CBS: U.S. Treasury Secretary and Trump Once Had Heated Debate on Tariff Policy

According to HTX market data, Bitcoin is currently trading at $78,025.63, down 0.02% over 24 hours; Ethereum is at $2,285.78, up 0.65% in the same period. Per CoinGlass data, BTC’s open interest-weighted and volume-weighted funding rates are 0.0075% and 0.0071% respectively; ETH’s two weighted rates stand at 0.0062% and 0.0064%, both in the neutral range. Compared to its previous negative levels, ETH’s bearish sentiment has eased notably, though overall rates remain lower than BTC’s. BlockBeats Note: CoinGlass converts funding rates from all platforms to an 8-hour timeframe. The funding rate is a mechanism designed to align perpetual contract prices with spot prices, involving fund exchanges between long and short positions; the benchmark rate is 0.01%, with rates above 0.01% signaling a bullish market and rates below 0.005% indicating a bearish market.

Beating AI News: Tencent’s Hunyuan has released research agent Hyra 2.0, shifting its focus from trial-and-error for individual problems to developing reusable scientific models, algorithms, and tools. To enable AI to conduct research over days or even weeks, Hyra 2.0 divides work into three agent layers: Overall (resource coordination), Mission (research planning), and Worker (experiment execution). Tasks are persistently stored; while waiting for training results, other research can proceed. Each conclusion must be accompanied by evidence such as code and logs, and is reviewed by an independent agent when necessary. Valid experiences are also accumulated as Skills for subsequent tasks. Tencent disclosed that Hyra 2.0 has trained two protein models from scratch: HyInv, a 1.96 million-parameter model that designs protein sequences based on target backbones, generating sequences 57 times faster than the widely used ProteinMPNN. In a warm-up test on a single H20 GPU, generating 128 sequences of ~400 residues takes only ~50 milliseconds. The other model, HyPLM-Flash, has 650 million parameters and was trained on 726 billion tokens. The weights, inference code, and technical reports of both models are now public. Hyra 2.0 also participated in optimizing the open-source RISC-V processor Xiangshan, boosting its SPEC CPU2006 performance per GHz by 5.79% while only increasing chip area by 0.3%. It also designed a five-finger "spiral ladder hand" with incrementally longer fingers; initially, this hand was ~11% slower than the best existing design, but after training with the same budget control algorithm, its simulated rotation speed increased by 28.6%. Currently, Hyra 2.0 is also involved in pre-training, post-training, and inference system optimization for Tencent’s next-generation Hunyuan model, aiming to help develop a stronger next-generation AI by improving R&D tools.

Japan’s largest Bitcoin treasury firm Metaplanet announced that between October 14 and 31, 2026, it repurchased a total of 96,447,400 common shares for roughly $160 million, via market purchases including ToSTNeT-3 transactions. Under a board authorization approved on October 13, the company can buy back up to 200 million common shares — equal to 14.7% of its issued common shares (excluding treasury stock) as of September 30 — with a maximum repurchase value of around $380 million. The repurchase program runs from October 14, 2026 to October 13, 2027.

Beating AI News: AI model aggregation platform OpenRouter has seen a sharp surge in token consumption. In the week of October 26, the platform processed approximately 194 trillion tokens, with its annualized throughput breaking 1 quadrillion (101?) for the first time. The total volume handled over the past 9 weeks exceeds the platform’s cumulative output from roughly 3 years and 3 months prior. One of the key drivers of this growth is AI Agents. OpenRouter’s official data shows that on February 6 this year, AI Agent token consumption surpassed human interaction for the first time. By the last week of September, the former had reached around 7 times the latter. Each step of an AI Agent’s execution typically requires resubmitting instructions, tool definitions, and prior operation records, with longer tasks involving more repeated context processing. OpenRouter’s growth has also drawn the attention of payment giant Stripe, which announced the acquisition of the platform in August this year, aiming to integrate cross-model routing, token billing, and cost optimization capabilities.

The Hong Kong Monetary Authority (HKMA) announced that the EnsembleTX project has entered a new phase, supporting 24/7 real-value tokenized transaction settlement to help banks and clients conduct global fund transfers, tokenized investment management, and treasury operations outside traditional settlement hours. HKMA Chief Executive Eddie Yue Wai-man stated that tokenized central bank money is being introduced as the "infrastructure layer" for round-the-clock settlement. The HKMA added that authorized dealers have completed secondary market pilot trades of tokenized Exchange Fund Bills, with the next step being to include regulated stablecoins in the relevant trials. Separately, HSBC Group CEO Georges Elhedery disclosed at the Hong Kong FinTech Week that the Hong Kong dollar stablecoin RedCoin has reached 50,000 users in about 10 days since its launch, with a target to expand to millions of users. RedCoin was launched on PayMe on October 23, supporting transfers and merchant payments, and over 3.3 million PayMe customers can open stablecoin wallets to use it. HSBC also plans to integrate RedCoin into its mobile banking platform, enabling subscriptions to tokenized investment products including Hong Kong dollar money market fund shares.

According to Bloomberg, Gerry Fowler, UBS’s chief European equity strategist and head of derivatives, said the market is underestimating the duration of the AI cycle, and AI trading remains robust. He noted that some hyperscale cloud providers now trade at valuation multiples below 20 times, while their revenue still grows 10% to 20% annually, reflecting a fairly pessimistic market outlook, which is why many investors are still willing to re-enter AI-related assets. Fowler believes massive AI spending is nearing the stage where it will start generating returns, with a potential inflection point in 2027: data center deliveries are expected to clear backlogs, driving revenue to gradually catch up with capital expenditures. He added that cash has not disappeared from the market, but has shifted to suppliers such as memory chip makers, which are now starting to return cash to investors. The pricing of memory chip companies also reflects what he views as an overly short AI cycle. As for the next set of beneficiaries, Fowler is more focused on global and European industrial firms that help build new factories. He noted that many infrastructure suppliers, such as cable and power equipment makers, have order books extending into 2027 and even 2028, leaving limited room for further upward revisions to earnings forecasts. ASML benefits from supplying equipment to memory chip makers expanding capacity, such as Samsung Electronics and SK Hynix, rather than directly to data centers; he said ASML and some European industrial firms are seeing upward revisions to their outlooks.