Crypto News & Market Updates

According to Bloomberg from Binance, generative AI giant Anthropic is implementing a highly controversial interview process that strictly prohibits job seekers from cheating using AI tools and has introduced a cultural approach known as "psychological therapy." CEO Dario Amodei revealed that he spends nearly one-third to 40% of his time building culture. Interview scoring has veto power, aiming to screen candidates who align with the safety mission and possess independent thinking skills. Candidates must pass up to five rigorous rounds of screening, including signing a Non-Disclosure Agreement (NDA). In high-elimination cultural interviews, interviewers use rapid Q&A to obtain evaluation signals, sometimes interrupting directly when there is no more incremental information. Some candidates reported that interviewers had no facial expressions during questions, causing significant psychological pressure for job seekers. The core of the interview is assessing the candidate's values and awareness of the threats posed by artificial intelligence. Interviewers will present professional ethical dilemmas and assess the candidate's genuine moral struggles. President Daniela Amodei pointed out that assessments do not pursue specific beliefs but evaluate whether candidates can debate rationally.

According to Grayscale, Ethereum (ETH), Solana (SOL), and two other blockchains are expected to benefit from the Transparency Act. The company pointed out that anticipated regulatory changes, including the Transparency Act aimed at establishing digital asset classification and regulatory rules, will drive the application scenarios of tokenized assets and decentralized finance (DEFI). Grayscale believes that the development of regulatory transparency will benefit leading chains such as Ethereum, Solana, BNB Chain, and the Canton network. The company also mentioned that other blockchains such as Avalanche (AVAX), Ethereum Layer 2 chains Base and Arbitrum (ARB), HyperLiquid (HYPE), and stablecoin-focused Tron (TRX) will also benefit from regulatory transparency. Although Bitcoin (BTC) does not natively support smart contracts, Grayscale believes it will also benefit from regulatory transparency as the industry's safest asset and primary collateral.

According to Bijie.com, in an interview with OKX on April 29, Xiao Feng revealed the reasons for early investment in Ethereum and founder Vitalik. He stated that Vitalik breaks the limitations of Bitcoin as a mere distributed ledger and is committed to building a new type of blockchain that supports application development. Xiao Feng mentioned that the decision at the time was made during a critical risk period of five months before mainnet launch; if the mainnet stalled, the tokens acquired would be meaningless. But he believes Vitalik's idea is groundbreaking and worth supporting even if the entire $500,000 is lost.

According to ChainCatcher, Push Labs Ltd and Push Virtual Assets Ltd, the UK subsidiaries of Aave Labs, have been approved by the UK Financial Conduct Authority (FCA) to register as local crypto asset trading service providers. Aave stated that these licenses enable Push to conduct regulated crypto asset and payment-related businesses in the UK and support the launch of zero-fee stablecoin deposits and withdrawals. In addition to crypto asset registration, Push is also authorized to issue electronic money under the Electronic Money Regulation. The two companies are assigned FCA registration numbers 1031720 and 1031721, respectively, and AML regulations are governed by the 2017 Anti-Money Laundering and Money Transfer Regulations.

BlockBeats reported on May 28 that crypto analyst Axel Adler Jr. stated that BTC inflows into exchanges and stablecoins leaving exchanges simultaneously send "risk aversion" signals, intensifying market selling pressure. Data shows that the net inflow on BTC trading platforms over the 30th has shifted from an extreme net outflow of 300,000 BTC at the end of March to an inflow of 103,000 BTC on May 26, indicating that more BTC is reflowing into the platform for sale. During the same period, BTC's price fell from $80,000 to $73,700. Meanwhile, stablecoins are flowing out of centralized exchanges at a record pace. The average 30-day net flow for stablecoins has shifted from $164 million daily inflows at the end of April to $153 million daily outflows on May 27. This means the liquidity available to buy BTC in the market is decreasing. Axel Adler Jr. pointed out that when BTC flows into exchanges and stablecoins simultaneously exit exchanges, a "supply increases, demand falls" unfavorable structure forms, which is a typical risk-averse market. It believes that if BTC net inflows continue to exceed +100,000 BTC, the market may face a deeper correction; Stability signals include BTC turning into net outflows again, or stablecoins returning to trading platforms.

According to Bijie.com, French AI startup Mistral is expanding its business footprint into the advanced manufacturing sector. The company has reached cooperation agreements with new clients Airbus and BMW Group, hoping to drive business growth through so-called "physical AI." Mistral primarily provides AI models and computing infrastructure to enterprise clients. On Thursday, the company announced that its current strategic goal is to apply AI technology to industrial engineering processes, specifically covering design, simulation, and quality control. Currently, the company has not disclosed specific financial terms reached with the aforementioned new customer.

According to Bijie.com, LlamaIndex announced that the open-source document parsing library LiteParse has been completely rewritten and released in Rust for version 2.0. The refactored core parser increases processing speed for small documents by up to 100 times and parses large documents by nearly 3 times, aiming to provide a locally running parsing foundation for AI agents and retrieval-enhanced generative RAG pipelines. LiteParse 2.0 supports Office documents such as PDF, DOCX, XXX, and PPTX. The parser projects text into 2D space according to the document layout, outputting structured text that retains the relative relationship between position and layout. LlamaIndex provides native package support across mainstream runtimes, allowing developers to quickly integrate via Python's pip, JavaScript's npm, and Rust's cargo. The new version compiles it into the WebAssembly format to enable local operation between browsers and edge computing nodes, but OCR functionality in the WebAssembly environment requires external callback injection.

Odaily Planet Daily reports that two members of the prediction market platform team clashed on X platform over KYC issues. The cause was that Josh Stevens, Polymarket's Vice President of Engineering, denied that the platform would fully implement KYC. Kalshi's head of enforcement@robertjdenault cited related posts saying Polymarket should have implemented KYC verification long ago, and said that Russians and Iranians trade freely on the platform, and insiders are also using it recklessly. If Polymarket truly wants to stop this, Either make all operational activities compliant (first KYC), or shut down. In response, Polymarket official Mustafa angrily called the Kalshi executive an "idiot" under the post.

According to Bijie.com, OpenAI, in collaboration with traditional business digital transformation platform Thrive Holdings, announced that it has developed an adaptive evolution tax agent (Tax AI) for its Crete Professional Alliance, which has a network of over 30 accounting firms. In terms of operational mechanisms, the errors corrected by accountants during daily tax filing are automatically converted into test cases by the system, which then the Codex model automatically generates code patches for the tests and submits them to programmers for review as merging requests (PRs). This closed-loop design compresses the previously lengthy software bug repair cycle to the second level, ensuring high compliance and security red lines while enabling rapid autonomous evolution of agents. In the pilot project, the agent processed a total of 7,000 complex tax invoices of types 1040 and 1041. After six weeks of deployment, the proportion of tax forms capable of achieving a 75% accuracy rate for correct fields surged from an initial 25% to 86%, with the highest field accuracy rate reaching 97%. In actual tax filing operations, the system helps accountants save about one-third of tax bill preparation time on average, and overall business throughput increases by approximately 50%.