加密城市 Crypto City

Machi Big Brother has significantly increased his bets again, with his Ether long position rising to 4,290 coins
The high-leverage Ether trading by Taiwanese entertainer and cryptocurrency investor “Machi Big Brother” Huang Licheng (Jeffrey Huang) has once again undergone a major change. According to the latest Hyperliquid account data, as of the evening of August 17, Huang Licheng holds about 4,290 Ether long positions, using 25x leverage. The position’s notional value is approximately US$8.183 million, with an average entry price of about US$1,894.8.
Source: HyperDash Huang Licheng holds about 4,290 Ether long positions, using 25x leverage. The position’s notional value is approximately US$8.183 million, with an average entry price of about US$1,894.8.

“Everything is broken—Bitcoin is burning.” This quote comes from Calle, the core developer of a Bitcoin cybersecurity red team. He shared this work on X (Decrypt has also compiled it): a volunteer group called Bitcoin Red Team uses open-source code that leverages China’s AI models to scan Bitcoin ecosystems at large scale. Within a short time, it turned up thousands of vulnerabilities, exposing old code that had gone years without scrutiny.
Nearly 5,000 vulnerabilities: 85 critical, 635 high risk
This team is led by developers using the pseudonym Calle (@callebtc). Their approach combines AI tools with human re-review. They primarily use Moonshot AI’s Kimi K3 from China, along with another Chinese model, GLM 5.2, and they also benchmark against models from OpenAI and Anthropic. In August 2026, the team reported a total of 4,962 findings across 390 projects: 85 were classified as critical, and 635 as high risk (this is statistics at a specific point in time; as scanning continues, the numbers are still rising, and some reports have already been updated to more than 7,900 items). Calle said the developers indeed have a large set of real critical and high-risk vulnerabilities on hand, but in line with responsible disclosure principles, they will not publish the affected projects and technical details for now. Among them, software related to the Lightning Network is described as being more fragmented than average because its architecture is complex.

Share price plunges more than 40% in a year; Convano cuts losses and exits Bitcoin
FASTNAIL, a Japanese nail salon brand, and its parent company Convano (TSE: 6574). Last year, after boldly announcing that it would include Bitcoin in its corporate asset reserves, it once sparked a wave of investor enthusiasm and its stock surged dramatically.
However, as the cryptocurrency market has pulled back sharply, as of the time of writing today (8/17), Convano’s share price has crashed to 110 yen—down a cumulative 46.26% compared with a year ago. On August 13, intraday trading even pushed it below the 100-yen level to 99 yen.
Recently, Convano also completed the process of liquidating all its held bitcoins and lost about 300 million yen, effectively declaring the end of this cross-industry experiment in crypto asset reserves—ultimately ending in a disastrous exit after cutting losses.

OCC conditionally approves a trust bank—World Liberty clears a key hurdle
The Trump-family-backed crypto project World Liberty Financial ($WLFI) has made major regulatory progress. On August 14, the Office of the Comptroller of the Currency (OCC) issued a preliminary conditional approval to World Liberty Trust, agreeing to its chartering as a national trust bank, which in the future will be able to provide digital asset custody, issue and redeem USD stablecoins, and related asset conversion services.
This approval is still provisional and conditional; World Liberty has not yet obtained formal authorization to open for business. The OCC clearly requires the bank to complete all pre-opening conditions in order to obtain final approval and begin operations officially. Before then, the regulatory authorities may still revise, suspend, or revoke this approval depending on the actual circumstances.

Mitsubishi UFJ Financial Group (MUFG) of Japan announced that it will use blockchain technology to improve the settlement speed of Japanese government bond trading. In the past, traditional settlement processes often took anywhere from 1 to 3 days; in the future, it is hoped that blockchain can enable “24/7, year-round” real-time delivery.
In the traditional financial system, settlement and clearing of government bonds typically takes 1 to 3 business days. To break through efficiency bottlenecks, Mitsubishi UFJ is working on a proof of concept and plans to conduct on-chain testing via the enterprise-grade blockchain network Canton in order to speed up settlement.

Coldcard vulnerability exploited on a large scale; more than 8,600 addresses affected
The Coldcard hardware wallet explosion—an ongoing large-scale asset theft incident continues to unfold. A recent study by Galaxy Research indicates that the attackers exploited a seed-generation vulnerability affecting some Coldcard devices, and used it to derive the private keys of a large number of impacted wallets. It has already been confirmed that at least 1,778.84 BTC ($BTC) have been stolen, involving more than 8,600 addresses. Based on the Bitcoin price during the incident period, total losses may have already surpassed $150 million.
Image source: Galaxy. The attacker exploited a seed-generation vulnerability in some Coldcard devices, and derived the private keys of impacted wallets in large quantities. It has now been confirmed that at least 1,778.84 BTC were stolen

Crypto wallet SafePal hit with personal data breach, nearly 40,000 customers affected
Recently, multiple cryptocurrency wallet providers have reported data breach incidents. The crypto wallet provider SafePal, which was once invested in by Binance, confirmed in a public announcement yesterday (8/16) that a plug-in used to track customer orders had an authorization vulnerability, allowing unauthorized parties to access certain customers’ personal data.
According to an internal investigation, the affected parties were customers who placed orders on the SafePal official website during the period from March 2, 2025, to April 11, 2026. A total of 39,798 people had their names, email addresses, shipping addresses, phone numbers, and purchase details exposed.

