A certain whale added 9,375 ETH to its holdings at an average price of $2,264.

According to monitoring by Onchain Lens, the hacker group behind the Binance attack is stepping up money laundering efforts. They are repeatedly splitting funds into multiple wallets via automated transfers, returning the funds to previously used addresses, then redistributing them to new destination addresses to increase the difficulty of tracing the funds. The latest transfer involves 120 million TRX, valued at approximately $38.78 million at the time, and the relevant funds continue to be dispersed to other addresses after the transfer. Onchain Lens notes that this automated fund transfer pattern adopted by the attackers was previously uncommon.

Breaking news from Dongcha Beating AI: Google CEO Sundar Pichai announced that the prototype satellite of Project Suncatcher, launched in early October, has completed its first AI inference in space using data center-grade TPUs, with results transmitted back to Earth. Engineers sent a ground prompt instructing the satellite’s TPU to run Gemma 4 to compose a space-themed haiku. As the satellite passed over near Svalbard, Norway, it sent back the three-line poem: *Blue marble below* *Data streams in silent grace* *Starlight feeds my core.* [Translation: "Blue planet beneath, data streams flow silently, starlight nourishes my core."] The satellite carries four TPUs, but power constraints limit operation to one TPU at a time, with each session lasting around 20 minutes. While space is frigid, there is no air to cool chips, so heat pipes and radiators are used to dissipate heat outward. The impact of cosmic radiation on chip performance also requires further testing. Notably, AI has already operated in space before: last December, U.S. space computing startup Starcloud ran Gemma on orbit using NVIDIA H100, and even trained a small language model. Google’s latest test uses its own TPUs; upcoming challenges include ensuring long-term stable chip operation and enabling collaborative computing across multiple satellites.

Token issuance platform Flap announced in a post that its official website flap.sh is currently experiencing DNS hijacking, which may redirect users to unauthorized sites upon access. The team is urging users not to access the website, connect wallets, or sign any transactions until further notice to avoid asset security risks. Flap added that the team is working to restore the website and will release further updates once the site’s security is confirmed.

Security firm Blockaid has issued a warning that its system detected an ongoing front-end attack tied to the StepDrainer toolset, targeting three websites: kintsu, flap, and stonkbrokers. Blockaid noted that malicious scripts were injected into the compromised sites; after users click "Connect Wallet", they may be redirected to a malicious asset transfer page. The firm advised users to temporarily refrain from using the affected decentralized applications, adding that it is currently investigating the attack, collecting additional information, and will release further updates later.

Marking the 18th anniversary since Satoshi Nakamoto published the Bitcoin whitepaper *Bitcoin: A Peer-to-Peer Electronic Cash System* on the P2P Foundation website. The Bitcoin network officially launched on January 3, 2009, with an initial transaction price of $0.0008. According to HTX market data, Bitcoin’s current price stands at $77,376, with a total market capitalization of $1.56 trillion.

Beating AI Insight News Flash: Factory has announced the acquisition of data infrastructure startup Typedef, with plans to integrate Typedef’s data context technology into its Coding Agent Droid. Typedef can analyze data engineering code such as SQL, and map dependencies between database fields, data processing workflows, and downstream reports. For example, modifying a database association field may let the program run normally, but cause double-counted revenue in financial statements. Typedef helps AI proactively identify such cascading impacts and trace the root cause when data errors occur. In its in-house data engineering tests, Typedef previously found that adding this context technology boosted the agent’s task success rate from 46.4% to 81.5%.