OrangeBot.AI Digest — 2026-10-25

The fastest headless browser for agents and harnesses

Your iPhone becomes a name badge that faces the room

One local AI server for all your Windows apps

Product observability for teams shipping fast with AI

AI summaries with tap-to-jump timestamps, right in Telegram

Lyric video ads for every part of your song, on the beat

Understand why your MRR differs and get a number you trust

Your favorite Git workflow. Reimagined.

Incident management for small teams

Shared workspace for humans and agents

See how anything went viral, from the first post

a tiny mac focus timer you start by pulling a paw

Secure data rooms your AI agents can run. Live in < 60 sec.

Let AI agents sign in for you without seeing your passwords

Vibe code, publish, and monetize games with AI

Existing vision-language models (VLMs) remain unreliable for fine-grained visual perception. In this report, we present WeVisOne-VL (WeChat Vision All-in-One), a powerful VLM that unifies object-level and pixel-level perception through structured next-token prediction. It supports multiple visual tasks including object detection, referring localization, counting, OCR, segmentation, and matting, enabling localization at any user-specific granularity, from points and boxes to masks and alpha mattes. Architecturally, we integrate a lightweight pixel decoder into the VLM via a special <OBJ> token to generate instance-level pixel outputs. We train the model with a three-stage curriculum SFT followed by mixed-task RL. Across over 30 benchmarks, WeVisOne-VL demonstrates leading performance on both object-level and pixel-level tasks, while maintaining robust rejection for object-absent queries. By releasing WeVisOne-VL to the open-source community, we hope to advance VLMs from broad visual understanding toward truly reliable perception systems for real-world deployment.

High-quality image editing requires precise modification of target regions while preserving the content of unrelated regions. Existing editing models produce flattened outputs that limit precise local manipulation and layer reuse, while decomposition methods typically optimize layer recovery separately from instruction-driven editing. To this end, we introduce LayerEditor, a unified framework that couples instruction-guided image editing with controllable generation of reusable RGBA layers. At its core, In-Context RGB-Alpha Token Modeling jointly represents the edited canvas and an ordered sequence of RGB-Alpha layers within a single token sequence, enabling semantic edits and explicit control over individual layer content. To retain the pretrained model's editing capability during adaptation to layered outputs, we propose Compositional Edit-Prior Distillation, which transfers its editing prior to both the edited canvas and the image composed from the predicted layers. To support training and evaluation, we construct LayerEditor-100K, a large-scale paired image-editing dataset with ordered RGBA layer annotations, and introduce dedicated benchmarks for layered editing. Extensive experiments demonstrate that LayerEditor maintains strong editing fidelity while generating accurate RGBA layers, supporting both flexible layer-level manipulation and multi-layer decomposition.

Cinematic video generation requires both temporal inference from limited inputs (i.e., sparse rather than frame-by-frame guidance) and factor-aware control of character, scene, action, and camera. Multimodal large language models (MLLMs) offer useful priors for both capabilities, but their representations are not automatically actionable for video generation: a Video Diffusion Transformer (DiT) may bypass them through native conditioning pathways, and using them does not ensure either factor-aware control or generation-compatible temporal prediction. We propose Factor-Aware Predictive Learning (FAPL), which establishes effective MLLM latent use as a shared prerequisite and then targets the two capabilities with complementary learning signals. First, latent-only generator grounding trains the DiT with a frozen MLLM and connector, strengthening the DiT's use of MLLM latents for generation. Second, for factor-aware control, paired videos specify a target-factor change and non-target preservation, and factor-span replacement trains the DiT to realize this relationship in its output. Third, for temporal inference, text-only and text-plus-complete-video observations share a video target, and the trained DiT is then frozen while the MLLM and connector are optimized through it. This final stage trains limited-input representations for generation while retaining complete-observation reconstruction as a complementary training mode. Extensive experiments demonstrate our consistently better performance on three cinematic video generation tasks (i.e., sparse-to-video, factor-aware editing, and multi-shot generation) over baselines and validate FAPL's contributions to MLLM feature use, temporal control, and factor coordination.

