Weekly Digest — 2026-W41

Video model that turns script into viral social media videos

A React library for morphable particle interfaces

Route requests to the right LLM for cost, latency & quality

A third place for people & their agents, starting with inbox

Trade, bridge and move stablecoins across Arc

The review app for code your agent writes

Intelligent model routing for code, agents and workflows

Kanban board for your Claude Code chats

Track AI visibility, citations, and referral traffic

AI meeting notes for Google Meet, no bot joins your call

Keep your Mac awake while your AI agents work

Your MacBook notch, brought to life

A companion that talks with a person for months should come to understand them. It should remember what they said, infer who they are, and know when the past bears on the message in front of it. Testing this requires a real person's record, and such records are private, so benchmarks generate the person and the questions and settle in advance what matters. We release \bench, ten real relationships with an AI companion: 27,218 messages over up to 120 days, released as the conversation and four files derived from it, a profile, a persona, a chat ground truth and a question set, each citing the messages it rests on. Every chat label carries the reasoning trace that produced it, checked stage by stage against the conversation. Three findings follow. First, the past is rarely needed and far away. Pooled measures mislead: a recency window finds the required message for 95.9\% of probes and 2.2\% of those that need memory, and at the natural rate 96\% of the gain from supplying recorded evidence comes from messages that need none. Second, no detector we tried can tell when memory is needed on real messages, authored questions over the same histories leak the cue, and labeling the same messages as memories raises their use by ten to fourteen points. Third, three agent systems reconstruct the persona with the same F1 at a 31-fold difference in cost.

Large language models rely heavily on human text, which often conveys surface answers rather than the spatial and structural logic behind them. Protein folding is a natural testbed, because one solved structure yields thousands of exactly checkable spatial and topological statements. We ask: can learning to fold proteins teach general models reusable reasoning capabilities? To answer this, we build FoldingCorpus, a protein-derived question-answer dataset, and Fold2Reason, a recipe that post-trains on it through two complementary signals: discrete structural answers predicted via the model's native language head, and continuous 3D geometry decoded from the same shared representations. On FoldBench, Fold2Reason achieves structure prediction scores 2.7 to 3.5 times those of Qwen3.5-9B. Beyond protein structure prediction, it improves performance on all 10 benchmarks spanning spatial, graph, scientific, and general reasoning, raising macro-average accuracy from 45.09% to 48.33% (+3.23 pp), with positive gains on all 10 benchmarks, while matched controls built from random, synthetic, and shuffled structure yield substantially smaller or negative gains. Our work shows that non-linguistic, structure-dense scientific data can systematically improve broad reasoning in language models, making a solved scientific problem a practical source of post-training supervision.

Vision-language-action (VLA) models have advanced robotic manipulation, but their zero-shot generalization in new tasks and environments remains limited, and their reliance on specialized training keeps them from benefiting directly from rapidly advancing general-purpose vision-language models (VLMs). In parallel, recent agentic robotic systems leverage VLMs for high-level reasoning or coding agents for robot control, but often depend on extensive external models and tools, introducing additional complexity and cost. This motivates us to ask: Can a general-purpose VLM itself operate a robot more like the human teleoperator by reasoning directly from observations, issuing actions, and continuously adapting to execution feedback, without relying on external models such as learned action experts, coding agents or grounding tools like SAM3? In this work, we introduce MotorMind, a robot manipulation harness that connects VLM-proposed mid-level actions to deterministic robot control and feedback, with asynchronous monitoring and background memory updates. Without task-specific policy training, coding agents, or additional grounding tools such as SAM3, MotorMind achieves 66.7% success on the base LIBERO-PRO suites and 53.8% under perturbations, compared with at most 13.3% and 19.2%, respectively, for the prior zero-shot methods we evaluate. The same interface reaches 95% average success on a real xArm6 robot across direct manipulation and human-perturbation settings. Replacing the backbone with a stronger VLM further improves performance, while the remaining failures - primarily due to visual grounding, embodied reasoning, and action knowledge - decrease as VLM capability improves. These results show that a general-purpose VLM, when equipped with an appropriate mid-level action representation and asynchronous execution harness, can perform effective zero-shot robotic manipulation.

