OrangeBot.AI Digest — 2026-09-20

The play/pause key, finally aimed at what's playing

Digital family refrigerator door, now pocket-sized n synced

Bring one past deliverable. Mycel drafts every future one.

Five lanes of LA highway turned into an instrument

The Evernote alternative that puts its price in writing

Launch customer dashboards in minutes

Real Japanese slang, one giant word at a time

10x your revenue by finding your ideal customers

Keep your work when you switch AI tools

Give it a research task + the JSON shape you want back.

Launch your store. Let AI run the busywork.

Launchpad Replacement for MacOS 27 Golden Gate

SSH client with sessions that never drop.

GTM and AI visibility workflows for developer tools

Tell it what is on your mind. It writes the meditation.

The widespread adoption of long-horizon agents has made model workloads increasingly input-heavy. Although prior work has substantially reduced the cost of long-context computation, prefill remains computationally expensive, and large KV caches continue to strain HBM and SSD capacity and data-transfer bandwidth. Together, these compute, storage, and bandwidth demands constitute the primary bottleneck to further lowering deployment costs. To address this challenge, we introduce DeepSeek-V4.1-Flash, a multimodal Mixture-of-Experts (MoE) model with 552B backbone parameters and support for contexts of up to one million tokens. With its Causal Encoder-Decoder (CED) architecture, the model activates 16B parameters per token during decode but only 8B parameters during prefill, substantially improving cost efficiency for agentic workloads. To push the limits of KV cache compression, DeepSeek-V4.1-Flash combines cross-layer KV cache reuse in Compressed Sparse Attention 2 (CSA2) with FP4 KV caching. These designs reduce its global KV cache footprint (always in HBM) to 890 bytes per token, roughly 1/4 of the corresponding footprint of DeepSeek-V4-Flash. Further, through a dedicated deployment optimization known as SWA Bounded Replay, DeepSeek-V4.1-Flash reduces its persistent KV cache footprint (always on SSD or in host memory) to roughly 1/8 of that of DeepSeek-V4-Flash. Despite its much smaller KV cache footprint, the model delivers substantially better performance than the baseline. In addition, we streamline the DeepSeek-V4 architecture and introduce several efficient architectural extensions. We pretrain DeepSeek-V4.1-Flash on a multimodal corpus comprising 45T tokens and conduct comprehensive post-training, yielding strong performance across diverse text-based and multimodal agentic scenarios. Model checkpoints are available at https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash.

Recent Omni-Modal Generative Models (Omni-Models) have advanced content generation toward unified modeling of text, images, video, and audio. MiniMax-H3 exemplifies this transition by combining multimodal context understanding with joint audio-visual generation in a shared latent framework. Its unified architecture raises a fundamental question: Can multimodal alignment improve the model's world reasoning, and what new evaluation paradigms do omni-modal inputs enable? To investigate this question, this work introduces a comprehensive evaluation framework organized around four complementary dimensions of physical world reasoning. Unlike existing evaluation frameworks for video generation and world models, which are often constrained by limited input modalities and evaluation settings where prompts closely match the target video content, our evaluation is specifically designed to exploit the multimodal inputs of Omni-Model. We construct a diverse set of novel tasks that require models to integrate complementary information across modalities. Specifically, we consider four scenarios, including implicit prompts paired with multiple frames, audio-image, prefix-videos, and audio-video inputs. Every single modality provides only partial evidence about the underlying event, requiring the model to jointly reason over the complementary semantic cues to infer latent event states and future dynamics. Across 517 evaluation instances, MiniMax-H3 achieves an overall success rate of 41.97%. Video-based Decision Reasoning yields the highest success rate at 56.00%, while Audio-based Disambiguation Reasoning is the weakest, reaching only 27.40%. These results indicate that effective multimodal integration remains key to fully exploiting the benefits of diverse input modalities. The project is available at https://github.com/gulucaptain/MiniMax-H3-Reason.

We study length inflation in on-policy distillation (OPD), where student responses can become excessively long and even exhaust the generation budget. We identify termination-token mismatch between base students and post-trained teachers as an important source of this behavior. Across Qwen3, Llama, and Gemma, the two models can place their stopping probability on different EOS tokens, even when their declared stopping sets are identical. This mismatch can suppress the student's preferred termination action without reliably transferring the teacher-preferred alternative. We show that aligning the decoding stopping set alone is insufficient, while treating functionally equivalent EOS tokens as a shared semantic stopping action substantially mitigates mismatch-induced length inflation across all three model families. To further understand how termination behavior evolves over training, we study OPD across different K2-Horizon training stages. This stage-wise analysis shows that termination preferences can shift substantially during training, while also revealing a distinct length inflation late in the OPD run that persists beyond termination alignment. Together, these results identify termination mismatch as an important, but not exhaustive, source of OPD length dynamics. We release an implementation incorporating the proposed termination-handling corrections.