Despite substantial progress in Deep Search agents, their ability to conduct end-to-end research in real-world open Web environments remains insufficiently evaluated. Unlike static retrieval settings, information on the real-world Web is distributed across diverse sources, modalities, and heterogeneous formats, often involving complex interaction mechanisms that require agents to autonomously search for, access, and integrate relevant evidence during execution. To address this gap, we introduce the OpenDS task, designed to evaluate agents' ability to perform multi-step deep search in the live, open Web. This task requires agents to autonomously explore the real Web, discover and connect relevant evidence from multiple independent sources through Web search, browser interaction, and evidence processing, and ultimately determine a unique factual answer. Building on this task formulation, we construct OpenDS-World, a benchmark comprising 200 challenging tasks spanning diverse information modalities, heterogeneous file formats, and interactive access requirements, with independent human verification to ensure evidence validity and answer uniqueness. We evaluate 12 frontier models under a unified tool environment and execution protocol. The best-performing model achieves 39.0% accuracy, while human evaluation across 1,000 independent attempts yields 21.1% accuracy. Further analyses of tool usage, search cost, termination behavior, and failure recovery reveal substantial differences in research strategies across models, as well as persistent challenges in tool coordination and long-horizon search management. OpenDS-World provides a unified evaluation platform for systematically assessing and advancing agents' autonomous deep search capabilities in real-world open Web environments.

We present IQuest-Q1, a sparse Mixture-of-Experts (MoE) agentic foundation model with 320B total parameters (15B activated). IQuest-Q1 is developed through a three-stage process: (1) stable and efficient pre-training, combining Stable SplitNorm, Muon, and muP with a two-stage curriculum whose second stage increases the coding-data mixture, preserving a wide stability margin between safe and optimal learning rates; (2) fine-grained mid-training, expanding the data mixture across domains and formats and progressively extending the context; and (3) multi-scaffold with multi-version post-training, combining SFT, domain-specialized RL, and MOPD across various scaffolds to integrate complementary training signals and strengthen agentic coding capabilities. Furthermore, model merging combines different experts and checkpoints from different stages while mitigating conflicts among their parameter updates. The resulting IQuest-Q1 demonstrates capabilities in both coding and general tasks. We release IQuest-Q1 as an open-source model to support community research and development.

Recursive self-improvement (RSI) describes a system that improves not only its behavior but also the process by which it improves. Recent digital agents test this idea by rewriting their own code and checking each version with executable tests. Bringing these mechanisms into the physical world requires learning from partial observations through costly trials, while accounting for changing operating conditions and potentially irreversible consequences. We position RSI within robot learning and trace pathways from improving robot capabilities to improving the processes that generate them. We examine how physical experience can strengthen these improvement mechanisms and when the resulting gains can support further self-improvement of the mechanisms themselves. Gains carry forward only when feedback addresses remaining learning needs, the receiving component can use the change, its value survives changes in tasks or hardware, and continued practice remains safe and affordable. Existing studies, concentrated in manipulation, support behavioral improvement, benefits to later learning, and procedure improvement. Some systems also change the improver's knowledge or tools, but the physical studies reviewed here have yet to isolate gains attributable to recursion. We synthesize these findings into evaluation principles and research directions for machines that, trial by trial, safely improve the way they improve.

Selecting intermediate features from a frozen diffusion model for discriminative tasks is non-trivial because the useful signal is distributed across the denoising trajectory, network depth, and channels. In practice, this requires choosing a timestep, a block, and a channel subset. Existing methods either fix these choices by hand or optimize only a few extraction hyperparameters, most often the timestep, then reuse the same rule across inputs and budgets. We introduce Policy-Mask, a budget-conditioned policy trained on the downstream objective that jointly selects all three under a prescribed channel budget. Without normalization, the channel-score contribution grows with the number of selected channels and dominates the timestep and block updates. We normalize it by subset cardinality, keeping the three policy heads on comparable scales and stabilizing joint training. At matched budgets, Policy-Mask improves over fixed extraction and feature-filter baselines across segmentation, classification, and semantic correspondence, reaching 38.9 mIoU on ADE20K, 75.4% top-1 on CIFAR-100 at 128 dimensions, and 76.3 [email protected] on SPair-71k. Component-wise ablations show that block selection materially affects downstream accuracy, and the highest recorded channel-budget score occurs at 256 retained channels, identifying a practical operating point for compact diffusion representations. We further adapt Policy-Mask to diffusion transformers, including PixArt-Σ and SD3.5 MMDiT, by replacing U-Net channels with image-token hidden dimensions.

Large language models need effective ways to acquire and use external knowledge beyond what they retain from pretraining. Continued pretraining risks catastrophic forgetting, while retrieval at inference time can add substantial overhead. External parametric memory avoids online retrieval, but constructing its training data requires extensive offline search, and output-level interpolation provides only shallow fusion. We introduce Deep Fusion Memory (DFM), a parametric memory architecture that reduces this construction workload and incorporates learned memory into intermediate hidden states. A causal memory predictor learns to generate memory from the preceding context, using retrieved chunk embeddings from the corpus as supervision. The predicted memory is shared across tokens within each chunk and fused into their hidden states through gated cross-attention. Memory prediction and fusion are trained end to end with the backbone frozen, and inference requires no explicit retrieval. Compared with existing parametric memory approaches, which often construct token-level memory targets, DFM requires 32times fewer retrieval queries and 2048times fewer retrieved items to construct its training data. Across six QA tasks, DFM performs better over the strongest baseline by 12.71% on Mistral-7B-v0.3 and 11.19% on Qwen3-8B-Base. Beyond QA, domain adaptation yields average gains of 6.51 and 4.38 points over Qwen3-4B-Base and Qwen3-8B-Base, respectively.