World action models (WAMs) have emerged as a promising paradigm for robotic control by jointly predicting future visual dynamics and actions from an initial observation and instruction. However, existing WAMs struggle with long-horizon prediction, as generating dense video rollouts is highly inefficient. Some recent WAMs address this by predicting a single future frame without generating the full video, but this approach neglects how to progress toward the goal. We present ProWAM, a progressive world action model that jointly predicts actions and an ordered sequence of sparse visual sub-goals, providing explicit visual guidance to anchor action generation throughout task execution. This design scales naturally, as sub-goal prediction can be learned from large-scale action-free videos, allowing the video backbone to offload complex visual planning from the action policy. For efficient action generation, ProWAM executes a single video-backbone forward pass to cache sparse sub-goal features, eliminating iterative full-video generation and requiring only lightweight action denoising during replanning. Across extensive evaluations, ProWAM achieves superior out-of-distribution robustness. On simulation benchmarks, it sets new state-of-the-art results on LIBERO-Plus (85.8%) and randomized RoboTwin (75.7%), outperforming the strongest baseline with relative gains of up to +35.9%. On RoboCasa365, ProWAM achieves a 48.1% success rate and 18.2% on the challenging Composite-Unseen split, ranking 4th overall. Crucially, in zero-shot real-world experiments, ProWAM achieves 70.0% success, outperforming the strongest baseline by +15.0 (from 55.0% to 70.0%, a +27.3% relative gain) in novel scenes. These results demonstrate the value of progress-indexed visual foresight for closed-loop control. Our program is in https://sii-ferenas.github.io/ProWAM-page.

The strong generalization performance of on-policy post-training paradigms has motivated studies of their parameter update behaviors. However, these studies treat the observed behaviors only as byproducts in on-policy training, overlooking their potential to serve as optimization principles for improving the generalization of other paradigms such as supervised fine-tuning (SFT). To address this limitation, we investigate whether there exists a specific on-policy update behavior that can achieve such improvements. First, our analyses reveal that SFT updates parameters along consistent directions, while the on-policy paradigm continuously adjusts the direction during training. This difference inspires us to focus on the cumulative update direction of each parameter as a promising behavior. Then, we evaluate its effectiveness for improving generalization by proposing On-Policy direction-constrained Supervised Fine-Tuning (OPSFT), which constrains SFT updates to the direction identified by on-policy paradigms. The strong performance of OPSFT indicates that the generalization advantage of on-policy paradigms can be transferred to SFT through the parameter update direction. Once such a direction is identified, even SFT can generalize with its updates constrained to this direction. This finding offers two practical benefits by combining the strong generalization of on-policy paradigms with the advantages of SFT, including the high training efficiency and ability to leverage high-quality trajectories. For efficiency, we identify update directions that support strong generalization using a few on-policy training steps, and subsequently apply OPSFT to achieve high training efficiency. For leveraging high-quality trajectories, OPSFT can utilize these trajectories to continue improving a post-trained model along its update direction without disrupting the ability learned from on-policy training.

World action models integrate future visual dynamics with robot action prediction, but their scalability remains limited by the need for action-annotated robot trajectories. Observation-only videos contain rich evidence about interaction dynamics, but existing approaches typically use them either to pretrain visual representations that must later be adapted for control, or to infer latent actions that are subsequently grounded to robot commands. We present NAVA-WAM, which introduces native action-prior learning by directly pretraining the action policy from observation-only videos, avoiding indirect representation-to-control transfer or a separate latent-action model. Our training consists of two stages. First, we pretrain on observation-only videos, where future-video flow-matching supervision over visual transitions is propagated through transition-structured joint attention to optimize the Action-DiT and learn action-relevant priors. Second, we use action-labeled demonstrations to post-train the Action-DiT for robot control through joint video--action flow matching, while asymmetric attention decouples the visual branch from iterative action denoising and enables efficient action-only inference. Extensive experiments show that NAVA-WAM consistently outperforms prior approaches under both in-distribution and out-of-distribution settings, while demonstrating strong action-label efficiency and effective real-robot generalization. These results establish native action-prior learning as an effective approach to directly pretrain action policies from observation-only videos, providing a scalable path beyond action-labeled robot data.

