According to @santimentfeed, Bitcoin (BTC) dropped to $72.8K before bouncing as clarity from a passed bill prevented a U.S. government shutdown. According to @santimentfeed, DeFi liquidations totaled about $30M during the move, underscoring elevated crypto volatility. According to @santimentfeed, their latest insight analyzes the volatility drivers and what traders should watch next. (Source)

According to @ai_9684xtpa, a previously dormant buyer resumed accumulating ETH after two years, not selling during drawdowns to 1522 or rallies to 4461, with addresses 0xF788BB5F380f7972B5963f2eF8bE6F1Df4104d25 and 0x5F44E9030bbc1c5dbc8122C85D82dB2661117f86 flagged as likely controlled by the same entity (source: @ai_9684xtpa on X; ARKM Intel explorer). The entity’s most recent prior ETH purchase was in 2024-02 at an estimated cost of 2631.88 (source: @ai_9684xtpa on X). Roughly seven hours ago, the wallet bought 4020.61 ETH on chain at an average of 2174, about 8.74 million USD, bringing total holdings to 5122 ETH with an average cost of 2269, close to spot (source: @ai_9684xtpa on X; ARKM Intel explorer). (Source)

According to @OnchainLens, an Ethereum whale deposited 8,329 ETH to Galaxy Digital after holding for four months and is down $19.69M on the position. According to @OnchainLens, the address initially withdrew 11,860 ETH from OKX for $46.26M and still holds 3,530 ETH worth about $8M, with figures cited from Arkham. (Source)

According to @ai_9684xtpa, a wallet described as smart money that previously realized $99.22 million in ETH swing profits withdrew 750 cbBTC from Coinbase over the past 8 hours, totaling $55.999 million at an average $74,665.8 (source: @ai_9684xtpa). The same thread reports 20,392 ETH were withdrawn from multiple exchanges over the past 9 hours valued at $47.17 million with an average $2,313.35, and an initial 500 cbBTC position was established worth $38.74 million at an average $77,484.46 (source: @ai_9684xtpa). Wallet addresses shared for tracking are 0xFB78AA8F38843629e89951D9db6FdC398d75e0A3 and 0xd4584Bf988C9E8994688b56484E2F74cEAEeFB20 (source: @ai_9684xtpa). Traders may view these reported exchange outflows and self-custody accumulation as dip buying that could tighten near-term spot supply around the cited cost bases, offering actionable levels to monitor for BTC, ETH, and cbBTC (source: @ai_9684xtpa). (Source)

According to @EmberCN, a whale’s ETH long on Hyperliquid was liquidated with an estimated 230M USD loss. Per @EmberCN, the same address is also stopping out a leveraged spot position, selling 50,000 ETH (about 112.8M USD) to reduce leverage. Per @EmberCN, the address opened a 148,000 ETH leveraged long at 2,883 USD roughly a week earlier, after which ETH fell toward 2.2K USD, leading to the liquidation and de-leveraging. (Source)

According to @ai_9684xtpa, Vitalik Buterin’s planned donation sale of ETH is ongoing, with 1,511.42 ETH sold on chain at an average price of $2,298.43 from a 16,384 ETH allocation, leaving 90.7% remaining; source: @ai_9684xtpa on X and Arkham Intelligence wallet 0xfEB016D0D14AC0Fa6d69199608B0776d007203B2. The sales align with Vitalik’s post outlining the withdrawal of 16,384 ETH to fund donations over the coming years and a subsequent transfer from his personal address to a multi-signature wallet; source: Vitalik Buterin on X x.com/VitalikButerin/status/2017145595819933745 and @ai_9684xtpa on X. Traders can monitor further distributions from the cited multi-signature wallet as the remaining ETH is sold over time; source: @ai_9684xtpa on X and Arkham Intelligence wallet 0xfEB016D0D14AC0Fa6d69199608B0776d007203B2. (Source)

