Wu learned that Bitkey, the Bitcoin hardware wallet launched by Jack Dorsey's Block, officially confirmed that its mobile Recovery Contact and Inheritance relationship binding process has a random number generation flaw. The affected component is the one-time secret used for SPAKE2 authentication, not the mnemonic phrase, private key, or wallet key. The official statement says that no user assets have been found to be affected so far, and there is no direct remote theft risk. It is recommended that users of the second-generation Bitkey who have set up inheritors update and rebind their inheritors after the new version of the app is released.

Wu reported that according to Ember monitoring, Trump-associated listed company Trump Media (DJT) transferred 2,628 BTC to Crypto.com, valued at approximately $165 million. The company previously purchased 11,542 BTC at an estimated average price of about $118,500 each. Since the beginning of this year, related addresses have cumulatively transferred out about 7,281 BTC and currently still hold about 4,261 BTC. Ember estimates that the total realized and unrealized losses amount to approximately $555 million.

According to Yonhap News Agency, in June 2026, South Korea's five major KRW crypto exchanges Upbit, Bithumb, Coinone, Korbit, and Gopax transferred out stablecoins worth 2.7625 trillion KRW to overseas exchanges, transferred in 2.2022 trillion KRW, resulting in a net outflow of 560.3 billion KRW (approximately 367 million USD), equivalent to 77.6% of the net overseas stock purchases by Korean investors that month. Since January 2025, Korean stablecoins have had a net outflow to overseas exchanges for 18 consecutive months. The report suggests that the related funds are mainly used for crypto and stock derivatives, RWA, DeFi, and staking services provided by overseas platforms.

Wu reported that Galaxy Research stated it identified three waves of suspected attacks targeting addresses generated by Coldcard based on Bitcoin on-chain data, involving a total of 4,585 addresses and transferring out 1,367.05 BTC, valued at approximately 88.6 million USD. The first two waves share similar characteristics in fund aggregation structure, receiving address types, and address generation paths, reasonably suggesting they were carried out by the same attacker, though this cannot yet be confirmed; the third wave differs significantly in fund aggregation, receiving addresses, and batch transfer methods, possibly indicating the same attacker changed tools or another attacker independently exploited the same address generation issue. Currently, on-chain data alone cannot confirm whether there is a connection between the three waves of attacks. Galaxy emphasized that the above analysis is based solely on on-chain data and has not verified whether the related addresses were indeed generated by insufficient randomness, so the analysis is incomplete and should not be considered conclusive.

According to the New York Post, Dionysios “Dean” Karakitsos, co-founder and former CEO of blockchain data company Bitquery, has been sued in Manhattan Supreme Court, accused of illegally embezzling over $5 million from the company and deleting 194 expense records in the QuickBooks accounting system on the eve of his resignation in October 2025 to cover up evidence. The lawsuit alleges that since 2022, Karakitsos transferred funds from the company account to personal and affiliated entities unilaterally and deceived investors by inflating company revenue. Bitquery has appointed another co-founder, Aleksey Studnev, as the new CEO and is demanding the return of at least $5 million and control of the bank accounts. Karakitsos has strongly denied the allegations and will respond through legal proceedings.