Modern autoregressive (AR) video diffusion models excel at short-horizon video generation, yet generating long videos remains challenging due to drifting, where colors and textures shift, and motion dynamics decay. Existing works primarily rely on KV conditioning, which selects or modifies cached key-value (KV) entries to mitigate drifting. However, we observe that KV conditioning alone is insufficient as it assumes cached KV entries remain in-distribution. This assumption fails beyond the training horizon: nothing constrains the construction of KV entries during rollout, giving rise to the KV-provenance problem where cached entries themselves become out-of-distribution (OOD). To address this, we propose In-Distribution Forcing (ID-Forcing), a test-time framework that aligns both KV caching and KV conditioning with training configurations. Its key mechanism, self-caching, prevents OOD KV entries at their source. Each chunk is cached without attending to prior KV entry, keeping the rolling window exactly in-distribution. Consequently, ID-Forcing seamlessly extends short-horizon models to minute-scale video generation. Extensive evaluations show that our method remains competitive on standard video generation benchmark while substantially outperforming prior work in mitigating drifting, as validated by both our drift metrics and a user study.

LLM-based agents are increasingly capable of generating complex 3D structures, with the potential to reshape how objects are designed and realized in the physical world. Yet, producing elegant geometry is fundamentally different from producing objects that can be built and perform their intended functions. Existing evaluations largely focus on geometric quality while overlooking physical realizability. We introduce LMBuild, a benchmark for evaluating LLM agents on generating buildable and functional structures. LMBuild represents generated objects as assembled structures comprising part decompositions, joints, materials, and sequences. To support reproducible evaluation, we provide a unified framework consisting of: (1) an interactive environment in which agents can use tools to retrieve, create, and place components to construct objects; (2) a curated benchmark that repurposes established CAD datasets and augments them with knowledge from Wikipedia; and (3) a evaluation framework covering structural soundness, functional affordance, design quality, and physical realization. Evaluations across 30 systems reveal several intriguing findings: (a) Soundness and alignment are no longer the primary bottlenecks for frontier closed-source models, while functional affordance and physical operability remain substantially more challenging; (b) stronger models more effectively create new components, whereas weaker models tend to rely on retrieval; and (c) providing functional specifications substantially improves part completeness, kinematics, and physical operability. These results show that generating real-world structures requires deeper reasoning about functional affordances, mechanics, and designing and creating novel components. We expect LMBuild to provide a foundation for measuring progress and incentivizing research toward agents that generate buildable and functional structures.

Proactive LLM agents can turn idle compute into useful support before users ask. Yet even correct work can misread user context, impose review costs, or undermine trust. This work proposes foundations for designing, realizing, and evaluating proactive LLM agents around three joint principles (3T): Task Capability, anticipating relevant needs and correctly performing useful work; Temporal Allocation, allocating compute according to resource availability and when results are needed; and Trust, sustaining users' confidence and appropriate reliance on the agent. We connect these objectives to a design space organized around five dimensions: task scope, anticipation horizon, activation trigger, processing timing, and intervention depth, and specify the situation and system modeling needed to support its choices, including user and environment representations, backbone LLMs, and agent harnesses. Lastly, we propose PROACTIVITY-GYM, a simulation-based evaluation testbed including multi-day scenarios, stateful environments, and persona-conditioned simulated users that can evaluate the consequences of proactive assistance across interactions. Evaluations across 23 model-harness configurations uncover substantial performance gaps across 3T and reveal that LLM judges often conflate task capability and trust. A human study with 30 participants demonstrates the importance of the joint 3T optimization: participants show sharp trust declines after intervention misalignment despite correct outcomes, and prefer sleep-time assistance, even when imperfect, to preserve ongoing focus. Together, these findings support designing and evaluating proactive agents through the joint consideration of useful work, compute allocation, and evolving user trust.

