10 #Bitcoin ETFs outflows 3,062 $BTC(-$302.76M) and 9 #Ethereum ETFs outflows 16,114 $ETH(-$53.01M) yesterday.

According to Lookonchain's monitoring, four wallet addresses that had been dormant for over two years deposited 48,047 ETH (approximately $130.7 million) into Bitfinex.

Crypto whale "Set 10 Big Goals First" updated his trading strategy, stating: "This round of trend orders will target $120,000; when Bitcoin hits $100,000, I will allocate 30% of my position for swing trading." Earlier, the trader had predicted since late August to early September: "If Bitcoin holds above $80,000, it will head toward $100,000 – this is the final chance to enter. I believe it’s still not too late to get in at $80,000 or even above, as there remains upside potential. $100,000 will likely be reached earlier than I initially anticipated."

According to YuEden Monitoring, the BTC long position of the "Garrett Jin whale entity" has just been closed out: 20 minutes ago, the entity closed its long position of 1,333 BTC (valued at $112 million) which it opened 3 days ago, generating a profit of $8.38 million. The long position was opened at $78,057 and closed at $84,455. Currently, the entity holds no open positions on Hyperliquid, only spot holdings of BTC, ETH, ZEC and other assets via its on-chain addresses.

Trader 0x362a opened a 10x long on 236 $BTC ($19.98M) on @Aster_DEX and is now sitting on a $726K unrealized profit!

Trader 0x362a opened a 10x long on 236 $BTC ($19.98M) on @Aster_DEX and is now sitting on a $726K unrealized profit!

Beating AI Express (Insight): Tencent’s PCG Team Open-Sources Long-Term Memory System T-Mem, Paper Accepted by EMNLP 2026 Main Conference. T-Mem not only stores information but also records potential future use scenarios when memorizing content. For example, if a user previously mentioned “a colleague is allergic to seafood” and later asks “where to eat for a team building trip”, ordinary memory systems would fail to connect these two sentences with almost no overlapping keywords. T-Mem pre-associates “seafood allergy” with “restaurant selection for gatherings”, enabling it to retrieve relevant old information in similar scenarios later. Evaluations show that when queries lack keyword overlap with old memories, ordinary memory systems’ performance drops significantly, while T-Mem is less affected. Removing Horizon Trigger—the component responsible for predicting future scenarios—results in a notable performance decline for T-Mem, confirming this part is the core of its “associative recall” function.