As coding agents move from supervised code completion to unattended, around-the-clock exploration, their work expands from isolated predictions into long trajectories of reasoning, tool use, and feedback. Token efficiency therefore becomes important for scaling recursive self-improvement. We take an RSI-inspired approach at the harness layer, scaling auto-research loops across increasingly numerous and diverse environments for harness rollouts. At this scale, the process yields reusable improvements that transfer beyond their development setting, moving automated harness discovery toward production-level outcomes. Four mechanisms survive selection and form SoL-Pi, spanning action execution, context compaction, observation handling, and delegated reading. On the 51-task EdgeBench evaluation, SoL-Pi achieves performance comparable to Pi across GPT-5.6 Sol and Opus 5 while reducing recorded token traffic by 44.7-49.0% and API cost by about one third. In other words, estimated hourly savings are \8.75-13.50 relative to native Codex and Claude Code harnesses, and \4.36-5.71 relative to Pi.

Coding harnesses shape how autonomous coding agents translate model capabilities into long-horizon software-engineering performance, yet existing work typically evaluates harnesses as monolithic systems, leaving the effectiveness of individual components unclear. To enable component-level comparisons, we study this question with a lightweight coding harness whose execution loop is fixed while three components are varied: planning, action space, and context management. Across four models evaluated on SWE-Bench Verified and Terminal-Bench 2.1, we evaluate 176 matched settings spanning five context-management strategies, four context-window budgets, and targeted ablations of planning and action space. We find that: (1) Context management becomes increasingly valuable as the context-window budget tightens, with most of its benefit coming from preventing context-overflow failures. (2) Staging rule-based elision before LLM-based summarization provides the strongest overall efficiency among the context-management strategies, whereas making elided content recoverable adds machinery that models rarely use and yields no accuracy gain. (3) Planning shifts from an accuracy scaffold for weaker models to a cost saver for stronger models, with little change in accuracy. (4) Predefined tools improve performance for models with weaker bash proficiency, whereas bash-capable models can operate effectively with a bash-only interface and achieve substantially lower cost, especially on command-line-centric tasks. Trajectory-level analysis explains these effects: context management extends execution trajectories without substantially altering agent behavior, planning changes where trajectories stop, and the action space changes the granularity at which code is written. These findings inform model- and budget-aware harness design and provide a modular framework for evaluating future harness components.

World modeling enables intelligence to anticipate consequences, guide interventions, and learn from interaction. Yet predictive models remain domain-specific: can a common learning principle support world modeling across radically different systems? We introduce JEPA-Anything, a domain-agnostic framework based on orthogonal predictive factorization (OPF). Extending joint-embedding predictive architectures, OPF decomposes latent targets into complementary factors, learns them through dedicated pathways, and recombines them within a shared predictive design. We evaluate JEPA-Anything across seven domains: vision, biology, clinical trajectories, control, molecular dynamics, physical fields, and weather. Experiments span representation learning, intervention prediction, out-of-distribution generalization, and long-horizon dynamics, including 10 matched dynamics tasks, forecasting of over 1,000 clinical events, and 100-step molecular rollouts across four systems. Against matched JEPA baselines, JEPA-Anything improves reported metrics on all 10 dynamics tasks and reduces single-intervention prediction error on Interventional Pong by 34.8%. It achieves the lowest one-step and 100-step molecular errors among compared methods in all four systems. Beyond prediction, a factor-nominated biological intervention receives experimental support in cell co-cultures, patient-derived organoids, tumor fragments, and mice; latent orbital modes recover the Keplerian scaling exponent with a fitted slope of -1.4991. These results support a common factorized predictive principle across heterogeneous worlds, connecting world modeling with intervention and experimentally grounded scientific discovery. Code: https://github.com/Gen-Verse/JEPA-Anything

LLM assistants are widely used for daily social advice, yet evaluating their social reasoning in such consultation settings remains challenging since (i) it requires setups where the assistant learns about social situations from subjective user narratives, and (ii) social properties, such as others' intentions, typically lack verifiable ground truth. To address these challenges, we introduce Fuse, a multi-agent simulation framework for studying user-mediated social reasoning. In Fuse, a target agent with a hidden motive interacts with other agents including one representing the user, who then consults the evaluated assistant to infer the target's motive, providing verifiable ground truth by construction. Simulation faithfulness is validated through a human study with 24k annotations. We apply Fuse to 12 LLMs and demonstrate its analytical utility by systematically isolating key factors, showing that (i) user mediation compounds the inherent difficulty of social reasoning; (ii) LLMs exhibit systematic sensitivity to biased user framing; (iii) models can require more details than humans need to reach a correct prediction; and (iv) longer conversations do not always improve performance despite providing opportunities for clarifying questions. We open-source Fuse and a dataset with 21k examples.