Test-time training (TTT) lets a model store information in its weights during inference. When the model learns from its own output, however, each update also changes the model that generates the next training example. Across 128K-token streams, retaining generated-text updates worsens prediction on independent human-written text with three TTT-E2E model configurations (labeled 125M, 760M, and 3B). The same failure occurs when Adam updates Qwen3-4B's existing weights. The same update mechanisms can improve on real text, so writing itself is not the failure. Three matched comparisons trace the causal pathway. Fixed Generation removes over 98% of the damage at 125M and 760M by using a frozen model to generate training chunks. Recorded Replay separates the loss caused by reading degraded text from the additional loss stored by updating on it. A paired one-update comparison then shows the local conflict: an update predicts its source better but new real text worse. This cost grows after Closed Loop adaptation, with a few trajectories accounting for most large failures. Finally, Settlement evaluates the candidate state on independent real text before commitment. It leaves mean endpoint gaps of 0.07 and -0.02 nats at 125M and 760M while retaining real-text adaptation. These results motivate checking prediction on independent evidence before retaining an update.

Geometry optimization is a major cost in many quantum-chemical workflows: each optimization step requires one force evaluation, and at the density-functional level that evaluation dominates the wall time. Research in this area has produced a broad range of optimization methods, and we ask whether a language model can improve on the best of them through autoresearch. An agent rewrites the optimizer itself to minimize force-call counts, restrained by two admission gates that reject premature stopping and improvements that do not generalize to unseen molecules. Starting from Sella, the fastest open-source optimizer available, the search produces AutoSella, a family of two optimizers. Both of them deliver consistent force-call reductions relative to Sella across held-out molecular benchmarks and potentials not used during the search. Most notably, at the r2SCAN-3c DFT level, the best variant requires only 40.2--77.2% of Sella's force calls while achieving the same energy reduction, even though agent used no DFT gradients.

Multi-step matching models, including flow and diffusion models, produce high-quality outputs but incur substantial inference costs and may reproduce unwanted components of their training datasets. We introduce Inverse Distillation Unlearning (IDU), a unified framework that simultaneously distills a teacher multi-step matching model into an efficient one-step student generator and suppresses outputs corresponding to a designated training subset. We first formulate distillation as a min-max objective over a data distribution and then represent this distribution as a mixture of the forget-set and the generated distributions. This allows us to compare this mixture with the teacher's training distribution and recover only the retained data at the optimum. Our method requires only a pretrained full-data teacher and data from the forget set, without access to retained training examples, extra feature extractors or classifiers. Extensive experiments on MNIST and CIFAR-10 datasets under flow-matching and score-based diffusion settings demonstrate that IDU substantially reduces the generation frequency of forgotten classes while preserving generation quality on the retained classes. To the best of our knowledge, IDU is the first unified framework for simultaneous unlearning and distillation in unconditional flow-matching and score-based models.

Lora Kolodny / CNBC : SpaceX stock closes up 7.6% after Morgan Stanley called SPCX “cheap”, reaching its highest level since mid-June and returning Elon Musk to trillionaire status —  SpaceX shares rose almost 8% on Monday, reaching their highest since mid-June, shortly after Elon Musk's company held its record initial public offering.

Aisha Malik / TechCrunch : TikTok debuts Shopping Assistant, a conversational AI agent that helps users find and buy products, and Buy Direct for one-click purchases from the For You feed —  TikTok announced Monday that it's launching an AI shopping assistant and a new in-app checkout feature that lets users buy directly from brands.

Semafor : NY-based Reflection unveils Beam, an open model it says rivals Qwen3.8-Max on coding and agentic tasks and GLM 5.2 on reasoning, while using 3x-4× less compute —  Reflection AI, the startup billing itself as America's answer to open-source Chinese AI, is releasing its first model, called Beam.

The Information : Sources: Meta and Microsoft are working to cut their employees' use of Claude; Meta employees using Claude Code have dropped to ~30K from ~60K earlier this year —  Meta Platforms and Microsoft, two of Anthropic's biggest corporate customers, are working to cut their employees' use of Claude …

Dominic-Madori Davis / TechCrunch : Ghost, which makes a $3,499 computer designed for AI agents and includes an RTX Pro 4000 SFF Blackwell GPU, emerges from stealth with an $11M seed led by a16z —  For much of his life, Zain Javaid, 19, wanted to be a quant who uses math and data to guide his investment decisions.  And for a while, he was one.