Apple has broken from its previous strategy; it is rumored that it trained a dedicated AI model for China.
According to a report by Reuters, Apple has trained a dedicated large language model (LLM) for the China market. Three people with knowledge of the matter said that Alibaba provided support during the training process.
This is the first time Apple has trained a China-specific model on its own. Previously, when Apple rolled out generative AI features in the China market, it relied on models provided by local Chinese vendors, because U.S. models such as Anthropic's Claude and OpenAI's ChatGPT have not been available for use in China.
Analysts say that in the future, if Apple has its own models, it could gain more control in China's highly competitive market, where local brands such as Huawei are taking market share.

Harvard stops trimming; holds 3.04 million shares of IBIT in Q2
Harvard University’s endowment fund stopped trimming its Bitcoin ETF in the second quarter. The latest regulatory filings show that Harvard Management Company, which manages the endowment fund, as of June 30 still held 3,044,612 shares of BlackRock’s iShares Bitcoin Trust (IBIT). The number of shares remained exactly the same as in the first quarter, and the market value at quarter-end was approximately US$101.4 million.
This change has drawn market attention mainly because Harvard previously reduced its Bitcoin ETF holdings for two consecutive quarters. In the fourth quarter of 2025, the fund trimmed its position by about 21%, and in the first quarter this year it cut it sharply by another 43%, bringing its IBIT holdings down to roughly 3.04 million shares. After entering the second quarter, Harvard did not continue selling or increase its stake, maintaining the allocation level as of the end of the first quarter.

Stripe to acquire OpenRouter; deal reportedly agreed at more than $7 billion
A U.S. financial payments giant, Stripe, has reportedly officially reached an agreement to buy the new startup OpenRouter—described as an “AI transit hub”—for more than $7 billion.
(Bloomberg) Citing sources familiar with the matter, it said that just months after OpenRouter reportedly began raising funds at a $1.3 billion valuation for this acquisition, the acquisition price is now far higher than the valuation—fully demonstrating the strong market demand for companies seeking the most cost-effective AI solutions.
If true, this acquisition will help Stripe, which began as a payments company, secure an even more solid strategic foothold in the rapidly growing AI industry.

Trump to meet crypto giants on August 19, with the SEC and CFTC chairs attending in tandem
The U.S. crypto market structure legislation has entered a critical juncture again. According to a report by Bloomberg, U.S. President Trump is expected to meet with senior executives from multiple cryptocurrency, prediction market, and fintech companies on August 19 at the Eisenhower Executive Office Building near the White House. Attendees include representatives from Coinbase, Ripple, Chainlink, Andreessen Horowitz (a16z), Paradigm, Kalshi, and others. Some reports also indicate that Kraken, Gemini, the New York Stock Exchange (NYSE), and Nasdaq have been invited.

Crypto cybersecurity lights up red again! On Thursday, well-known cold wallet maker Trezor confirmed that a system breach at its partner logistics provider, ShipMonk, led to the leakage of confidential personal information for nearly 14,000 customers.
According to Trezor, a total of 11,742 customers’ names, email addresses, phone numbers, and delivery addresses were fully leaked; in addition, another 1,947 customers’ names, residential cities, and emails were also exposed. Those affected total nearly 14,000 people, spanning multiple countries including the United States, the United Kingdom, Sweden, Colombia, Brazil, Italy, and Portugal.

Bitcoin reserve companies may be removed from mainstream indexes. According to a report by The Block, index provider MSCI has proposed a new classification approach that could remove companies such as Strategy (MSTR) and Metaplanet from its global indexes. JPMorgan estimates that simply removing Strategy from the MSCI index could trigger passive outflows of about $2.8 billion. In response, Strategy has publicly rebutted the proposal.
New screening mechanism: “Non-operating companies” face five financial tests
MSCI has designed a two-step screening process for “non-operating companies.” First, it checks whether the asset structure contains sufficient operating assets. If it fails that test, it then reviews five financial ratios, including: operating assets are less than 20% of total assets; operating expenses are less than 5% of total assets; operating cash flow is negative; changes in fair value of non-operating items exceed 5% of total assets; and capital dependence exceeds 20%. If at least four of these are triggered, the company is classified as non-operating. The companies named this time—potentially at risk of being removed from the MSCI ACWI IMI Index—include Strategy, Metaplanet, and uranium company Yellow Cake, along with three others such as SharpLink placed under observation. MSCI emphasized that only “sustained changes in business structure” would trigger reclassification, and a one-time failure to meet the criteria would not count. The consultation period runs until September 30, and the results are expected to be announced as early as October 16. Chain News previously reported in January this year that MSCI had decided to keep MicroStrategy in the index; now it has restarted this review.