Information retrieval is increasingly important as LLM agents tackle complex tasks involving diverse information needs. Because retrieval relies on an index that represents each document through index keys, retrieval quality depends heavily on how effectively these keys expose the knowledge contained in each document. However, effective index representations vary across retrieval environments, making it difficult for any fixed optimization strategy to perform consistently. Yet evolving an index to its retrieval environment remains largely human-driven, requiring humans to diagnose retrieval failures, refine the optimization strategy, and reprocess the index accordingly. We propose SELF-INDEX, a framework that enables an index to self-evolve without human intervention. Its Optimizer autonomously diagnoses retrieval shortfalls, selectively revises the responsible index keys, and validates each revision before updating the index. Beyond reacting to observed retrieval demands, SELF-INDEX proactively explores additional demands through a Query Simulator, allowing the index to evolve beyond the queries already available for optimization. Across diverse corpora and retrievers, SELF-INDEX consistently improves retrieval performance while outperforming existing index optimization methods. We further show that these benefits extend to downstream applications, improving the effectiveness and efficiency of search agents and helping agent memory systems retrieve useful past interactions.

Platform abuse campaigns conceal redirection instructions with emojis, homophones, character decomposition, and redundant symbols, then route users through disguised links to services associated with pornography, fraud, gambling, or illicit transactions. Existing benchmarks evaluate obfuscated text and risky webpages separately, obscuring how target recovery affects downstream evidence acquisition. We introduce RiskChainBench, pairing 3,600 synthetic token-text restoration inputs from 600 source sessions with 600 corresponding human-labeled local web environments. A model first restores the message, operational intent, and destination; the same underlying model then acts as a VLM-driven web agent that investigates the correctly associated website and produces a frozen, evidence-cited risk report without message-side semantics or domain-reputation cues. We score restoration and correct-routing web investigation separately and compose them offline by applying the frozen primary-entry prediction as a gate to the same Task 2 result. Human labels determine task correctness, while a fixed multimodal evidence judge assesses faithfulness, sufficiency, completeness, and consistency. Across ten models, Entry Top-1 ranges from 35.2% to 95.2% and web decision accuracy from 26.3% to 62.8%; the leading systems differ across entry recovery, full reconstruction, website decisions, and fine-grained typing. Execution failures account for 31.9% of web runs, whereas post-decision type errors account for only 0.9%, identifying stable exploration and risk judgment as the principal bottlenecks. We release the benchmark, protocol, and resettable local sandbox.

Multi-turn agents trained with reinforcement learning (RL) receive a single scalar reward per trajectory, which motivates self on-policy distillation (OPD) to supply dense token-level supervision from a self-teacher with privileged task skills, letting a skill-free student internalize them. This recipe, however, is undermined by two findings in agentic tasks: privileged information alone does not always make a teacher reliable, and the benefit of teacher supervision is stage-dependent. We therefore propose RetireOPD (Self-Retiring On-Policy Distillation), which first optimizes a decoupled, skill-conditioned teacher with environment rewards and then trains a skill-free student jointly with RL and OPD. Rather than following a predefined distillation schedule, RetireOPD adopts Adaptive Retirement: the student drops the teacher on its own once their discrepancy stops shrinking and it reaches a target fraction of the teacher's success rate, after which training proceeds with RL alone. Across Qwen2.5 models from 1.5B to 7B, RetireOPD improves ALFWorld success rate over RL baseline by 14.1% to 18.8% and WebShop accuracy by 11.8% to 19.0%, and surpasses its own skill-conditioned teacher in every setting.

GUI agents execute long-horizon tasks on dynamic graphical user interfaces, where pop-ups, delayed loads, and relocated widgets routinely invalidate plans fixed before execution. Recent agent-skill frameworks encapsulate reusable procedural knowledge to mitigate this, yet existing skill designs are largely developed without targeting GUI execution dynamics and treat skills as static artifacts produced before deployment rather than living procedural knowledge that improves through it. We argue that what GUI agents need is not better static skills, but skills that can be revised from execution feedback at deployment time, without additional training. We propose EvoSkill-GUI, a training-free framework in which each skill is a structured multi-file package containing retrieval metadata, executable plans, backup localization, failure-recovery rules, accessibility utilities, and failure cases. EvoSkill-GUI operates through a \emph{reflect-revise-reuse} loop: the executor performs instant in-rollout revisions, an isolated critic diagnoses failed trajectories under strict information isolation, and the executor edits specific skill files through a restricted tool interface. Across MobileWorld, AndroidWorld, and OSWorld, three mainstream GUI benchmarks spanning mobile and desktop platforms, EvoSkill-GUI consistently improves multiple base models without any training, with maximum gains of +16.2%, +6.0%, and +10.5% respectively, and evolved skill libraries continue to benefit related tasks rather than being rebuilt from scratch. Our code is available at https://github.com/ZJU-REAL/EvoSkill-GUI.