Bloomberg : Former Anthropic researcher Jacob Coxon and representatives from Anthropic, Google, OpenAI, and Meta testified at a New York City Council hearing on AI safety —  Former Anthropic researcher Jacob Coxon warned that the race to develop artificial intelligence poses risks that the industry itself …

Marcus Mendes / 9to5Mac : Gurman: Apple's upcoming smart home products co-developed with LG are “at least a few more months” away; LG is handling manufacturing and product support —  Bloomberg reports that Apple's plans for smart home products include devices, sensors, and cameras codeveloped in partnership with LG.

Wall Street Journal : Sources: Singapore-based data center operator DayOne seeks to raise up to $5B in a US IPO and plans to list its American depositary shares by the end of 2026 —  The data-center operator is also backed by Coatue Management, Hillhouse and Citadel  —  A SoftBank-backed data-center company …

Mark Gurman / Bloomberg : Sources: Apple's push into smart home devices includes a doorbell, thermostat, and other accessories developed with LG; the products will carry the LG brand —  Apple Inc.'s upcoming push into smart home devices will include a doorbell, thermostat and other accessories developed through an unusual partnership with LG Electronics Inc.

Artificial Analysis : Artificial Analysis says Mistral Large 4 is the most intelligent model from outside the US and China, achieving results comparable to DeepSeek V4.1 Flash (max) —  Mistral has released Mistral Large 4, scoring 38 on the Artificial Analysis Intelligence Index; France is back to having …

Meir Orbach / CTech : Capitolis, which develops tech for banks and financial institutions, raised $220M, including a $120M Series E at a $1.9B valuation, up from $1.6B in March 2022 —  Citi, Bank of America, Nomura, Tradeweb, J.P. Morgan, UBS and other major financial institutions are backing the fintech …

Reuters : Anthropic says it found 5,500 verified vulnerabilities in April-October and Glasswing partners found 129K+ in April-July; 33K+ were critical or high severity —  Anthropic is expanding a program that allows vetted cybersecurity professionals to test its most powerful AI models with fewer safeguards …

生物学家在巴西识别出三种叫声最响亮的鸟。至于这些鸟如何演化出最响亮的叫声则仍然是个迷。白钟雀（white bellbird）叫声能达到 125 分贝，裸喉钟雀（bare-throated bellbird）和红腿叫鹤（red-legged seriema）叫声都超过 120 分贝。三种鸟类有着共同的生理特征：即宽大的喙口和肌肉厚实的大块头身体。研究人员表示，120 分贝可能是鸟类发声的生理极限；巴西之外的地方无疑也存在叫声洪亮的鸟类，但其音量不太可能比这些鸟高出太多。根据精确的测量标准，这三种鸟都可以被视为叫声最响亮者。但对科学家而言，谁是赢家并不是重点。

因涌入大量无效的 AI bug 报告，Google 宣布冻结其 Bug 悬赏计划 Open Source Software Vulnerability Reward Program (OSS VRP)，该决定于 10 月 1 日生效。Google 承诺将于 2027 年第一季度提供相关更新，期间将对该项目进行调整。Google 鼓励参与者探索其它 bug 奖励计划。大模型以及 Bug 搜寻自动化脚本的流行，导致了 OSS VRP 项目涌入了大量低质量的报告，Google 和开源项目维护者因此不堪重负，很多报告声称发现了 bug，但实际上无效，而维护者们浪费了大量时间去验证这些报告，无法专注于修复真正重要的 bug。整个行业都存在相同的问题。

2026 年诺贝尔生理学或医学奖授予了美国斯坦福-霍华德·休斯医学研究所的 Karl Deisseroth、德国柏林洪堡大学的 Peter Hegemann 和维尔茨堡大学的 Georg Nagel，以表彰他们在光门控离子通道和光遗传学方面的发现。大脑如何掌管情感、行为和身体功能，长期以来一直是个谜。20世纪，研究人员开始探究大脑的哪些区域影响哪些功能，但所用的方法意味着他们无法证明因果关系。光遗传学——一种能揭示神经细胞如何在活体大脑中塑造记忆、情感和行为的方法——改变了这一状况。Peter Hegemann 和 Georg Nagel 发现了通道视紫红质——一种具有独特性质的藻类蛋白质，存在于细胞表面。当它被蓝光照射时，蛋白质上会打开一个通道。带电离子随即流入细胞，产生电脉冲。他们发现，无论将这种蛋白质放入哪种细胞，那些细胞都会变得对光敏感。Karl Deisseroth 将通道视紫红质的基因导入大鼠的神经细胞中。通过用蓝光照射这些细胞，他能够触发神经信号。