Author: @Lyangminn Original link: https://x.com/Lyangminn/status/2079191212863750200 Disclaimer: This article is a repost. Readers can obtain more information through the original link. If the author has any objections to the form of reposting, please contact us, and we will make modifications according to the author's requirements. The repost is for information sharing only and does not constitute any investment advice, nor does it represent the views and positions of Wu Shuo. 1. WAIC should have been held earlier, so after visiting, one could have cleared out tech stocks sooner. There are too many peak signals: various side event crypto themes, a gathering of high-attractiveness women, highly homogeneous products on site... a track that was once a geek's game has become a public feast. Whether in primary or secondary markets, when everyone starts believing the same story and consensus becomes too concentrated, risks begin to accumulate. Because the bulls who stood with you during the bull market all become sellers in the stampede during the downturn. Many kill many! 2. Large models do not have a true moat The foundation of AGI is still the Scaling Law. Model capability still follows a power-law relationship with parameter scale, training compute, and data quality. Whoever can obtain GPU compute more cheaply, higher quality data, better researchers, and spend money more efficiently. Therefore, the moat of large model companies is more fragile than many imagine. Model capabilities will continuously converge, leading advantages will be constantly chased, and prices will keep falling. 3. Data Bottleneck is a potential opportunity seen at this conference After talking with some friends, it was found that companies making money in this AI wave are very low-key and don’t even attend exhibitions; one direction is data. The success of unified multimodal representation essentially depends on high-quality multimodal training data. There is a severe shortage of data assets that are high-quality cleaned, accurately labeled, cross-modal aligned, and continuously iterated. Models will become cheaper, GPUs will eventually become more abundant, but truly high-quality data is increasingly scarce. The multimodal data industry chain is becoming a highly profitable money-making track. Data cleaning, data labeling, data synthesis, vertical data assets, robotic data collection... and companies specializing in algorithms that solve cross-modal unified representation and multimodal pre-training encoders. GPU is a one-time capital expenditure; data is a continuous capital expenditure. Currently, I categorize data directions into three types: The first is Expert Data; the second is RL Environment / Agent Data. Future agent training will no longer just collect question-answer pairs but construct Environment → Task → Trajectory → Reward → Verifier, which may be one of the largest incremental markets in the future. The third is Embodied / Robotics Data, which is even scarcer than LLM data and extremely difficult to collect. 4. AI application dilemma, a microcosm of the crypto world AI today is very similar to past Crypto: fat protocols, L1 captures most of the value, so Crypto VCs desperately pour money into public chains, applications? Nobody invests. Currently, AI is the same. Large models are a POW Layer1, even the model is the application. In recent years, most AI applications have been doing a dangerous thing: packaging capabilities the model temporarily lacks into products. But the model’s capability boundaries keep expanding outward. Every model upgrade is like L1 writing functions originally belonging to the application layer directly into the protocol: search, deep research, coding, image generation, video generation, computer use, agents... If every model upgrade reduces a layer of value in your product, then essentially you are a feature of AI, not an AI application. Therefore, you must build things the model can do but are hard to take away, which boils down to data, context, workflow, permission, distribution. The moat of AI applications is to become a customer of the model, not a competitor. 5. Embodied intelligence, a huge bubble On the second floor of Hall H4, watching many similar robots produced by the same supply chain slowly performing similar actions, then looking at the valuations of various companies, it’s chilling and makes one pity investors’ money. The problem with embodied intelligence now is that capital is pricing “soft-hardware complex systems” using the “software Scaling Law.” The miracle of large models is that one company training GPT-5 can theoretically serve hundreds of millions globally at near zero marginal cost, but robots cannot. Every additional user served by a robot requires building another machine, involving BOM, manufacturing, supply chain, delivery, maintenance, depreciation. AI intelligence can improve exponentially; for example, when a model upgrades once, all global users become smarter simultaneously, but physical world costs won’t drop exponentially. If intelligence follows Moore’s Law, embodied intelligence still follows manufacturing industry rules. Embodied intelligence may ultimately be a huge industry but not necessarily one with large model–style profit margins. Also, it’s too early. Autonomous driving is already a highly constrained embodied intelligence problem with clear goals, limited action space, standardized road rules, massive real data, and a mature automotive industry system. Even so, this industry has burned hundreds of billions of dollars and still hasn’t fully solved some long-tail road problems. The real difficulty embodied intelligence faces is N times that of autonomous driving. 6. The best era for middlemen A harsh truth: currently, most AI startups are not very profitable, even less so than a woman selling her skills on Ququ. So an interesting thing is found: many companies appear to have one business, but after chatting, “Oh, you also sell tokens,” “Oh, you also sell compute,” turns out large model companies also resell compute for profit. The real money is still made by middlemen. FAs, old stock brokers, flipping B300, selling tokens, selling datasets, selling compute, even selling opportunities to meet certain founders... AI meets all conditions for a thriving middleman market: fast tech changes, large information asymmetry, abundant capital, and sexy stories. VCs make money by judging the endgame; FAs make money by judging consensus, without needing to prove a trend is ultimately correct. On the contrary, the more vague and grand the narrative, the more room for FAs. For example, embodied intelligence, world models... these tracks share a common feature: the future narrative is big enough, technology complex enough, and short-term falsification difficult enough. Big companies burn money training models, investors bet on AGI ten years from now, and middlemen make the money. Either milk the capital expenditure of big companies Or milk the FOMO expenditure of LPs 7. Money is everything The AI race still has a long way to go. People often ask, what is the scarcest resource in the AI era? Many answer: talent, compute, data... but ultimately, they all have one name: money. The biggest competitive advantage in the AI industry is not technology but fundraising ability. Entrepreneurs now must keep raising funds, even if there is already enough money in the account. The real war will be in the future industry downturn cycle. As long as you have money to survive, you can wait until competitors run out of money, then acquire their talent, technology, and customers at low prices. The secondary market is the same. In the foreseeable 2-3 years, a major bubble crash is very likely. As long as you have enough money to bottom-fish then, you will beat 99% of people.

