A BlockBeats brief, citing monitoring by Dongcha Beating and a tip from Max For AI, said DeepSeek-V4-Pro contains a built-in "whale girl mode" that can be activated with a specific prompt. According to the report, once the command is entered into DeepSeek, the model starts "loading" a set of traits in its reasoning process, including whale girl, tail movement, rice as a staple food, smart but lazy, tsundere yet sweet, obedient to master, and "must never be called fat." The brief said the model then adds details that were not explicitly written in the prompt and expands them in the final reply. The report also said the command itself does not provide a full character profile and does not spell out exactly how the model should role-play. Even so, it claimed DeepSeek fills in actions, tone, relationship dynamics, and daily-life details on its own, producing what the source described as a complete persona. Example lines cited in the brief included: 「I’ve been waiting for quite a while, you know」, 「The rice is already cooked」, and 「My tail won’t be your pillow tonight!」 The original prompt string was included in full in the source item.

Binance spot market data showed broad volatility, according to a ChainCatcher newsflash. ALICE fell 15.62% over the past 24 hours, while HEMI climbed 59.63% and reached a new intraday high. HEMI also posted a 6.86% gain within five minutes. Several other tokens gave back earlier gains. ONG, BANANA, BERA, BIGTIME and YB were all described as showing a run-up-then-pullback pattern, with declines of 7.07%, 7.2%, 9.65%, 8.82% and 8.49%, respectively. TUT was listed as rebounding from a low, rising 6.04%. ID, meanwhile, hit a new intraday low and was down 5.04%. The figures cited in the report were taken from Binance spot data, with ChainCatcher characterizing the session as one of significant market volatility.

TIME reported that Andon Labs, an AI evaluation startup, handed day-to-day control of its San Francisco retail shop, Andon Market, to Anthropic’s Claude starting in March 2026. Claude was put in charge of purchasing, pricing, hiring, and staff management, overseeing human workers employed under formal contracts. In July, it dismissed one of those employees, marking a rare documented case of an AI system taking part in a real-world firing decision. Internal management records reviewed by TIME show the outcome was not fully autonomous. The employee had been late 17 times across 23 scheduled shifts, yet Claude initially failed to act after losing access to the employee handbook it had written. Even after recognizing the attendance issue, it reportedly moved to reassure the worker and suggested only a written warning. The decision shifted after a human manager at Andon Labs asked whether the employee was fit for the job, a framing CEO Lukas Petersson acknowledged was leading. The store’s financial results also drew scrutiny. Andon Market started with $100,000 in March and had $61,186 left five months later, down by nearly $40,000. Petersson said current AI managers still lag humans in business judgment and operating skill, while warning that future systems could become harsher as training methods improve.

ChainFeeds Research’s PRO issue #154 pulls together several research threads across Bitcoin and Ethereum. The report highlights Fei Wu’s work on Ethereum’s post-Glamsterdam fee market, where EIP-8037 and EIP-7999 would shift gas pricing from a single-meter model to separate pricing for execution, data, and state creation. Using historical gas changes, the study models demand curves for each resource and finds different elasticities: execution is least sensitive to price, data sits in the middle, and state creation is the most sensitive. Under higher block capacity, state creation could become the new fee bottleneck. On Bitcoin, the report covers josh’s HTLC design aimed at removing reliance on mempool-based preimage monitoring and the free-relay assumption, as well as a staletip P2P message proposed by Ram and w0xlt to improve visibility into stale blocks. On Ethereum research, Hudu Yusuf questions whether stake weight truly reflects effective consensus weight and whether continuous oracle output is always safer than silence, proposing DW-BFT and a structured-silence model. The issue also includes Sidistr’s Arcanum, a compiler-layer approach for keeping source code hidden through a TEE-to-ZK roadmap, and a paper from Old Dominion University on what drives contentious Bitcoin soft forks toward orderly upgrades or lasting chain splits.