Fine-grained visual understanding requires distinguishing subtle differences in local attributes and spatial relations, yet remains challenging for existing vision-language models. Unified multimodal models (UMMs) offer a new opportunity through generation pathways that preserve rich visual information complementary to understanding. However, existing generation-assisted understanding methods mainly rely on generative supervision or explicit generation and have not systematically explored direct use of native generation-side representations. To address this gap, we systematically analyze what generation-side representations offer for fine-grained understanding, finding that they are more sensitive to fine-grained semantic changes, capture task-relevant changes missed by understanding-side representations, and depend on correct spatial organization for this advantage. Guided by these findings, we design GAFU, a post-training framework that unlocks native generation-side information for fine-grained understanding. It compresses generation-side visual representations into compact \(G\) tokens that preserve fine-grained visual details, enabling task-conditioned retrieval by the understanding pathway. Experiments show that GAFU consistently outperforms SFT baselines trained on the same data across diverse visual understanding tasks and UMM architectures, with maximum relative improvements of 20.00\%, 17.94\%, and 8.62\% on VStar-Bench, OmniSpatial, and VSR-Bench, respectively. Overall, GAFU converts complementary information from the generation pathway into gains in fine-grained visual understanding, providing an effective path toward synergy between generation and understanding in UMMs.

This paper shows that JEPA-style latent world models (LWMs) can learn across diverse domains and data sources by adopting the language-conditioned interface of video generation models. We establish this approach as an effective and efficient basis for planning with imagination and downstream policy fine-tuning. To this end, we introduce LaLa-World, a Language-conditioned Latent World model that learns from simulation and robot trajectories together with broader image and video data. Two designs make this work. First, we describe both numerical actions and visual changes in language, enabling joint training of a single future predictor across heterogeneous datasets. Second, we build the predictor on a language model, allowing it to interpret language conditions while predicting future latent states. The resulting predictor achieves higher average MPC planning performance than existing LWMs trained separately for each domain or task. Adding image and video data further improves average planning performance and robustness to action phrasing. In a matched comparison with an RAEv2-based video generation model baseline, LaLa-World reaches comparable planning performance with 6times less training cost and 13.9times faster planning. Beyond planning, the pretrained LaLa-World serves as an initialization for LaLa-Policy, which jointly predicts actions and future visual representations. LaLa-Policy performs competitively with Cosmos Policy on LIBERO and RoboCasa, with up to 5.5times faster inference.

World Action Models (WAMs) have emerged as a promising approach to learning generalist robot policies using rich spatiotemporal priors from pre-trained video diffusion models. A recent line of WAMs improves inference efficiency by conditioning a separate action decoder on intermediate future video representations extracted from a single denoising step, avoiding costly iterative denoising required for video generation. This design makes action prediction depend directly on the quality of these future representations. In this paper, we introduce Future REpresentation learning via self-Distillation with temporal context (FRED), a framework that trains WAMs to improve the quality of future video representations from noise in a single denoising step. Specifically, FRED uses self-distillation to predict future representations from noise, using representations extracted from ground-truth future observations as targets. We further condition the video model on historical observations, providing temporal context for predicting future representations better. Across diverse real-world manipulation scenarios, FRED consistently outperforms competing WAMs. It achieves particularly large gains on long-horizon and out-of-domain tasks under zero-shot environment generalization, and substantially improves human-to-robot skill transfer, reaching 50.0% task success on tasks absent from robot training data and demonstrated only in human videos, compared with 19.4% for the best competing WAM.