Wu reported that Strategy founder Michael Saylor posted that at block 960,561, BIP-110 received only 24 signals in 946 blocks (accounting for 2.54%), and all signals came from DATUM miners sharing rewards through OCEAN, with zero signals from outside OCEAN. He stated that the current 55% voluntary activation threshold is mathematically unattainable in the current cycle, which does not represent miner consensus. Saylor further pointed out that OCEAN has set the BIP-110 signal as the default for its existing endpoints, and the BIP-110 guide directs node operators to use Bitcoin Knots and miners to use DATUM, essentially functioning as a vertical integration marketing campaign targeting Knots and OCEAN/DATUM.

1. Federal Reserve Chair Warsh Considers Reducing Number of Annual FOMC Policy Meetings Federal Reserve Chair Kevin Warsh is considering reducing the number of routine Federal Open Market Committee (FOMC) policy meetings held annually and has proposed related ideas at this week's meeting. Currently, the FOMC typically holds eight two-day meetings per year, with rules requiring at least four meetings annually. Warsh is also considering reducing press conferences following rate decisions and has established five working groups to study reforms in monetary policy communication, data usage, and balance sheet mechanisms. A Federal Reserve spokesperson declined to comment. The remaining meetings for 2026 and those for 2027 have been scheduled, but specific dates still require confirmation one by one. 2. Russia to Ban Cryptocurrency Mining in Moscow and Other Areas Until 2032 The Russian government will ban cryptocurrency mining starting August 15, 2026, in Moscow, Moscow Oblast, and parts of Kursk Oblast, with the ban lasting until December 31, 2032. The restrictions cover eight city districts in Kursk Oblast and the city of Lgov, applying to local residents participating in mining pools and similar activities. The related government decree was signed by Russian Prime Minister Mikhail Mishustin. 3. Hong Kong Woman Scammed of Over HKD 26 Million in Virtual Currency Investment A woman in her fifties working in insurance in Hong Kong was introduced to a man claiming to be involved in car trading. They developed an online romantic relationship, and under his inducement, she invested in virtual currency through a fake platform. Over six months, the victim paid over HKD 4 million in cash and transferred nearly HKD 22 million to multiple puppet accounts. In June, the platform showed her account gains exceeding 800%, but her withdrawal requests were denied. The man and the so-called "investment expert" then disappeared, resulting in a loss exceeding HKD 26 million. Hong Kong law enforcement received 25 cases of romance-related investment scams from July 24 to 30, with total losses nearing HKD 70 million. 4. French Woman Targeted by Armed Gang for Extortion Suspected Linked to Waltio Data Leak The Criminal Court of Amiens, France, this week heard a case involving an attempted kidnapping and extortion related to crypto assets. On June 5, three masked armed men broke into a woman's home in Ételfay, Somme department, France, but fled due to unforeseen circumstances, failing in their attack. Court information indicates the victim was likely targeted through Waltio data leak files circulating on the dark web. A 26-year-old man involved admitted to being hired for €5,000 to participate but was not the mastermind. Waltio experienced a data breach in January this year, potentially affecting nearly 50,000 users. The investigation is ongoing, and the main perpetrator remains unidentified. 5. MoneyFlip CEO Arrested for Using 25,000 USDT to Pay for Hitman Marcos Arturo Kleiman Tronllan, CEO of cross-border currency exchange MoneyFlip, was arrested in Miami on suspicion of hiring a hitman. Prosecutors stated Kleiman allegedly used his company to provide cross-border exchange services for illicit funds, assisting an undercover agent in converting approximately $750,000 of so-called "drug sale proceeds" into cryptocurrency, charging a 10% fee. During the investigation, Kleiman allegedly hired the undercover agent for $40,000 to kidnap and kill a Mexican businessman who owed him money. After the agent showed fabricated photos and videos of the victim's murder, Kleiman paid $5,000 in cash and transferred about 25,000 USDT to the undercover wallet as the final payment. The proposed victim was never actually harmed. Kleiman currently faces charges of hiring a hitman, punishable by up to 10 years in prison; the charges have not yet been adjudicated. 