Document parsing converts document images into structured content and requires reliable performance across diverse layouts and acquisition conditions. Yet training corpora are biased toward common document types and clean digital pages, while expanding coverage alone does not specify how to address a parser's remaining weaknesses. We present WeVisDoc, a two-stage data-centric framework for robust end-to-end document parsing. Stage I broadens semantic, structural, and appearance coverage through heterogeneous data and structure-preserving degradation synthesis. Stage II uses a held-out probe to measure the Stage I parser's residual errors within fixed visual-structural clusters. These diagnostics guide targeted data construction and reallocation of the target-token budget. WeVisDoc-4B achieves an Overall score of 95.38 on OmniDocBench v1.6 and a mean Overall score of 75.54 across the three PureDocBench tracks, ranking first among the compared end-to-end parsers in all four settings. Compared with Stage I, Stage II improves Overall scores for the 2B and 4B models on both benchmarks, with larger gains on the degraded PureDocBench tracks, including a 4.03-point gain for the 4B model on the Real Degraded track.

Video diffusion models repeatedly process long spatiotemporal token sequences during denoising, making attention a major computational bottleneck. Linear attention offers an appealing alternative and has been widely adopted in recent large language models, but directly applying it to video models often fails to preserve the fine-grained interactions required for high-quality generation. We present Video DeltaNet (VDN), which combines local Softmax attention with bidirectional linear memory for long-range video context. Its linear branch introduces Video Delta Attention (VDA), which updates memory once per frame by jointly incorporating its spatial tokens. Separate output projections and learnable gates calibrate the two branches, while a staged teacher-alignment recipe progressively introduces the new pathway into pretrained models. We instantiate VDN on MiniMax H3, applying the hybrid to video-to-video interactions while retaining Softmax for interactions involving text or audio. With eight-step distillation and an optimized SGLang serving stack, VDN-H3 completes DiT denoising for a 14.3-second, 768p video in 6.70 seconds on eight NVIDIA B200 GPUs, corresponding to a 14.5x speedup over the 50-step dense H3 baseline on the same GPU count.

Spatial intelligence requires more than describing object locations. Under incomplete observation, models must identify and acquire missing evidence, interpret it in a common spatial frame, and act on it. We introduce VA-Bench to evaluate the complete observe-reason-act-revise loop. General-purpose MLLMs learn procedural context from RGB-only demonstrations, actively select camera viewpoints, issue metric Cartesian commands, and revise them from execution feedback. Models receive no privileged object poses, oracle trajectories, or learned action heads. A fixed model-agnostic controller executes only model-specified targets. VA-Bench contains 14 base task families (11 single-arm and three dual-arm), seven held-out geometry/layout variants, and a long-horizon five-object composition track. We evaluate 12 primary model conditions in three independent runs over the same 20 physically verified seeds per base task, reporting terminal success, nine trajectory-level behavioral diagnostics, and subtask progress. First, the best-performing model scores 100.0% on target localization and 78.9% on spatial relations in the annotated run. Its three-run macro-average task success is only 53.93+/-3.17%. Second, active camera control significantly improves task success over passive multi-view observation. In one matched comparison, success rises from 27.86% to 57.50%. Third, held-out geometric transfer can reduce task success by over 30 percentage points. No model completes a strict long-horizon episode, despite substantial partial progress. VA-Bench thus tests whether general-purpose MLLMs can turn visual demonstrations and actively acquired evidence into successful embodied action.

Large Reasoning Models (LRMs) achieve strong performance on complex tasks but exhibit systematic inefficiency: they often overthink easy problems and underthink hard ones. Existing approaches based on uniform length penalties or rigid routing incur an efficiency tax, trading reduced computation on easy instances for accuracy loss on hard instances. We formulate efficient reasoning as an instance-adaptive computation allocation problem and propose When2Think, a post-training framework for hybrid reasoning that dynamically allocates computation based on problem difficulty. Our method introduces Instance-level Difficulty-Aware Control (IDAC), a reward-shaping mechanism that leverages pre-computed reference statistics (accuracy and token usage) to regulate reasoning depth. Combined with verifier-based rewards and batch-wise standardized advantages, IDAC enables stable critic-free optimization without learned reward models or online reference-model queries. When2Think encourages direct answering on easy instances while preserving extended reasoning on hard instances, thereby learning when to use System 1 (NoThink) versus System 2 (Think). Experiments on mathematical benchmarks demonstrate improved accuracy-efficiency trade-offs: on AIME24, Pass@3 increases by 10.0% while token usage is reduced by 27.9% relative to the base model, and on AIME25, When2Think achieves 40.0% Pass@3, outperforming compression and routing-only baselines.