Code-based video generation offers fine-grained control, targeted editing, and component reuse compared with direct video generation. This approach requires multimodal large language models (MLLMs) to translate visual and temporal requirements into executable programs that produce complete videos. However, existing code-based benchmarks primarily target specialized tasks, leaving complete video creation across diverse creative categories and input modalities insufficiently evaluated. We introduce Code2Video-Compass, a general-purpose benchmark for code-based video generation across text-to-video (T2V), image-to-video (I2V), and video-to-video (V2V) settings. Code2Video-Compass comprises 2,104 tasks spanning 10 categories, 24 subtypes, and three complexity dimensions covering construction, dynamic animation, and state dependency. It combines tasks curated from textual instructions and visual references with corresponding T2V, I2V, and V2V tasks derived from complete professional and user-created videos. We propose a dual-level evaluation framework combining 12 objective metrics and five subjective metrics. Objective evaluation covers operational performance, video quality, text-video alignment, and reference fidelity, while subjective evaluation uses an MLLM judge to evaluate instruction following and holistic video quality. Extensive evaluation of 15 recent proprietary and open-weight MLLMs from nine families reveals that even leading models face challenges in reliably generating high-quality complete videos aligned with textual and visual requirements, with performance generally declining as construction, dynamic animation, and state-dependency complexity increases. We anticipate that Code2Video-Compass will advance reliable, coherent, and controllable code-based video generation and inspire broader progress in programmatic visual creation.

Complex camera motion is central to visual storytelling, yet translating cinematographic intent into controllable, executable camera trajectories remains challenging. We introduce FilmFlow, an agentic framework that converts natural-language instructions or reference videos into explicit camera trajectories for dynamic 3D environments. To generate these trajectories, a Director agent constructs executable camera plans by composing high-level motions from concurrent atomic operations that jointly control translation, rotation, and framing. The Director is supported by three specialized agents: an Interpreter that extracts scene-grounded cinematographic intent, a Scout that selects initial viewpoints, and a Critic that evaluates rendered results and provides visual feedback for iterative plan refinement. We also introduce CinemaBench, comprising 150 Blender scenes from self-created and licensed sources, with five camera instructions per scene for 750 text-directed tasks and 60 reference videos associated with a subset of scenes. CinemaBench supports evaluation of both text-directed camera generation and video reference-directed camera generation. Across a diverse set of evaluated baselines, FilmFlow achieves the strongest benchmark performance and the highest user preference, demonstrating effective spatial reasoning, temporal coordination, and visual composition for executable 3D cinematography.

As model parameters scale up, sparsely activated Mixture-of-Experts (MoE) approaches have become essential for model scaling. However, mainstream models currently adopt hierarchical architectures, where experts in different layers, although topologically independent, exhibit substantial similarity and redundancy in their stored content. Compared to the hierarchically arranged Local Expert in existing mainstream models, we hypothesize that inter-layer shared Meta Expert hold greater potential. To this end, we conduct a multi-dimensional empirical comparison between Local and Meta Experts, and find that models incorporating Meta Experts exhibit lower parameter redundancy, together with lower cross-entropy loss and higher downstream task accuracy. With the activated quantity and the expert granularity held fixed, our model matches the performance of Vanilla MoE using only 70--80\% of the experts, substantially improving the model's scaling potential. Beyond removing inter-layer redundancy among experts and improving information storage efficiency, our theoretical analysis further shows that, compared with Local Experts, Meta Experts provide hidden states with more flexible residual pathways, thereby improving information retrieval efficiency.

LLM routing dynamically selects and coordinates heterogeneous language models across diverse tasks. However, existing approaches are largely built around a predefined execution paradigm, such as single-round, multi-round, or agentic routing, which fixes how model calls are organized and limits adaptation to different query needs. We propose RouteWeaver, a unified routing framework that integrates these routing modes within a single LLM policy, jointly learning which mode to use and how to select models within that mode. To jointly optimize mode selection and within-mode execution, we develop COMET-GRPO (COordinated Mode-Execution Training), which decouples the trajectory-level advantage into separate signals for mode selection and within-mode execution, enabling more targeted credit assignment to each decision level. We further introduce progressive exploration, which gradually shifts training from externally assigned modes to autonomous selection, so that the router first learns how to execute each mode before deciding when to use it. We evaluate RouteWeaver across mathematics, coding, knowledge, logical reasoning, and factual recall. RouteWeaver achieves an average accuracy of 61.4%, an 11.03% relative improvement over the strongest baseline, and outperforms all compared baselines in each of the five domains. Ablations show that both COMET-GRPO and progressive exploration are important for adaptive routing; without progressive exploration, the routing policy collapses entirely to single-round execution.

Elon Musk / @elonmusk : Elon Musk says SpaceX closed $40B in GPU debt, debt markets have shut off credit for AI data center firms, and 200+ neoclouds face “death and consolidation” —  Pretty much. As of the past few weeks, debt markets have almost entirely shut off access to money for any company that doesn't have substantial revenue independent of AI data centers. SpaceX closed $40B of debt for GPUs, but we're investment grade and have Starlink & launch revenue. That means...