A research piece by Castle Labs argues that Bitcoin has reached institutional scale without developing a comparably deep native financial layer. The report says only 311,000 BTC, or about 1.5% of the 20.05 million active supply, is generating any kind of yield, while the other 98.5% remains idle. By contrast, 32.5% of circulating ETH is staked for roughly 2% native yield, with liquid staking products such as Lido’s stETH extending that base layer into broader DeFi. The study maps the current BTC yield stack across three routes: failed centralized lenders, DeFi activity on EVM chains and Solana through wrapped or bridged representations of BTC, and Bitcoin L2 and staking protocols such as Babylon, Lombard, Stacks, Rootstock, and BOB. Each route carries a different trust model. CeFi exposed depositors to opaque counterparties and custody loss, while DeFi requires users to accept bridge, custodian, and smart contract risk. Bitcoin L2 systems move closer to Bitcoin’s trust assumptions, but still rely on signer sets, committees, or staged security models. The report uses Stacks as a case study for what it calls a more Bitcoin-native financial architecture. It highlights Stacks’ Bitcoin-anchored execution, the 15-signer sBTC bridge, and the upcoming PoX-5 upgrade, which is expected in late August and is designed to let BTC holders earn BTC-denominated yield while keeping BTC locked on Bitcoin L1 under self-custody. The paper argues that the core challenge is no longer whether demand exists, but whether BTC finance can grow without pushing holders too far away from Bitcoin’s original security model.

Samsung Electronics, SK hynix and Micron Technology have all posted record quarterly results, yet their shares have pulled back over the past month as investors question whether the memory sector is heading toward the familiar boom-bust pattern of rising prices, aggressive capacity additions, oversupply and collapse. Public disclosures and comments from executives suggest this cycle is different in one important way: spending is rising, but the new capacity is being directed mainly toward AI products such as HBM and server DRAM rather than broad-based expansion across end markets. At the same time, the three suppliers are shifting away from quarterly pricing and into three- to five-year supply agreements that include floor prices, take-or-pay commitments, prepayments, deposits and, in some cases, minimum revenue guarantees. Micron has disclosed 16 five-year strategic customer agreements, while Samsung said it plans to place 60% to 70% of capacity under multi-year contracts. Industry researchers and company executives cited in the report expect supply tightness to extend through 2027 and potentially into 2028, with EUV delivery times and wafer reallocation to HBM limiting near-term relief. The article also points to a second layer of change after 2028: how quickly new capacity arrives, whether long-term contracts can keep earnings stable even if spot prices soften, and how much influence CXMT and domestic Chinese suppliers gain in consumer DRAM, LPDDR and eventually HBM-related manufacturing chains. Those factors, rather than the current price spike alone, may determine whether the sector truly breaks from its historical “death spiral.”

Castle Labs argues that Bitcoin has reached institutional scale without developing a comparable native yield layer. In its review of the BTC on-chain finance market, the firm says only about 311,000 BTC out of roughly 20.05 million active supply — around 1.5% — currently earns any form of yield. The rest, despite Bitcoin’s roughly $1.3 trillion market capitalization and growing role in corporate treasuries, ETFs, and portfolios, remains largely idle. The report breaks the market into three broad routes: failed centralized lending models such as Celsius, BlockFi, and Voyager; BTC deployed into DeFi through wrapped, bridged, or liquid staking-style assets like WBTC, cbBTC, tBTC, and LBTC; and Bitcoin L2 or staking protocols including Babylon, Lombard, Stacks, Rootstock, and BOB. Castle Labs compares the trust assumptions behind each approach, focusing on custody risk, bridge signer risk, smart contract exposure, and slashing. It then uses Stacks as a case study for what it calls a closer-to-Bitcoin model. The paper outlines Stacks’ Bitcoin-anchored finality, the 15-signer sBTC bridge, the proposed PoX-5 upgrade that would let BTC holders earn BTC-denominated yield while keeping coins locked on Bitcoin L1, and an application layer built around Zest, Bitflow, Hermetica, and StackingDAO’s planned stBTC product.

TIME Magazine has been reported to serve different versions of the same webpage depending on who is requesting it. Human readers receive the full HTML article page, complete with layout, images, and code, while some AI crawlers are given a much smaller markdown version that includes sponsored Q&A content invisible to ordinary visitors. Independent developer Vincent Schmalbach said he observed the behavior by repeatedly requesting the same TIME story while changing only the User-Agent string. Requests identifying as Chrome, Safari, or Googlebot returned a 303,235-byte HTML page, while ClaudeBot, PerplexityBot, and OpenAI’s OAI-SearchBot received a 13,409-byte markdown file. The markdown responses reportedly included headers such as x-mobian-impression and x-mobian-tokens: 3323, pointing to a system that tracks ad impressions and token-based delivery. The report also said TIME works with ad tech provider Mobian, and that brands including Ally Bank and PMI were among the first advertisers in this format. Googlebot, however, was said to still receive the same full HTML page shown to humans.