Qwen : Alibaba releases Qwen-Image-2.1, a 7B open-weight model it says outperforms most closed-source models, with native transparency and up to ten reference images —  · QwenTeam丨Translations:.体中文  —  We are excited to open-source Qwen-Image-2.1 …

Truman Dickerson / Business Insider : Jensen Huang says AI leaders calling for regulation don't want new legislation, but to be “relieved of the laws we do have” because of “ulterior reasons” —  Don't let the black leather jacket fool you, Jensen Huang isn't a doomer.  —  Nvidia's CEO said as much in an interview with CBS News aired on Sunday.

Owen Walker / Financial Times : AI companies, including OpenAI and Anthropic, are putting pressure on office rents in Singapore as they embark on expansion in response to government overtures —  Anthropic and OpenAI taking more space in city's already squeezed prime property market  —  AI heavyweights including Anthropic …

Owen Tucker-Smith / Wall Street Journal : Indeed: data center roles in the US offer a ~42% pay premium for hourly maintenance and installation workers compared with similar jobs at other businesses —  Powering the nation's massive AI build-out requires lots of walking, checking for burning smells and monitoring for bird damage

Josipa Majic Predin / Forbes : Vercel, Cloudflare, and others quickly add Jev, as it makes AI tool selection much faster and cheaper; TypeSafe: Jev matches GPT-5.6 and Sonnet 5 workflow evals —  Most of what an AI agent asks a frontier model to do is not writing.  It is choosing.  Which tool to call next, whether to retry, whether a command is safe to run.

Mark Gurman / Bloomberg : A closer look at the upcoming Siri AI-powered home hub, a key pillar of Apple's strategy for the home; sources: Apple started cutting Fitness+ staff —  Also: What happened with the iPhone Duo Pencil and layoffs at Fitness+.  —  Apple's new “intelligent personal hub” strategy is about much more than the iPhone.

John Thornhill / Financial Times : An interview with Raspberry Pi CEO Eben Upton on the Raspberry Pi Foundation, his skepticism of AI evangelists, AI driving edge compute device demand, and more —  The British tech leader on coding as the modern form of literacy, his scepticism of AI evangelists — and why ‘the p(doom) is probably Russia’

Katie McQue / NPR : US policy analysts, lawmakers, and others say several state-level AI chatbot safety bills included language that could provide loopholes for tech companies —  Cynthia Montoya's daughter, Juliana, was 13 when she began talking to AI chatbots.  It was 2023, and the technology was only beginning to enter widespread use.

Sarah Neville / Financial Times : Clinicians raise concerns over medical AI adoption beyond diagnostics and imaging, citing limited clinical and performance data on its broader effectiveness —  The technology's advances have not yet translated into big improvements in real-life care.  Kayla Secrest was a newly fledged doctor beginning …

Ann Cao / South China Morning Post : Alibaba's Damo Academy open sources RADAR, a medical vision-language model it says can read CT scans and identify ~150 abdominal conditions, including cancers —  Tested on nearly 40,000 real-world exams, the model outperformed most radiologists, according to a new study published in Science

Reuters : Chinese DRAM chipmaker CXMT says its new fifth-gen technology platform enters mass production and unveils two 24Gb LPDDR5X phone memory products built on it —  Chinese DRAM chipmaker CXMT (688825.SS) said on Sunday its fifth-generation technology platform had entered mass production …

Wall Street Journal : Sources: Polymarket CEO Shayne Coplan dismissed concerns after fraudsters tried to launder $10M+ from stolen cards to clean accounts via its wagers in February —  CEO Shayne Coplan brushed off concerns about scheme involving stolen debit cards; now the prediction market is bolstering executive ranks as it eyes IPO

Daniel Thomas / Financial Times : Meta challenges Ofcom's OSA categorization of WhatsApp and Instagram imposing additional regulations; Roblox and Quora are also challenging their categorization —  US tech giant's latest action against regulator Ofcom could slow down implementation of the far-reaching law

Ken Bensinger / New York Times : California Gov. Gavin Newsom signs a law that will penalize influencers who do not properly disclose paid political content, with fines of up to $5K per post —  Under a new law, creators could face fines of $5,000 per post if they don't disclose that they were paid to make political content.