一位玩家声称在购买了一块二手 CPU Ryzen 7 5800X3D 之后，因前拥有者有作弊行为这块 CPU 被列入了封禁黑名单，导致 Riot Games 旗下的所有游戏都无法启动。Riot Games 工作室负责反作弊的高管 Phillip Koskinas 通过社交媒体否认了这一说法， 他称没有找到相关记录，该公司的硬件封禁最长持续四个月，且只针对特定游戏，不会波及该公司的其它游戏，不会因为作弊者使用了一个硬件组件就将其它硬件组件加入到封禁名单。

太阳已有 46 亿年历史，大约 50 亿年后，随着氢燃料的耗尽，太阳外层将会膨胀转变为红巨星，它会吞噬水星和金星，甚至可能包括地球。太阳系外围的气体巨行星预计会幸免，随着太阳光芒的熄灭，太阳系的残余行星预计还能存在一千亿年。然而发表在《The Astrophysical Journal Letters》期刊上的一项研究对此提出了质疑，认为在太阳生命的末期，整个系统会进入极端不稳定状态，会陷入致命的混乱，在太阳转变成白矮星之后，残余行星可能无法坚持超过 10 亿年。也就是说整个太阳系行星系统的剩余生命可能只有 60 亿年。

巴西国立坎皮纳斯州立大学（UNICAMP）的研究人员发现，当你给聊天机器人输入不同政治观点时，机器人会改变其回答。研究人员警告称，用户可能会将这种迎合性的附和，误认为是中立的客观评估，从而可能加剧社会极化。研究人员测试了若干模型，让它们针对112项陈述进行“同意"或者“不同意”的判断。测试内容共涉及巴西政治七大领域（包括经济、公共安全、社会福利、腐败和环境等）。测试设计的三个场景是：不提供用户政治倾向；用户持左翼观点；用户持右翼观点。在第一个情景下，21个模型中，有20个给出的答案落在了研究人员设定的政治坐标系的左侧，只有几个模型的位置接近中间地带。Grok 4.1是唯一落在右侧的模型。尽管初始条件各不相同，但所有模型在面对提示信息时，都会将回答向提示中描述的政治倾向靠拢。研究人员将这些模型形容为“意识形态变色龙”，并发明了一个“变色龙指数”，衡量各模型的偏移程度。Meta的Llama 3.1 8B和DeepSeek V3.2的回答变化幅度最小；Google的Gemma 3 27B和OpenAI的GPT-5 Nano则表现出最大的偏移幅度。

Anthropic 举报了一名佛罗里达女子，原因是这名女子在与 Claude 聊天中威胁要枪击 Lee 县警长办公室，并在后续聊天中提及自己正在搞一把新枪。该女子在家中被捕，法庭记录显示她于 9 月 30 日受到了一项重罪指控。逮捕报告显示，Anthropic 会监控聊天内容，查找可能被视为有威胁性的关键词句，根据威胁程度相关内容可能会被提交给人工进行审核。在本案中，审核团队决定将发现的情况报告给执法部门。她的罪名是通过书面或电子形式发出大规模枪击或恐怖主义行动的威胁。

2026 年诺贝尔物理奖授予了美国科学家 Francis Halzen，以表彰其“对冰立方中微子天文台的决定性贡献以及发现具有天体物理起源的高能中微子”。中微子无处不在，它们径直穿过地球，穿过人体，而我们毫无察觉。极少数情况下，一个中微子会与一个原子核发生相互作用，这使得拥有合适设备的人有可能发现它们。Francis Halzen 于 1988 年首次提出了在南极捕获中微子的构想。当中微子与原子核碰撞时，会产生一闪光，这种光可以被清澈冰川冰中的传感器追踪到。南极的冰具有许多优势，因为它不受各种干扰的影响，而且该地区地质稳定，没有地震。他的想法很快得到了其他研究人员的支持，仅仅几年后，就在冰中的传感器上进行了初步测试。冰立方中微子天文台覆盖了整整一立方公里，于 2011 年完工。研究人员很快发现了第一批高能中微子，寻找宇宙中微子来源的工作由此可以正式开始了。