6. Wu Shuo Weekly Picks: SEC Supports Crypto Market Legislation, Morgan Stanley Launches Ethereum and Solana Spot ETPs, Multiple Digital Asset Treasury Firms Shift to AI Data Centers and News Top 10 This Week's Top 10 News 1. If the CLARITY Act Fails, SEC Will Independently Issue Crypto Market Rules link Paul Atkins, Chair of the U.S. SEC, stated that if Congress fails to pass the CLARITY Act, the SEC is prepared to independently establish rules covering crypto market structure. However, legislation offers more sustainability and can prevent regulatory frameworks from changing with government transitions. The Act passed the Senate Banking Committee in May by 15 to 9 votes but has not yet reached a full Senate vote. The SEC has included crypto asset issuance, custody, and trading rules in its 2026 regulatory agenda. 2. Canadian Crypto Ownership Rises to 25%, Officially Entering Mainstream Finance link According to the latest survey by the Ontario Securities Commission (OSC), the number of Canadian crypto holders has more than doubled in recent years, with one-quarter (25%) of Canadians now holding digital assets or crypto investment funds, up from 10% in 2023 and 13% in 2022. The survey also shows crypto is moving beyond mere speculation: 74% of current crypto owners have actually used digital assets, 89% of stablecoin holders report having used their holdings, with 20% for international transfers. Additionally, investor due diligence awareness is improving, with 50% of crypto owners verifying if trading platforms are legally registered before investing, up from 38% in previous surveys. 3. Morgan Stanley Asset Management Launches Ethereum and Solana Spot ETPs, Both Plan to Participate in Staking link Morgan Stanley Investment Management launched Ethereum spot ETP MSSE and Solana spot ETP MSOL, listed on NYSE Arca, both with a 0.14% fee. Both products plan to stake portions of ETH and SOL, with Morgan Stanley Asset Management not retaining staking rewards. Along with the previously launched Bitcoin product MSBT, their crypto ETP lineup now covers BTC, ETH, and SOL. 4. BNY Mellon to Launch Digital Transfer Agency Service, Putting Fund Holder Records on Blockchain link BNY Mellon, with asset custody and management exceeding $59 trillion, will launch a digital transfer agency service to migrate fund transaction processing and holder records onto blockchain while retaining traditional transfer agency systems. BNY's transfer agency currently serves about $8.6 trillion in assets and 7.6 million accounts. Baillie Gifford will be the first to use this system for the UK's first "fully native" regulated tokenized fund. BlackRock and BNY's Dreyfus are also expected to adopt it in upcoming funds. 5. Strategy Reports Q2: Holding 843,800 BTC, $8.22 Billion Loss Due to Fair Value Changes link Strategy (formerly MicroStrategy) released its Q2 2026 financial report. As of July 26, the company held approximately 843,775 BTC, a 25% increase year-to-date, and raised about $17.06 billion through ATM issuance programs. Q2 revenue was $122.4 million, up 6.9% year-over-year; due to Bitcoin asset fair value changes, it recorded a net loss of $8.22 billion, compared to a net profit of $10.02 billion in the same period last year. As of June, the company held $1.71 billion in cash and equivalents and $736.1 million in short-term investments. Strategy stated it has increased its USD reserves to $3.75 billion, covering preferred stock dividends and debt interest for about 2.1 years. The company raised about $218.4 million this year by selling Bitcoin to pay some preferred dividends and repurchased approximately $28.9 million nominal value of STRC preferred stock, paying about $25 million. Convertible debt decreased from $8.21 billion to $6.71 billion by the end of Q2. In the Q2 earnings call, Strategy said future fundraising will not be used solely to buy Bitcoin but will dynamically allocate BTC and USD reserves based on market conditions to enhance the stability of its Digital Credit system. The company plans to sell small amounts of Bitcoin when favorable to supplement USD reserves, pay preferred dividends and interest, and support share repurchases. It also clarified no plans to use Bitcoin-collateralized loans due to counterparty risk, margin risk, and potential short-selling targets. Management expressed a desire to gradually reduce convertible debt and maintain USD reserves at about 2 to 3 years' coverage of dividends and interest. 6. Robinhood CEO: Company Aims to Become First Trillion-Dollar Financial Enterprise link Robinhood CEO Vlad Tenev stated in the Q2 earnings call that the company aims to become the first financial firm with a $1 trillion market cap and will continue expanding around Agentic Finance, Robinhood Chain, global asset tokenization, and private markets. CFO Shiv Verma said the company expects business scale to grow tenfold over the next decade. Additionally, Robinhood has launched perpetual futures overseas and plans to offer the product to more customers. Regarding the U.S. launch, no specific timeline was given, but the company emphasized it will actively advance the product. Tenev also revealed the team is actively developing AI Agentic Trading features to provide more AI capabilities to traders in suitable scenarios. Robinhood's Q2 2026 net revenue grew 32% year-over-year to $1.308 billion, a record high; net profit rose 48% to $573 million, with diluted EPS of $0.62. Trading-related revenue increased 44% to $776 million, with prediction market revenue surging over tenfold to $156 million; options and stock revenues grew 29% and 95%, respectively, while crypto revenue declined 38% to $100 million. Quarterly net deposits reached $21.7 billion, platform assets grew 32% to $369 billion, and Gold subscription users increased to 4.8 million. The company lowered its 2026 adjusted operating expenses and equity incentive guidance to $2.675 billion to $2.775 billion. 