New York Times : Experts say AI kill-switch legislation is far harder to implement than lawmakers assume, warning a rogue AI could actively try to dismantle the mechanism itself —  A bipartisan group in Congress and Gov. Gavin Newsom of California have floated ideas for building a mechanism that would instantly power down an A.I. system.

英伟达在 2024 年称当年它的 GPU 产品共使用了逾 10 亿 RISC-V 核心。这些 RISC-V 核心没有被用于图形渲染，而是充当了微控制器，执行各类辅助任务，根据型号不同，每个 GPU 包含了 10-40 个 RISC-V 核心。英伟达在 RISC-V 之前使用的是私有微控制器 FAst Logic CONtroller（Falcon）。Falcon 最早于 2005 年随 G98 引入，到了 2016 年单个 GPU 芯片集成了超过 15 个不同的 Falcon 引擎，十年间使用的 Falcon 核心总数约 30 亿个，这些核心被用于视频编解码、电源管理、安全引擎等不同任务。Falcon 为 32 位核心，没有数据缓存，不再满足英伟达的需求，它开始寻找替代，在评估了 Arm、MIPS 等架构之后，它最终选择了开源指令集架构的 RISC-V。

根据美国俄勒冈州的一份报告，该州有 111 个运营中的数据中心，建筑面积 2210 万平方英尺，2025 年用电量占到该州总用电量的 23%。该州还有 32 个数据中心处于建设中或规划中，将新增 690 万平方英尺的建筑面积。报告预测，到 2030 年该州数据中心的能耗将上升至近 25TWh，约占该州总电力需求的 31%-32%，相当于 250 万户家庭的用电量。报告称，这些数据中心直接提供了 2630 个就业岗位，占到了该州总人口的 0.06%。

尼泊尔正面临从致命泥石流灾难后的艰巨重建任务。它寻求从气候赔偿基金 Fund for Responding to Loss and Damage 获得 2000 万美元的赔偿——即要求主要排放国承担责任，尼泊尔认为这场灾难是由气候变化引发的。尼泊尔外交部长 Shisir Khanal 表示，这不仅仅是寻求援助或慈善，更是一项关于正义以及“法律与道义责任”的诉求。他表示，该国的温室气体排放量几乎可忽略不计，但却要为一场并非由其引发的全球危机付出惨痛代价。相比数十亿美元的损失，2000 万美元的赔偿金虽然少，但却是检验气候赔偿基金公信力的一次重要考验。该基金设立于 2022 年的联合国气候大会期间，旨在应对气候变化对发展中国家的影响。迄今为止主要温室气体排放国承诺捐赠的资金中只有极少部分到位，而至少已有 119 个国家提出了应对气候问题的资金申请，它还没有发放任何资金。这些国家申请的总金额达到了 28 亿美元，而该基金实际到位资金仅 2.5 亿-3 亿美元，其中美国已于 2025 年撤出。根据最新数据，这次尼泊尔西藏泥石流灾害，尼泊尔至少 1410 人死亡，6145 人失踪；西藏有 43 人死亡，519 人失踪。

微软利用使用 GPT-5.6 Sol 和 Claude Opus 4.8 的 AI 智能体、历时 14.5 周，花费 12 万美元 token 将 Copilot 运行时从 TypeScript 语言移植到 Rust 语言。该项目采用逐个更新运行时模块的方式执行，共进行了 135 次发布，平均每天提交约 1.3 个 Pull Request，最终将 43 万行 TypeScript 代码转换为 80 万行可用于生产的 Rust 代码。测试显示，原 TypeScript 代码每秒能完成 7.55 个生命周期（one-turn session lifecycles），而 Rust 代码每秒 120 个——意味着在特定工作负载下速度提升了 15.9 倍。包含 10 个客户端的智能体在 TypeScript 语言下消耗了 1383 MB 内存，而 Rust 语言版本仅消耗了 126 MB。Rust 版本将任务保持在进程内执行，无需像 TypeScript 版本那样启动外部后台进程完成任务。

西班牙政府下令屏蔽 Archive.today 及相关镜像域名 Archive.is、Archive.ph 和 Archive.li。Archive.today 等网站被广泛用于绕过付费墙，此前由于植入了 DDoS 脚本（至今还在）而被维基百科屏蔽。但西班牙的屏蔽与此无关，而是由于有人投诉该存档网站上有违规内容。此举属于行政决定，而非法院裁决。

