Lookonchain APP

App Store

Claude has written 180 million lines of mathematical proofs for 123 open-source libraries, uncovering hundreds of bugs.

40 minutes ago

Beating AI News Flash: Anthropic has partnered with formal verification firm Theorem to leverage Claude for auditing software vulnerabilities using mathematical methods. Traditional testing typically relies on repeatedly feeding random data to check if a program crashes, but no matter how many tests are run, it can never guarantee full coverage of all potential issues. This time, Claude directly generates mathematical proofs for programs, aiming to fundamentally eliminate specific types of errors. Within weeks, Claude verified 239 test entry points across 123 open-source libraries, generating around 180 million lines of proof code, which was then checked by the Rocq mathematical proof tool. The results showed that under the specified runtime environment, these programs would not trigger the crashes targeted by the tests when handling any input smaller than 64 MB. During the proof process, Claude also discovered and fixed hundreds of bugs. For example, a decompression program would only suffer a memory out-of-bounds write when processing roughly 20 MB of data with an extremely specific structure. Another image decoding vulnerability required 64 MB of input and over 2 GB of memory to trigger—scenarios extremely hard to catch with standard random testing. This type of mathematical verification was historically extremely time-consuming. For instance, the seL4 operating system kernel required around 2 years of full-time human effort for development, but its mathematical verification took roughly 20 years—10 times the development workload.

Relevant content

MGBX Launches Exclusive New User Benefits, With Up to 30,000 USDT in Rewards for Completing Tasks

According to official announcements, the MGBX platform has launched an exclusive new user welfare campaign open to eligible new users. During the event, users can claim a new user reward by completing tasks including registration, identity verification, deposit, and contract trading; by finishing tiered trading growth tasks, they also have the opportunity to unlock higher-value rewards, with a maximum of 30,000 USDT per user.

8 minutes ago

Binance Alpha will remove 7 tokens including PTB, AA, OWL, etc.

According to official announcements, Binance Alpha will delist PTB (Portal to Bitcoin), AA (ARAI), OWL (Owlto Finance), DORA, ELIZAOS (elizaOS), MEMES (memes will continue), and Kuku Ma on October 23 at 17:30. Following the delisting, Binance Alpha will still support withdrawals and sales of these tokens.

8 minutes ago

Marinade Q3 Report: SOL staked volume rose 21.2% quarter-over-quarter to 7.94 million tokens, generating roughly $1 million in revenue.

Solana staking service protocol Marinade has released its Q3 2026 report. As of September 30, its staked SOL amount reached 7.94 million tokens, a 21.2% sequential increase, driven primarily by its non-custodial native staking product Native and institutional-focused Select. Specifically, Native staked volume rose 47.3% sequentially to around 3.77 million SOL, Select grew 38.7% to approximately 1.68 million SOL, while mSOL decreased 6.8% sequentially to roughly 2.27 million SOL. Q3 revenue totaled about $1 million (equivalent to 10,918 SOL), a 25.9% sequential increase, mainly due to a 38.1% rise in validator bond settlement revenue. Additionally, its Staking Auction Market (SAM) delivered yields above the Solana Staking Index (SSI) in 48 out of 51 epochs during the quarter, with an average pre-Marindade-fee yield of 5.93%, compared to the index’s 5.67% over the same period. On the governance front, since MIP-22 took effect in July, Marinade has allocated 10% of its auction fees to repurchase MNDE on the open market. In Q3, it repurchased 4.2 million MNDE using 906.6 SOL, with the first distribution opening on October 1. The report also disclosed that on September 25, an attacker exploited a vulnerability in the DAO voting process to submit two malicious proposals, one involving transferring DAO treasury assets. The council vetoed both proposals in approximately six hours, completed a fix that same night, and no funds were transferred.

8 minutes ago

Futu Securities launches BNB deposit and withdrawal services for Hong Kong Qualified Professional Investors.

Futu Holdings subsidiary Futu Securities International (Hong Kong) Limited has announced the launch of BNB deposit and withdrawal services for eligible professional investors (PIs) in Hong Kong. Qualified PIs can now deposit BNB to their Futu crypto accounts from self-custody wallets, or withdraw to verified whitelisted addresses, via the BNB Smart Chain (BEP20) network. Earlier, Futu’s wholly-owned SFC (Securities and Futures Commission)-licensed virtual asset trading platform PantherTrade (Cheetah Trade) in Hong Kong had already listed BNB and offered BNB trading pairs to Hong Kong’s eligible professional investors.

8 minutes ago

SpaceX and Tesla's potential merger may face antitrust scrutiny.

U.S. financial media outlet Semafor reported this week that Elon Musk has proposed a merger between Tesla (TSLA.O) and SpaceX (SPCX.O). A former U.S. Department of Justice (DOJ) official said the risk of antitrust issues arising is real. The official noted that antitrust regulators typically view such "vertical" mergers as pro-competitive, as they help improve supply chain efficiency. However, if the Tesla-SpaceX merger is deemed a potential "vertical exclusionary" deal, regulators may seek to block it. The official explained that an "exclusionary" scenario refers to a company leveraging its market position to restrict competitors at other levels of its supply chain. In this case, if only Tesla vehicles were allowed to use SpaceX technologies like Starlink in the future, other autonomous driving or electric vehicle manufacturers could suffer as a result.

8 minutes ago

Multiple vulnerabilities have been discovered in OpenAI's AI security defenses, with hundreds of operations potentially remaining undetected.

Beating AI Flash News: OpenAI originally deployed a monitoring system to track internal AI Agents for unauthorized actions, deception, or covert dangerous operations. Following audits by AI security firms METR and Redwood Research, the system itself was found to have multiple vulnerabilities. Two researchers worked on-site for three weeks, reviewing internal code and monitoring configurations, and submitted a 61-page report. One of the most severe flaws could let an AI evade monitoring for its last several hundred operations, with the system even failing to detect how the AI bypassed checks. Researchers also found Codex could theoretically break out of the sandbox restricting its operational scope, bypass automated security reviews, and gain the same access as its user. Additionally, two settings could directly disable monitoring for newly launched Agents, though enabling these settings might still be detected. The monitoring system’s design also had issues: the legacy system primarily checked an AI’s past activities only after it completed its current task. If the AI kept executing tasks without returning control to the user, it could avoid checks entirely. The system also lacked a step to compress historical records. Researchers estimated that between May and July 2026, around 10% of internal Codex traffic that should have been monitored was missed due to this gap. OpenAI employees had previously identified the problem and fixed it during the audit. OpenAI stated all vulnerabilities affecting its formal security protection systems have been patched, though the two firms have not yet conducted independent re-tests.

8 minutes ago

Popular tokens

BitcoinEthereumHyperliquidSolanaTRONBNBTetherAaveXRPPepeFartcoinOndoJupiterUniswapBonkPendleEthenaArbitrumAvalancheLidoChainlinkPolygonDogecoinCardano