MiniMax rallied nearly 11%, with some longs turning from losses to profits.

Beating AI News Flash: A post on Shen Ma’s WeChat Moments reveals that Pang Tianyu, Chief Research Scientist of Tencent’s Hunyuan and Head of Multimodal Reinforcement Learning Technology, has left his role and is likely to launch a startup next. Pang, a Tsinghua University Computer Science PhD advised by Professor Zhu Jun, previously served as Senior Research Scientist at Singapore’s Sea AI Lab. He led multimodal reinforcement learning research at Hunyuan, covering areas including image generation and vision-language models. In June, Tencent open-sourced the multimodal reinforcement learning framework UniRL, with Pang listed as a contributing author; the framework was updated to v0.2.0 in October. In July, Tencent merged Hunyuan’s Large Language Model Division and Multimodal Model Division to form the Basic Model Division, overseen by Yao Shunyu. That same month, Hu Han, Head of Hunyuan Multimodal Understanding, left to start a business. In September, former OpenAI researcher Tian Yonglong took over as Head of Hunyuan Multimodal.

According to Bitget market data, Unitree Technology’s token has continued to rally in the afternoon, with its current gain exceeding 6%.

Goldman Sachs has released a US economic analysis report noting that the Democratic Party’s probability of securing a majority in the House of Representatives exceeds 90%, while control of the Senate remains uncertain. If the Democrats take the House, fiscal deadlines including the December 11 expiration of spending authorization and the 2027 debt ceiling will become key legislative battlegrounds. Goldman Sachs projects that the Democrats may leverage forced events such as government shutdown risks or the debt ceiling to push for policy concessions. In a Democratic victory scenario, while major fiscal policy changes are unlikely over the next two years, spending will likely rise slightly, with 2027 expenditure projected to increase by roughly 0.25% of GDP. On tariffs, Goldman Sachs believes that while the Democrats aim to roll back some tariffs, Trump will likely release the results of the Section 301 investigation and impose new tariffs after the midterm elections. The impact of these new tariffs will outweigh the Democrats’ partial cuts, potentially leading to a net increase of just under 2 percentage points in the US’ effective tariff rate. In the AI and data center space, attention to data center issues is expected to cool after the election, with little likelihood of a nationwide construction pause. Congress’ focus will shift to AI regulation, where a moderate compromise including basic federal oversight and corporate self-testing may be reached. If the Democrats win either chamber of Congress, investigations into large tech companies are anticipated, and stricter AI restriction bills may also be introduced at the state level.

According to MLM On Chain monitoring, Paradigm has un-staked 2.55 million HYPE tokens via 15 separate wallets over the past two days, worth roughly $224 million. This accounts for 17.85% of its current 14.26 million HYPE tokens staked on Hypercore, which are valued at approximately $1.26 billion. The monitor noted that Paradigm also un-staked large volumes of HYPE several weeks ago, before aggregating the tokens into new wallets and re-staking them, so a similar operation is expected this time. The specific arrangements following this un-staking remain unconfirmed as of now.

According to TradingBeats' monitoring, ZEC has rebounded continuously from its intraday low of $1,210.8 after a pullback today. As of press time, ZEC is trading at $1,272, up around 2.62% since 00:00 UTC today and 5.05% from its intraday low, with a peak of $1,276 hit. During the rebound, two whales successively opened or re-established ZEC long positions, totaling around $23.268 million in buying, and their combined positions have an unrealized profit of roughly $429,000. The address starting with 0xf081 bought around $7.564 million at an average price of $1,232.89, 1.82% above the intraday low, and currently holds an unrealized profit of around $239,000. Another whale, address 0xbf732, re-purchased around $15.703 million during the same period at an average price of $1,256.28, 3.76% above the intraday low, with an unrealized profit of roughly $190,000. Additionally, this account also bought around $1.508 million worth of XMR.

According to Solana’s official GitHub page, the Solana Foundation has merged the SIMD-0567 proposal into its main branch. The proposal aims to replace the existing Associated Token Account (ATA) program with an optimized version called p-ATA, built on Pinocchio, to reduce compute unit (CU) consumption and boost execution efficiency. Per the proposal, p-ATA uses a no-standard-library design, avoids heap memory allocation, parses instruction and account data via zero-copy, and minimizes dependencies. Developers’ prior benchmark tests show p-ATA cuts transaction execution time by roughly 35% compared to the legacy program. When a Rent Sysvar account is provided, CU consumption drops further, though actual execution time rises slightly, with transaction-level latency remaining around 1%. The proposal went through multiple rounds of technical review, and was officially merged 17 hours ago after completing formal verification of the new program, updating the program’s binary hash, and finalizing the migration process. It’s important to note that merging the proposal does not equate to a mainnet program replacement; the new program still needs to be deployed in line with the established migration workflow.