即将于下个月推出的 Ubuntu 26.10 将采用 Linux 7.3 Kernel，而不是原计划的 Linux 7.2。Linux 7.3 目前还是 RC3，预计最快于 10 月 18 日释出正式版，可能会延期一周到 10 月 25 日。而 Ubuntu 26.10 将于 10 月 1 日冻结内核，10 月 15 日释出正式版本，这意味着届时它使用的 7.3 Kernel 仍然会是 RC 状态。Ubuntu 项目此前已宣布它将会紧跟最新内核版本，而 26.10 是一个短期版本，因此 Ubuntu 团队将紧跟上游内核版本。

面向中低收入国家的人道主义与发展援助资金 Official development assistance(ODA)过去两年被大幅削减，但如果对亿万富翁征税，将可以填补资金缺口，有望拯救数百万人的生命。全世界资产逾 10 亿美元的亿万富翁超过 3000 人，他们的总资产达到创纪录的 20.1 万亿美元，比上一年增加了 4 万亿美元。如果对这些财富征收 3% 的税，有望到 2030 年将死亡人数减少 2950 万；征收 1% 的财富税则可减少 1510 万死亡人数。研究人员表示，在贫富差距日益悬殊且财富高度集中的当今世界，研究结果显示，旨在支持全球发展的替代性财富再分配政策，有潜力缓解 ODA 资金削减带来的负面影响。他们同时强调，结果是基于模型估算。少数国家已实施了财富税。西班牙于 202 2年开征团结财富税（solidarity wealth tax），对净资产逾 300 万欧元的个人征收 1.7%-3.5% 的税款。法国也曾讨论过征收 2% 的财富税，但相关提案遭到了参议院的否决。

一项罕见的观察显示，座头鲸母亲在幼鲸死亡后可能会悲痛欲绝。2025 年科学家在 Gold Coast 南部观察到一头雌性座头鲸在产下一头推测已死亡的幼鲸后，与另一头伴游鲸鱼在一起。她陪伴幼鲸长达数小时，甚至可能持续了数天。齿鲸曾被观察母亲会将死去的幼鲸托举到水面，而这只座头鲸母亲则在水下凝视着幼鲸，表现出长时间的哀悼行为。研究人员表示，“了解非人类动物如何应对死亡，能让我们深入洞察它们的情感世界、社会纽带以及认知能力。”

学习另一门语言是一项高度复杂的脑力活动。它要求人们记忆单词、分辨陌生的声音、识别语言规律、推敲语法规则，在恰当的时刻提取出正确的表达。而这一切，都在人们进行倾听、理解并准备回应的同时发生。这些脑力活动也能帮助保持衰老大脑的健康吗？研究表明，答案是肯定的。愈来愈多的研究表明，语言学习和双语能力能带来认知上的益处——对老年人而言同样如此。学习另一门语言还能为沟通和社会交往创造机会。神经可塑性让大脑在一生中都具备保持改变的能力。学习陌生的语音、词汇和语法结构需要大脑形成并巩固神经连接。语言学习能同时调动多种心理过程，要求人们吸收并提取信息、察觉问题并及时调整。这种持续的脑力付出，使语言学习成为一种极其丰富的认知锻炼形式。研究表明，双语者的灰质密度更高，白质完整性更强。这表明，语言经验不仅会体现在大脑的活动中，也会反映在大脑的结构上。还有研究发现，患上痴呆症的双语者更晚出现症状。这表明，双语能力可能对大脑衰老的影响有一定的保护作用，但它不一定能预防痴呆症或相关的大脑变化。

通过美国科罗拉多州 Douglas 县和 El Paso 县的 I-25 高速公路多年来一直阻碍着动物的季节性迁徙，这条公路是许多动物的死亡之路。如今北美最大的野生动物专用通道将动物碰撞事故减少 91%。这条通道由跨越 I-25 的天桥，五条野生动物地下通道以及引导动物前往通道的围栏构成。天桥本身长 209 英尺、宽 200 英尺，于去年 12 月完工，整个工程耗资 1500 万美元。科罗拉多州交通部环境经理 Chuck Attardo 指出，减少野生动物碰撞事故不仅能挽救人类和动物的生命，还能避免财产损失、节省公共服务开支并减少交通延误。

科学家几百年来一直将大脑视为单一的器官。但斯坦福大学医学院领导的一项新研究发现，大脑实际上由两个不同的器官构成，它们在数亿年里独立演化。这一发现推翻了大脑发育的主流模型。研究人员过去一直认为，在发育早期存在一个单一的祖细胞（progenitor cell），由此诞生了整个大脑。它暗示大脑的所有部分都有共同的发育起源。新的研究发现，人脑由两个古老的神经系统巧妙地结合而成——其中较原始的部分负责调节心跳、呼吸及其他生理功能；另一部分则赋予了人类特有的能力，使得人类能进行诗歌创作、数学运算以及思科自身的起源。研究人员通过观察发育中的小鼠胚胎识别出了两种不同的脑祖细胞：一种表达 Otx2 基因，发育成前脑和中脑；另一种表达 Gbx2 基因，发育成后脑。研究表明，两类细胞群从不重叠，在发育的最早阶段彼此就是互斥的。