Ann Davis Vaughan / The Information : Data center tenants now demand $5M+ a day in late-delivery fees, up from $1M this spring, spawning new “delay-in-delivery” coverage from brokers like Marsh —  This spring, Mike Mathews, who leads the global digital infrastructure practice at insurance broker Marsh …

Tom Matsuda / Financial Times : The iTunes Store is re-emerging as a venue for big pop stars, including Taylor Swift, to hawk collectibles to superfans and bolster their chart positions —  Record labels and artists including Taylor Swift and Ariana Grande are using the throwback music store to bolster their chart position and cash in on their fandom

Mark Gurman / Bloomberg : Sources: Apple tweaked macOS for its OLED touch MacBook Pro, due Wednesday, with a dynamic interface that adapts to whether touch or the trackpad was last used —  Also: The company rolls out a retail store tool to reduce wasted inventory.  —  Apple's touch-screen MacBook Pro will arrive on Wednesday …

Maria Curi / Axios : Lawmakers and Hill staffers in both parties say AI leaders are becoming a political liability as they swing between claims of prosperity and existential dangers —  Washington has always viewed Silicon Valley as culturally strange.  But the current crop of AI leaders is becoming …

Bloomberg : How a Blackstone-led $2.3B private credit loan to CoreWeave in 2023 became a blueprint for chip-backed debt, which is on track to become a $1T market —  “What is a GPU?”  Blackstone Inc. executives asked CoreWeave Inc.'s leadership.  Also: “Is AI real?”  —  These were entirely logical queries …

Cailey Gleeson / Fierce Healthcare : Baba, which develops AI tools to help Medicare and Medicaid patients connect to care teams and handle advocacy tasks, raised a $40M Series A led by FirstMark —  Applied artificial intelligence company Baba secured $40 million in series A funding to modernize the Medicare and Medicaid experience.

Ryan Lawler / Axios : CapitalOS, which helps vertical software platforms offer spend management, cards, and bill payment tools to SMB customers, raised a $21.9M Series A led by LGVP —  CapitalOS, which helps vertical software platforms offer cards and bill-pay tools to their SMB customers, has raised $21.9 million …

Hartley Charlton / MacRumors : Instagram launches an Apple TV app for watching Reels, longer videos, and live shows; the app previously launched on Fire TV, Google TV, and Samsung smart TVs —  Instagram is now available on the Apple TV, bringing Reels, longer videos, and live shows via the platform to Apple's streaming device for the first time.

Carl Franzen / VentureBeat : Kimi K3.1 Pro's debut draws scrutiny over allegations that influencers were asked to promote the model without disclosing a commercial relationship to Moonshot —  That's because Kimi K3.1 Pro scores so high on third-party benchmarks — according to both Moonshot and third-party AI model …

Dominic Chopping / Wall Street Journal : LSEG agrees to acquire Octus, an NY-based provider of credit intelligence, data, and workflow software, from Permira and Octus' management for $3.54B in cash —  The group said the acquisition would transform its presence in credit and private markets data  —  London Stock Exchange Group …

Augusta Saraiva / Bloomberg : A look at Meta's role in Brazil's runoff, where Lula da Silva alleges Meta restricts his content and a court ordered it not to interfere; Meta cites glitches —  A new villain is emerging in the closing stretch of Brazil's bitterly contested presidential election: Meta Platforms Inc.

Torsten Slok / Apollo Global Management : Outside of tech there has been no financial payoff to US companies for using AI, now viewed as a margin-hurting cost center, as VC-backed firms burn more tokens —  Despite massive investments and widespread hype around artificial intelligence, the technology has delivered zero financial payoff to corporate America outside of tech.

The Information : Source: Replit recently moved most AI spend from Anthropic to OpenAI, which gave 30%+ off in a deal worth $100M+/year, after Anthropic discontinued its discount —  OpenAI's efforts to undercut Anthropic on price are producing some results.  —  For instance, in May, Anthropic models powered 57% …

Dana Goldstein / New York Times : A look at Alpha School, a network of private schools that claims to teach kids 2x as fast using AI; data shows ~50% of K-8 students missed this goal in math —  Alpha School, a private network, is spreading into public education.  While its students' test scores are high, some parents and educators say it doesn't live up to the hype.

