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

According to crypto journalist Eleanor Terrett, JPMorgan Chase, AMD, Applied Materials, Coca-Cola, Adobe, Dell, and SanDisk have opposed OKXICE's tokenization of their shares, bringing the total number of opposing companies to 33.

According to an official announcement, Binance Wallet has unveiled details of the PREDICTPrime Sale: each user has a subscription cap of 6 BNB; PREDICT Keys (BSC) will be distributed post the Pre-TGE phase; PREDICT tokens (BSC) will be airdropped to Binance Alpha accounts on TGE day. Earlier, Binance Wallet announced it would launch predict.fun’s Prime Sale Pre-TGE event from 4 PM to 6 PM on October 29, with eligible users required to participate using Binance Alpha points.

Bitget, in collaboration with Reality and Keyring, has released a research report exploring the synergies between tokenized stocks, perpetual contracts, and decentralized lending, while evaluating the potential returns and risks of related strategies via simulations. The report notes that tokenized stocks are expanding beyond traditional stock trading to additional use cases including collateralized lending, asset management, and systematic investment strategies. Based on Reality’s rToken, the research combines stock perpetual contract hedging and on-chain lending mechanisms to explore stock basis yield strategies. Historical simulation data indicates that between December 15, 2025, and October 19, 2026, the strategy index posted an annualized net return of approximately 7% in-sample and 4% out-of-sample, with a gross funding rate premium of 11%–14%. The report also highlights that financing costs, transaction execution fees, market liquidity, and risk management are key factors impacting the actual returns of these strategies. It is learned that Bitget and Keyring plan to jointly develop a stock basis index and build an on-chain lending market leveraging Reality’s rToken. Specifically, Reality provides tokenized stock asset exposure, while Keyring is responsible for developing the index methodology and constructing the related lending market. The index aims to capture funding rate and basis yield opportunities by combining rToken positions with corresponding stock perpetual contract hedging positions, offering a new strategic path for the application of tokenized stocks in on-chain financial scenarios.

Beating AI News Brief: AI startup Xplore has launched RWE-Bench, a benchmark to test large language models (LLMs)’ ability to solve real-world engineering problems. 11 models were evaluated on 20 tasks spanning mechanical, electrical, and aerospace engineering. GPT-6 Astra ranked first with a 37% single-attempt average pass rate, followed by Claude Opus 5.5 at 29% and GPT-6.1 Sol at 25%. Five models, including DeepSeek V4.1 Flash, GLM-5.3, and Grok 4.7, failed all tasks. The tasks covered engine crankshaft design, circuit board routing, and turbine blade thermal simulation: crankshaft design requires balancing size, weight, strength, and fatigue life; routing an 8-layer circuit board with nearly 300 installed components demands connecting 71 signal lines while controlling timing errors and impedance; turbine blade thermal simulation needs accurate calculation of heat transfer between high-temperature airflow and the blade. Each model used its respective Agent toolchain to operate professional engineering software directly. Each test ran for a maximum of 24 hours, with 5 independent attempts per task. Submitted designs were automatically validated via engineering software and physical solvers; any unmet hard requirement resulted in a failure. Out of 1,100 total tests, 4 tasks remained unsolved. However, after allowing 5 attempts per task, Claude Opus 5.5’s task pass rate rose to 70%, overtaking GPT-6 Astra’s 60%. Xplore aims to turn such engineering tasks into reinforcement learning training environments, enabling AI to improve engineering capabilities through repeated design, simulation, and error correction.

Unitree Technology announced that its operating revenue for the third quarter of 2026 reached 665 million yuan, up 69.93% year-on-year; net profit attributable to shareholders of the parent company stood at 113 million yuan, down 17.37% year-on-year; non-recurring net profit was 105 million yuan, a year-on-year decrease of 17.42%. For the first three quarters of 2026, its operating revenue totaled 1.817 billion yuan, up 55.71% year-on-year; net profit attributable to the parent company was 387 million yuan, a year-on-year increase of 270.54%; non-recurring net profit came in at 349 million yuan, down 18.78% year-on-year.

Beating AI Express News: Liquid AI has launched its third-generation base model LFM3, rolling out six small models each with around 2 billion parameters, covering text, vision, speech and decision-making tasks. The 1.7B text model supports a 256K context window, with 4-bit quantized weights weighing only about 1GB. All six models support single-GPU fine-tuning, allowing enterprises to train them on their own data before deploying to mobile devices, personal computers or private servers. Liquid AI redesigned the model architecture, combining short convolutions, local attention and global attention, and specialized training for agent tasks such as tool calls and web operations. Compared to the previous LFM2.5-2.6B, the new models have around 35% fewer parameters but deliver significant performance improvements on some tasks. In official tests, LFM3-1.7B scored 38.42 on the WebArena-Verified browser operation task, versus just 3.02 for the prior generation, and outperformed Qwen3.5-9B (28.94), which has roughly 5 times more parameters. In the AutomationBench enterprise workflow test, its task success rate reached 41.29%, while Qwen3.5-9B scored 15.10%. However, results reversed for complex programming and long text tasks: in the SWE-bench Verified programming test, LFM3-1.7B scored 43.4, lower than Qwen3.5-9B's 51.2. In the 128K long text test, the two models notched scores of 65.6 and 87.1 respectively. Liquid AI also explicitly noted in its model card that this small model is not recommended for complex agent programming tasks.