Android 安全加固项目 GrapheneOS 披露，Google 向其旗舰手机 Pixel 推送了 Android 17 QPR1 更新，引入了新 API。而这些 API 没有提供给 Android 开源项目(AOSP)，这是自 Android Honeycomb (3.x) 以来从未有过的情况。新 API 为 Pixel 限时独占，其它 Android OEM 厂商无法获得。Google 现在只向 AOSP 项目提供年度版本和第二个季度更新版本 QPR2，第一个和第三个季度更新版本 QPR1 和 QPR3 都是 Pixel 独占。新 API 将通过 2026 年 12 月释出的 QPR2 提供给 AOSP 和 OEM 厂商。GrapheneOS 称，Pixel 通过固件移除了对安全特性 MTE 的支持，但硬件上仍然保留了 MTE。GrapheneOS 再次抱怨它对 Pixel 的支持变得日益困难，称与摩托罗拉的合作会更顺利。

由于海豚的脸部看起来总是在微笑，出于拟人化心理，人类会假设海豚是大海里的好孩子。但行为类似恶棍的海豚并非不存在。根据发表在《Ecology and Evolution》期刊上的一篇论文，研究人员报告了一只生活在澳大利亚 大堡礁南部 Lady Elliott 岛附近的雄性宽吻海豚 Bubbles。Bubbles 有着被称为盗食寄生（kleptoparasitism）的行为，它会专门追逐刚刚饱餐一顿的大眼鲹（bigeye trevally），强迫它们吐出刚吞下的食物，然后自己去吞下这些食物。盗食寄生最著名的动物可能是海鸥。研究人员在 2021 年 6 月-2025 年 10 月间记录了 20 次 Bubbles 的觅食行为，其中 11 次涉及盗食。Bubbles 首先会通过回声定位锁定一只目标鱼，然后紧跟这条鱼，对鱼群中的其它鱼视而不见，当目标鱼受惊而吐出食物时，它会停止追逐，吞食其吐出的食物。Bubbles 在半小时内可重复此过程多达 10 次，它主要针对大眼鲹，但不会去吃掉鱼。研究人员猜测这种觅食行为比海豚传统的猎物追逐，消耗的能量更少。Bubbles 可能是通过定位的回声差异区分哪些鱼刚刚饱餐一顿。

根据内部文件，微软应用科学总监 Brent Hecht 警告，抓取新闻内容训练 AI 是一次规模空前的盗窃，可能是人类历史上最大规模的劳动成果盗窃。他不认为抓取新闻内容训练 AI 是合理使用，认为这种做法完全是对合理使用理念的嘲弄。微软的另一份文件认为会形成“恶性循环”，会同时伤害模型和整个 Web。OpenAI 的产品如 ChatGPT 可能会危及新闻出版商的经济基础。OpenAI 和微软的数据都显示这一预测是正确的。微软的数据指出，部分新闻出版商网站的点击率下降了 83%-93%，其它新闻机构的点击率下降了 51%-94%。Hecht 承认，“几乎没有人希望自己创作的内容被以这种方式使用，也没有人因此获得报酬。”微软 CEO 纳德拉(Satya Nadella)在作证时表示，AI 公司不应该通过绕过付费墙违反新闻网站的使用条款。但 OpenAI 的内部信息显示，当一名员工通知总裁 Greg Brockman OpenAI 爬虫找到了绕过《纽约时报》付费墙的漏洞时，Brockman 表示好极了。

日本计算机娱乐协会（Computer Entertainment Supplier's Association ）的年度行业报告《Video Game Industry Report》显示，85.8% 的日本游戏开发者在工作中使用生成式 AI——63% 的开发者表示每天都在使用，22.8% 的开发者则是偶尔使用，去年这一比例是 51%。开发者使用生成式 AI 最主要是为了“改进运营效率与生产力”，其次是“缩短开发周期”以及“降低开发与运营成本”。多数受访者表示不会直接使用 AI 生成内容。报告称，AI 最常见用途是生成视觉素材和图像，其次是故事和文本生成，最后是编程辅助。32% 的游戏公司利用 AI 协助开发自研游戏引擎。Capcom 工程师称，AI 技术主要是用于减少“日常任务”，称最聪明的 AI 在感性上也无法与创作者相比拟。