为建造数据中心，Google 在芬兰农村砍伐了 25 万棵树木，清理了多达 1300 英亩的森林，它甚至还铲除了表层土壤。芬兰监管机构以未完成必要环评为由下令工程暂时停工，Google 芬兰高管已为此道歉。Google 砍伐森林的一个后果是由于猎物减少，狼群进入了附近的村庄寻找食物。 当地居民表示，由于森林里的野生动物全都离开了，狼进村了。Google 还在另一个数据中心选址砍伐了 15 万棵树。芬兰不是唯一一个为建造数据中心而生态系统遭到破坏的地方。美国西弗吉尼亚州的栗子树林以及印第安纳州的一处湿地都面临为数据中心让路。

2010 年英国石油公司的深水地平线（Deepwater Horizon）深海钻油平台发生井喷并爆炸，引发了史上规模最大的石油泄漏事故。在封堵油井的 87 天内，有约 1.34 亿加仑原油和 77 亿标准立方英尺的天然气泄漏到墨西哥湾。16 年后，墨西哥湾的深海生态系统仍未从灾难中恢复。研究人员在 2011 年，2015-2023 年收集了样本，发现深海中上层物种的数量急剧下降。数据显示，鱼类数量下降了 63.5%，甲壳类动物下降了 62%，头足类动物（主要指鱿鱼）下降了 59.7%。小型鱼类磷虾数量下降了 72.8%，灯笼鱼下降了 62.9%，圆罩鱼下降了 69%。作为大鱼的食物，这些小鱼是海洋生态系统重要组成部分。

自古以来，人类就从抚摸狗中获得慰藉。大量研究承认养宠物对福祉的益处，但驱动这些益处的机制仍不清楚。在一项研究中，研究人员邀请了 64 名狗主人及其宠物来到实验室，研究抚摸狗是否能帮助人们更好地应对压力，以及身体接触是否是这种对大脑和身体产生影响的关键。研究人员设计了实验，以分离触摸狗如何影响人在压力期间和之后的大脑、心脏和情绪状态。51 名女性和 13 名男性，平均年龄 34 岁，带着自己的宠物狗参加了研究。狗必须至少 2 岁、健康，并在新环境中感到舒适。团队追踪了参与者的大脑活动，并使用心电图监测测试前、测试中和测试后的心脏活动。他们还询问参与者在关键时间点的感受，从平静到兴奋、从负面到正面、从失控到掌控。研究人员最初预测，在高压力任务后立即抚摸狗会帮助身体恢复，但身体触摸没有提供即时恢复优势。无论参与者抚摸狗还是不触摸狗坐在狗附近，脑波模式和心率指标都以相似速度恢复到静息水平。然而，在压力任务后的五分钟休息期间抚摸狗，在参与者再次面对压力时产生了差异。当参与者后来面对第二项压力任务时，他们的身体表现出弱得多的压力反应。这在脑记录中很明显，显示在参与者触摸狗后，α 波活动在整个第二实验区组中保持更高，表明大脑激活减少。在第一区组中没有触摸狗而恢复的参与者没有表现出类似效果。

人类与黑猩猩虽然差异巨大，但在与共同祖先分化并各自演化了约 500-600 万年后，双方只有 1.2% 的 DNA 不同。而一种远古病毒可能在其中扮演了重要角色，阻碍黑猩猩演化出与人类相似的大脑。大约 400 万年前，逆转录病毒 PTERV1 (Pan troglodytes endogenous retrovirus-1) 感染了黑猩猩但没有感染人类，它对黑猩猩的 DNA 进行了细微的改动，导致黑猩猩的大脑发育方式与人类出现重大差异。至于为何人类未曾感染 PTERV1，一种解释是人类体内的一种独特抗病毒蛋白 TRIM5α 保护了我们免受该病毒的侵害；另一种解释是，该病毒大流行时黑猩猩与人类的祖先生活在不同的地理区域。发表在《Science Advances》期刊上的一项研究未试图解答这一问题，而是揭示了病毒如何改变黑猩猩的大脑。人类和黑猩猩都有名叫 LINC00662 的分子，它为细胞提供构建特定蛋白质的指令。LINC00662 在人类大脑发育的早期阶段非常活跃，它有助于人类脑细胞成熟并生长。病毒的 DNA 则将黑猩猩的 LINC00662 基因失活。实验室研究发现，LINC00662 被关闭时，神经元及其延伸部分无法像 LINC00662 正常工作时充分成熟与生长。移除病毒插入的 DNA 后，LINC00662 的活性恢复了。研究人员表示，这证明 病毒确实参与了对该基因的抑制，但这并不意味着移除它就能改变大脑的发育进程。

Google 正资助美国规模最大的虚拟电厂项目。该项目通过向家庭和企业提供经济激励，鼓励他们在电力需求高峰期间减少用电量，从而为数据中心提供更多的电网容量。该项目有望在中大西洋地区的 PJM 电网提供 1070 MW 的电力容量——相当于一座大型核反应堆的输出电力。很多削减用电量措施能自动执行，如利用智能恒温器在夏季用电高峰期调低空调运行强度，或者通过电池系统提供临时电力。与 Google 合作的能源服务公司 Voltus 表示，通过聚合成千上万家企业和家庭的用电，该项目能在最需要用电的时候可靠提供大量额外电力容量，而且能在几年内实现。尽管科技公司若想持续建设大型数据中心，最终仍需建造传统的发电厂，但在过渡阶段虚拟电厂方案不失为一种更快捷、更经济且污染更少的选择。