According to Lookonchain, Machi Big Brother was liquidated another 10 times, with the 250K USDC he deposited one day earlier now down to 57.7K USDC (source: Lookonchain). According to Lookonchain, the pattern is deposit funds, double down on longs, get liquidated, repeat, highlighting persistent overleverage risk (source: Lookonchain). According to Lookonchain, this sequence underscores how averaging into leveraged longs can rapidly erode capital when positions are repeatedly liquidated (source: Lookonchain). (Source)

According to @lookonchain, Trend Research and Garrett Jin are selling ETH to repay loans and avoid liquidation amid the ongoing drawdown. According to @lookonchain, they deposited a cumulative 316,185 ETH (about $738M) to Binance over the past three days with the intent to sell. According to @lookonchain, these exchange inflows indicate active deleveraging by large holders and immediate sell-side supply. (Source)

According to @ai_9684xtpa, on-chain tracking of address 0xFaf1358Fe6a9FA29A169Dfc272B14E709f54840f shows Trend Research sent 35,000 ETH to Binance in the past 14 hours and a further 20,000 ETH two hours later, with suspected single-day sales reaching 65,000 ETH; since Feb 1, cumulative deleveraging sales total about 160,000 ETH (approximately USD 367 million) with an estimated realized loss of about USD 128 million based on a cost near USD 3,104 and deposit price near USD 2,299, and the source reports an updated liquidation range around USD 1,686–1,974 with remaining holdings near 498,000 ETH. (Source)

According to @ai_9684xtpa, the 1011 whale who previously faced 230M dollars in liquidations allegedly sold 35,000 ETH in the past 8 hours and appears to have fully closed the 148,000 ETH accumulated on Jan 26–27, realizing a 76.41M dollar loss, source: @ai_9684xtpa. According to @ai_9684xtpa, during the rapid ETH drop between 00:38 and 01:22, the whale deposited another 78.31M dollars worth of ETH to Binance, and since Feb 1 has transferred a total of 156,185 ETH to exchanges valued at about 370M dollars, source: @ai_9684xtpa. According to @ai_9684xtpa, in the past 11 hours the whale received 806.47 BTC from B2C2 and Coinbase and now holds 40,411 BTC and 783,524.43 ETH, with on-chain assets exceeding 4.95B dollars and BTC exposure roughly triple ETH, source: @ai_9684xtpa. (Source)

According to @OnchainLens, on HyperLiquid, wallet 0x8d0 deposited $3M USDC and opened a BTC long with 20x leverage after previously losing over $11M, source: @OnchainLens. According to @OnchainLens, wallet 0x218 deposited $5.2M USDC and opened a BTC short with 14x leverage after previously making $10M, source: @OnchainLens. (Source)

According to @EmberCN, Trend Research cut exposure by selling 153.5K ETH at an average $2,294 and repaid $266M USDT to deleverage. According to @EmberCN, their ETH borrowing positions show liquidation levels between $1,685 and $1,855, clustering near $1,800, and they still hold 498K ETH with a $3,180 cost basis. For traders, the $1,800 zone cited by @EmberCN as a liquidation cluster is a key level to monitor, while the deleveraging reported by @EmberCN lowers stated leverage but the large remaining holdings keep exposure significant if price approaches that band. (Source)

According to @OnchainLens, Trend Research transferred another 15,000 ETH (about $33.08M) to Binance, bringing total deposits to 153,588 ETH (about $352.43M) explicitly to sell and repay a loan. According to @OnchainLens, these on chain data-confirmed exchange inflows indicate immediate sell-side supply from a large holder on Binance. According to @OnchainLens, the transfers are tracked via the Arkham Trend Research entity, highlighting the need to watch exchange inflows and liquidity conditions in ETH pairs. (Source)

According to @DeepLearningAI, Zhipu AI launched GLM-Image, an open-weights image generator designed to produce clearer, more accurate text inside images, source: @DeepLearningAI. It uses a two-stage pipeline that separates layout planning from detail rendering to improve text fidelity, source: @DeepLearningAI. The model outperformed both open and some proprietary competitors in text rendering tasks, source: @DeepLearningAI. (Source)