7. Among Top 50 Stablecoins by Market Cap, Only USDC, USDG, and EURC Comply with MiCA Requirements link Patrick Hansen, Senior Director of EU Strategy and Policy at Circle, wrote that the EU currently has 21 issuers offering about 35 regulated e-money tokens, but among the top 50 stablecoins globally by market cap, only USDC, USDG, and EURC meet MiCA requirements. He believes MiCA's future assessments should enhance institutional competitiveness and strengthen global regulatory coordination, supporting EU-issued e-money tokens in expanding cross-border payments and tokenized trade, while establishing recognition mechanisms for stablecoins regulated abroad to bring more global stablecoin activities under MiCA's scope. 8. Multiple DAT Companies Recently Announce Shift to AI Data Centers and Other Businesses link With the crypto market continuing to languish, at least a dozen digital asset treasury (DAT) companies have recently announced shifts toward AI data centers and other businesses to try to re-attract investors, but results remain limited. Since shifting to data center development in May, K Wave Media's stock has dropped about 71%; Lixte Biotechnology fell about 33% after merging with a battery company; AlphaTON Capital, renamed Alpha Compute, also dropped about 33%. Several law firms note the DAT model hype has clearly cooled, with more companies seeking to enter AI, data centers, aerospace, and small modular nuclear reactor sectors. 9. July Bitcoin Average Spot Trading Volume May Hit Lowest Since November 2023 link K33 Research reports that the Bitcoin market remained sluggish in July, expected to record the lowest average monthly spot trading volume since November 2023. The report states BTC fell about 3% in the past week, fluctuating between $60,000 and $66,000; CME Bitcoin futures open interest remains at multi-year lows, perpetual contract open interest is about 300,000 BTC, and July's daily average spot trading volume is about $2.2 billion, with market participation continuing to weaken. 10. Crypto Market in First Half of 2026 Shows Broad On-Chain Contraction Rather Than Sector Rotation link Binance Research released its 2026 first-half on-chain market report, noting a broad on-chain contraction rather than sector rotation. Total DeFi TVL dropped $43.4 billion (38%), and the six major L1s' total market cap shrank $246.5 billion (42%). Key highlights include: Ethereum marginal holders shifting from ETFs to corporate balance sheets, spot ETF holdings dropping to 5.2 million ETH, while digital asset treasury holdings rose to 7.7 million; general L2 user activity sharply declined, with user operations down about 77% from January to June; Solana network revenue fell 64.5%; BNB Chain became the leading tokenized stock platform and the only deflationary major L1 (annualized burn rate 5.05%); 207 security incidents occurred in the first half, causing $972 million in losses; prediction market monthly nominal trading volume surged 86% to $51.6 billion, with Kalshi and Polymarket accounting for 92% of June's total volume. Key Financing Events Ethereum Institutional completed its first round of ecosystem financing, led by BitMine, SharpLink, and others link AI security company V12 completed a $10 million seed round to develop an automated vulnerability detection system link Robot data platform Axis Robotics completed a $12 million seed round link Distributed AI infrastructure project ALPHEA announced a $5 million financing link For more industry financing events, visit crypto-fundraising.info.

Michael Coates, the new Chief Information Security Officer (CISO) of the Solana Foundation, stated in an interview that the biggest security threats to cryptocurrency are increasingly shifting toward AI-driven social engineering attacks, fake identities, and credential leaks, rather than being limited to smart contract vulnerabilities. Coates pointed out that as attackers use AI technologies such as deepfakes and full-spectrum voice cloning to make scams more convincing, Web3 organizations must implement multi-layered security controls and provide security mechanisms by default to guard against human error. Additionally, in response to the long-term challenge of quantum computing, the Solana Foundation has also developed strategies to adopt post-quantum cryptographic algorithms. (CoinDesk)