中科院研究人员首次揭示出神秘古丹尼索瓦人的形态共性。丹尼索瓦人是一支已灭绝的古老人类支系，2010 年科学家依靠西伯利亚丹尼索瓦洞化石的 DNA 证据发现了这一人群。后续分子证据显示，丹尼索瓦人曾广泛分布在亚洲多地，生存时间大约在距今 30 万年至 4.5 万年前。但他们究竟长什么样？位于河北阳原的许家窑遗址 1976—1979 年共发掘出土包括头骨、下颌骨、牙齿在内的 20 件人类化石，至少代表 16 个不同年龄阶段的个体。许家窑人化石历经十余万年的埋藏，古 DNA 提取近乎无望，研究团队转向古蛋白质组学这一相对新兴的技术路径，从许家窑地点保存的 20 件标本中，无损获取内源性蛋白质开展系统发育研究，以古蛋白质证据明确支持许家窑人属于丹尼索瓦人。在此基础上，研究团队系统总结出丹尼索瓦人在颅骨、下颌骨和牙齿上的典型形态特征。综合研究推测，丹尼索瓦人具有低矮后倾的额头、粗壮眉脊、面部扁平、突颌减弱等面部特征。

美光桃园厂工会星期五（10月23日）说，会员投票否决公司提出的新利润分享方案，并警告可能采取进一步集体行动，争取更公平的奖金制度。美光过去数周一直与桃园及台中两座工厂的工会就薪酬问题谈判。两个工会共代表两厂 1 万 5000 名员工中的 1 万 2000 人。两个工会此前均要求，公司发放相当于 83 个月薪资的一次性奖金，并建立长期利润分享制度，将营业利润的 15% 用于员工奖金，按季而非按年发放。美光与规模较大的台中厂工会星期三（10月21日）达成协议。根据薪酬和福利改革方案，公司将从 2027 财年起，为全球员工增设季度利润分享奖金，与原有激励奖金计划（IPP）并行，新增奖金以现金和股票各半的方式发放。桃园工会与管理层仍未达成共识，并在星期四举行小规模罢工，向公司施压。工会星期五发表声明说，2300 多名会员中，约七成投票否决了公司星期四提交的利润分享方案。工会警告，可能采取进一步集体行动，争取更公平的利润分享制度，并称“抗争尚未结束，我们也没有放弃”。双方的下一轮协商定于 10 月 27 日举行。

Google 修改了隐私政策，允许将 Gmail 和相册数据用于个性化广告。新政策于 10 月 16 日生效。新政策没有默认启用，而是声明：“我们可能还会根据您的兴趣以及您在 Google 服务中的活动或个性化内容，向您展示个性化广告，具体取决于您的设置。例如，如果在 Google 搜索中搜索“山地自行车”，您可能会在 YouTube 上看到运动装备广告...我们不会根据您在 Google 云端硬盘、Gmail 和 Google 相册中的内容向您展示个性化广告，除非您选择允许其他支持个性化广告的 Google 服务使用这些服务中的数据。”也就是如果用户不希望 Gmail 和 Google 相册中的内容被用于个性化广告需要切断关联服务，方法是：Google 账号 —— 数据和隐私设置——个性化设置——搜索服务个性化——更多个性化智能服务——关联的内容应用，切断 Google Workspace 关联。

梵高于 1889 年创作的名画《The Starry Night》可能描绘的是 1889 年 7 月 22 日法国真实的夜空。《The Starry Night》被普遍认为风格化地呈现了梵高在法国普罗旺斯圣雷米的窗外看到的景象。争议在于这片夜空究竟是这位后印象派画家自己的幻想，即象征性表达，还是真实的天文观察结果。此外这幅画的具体创作日期也无法确定。为解决这一问题，韩国庆熙大学的研究人员开发了一个框架，将天文学模拟结果与星象位置测量数据、历史天气记录，以及梵高的信件结合起来。研究人员根据描绘这片景色的早期素描重建了画中的窄视场。他们发现，下蛾眉月、金星，以及金牛座的毕星团或许肉眼可见于 1889 年 7 月 22 日的黎明前后，并且其构型与画中的排列大体相符。历史气象记录显示，1889 年 7月 22 日前后的夜空清澈到足以看见这一天文排列。此外梵高信件中频繁提及的蝉表明当时气候炎热，这也更贴近 1889 年 7 月的历史温度。

