundry services. As early as May this year, the U.S. Department of Commerce announced that nine companies expressed intentions to invest a total of $2.013 billion in the quantum industry, indicating that policy is gradually shifting from foundational research toward manufacturing and large-scale hardware.
Forecast Bitcoin to reach $400,000 by 2030! Coinbase CEO: The bottom of this cycle has already appeared
Coinbase CEO is bullish on Bitcoin, saying the bottom of this price cycle has already appeared After Bitcoin ($BTC) underwent a major correction this year, Brian Armstrong, co-founder and CEO of the cryptocurrency exchange Coinbase, has once again publicly turned bullish on the outlook. He believes the price bottom of this market cycle for Bitcoin may already have appeared, and that in the next one to two years it could return to an upward trend. In an interview with Bloomberg Television on September 10, Armstrong said that he personally believes Bitcoin has already passed the price bottom of this cycle. As the next Bitcoin halving gradually approaches, the price may move upward again in the next one to two years.
Micron unveils the highest bonus ever! A million-dollar thank-you bonus for Taiwan employees; production-line pay could reach up to 68 months
Memory giant Micron announced on the 11th that it will distribute compensation for the 2026 fiscal year (FY26) to more than 60,000 employees worldwide. The scale is the largest since the company was founded. For eligible employees in Taiwan, each person can receive a NT$1,000,000 thank-you bonus. For production-line personnel, total compensation is equivalent to 35 to 68 months of salary. Related coverage: Micron prepares to counter a strike! Urgently rolls out the highest performance bonus in history, with the CEO appearing in a video to show support for Taiwan Micron Taiwan to distribute a million-dollar thank-you bonus! Additional performance bonus up to 500% Micron stated that employees who join in Taiwan before August 29, 2025 can receive a NT$1,000,000 thank-you bonus per person. For new employees during the fiscal year FY26, the bonus will be paid proportionally based on their time on the job. In addition to cash incentives, the annual performance bonus can reach up to 500% of the base target. There are also additional stock awards; for entry-level engineers, the average total compensation is about NT$2.9 million. If stock awards are included, total compensation can reach NT$3.4 million. For production-line personnel, compensation is equivalent to about 35 to 68 months of salary.
Jensen Huang Refutes the “AI Will Destroy Humanity” Video With Over 100 Million Views: They Spread Panic to Sell New Products
Huang Renxun: The recent AI doomsday panic is to sell new products At the latest Goldman Sachs Technology and Media Summit, NVIDIA CEO Jensen Huang said that the outside world’s fear of AI is essentially a topic that the cybersecurity industry has stirred up in order to create new business opportunities. In the investment world, Huang Renxun has long been seen as one of the so-called AI barometers. Now, however, his remarks may face the risk of the main narrative going off track—especially as public concerns that AI may cause large-scale destruction continue to intensify. According to (Axios), Huang Renxun said on a forum, “The reason everyone talks so much about cybersecurity today is because this industry is about to release some new products.” He also asked, “What other way is better at creating demand than manufacturing a problem? Who wouldn’t want their market to fall into a frenzy, with consumers lining up to pay for purchases?”
Don’t just watch the Fed! Experts reveal Bitcoin’s “turning-point code”: Nonfarm hits hardest, Core CPI lasts longer
Markets often treat the Federal Reserve (Fed) interest rate decision as the most important macro trigger for Bitcoin, but Coin Metrics’ latest research points to a different answer: if you look only at the first 30 minutes after a news release, the U.S. employment data is actually the piece of information most likely to cause BTC to abruptly change direction. In a comparison published by Coin Metrics on September 8 (Bitcoin’s Shifting Macro Identity), it analyzed BTC volatility around the nonfarm payroll report, the Core Consumer Price Index (Core CPI), and FOMC decisions during the period from January 2025 to September 2026. The results show that within the first 30 minutes after the jobs report is released, the median BTC absolute price change is about 2 times that of a typical non-event period. Core CPI shortly follows at about 1.8 times. By contrast, the FOMC rate decision—what the market pays the most attention to—saw volatility in both observation windows remain roughly at normal baseline levels.
Driving the tokenization of agri-assets! Indian agricultural finance platform partners with Avalanche to put $2 billion in grain receipts on-chain
Indian large-scale agricultural warehousing and storage merchant trials RWA; grain receipts are moved to Avalanche India’s large agricultural warehousing and agri-finance platform, Arya.ag, is testing the tokenization of grain warehouse receipts. Using Avalanche technology, it builds a dedicated Layer 1 blockchain so that farmers, storage operators, and financial institutions can verify ownership and collateral status of agricultural products within the same system. The initiative was publicly introduced by Indian tech entrepreneur Nandan Nilekani at the Global Fintech Festival in Mumbai. It is currently in the testing phase, with three major banks participating, though their names have not yet been disclosed.
Up to US$3B in size! Tether launches a private credit fund, with USDT targeting the SME funding gap
Tether launches a private credit fund, targeting the funding gap for small and medium-sized enterprises Global stablecoin leader Tether and London private credit asset management firm Fasanara Capital have recently announced the joint launch of a private credit evergreen fund called “StableFund.” The StableFund is jointly seeded by the two companies with an initial US$400 million, with a future goal of raising up to US$3 billion in funding from third-party investment institutions. As the global private credit market is expected to grow from an estimated size of around US$3 trillion to US$5 trillion by 2029, the purpose of establishing StableFund is to ride this trend and direct large capital toward small and medium-sized enterprises that have long been overlooked by traditional finance.
U.S. PPI surges beyond expectations! Bitcoin drops below $77,000 as liquidations across the market exceed $560 million
U.S. PPI up 5.4% year over year as inflation concerns flare up again The latest U.S. inflation data once again brought pressure to the market. On September 10, the U.S. Bureau of Labor Statistics (BLS) released figures showing that the August Producer Price Index (PPI) rose 0.4% month over month, higher than July’s 0.1%; the year-over-year rate climbed from 4.8% in July to 5.4%, indicating that producer-side inflation has accelerated again. Looking at the details, the biggest price pressure in August came from goods. Final demand goods prices increased 1.1% in a single month, with more than three-quarters of the rise attributable to energy prices; final demand energy prices rose 4.2% month over month.