BTCPay Server suffers critical vulnerability exploit; official team urgently requests users to upgrade and replace credentials.

Blockchain technology firm StarkWare announced that after a week of AI-assisted programming, its estimated computing cost for quantum-secure Bitcoin transactions has fallen from roughly $320 to $66, a reduction of about 79%. Previously, a related transaction completed by StarkWare in August required around 3,100 hours of computing time, used approximately 100 GPUs, and carried a computing cost of $320. Following this, StarkWare launched a code optimization contest, where participants leveraged AI tools to enhance search efficiency, leading to a roughly 5x improvement in computing speed. The solution uses a hashing mechanism to help Bitcoin counter potential quantum computing threats, and does not require a soft fork upgrade to the Bitcoin network. That said, the $66 figure is currently only an estimate based on test results; the improved code has not yet been proven to successfully power another transaction mined on the Bitcoin network.

，甲骨文（Oracle）发布公告，宣布扩展其 Digital Assets Data Nexus 平台，新增支付执行集成、可配置的钱包与智能合约控制，以及 AI 驱动的交易监控与风险分析功能。此次升级支持通过 ISO 20022 消息标准，将银行现有支付系统与数字资产交易执行连接起来，帮助银行在传统货币与数字货币之间建立统一的资金流转运营模式，并提升账本可视性、合规控制和运营效率。该平台还将支持与 Oracle Banking Payments 及 Swift Ledger 集成，覆盖代币化存款、稳定币、CBDC、数字债券和现实世界资产等应用场景。甲骨文表示，相关功能计划于 2027 财年推出。

According to TradingBeats, NEAR is trading at roughly $4.27, up from around $2.36 a week ago, marking a total gain of approximately 80.6% — outperforming BTC by 70.6 percentage points over the same period. This rally has enabled two early Hyperliquid whales with large NEAR positions to generate over $10 million in unrealized profit from their single long positions alone. Together, the two hold about 10.982 million NEAR tokens, with a position value of roughly $46.96 million, and combined unrealized gains of around $23.204 million. Both positions were established before NEAR’s sharp rally on September 16, capturing the bulk of this upward move. Specifically, the address starting with 0x30af holds approximately 5.1432 million NEAR long positions, worth $21.99 million at an average price of $1.9473, with unrealized gains of roughly $11.977 million (+1199%). This position was built around August 26, when NEAR was still below $2, with its final addition made on September 4 and held since. Additionally, this address also holds around 1.4871 million INJ long positions, with unrealized gains of about $3.294 million. Prominent OG whale mk4 (0x773) holds about 5.8389 million NEAR long positions, valued at roughly $24.97 million at an average price of $2.353, with unrealized gains of approximately $11.227 million (+823%), opened on September 6. In addition, this address holds short or hedging positions in seven other assets including BTC, worth around $14.741 million. After the unrealized losses from these positions are offset by NEAR’s long gains, the address’s total unrealized profit stands at roughly $9.839 million.

Glassnode’s report indicates that Bitcoin (BTC) holders have been steadily realizing profits recently, yet the overall volume remains relatively low. Over the past seven days, the market’s net realized profit totals around $5.1 billion. The on-chain analytics firm notes that the current profit realization scale is closer to levels seen in late 2023, rather than those recorded during previous market peaks.

Tomorrow marks the quarterly settlement of crypto options. Data from Deribit shows that crypto options with a total notional value of $17 billion will expire, broken down as follows: · Bitcoin (BTC) options have a notional value of $14.9 billion, a put/call ratio of 0.76, and a max pain point of $78,000; · Ethereum (ETH) options carry a notional value of $2.1 billion, a put/call ratio of 0.63, and a max pain point of $2,300.

Beating AI Express News: The Information cited two people with direct knowledge that DeepSeek’s annual revenue has reached approximately $1 billion. Just a few months ago, this figure was less than $500 million, meaning it nearly doubled in a short period. Revenue mainly comes from model APIs; its free chat app still has no ad or subscription income. Part of this growth stems from price hikes. DeepSeek previously significantly raised prices for some APIs, with increases ranging from 2.3 to 4.5 times across different models and time periods. Liang Wenfeng recently told investors that the number of clients did not drop after the price hikes, and demand remains strong. Previously disclosed financial data shows that DeepSeek’s API business achieved a gross profit margin of 82.9% in the first seven months of this year. DeepSeek is also advancing its second round of financing, planning to raise approximately 50 billion yuan, with a valuation of around 500 billion yuan. This plan was previously reported, and the latest update is that the company hopes to complete the financing by the end of October. DeepSeek is also preparing for an IPO on the Shanghai Stock Exchange’s STAR Market. However, the company currently still allocates most of its computing power to model training. Liang Wenfeng said that over 70% of computing power is used for training new models, with less than 30% reserved for inference. DeepSeek is also trying to run more small models directly on gaming graphics cards, while keeping high-end chips for training.