Lookonchain APP

App Store

MetaMask responds to security incident: No signs of direct threat to wallets have been found so far.

2026.10.01 07:57:32

MetaMask has released a security update on X, disclosing that some of its infrastructure has been impacted by a security incident. No signs of direct threats to MetaMask wallets have been detected so far. As a preventive measure, MetaMask stated that it is coordinating with relevant clients, partners, and security advisors to proactively exit affected validators in its non-custodial staking business, and will announce further progress at an appropriate time. MetaMask emphasized that this incident relates to infrastructure and does not represent current security risks for MetaMask wallets.

Relevant content

Two crypto whales have placed $20 million worth of HYPE buy orders, with both looking to double their long positions.

According to TradingBeats' monitoring, HYPE is trading at around $93.58, up roughly 18.3% in 24 hours. Following the rally, trader Boomer and another large bull still hold a total of approximately $19.967 million in ladder buy orders, planning to add positions in batches after a pullback. Boomer’s addition interval is between $80 and $88, with the highest buy price about 6% lower than the current level. The 40 new buy orders he placed at 10:00 today are intended to purchase around 110,900 HYPE, with an order value of roughly $9.247 million. Currently, Boomer holds about 116,600 HYPE long positions, valued at approximately $10.913 million, with an average entry price of $89.75, generating an unrealized profit of around $446,000. If all orders are filled, his position size will nearly double. Another large whale with an address starting with 0x469e has set 84 additional buy orders in the $80–$88.469 range, planning to buy around 127,300 HYPE for approximately $10.721 million, with the highest buy price about 5.5% lower than the current price. The account currently holds around $3.046 million in HYPE long positions, with an average entry price of $91.03, and the pending addition amount is about 3.5 times the value of its existing position.

1 seconds ago

Google shuts down Jules: The project was still being reworked into its 2.0 version in May, and has now pivoted to Antigravity.

Beating AI News Flash: Google announced that its AI coding tool Jules will be shut down on December 31, and switch to read-only mode starting November 11. Jules focused on cloud-based asynchronous programming: after developers assign tasks, the AI would fix bugs, write tests in the background, and submit code changes. In May this year, the Jules team recruited beta testers for Jules 2.0 on Reddit, stating they were rebuilding the product from scratch, planning to launch a new web interface, API, and CLI, and even expanding its scope from assisting with coding to helping users decide what products to build. At that time, some questioned whether Jules and Antigravity were in internal competition. The team firmly denied this, noting the two tools were complementary, and also revealed that Jules’s programming components originally used Antigravity’s Agent and execution framework. However, just five months later, Google decided to shut down Jules, redirecting its asynchronous background execution R&D experience to Antigravity. In May this year, Google also announced it would shift its personal Gemini CLI to Antigravity CLI. Multiple parallel AI coding product lines are now converging toward Antigravity.

1 seconds ago

Google Antigravity launches its plugin marketplace, integrating Figma and Jira, and enabling online fault troubleshooting.

Beating AI Express News: Google’s coding agent product Antigravity has launched a Plugin Marketplace, which centrally offers plugins developed by Google and third-party developers. Previously, connecting an agent to external services typically required manual configuration of MCP, API keys, and related commands. Now, users can directly search for and install plugins with one click, and installed plugins are synced across the desktop app and CLI. The first batch of key partner showcases includes Atlassian, Figma, Canva, Datadog, CrowdStrike, Unity, and others. The agent can perform functions such as querying Jira tasks, Confluence documents, and Bitbucket code; reading Figma design drafts to generate code, as well as writing content back to the design canvas; analyzing logs, monitors, and call chains via Datadog to troubleshoot online outages. The Unity plugin allows the agent to directly participate in 3D project development and simulation.

1 seconds ago

The on-chain asset flow on OKX is its own operation, and the platform has completed security checks.

After verifying with OKX officials, the 172 million USDT (valued at approximately $171,846,059) monitored by Whale Alert as being transferred from OKX to an unknown wallet is an internal OKX transfer operation. OKX stated that regarding today’s industry security incident, it has completed a full investigation, found no signs of intrusion, the platform is operating normally, and user assets remain secure.

1 seconds ago

Approximately 1,077 BTC was transferred from an unknown wallet to OKX.

According to Whale Alert monitoring, 1077 BTC were transferred from an unknown wallet to the OKX exchange, valued at approximately $84.72 million.

1 seconds ago

Microsoft: North Korean Hackers Use AI to Mass-Generate Phishing Documents, With Fake Recruitment and Malicious Code Development Already Implemented

According to Yonhap News Agency, a report released by Microsoft Threat Intelligence on the 29th reveals that North Korean hacker groups are leveraging artificial intelligence (AI) to accelerate cyberattacks targeting South Korean users, expanding their tactics beyond customized phishing documents, fake job postings, and identity spoofing to include malicious code development and software supply chain attacks. Microsoft assessed the AI adoption levels of eight North Korea-related hacker groups: three have reached the highest "operationalized" stage, integrating AI into their daily attack workflows; four are in the "acceleration" phase, and one is in the "expansion" phase. The report notes that some groups generated over 100 targeted PDF lure documents for different targets between August and September this year, with cases of producing dozens of documents in just minutes. Between June 8 and 12, fake job offers were sent to AI, blockchain, and cybersecurity professionals using three spoofed companies. Additionally, Microsoft analysts found that North Korean hackers may be using AI programming tools like Cursor to assist with code generation, debugging, and evading detection. North Korea's Reconnaissance General Bureau-affiliated groups "Onyx Sleet" (referred to as Andariel by South Korea) and "Pearl Sleet" have been using the open-source AI penetration testing framework ARTEX in real-world operations since late October. Microsoft concluded that while these groups have boosted attack speed and efficiency through AI, they have not yet achieved fully autonomous attacks, and recommended that enterprises strengthen identity verification for job applicants, phishing protection, and malicious code detection.

1 seconds ago

Popular tokens

BitcoinEthereumHyperliquidSolanaTRONBNBTetherAaveXRPPepeFartcoinOndoJupiterUniswapBonkPendleEthenaArbitrumAvalancheLidoChainlinkPolygonDogecoinCardano