According to @OnchainLens, a wallet labeled Bitcoin OG (10/11) deposited an additional 15,000 ETH to Binance to sell and repay a loan, bringing total deposits over the past seven hours to 35,000 ETH. According to @OnchainLens, the entity also withdrew $82.39M in USDT and USDC to repay the loan, citing Arkham Intelligence's on-chain entity tracker. (Source)

According to @ItsDave_ADA, live on-chain data over the last hour shows real DEX swaps, real transactions, and resulting liquidity changes on Cardano, captured directly from the syncing ledger with no simulations or replays (source: @ItsDave_ADA on X). According to @ItsDave_ADA, his DEX battle view reveals that competition on Cardano is active—liquidity competes, routing competes, and execution quality matters—and DEXs contending for flow is a good signal the network is being actively used (source: @ItsDave_ADA on X). According to @ItsDave_ADA, he upgraded his arbitrage dashboard to combine swap execution, pool state changes, monitoring, and analytics in one place, emphasizing that Cardano’s determinism makes state transitions predictable, traceable, and verifiable for reliable operations over time (source: @ItsDave_ADA on X). According to @ItsDave_ADA, these observations highlight real-time liquidity routing and execution dynamics traders can monitor directly on-chain to evaluate DEX performance on Cardano (source: @ItsDave_ADA on X). (Source)

According to @santimentfeed, Bitcoin discussions have shifted from calling for a bounce to speculating how low BTC could go, with social metrics showing a rise in mentions of lower price ranges versus higher ones (source: @santimentfeed). According to @santimentfeed, its chart tracks mentions of 50k-59k versus 90k-99k ranges, where higher-range mentions signal greed and lower-range mentions signal fear (source: @santimentfeed). According to @santimentfeed, this fear-heavy crowd positioning supports a contrarian case for a short term relief rally while many retail traders assume sub 60k BTC is a given (source: @santimentfeed). According to @santimentfeed, traders can monitor sentiment spikes as potential inflection cues within this framework (source: @santimentfeed). (Source)

According to @karpathy, enabling FP8 training in the nanochat GPT-2 reproduction delivered a 4.3 percent improvement in time to GPT-2, reducing training to 2.91 hours on a single 8x H100 node. According to @karpathy, at spot pricing an 8x H100 run can cost about 20 dollars, while a previous 3.04 hour run cost about 73 dollars, highlighting roughly a 600 times cost reduction versus OpenAI’s original GPT-2 training. According to @karpathy, FP8 on H100 offers 2 times theoretical FLOPs but practical gains are limited by scaling conversion overhead, partial non compute bound training, and small GEMMs at GPT-2 scale, yielding about 7.3 percent per step speedup and roughly 5 percent net after adjusting the training horizon. According to @karpathy, torchao reported a 25 percent FP8 speedup on Llama3 8B, implying larger models may benefit more, and he expects further gains by selectively applying FP8 to layers and tightening numerics. According to @karpathy, additional wins came from Flash Attention 3, the Muon optimizer, gated residual and skip connections, and value embeddings, and he published a reproducible setup and a time to GPT-2 leaderboard on GitHub. (Source)

According to @lisaabramowicz1, private credit is taking a big hit from its software exposure, citing Barclays as the source. Barclays data shows software now accounts for about 20% of BDC investments, up from around 10% in 2016, source: Barclays via @lisaabramowicz1. Given this concentration reported by Barclays via @lisaabramowicz1, traders in BDC equities and loans should closely track portfolio updates for software-related non accruals and realized losses. (Source)

According to @rseroter, Google Cloud’s BigQuery Conversational Analytics paired with the TimesFM foundation model can answer data questions with a single prompt, replacing workflows that previously required deep SQL and multiple tools (source: @rseroter). The capability enables zero-shot time series forecasting, requires no model building or cold start handling, and supports chat-driven data preparation and visualization for BigQuery users (source: @kweinmeister via Google Cloud Medium). For data teams, this lowers the barrier to forecasting and accelerates exploratory analytics directly within BigQuery’s environment (source: @rseroter; source: @kweinmeister via Google Cloud Medium). (Source)