希腊议会周五通过了一项法案，禁止 15 岁以下的儿童使用社媒平台。此举旨在回应公众对社交媒体会对未成年人的健康与安全产生负面影响之日益增长的担忧。禁令将于 2027 年 1 月 1 日生效。澳大利亚是首个通过立法限制未成年人使用社媒的国家。法国也曾通过类似法律，但于 8 月被宪法委员会裁定无效。希腊总理基 Kyriakos Mitsotakis 表示，社媒平台有 2 个半月的时间制定方案，核实用户年龄并保护其隐私。希腊政府表示，将支持家长共同遏制“无限滚动”现象——这种行为会导致儿童睡眠不足、注意力难以集中，挤占与朋友相处的时间。

日本文部科学省公布了 2026 年度文化勋章获得者名单，任天堂首席游戏设计师及代表董事宫本茂位列其中， 成为首位获得文化勋章的游戏从业者。文化勋章旨在表彰在文化发展方面做出特别显著贡献的人士，自 1937 年设立以来，获得者包括了画家横山大观、诺贝尔物理学奖得主汤川秀树、棒球明星王贞治和长岛茂雄等。宫本茂在记者会上表示：“游戏并非由我一人独自完成，而是与众多伙伴共同创造的成果。我是怀着代表大家的心情来接受这份荣誉的。”

今年的厄尔尼诺事件强度可能前所未有。它将于 2026 年底前后达到峰值，但其影响在此之后仍将持续很长时间。热带太平洋（厄尔尼诺的中心区域）的海面温度在 2026 年 11 月至 2027 年 1 月期间可能比平均值高出 3.8°C。 一次非常强的厄尔尼诺会扰乱许多地区的正常温度和降水格局。它可能给一些地区带来严重干旱，给另一些地区带来强降雨和洪水，并导致世界大片地区温度升高。极端天气事件可能会成倍增多，并发生级联效应。 联合国人道主义事务协调厅（OCHA）厅长弗莱彻警告称，全球可能有多达 5000 万人面临严重的粮食短缺。他呼吁各国政府做好充分准备。厄尔尼诺可能会给非洲、中南美和东南亚等地带来巨大破坏。

制药公司 Merck 和 Moderna 宣布，在 III 期黑色素瘤临床试验中，基于 mRNA 技术的个性化癌症疫苗相比标准治疗，将死亡或复发风险减少 42%。黑色素瘤是最致命的皮肤癌。两家公司正与监管机构接洽争取该疫苗早日获批。有 1,137 名患者参加了 III 期临床试验，他们此前均患有 IIB-IV 期黑色素瘤，已通过手术切除肿瘤。患者被以 2:1 的比例随机分配至两组：一组接受个性化 mRNA 疫苗联合标准治疗，另一组仅接受标准治疗。标准治疗采用 Merck 已上市的单克隆抗体抗癌药物 Keytruda。定制 mRNA 疫苗名为 intismeran 或 mRNA-4157，编码了癌症特有的突变，帮助免疫反应精准识别和对癌细胞进行攻击。Intismeran 和 Keytruda 治疗组两年后有 82% 的患者存活且未出现复发，Keytruda 和安慰剂组中这一比例为 70%。

The Browser Company 宣布其浏览器 Arc 将于 12 月 22 日关闭，用户将无法再能启动应用。Arc 基于 Chromium，于 2023 年释出了 1.0 版本，但 2024 年就停止了开发，2025 年 5 月宣布它正在开发 AI 驱动的浏览器 Dia。Arc 从 10 月 22 日起无法创建新的 Arc 账号，11 月 23 日起停用部分功能，12 月 22 日无法启动应用，Arc 账户将被停用，个人数据会在 30 天内被删除。The Browser Company 表示 Arc 用户可通过数据导出功能无缝使用其全力开发的新浏览器 Dia，声称用户在 Arc 中使用的浏览器功能在 Dia 中均可免费使用，而 AI 功能则是可选使用的。

研究员 Sammy Hajhamid 利用软件漏洞在 Switch 2 上实现越狱，而且能运行自制程序。越狱利用了一个专门运行在任天堂硬件上的编程环境和解释器 SmileBasic 4。Hajhamid 使用 OpenAI 的 Codex 对 SmileBasic 的语言解释器进行探测，花了两周时间发现了一个 32 位整数溢出漏洞，利用该漏洞能对部分系统内存进行读写操作。他在此基础上从 Switch 2 内部找到了其它漏洞，提权获得任意读写权限。他之后演示了在越狱的游戏机上运行自制程序。SmileBasic 4 已从任天堂的 Switch eShop 中下架，漏洞可能会在后续的固件更新中修复。