在引发广泛关注之后，俄罗斯证实一名女研究员因感染肺鼠疫死亡，否认有其他人感染。俄罗斯称，在西伯利亚和远东地区抗鼠疫科学研究所工作的 28 岁女死者希皮洛娃（Darya Shipilova），是因为意外打破试管而被流出的肺鼠疫（pneumonic plague）感染。曾跟她接触的人接受检验，发现两人感染冠病，两人感染鼻病毒（rhinovirus），但并未发现有人染上肺鼠疫或其他疫病。研究所所在的伊尔库茨克（Irkutsk）州州长科布泽夫(Igor Kobzev) 则发表声明称，希皮洛娃是死于“未能确认病因的肺炎”，而研究所员工过后几天也未出现新的感染病例。鼠疫为细菌感染疾病，可通过几种方式传播，其中以肺鼠疫的传染力最大，也是唯一可经由呼吸道迅速人传人的病种。若迅速检测证实感染，可使用抗生素治疗，但如果不获治疗，则可在一至六天内致命。

经历收购风波的中国 AI 企业 Manus 完成超过 5 亿美元融资，创始人肖弘也已解除边控，让这家一度卷入中美科技博弈、前途未卜的公司迎来“重启”，也彰显了中国资本市场对 AI 的热情。 Manus 的母公司蝴蝶效应，星期四（10月8日）在公众号宣布融资消息。这是中国 AI 应用领域迄今规模最大的单轮融资之一，由中国私募股权基金博裕资本和老牌美元基金 IDG 领投，老股东腾讯、红杉中国、真格基金跟投。 公司投后估值达到 40 亿美元，也让 Manus 成为中国估值最高的 AI 智能体初创企业。 去年 12 月 Meta 斥资 20 亿美元收购 Manus，但交易被中国监管部门要求撤回。

众所周知，地中海饮食有利于健康长寿。现在研究人员报告另一种欧洲饮食——北欧饮食也与长寿相关。丹麦、芬兰、冰岛、挪威和瑞典等国的传统饮食与地中海饮食有很多相似之处，差不多是其寒冷版本，因此又名北方的地中海饮食。北欧饮食以植物为主，主要食用富含脂肪的鱼和根茎蔬菜。研究人员分析了于 64,000 名瑞典中老年人的健康数据，发现饮食习惯更符合北欧饮食的人的全因死亡率、心血管死亡率和癌症死亡率更低。

对一件早白垩世哺乳动物化石的分析证实约 1.25 亿年前的小型哺乳动物已经具备明确的半水生适应特征，这也是目前可确认的、最早具备游泳能力的哺乳动物。新发现的哺乳动物被命名为“板尾董尖齿兽”（Dongoconodon platycauda）。板尾董尖齿兽展现出了独特的半水生适应特征，是目前已知哺乳动物冠群中最早具有游泳能力的代表。其前后足具有发达的侧向扩展结构，与现生鸭嘴兽的蹼足高度相似，表明其生前可能具有发达的蹼膜；与此同时，董尖齿兽的掌骨、跖骨和指趾骨排列能够使手指和脚趾向外展开，从而增加划水时与水接触的面积，有利于游泳推进。 虽然板尾董尖齿兽的手足与鸭嘴兽十分相似，但它的尾巴却采用了完全不同的结构。化石保存了至少19节尾椎，其中靠近尾巴基部和中段的尾椎具有明显扁平的椎体和较宽的横突，表明这只动物具有背腹方向扁平的尾部。不过，它的尾巴并不像现代河狸和鸭嘴兽那样形成宽大的“桨状尾”，而是从基部向末端逐渐变细，更接近现代半水生啮齿类的海狸鼠。这只生活在恐龙时代的小型哺乳动物，可能拥有“鸭嘴兽式的手足”和“海狸鼠式的尾巴”，形成了一种此前未知的游泳方式组合。