The impact of generative AI on the content industry may be spreading beyond online articles, images, and videos into the traditional publishing market.
Venture capital firm a16z recently used the quite direct term “Bookslop” to describe the AI-generated books that are rapidly flooding Amazon’s self-publishing market.
More notably, these works aren’t just being produced in large quantities with little interest: books containing text that can be detected as AI-generated now account for about 40% of the sales volume in the research sample observations.
A study of 14,000 Amazon self-published novels finds that AI books have already formed a real market
a16z cites a study released in July titled “Generative AI floods and dilutes the market for books.”

The era of “mindless traffic farming to make money” is over! X has announced major reforms to its creator revenue-sharing system. The platform will no longer accept new applications. Old creators who have already passed the revenue-sharing review may reapply for a brand-new Original Content Rewards Program (“Original Content Rewards Program”). Platform resources will be fully directed toward original creators who can bring exclusive news, demonstrate deep professional capabilities, and share genuine real-world experience. This time, the official also released the “no-brainer definition standard” that Musk believes in—cracking down on plagiarism, reposting, automated bots, and low-quality, trash posts.

Chicago Federal Reserve Bank President Austan Goolsbee recently appeared on a Wired program to personally answer viewers’ questions about the state of the U.S. economy today and what role the Federal Reserve should play. Goolsbee emphasized that the Federal Reserve is an independent institution of the U.S. government, committed to two goals: maintaining price stability and promoting employment. He explained in plain terms how the Fed operates, the channels through which money circulates, why the stock market has been rising, why home prices are so high, and even how to spot counterfeit U.S. dollars—sharing a range of interesting facts.
What exactly is the Federal Reserve for?

Acquires a CFTC-compliant exchange to accelerate Fanatics' expansion into prediction markets
Global sports platform Fanatics has announced the acquisition of U.S. financial services company BGC Group's Water Street Labs and CX Clearinghouse, obtaining the licenses for a designated contract market (DCM) and derivatives clearing organization (DCO) approved by the U.S. Commodity Futures Trading Commission (CFTC), thereby establishing its own compliant prediction market platform.
The transaction value has not been disclosed yet. Both parties expect to complete the closing after completing regulatory procedures. Fanatics will use this infrastructure to launch its own prediction market exchange, allowing users to directly trade contracts for a wide range of events, while expanding the product footprint of its Fanatics Markets. BGC will collaborate with Fanatics to develop new data products that combine prediction markets with traditional financial markets.

Kamakura Inter FC obtains sports stadium NFT technology patent
Kamakura International FC, commonly known as Kamakura Inter FC, announced on July 24 that it has obtained a technical patent related to the “Kamakura Stadium NFT Project.” The patent (Special Patent No. 7891699) was first filed on September 17, 2025. Its core content is to automatically link physical areas within the stadium to NFT holders through the blockchain, so that the digital entitlements held by fans correspond to specific locations on the field.
Based on the patent design, the system can split specific spaces such as stadiums into multiple blocks, and record the virtual rights associated with each block on the blockchain. When a match or event takes place in a certain area, AI, sensors, or other detection devices can identify where the event occurred, and then automatically issue rewards or perks to the rights holder for that area via smart contracts. This design directly links on-field events, digital ownership, and fan engagement, turning the stadium into an interactive Web3 asset environment.

In the past, if someone claimed they ran a company with annual revenue of NT$ tens of millions, yet employed no staff, most people would probably think it was either fraud or some kind of financial miracle.
After generative AI rapidly became widespread, companies like this started appearing around the world.
In the past, entrepreneurship required a team, capital, and large amounts of manpower. But as AI gradually takes over tasks like software development, customer service, operations, and marketing, the “one-person company” has shifted from being a rare case to increasingly becoming a startup trend worth paying attention to.
But everyone’s questions are remarkably consistent: Is it really true? Where are the real examples?

SpaceX’s latest quarterly earnings report sparked lively discussion on the (All-In Podcast). The hosts were especially bullish on Starlink, believing that as Starship and the new-generation V3 satellites begin expanding network capacity, Starlink’s potential is no longer limited to serving remote areas—it is directly entering the global telecommunications and network infrastructure market.
Musk later responded personally, saying the market still severely underestimates Starlink’s potential.
He said that as AI and robots become widespread, global bandwidth demand will increase significantly because the amount of data that AI and robots need to transmit is several orders of magnitude higher than that used by humans.