OrangeBot.AI Digest — 2026-10-02

AI that helps you learn anything and everything.

An AI tutor on an infinite whiteboard, not a chat thread

Open-source dotted maps and 3D globes for designers

Beautifully native iPhone app for Fastmail

Customer email your agent can operate.

A voice assistant in your Mac's notch that does the task

Notion-style block editor for React. MIT, UI you own

Sync branches between GitHub repos, right from Chrome

The open-source video editor your AI agent can drive

Use any AI with Claude Code, switch accounts mid-task

Open Source Software Factory

Your clipboard but better

Searchable memory of your screen and meetings, on your Mac

A design studio for your email signature

The AI agent that writes your LinkedIn posts in your voice

Streaming video LLMs must retain evidence before its relevance to future tasks is known and respond when sufficient evidence becomes available. The challenge is to form reusable factual memory without compromising real-time perception. We introduce OneStreamer, which jointly learns query-independent evidence recording and task response through a shared proactive generation process. Its Proactive Hierarchical Caption Memory (PHCM) produces time-grounded local-detail captions and summaries of completed events. Streaming caption targets supervise the interpretation of observed video prefixes during training. At inference, model-generated records complement a recent visual window, providing reusable factual context without revisiting historical visual features. Proactive State Transition Learning (PSTL) reduces the dominance of repeated waiting states by preserving supervision at all output anchors and selecting representative state-change and state-persistence tokens. We further develop a streaming data synthesis pipeline that aligns output content and timing with available evidence. Combining the resulting streaming captions and QA with cleaned open-source data yields OneStreamer-1M, a broad-coverage streaming video interaction dataset with over one million records spanning diverse tasks. Our 4B model achieves the best results among the compared methods across all eight evaluated streaming video understanding benchmarks. Ablations show that retaining generated captions improves historical QA without degrading real-time perception. PSTL also outperforms dense state supervision while supervising only 27.5% of annotated state tokens. Together, these results support proactive generation as a shared learning interface connecting perception, memory formation, and timely response in streaming video interaction.

On-policy learning has been argued to reduce catastrophic forgetting, produce sparser parameter updates, and improve generalisation. However, existing comparisons between supervised fine-tuning and reinforcement learning vary many factors simultaneously, making the contribution of rollout policy difficult to isolate. We study the effect of rollout policy in a controlled strong-to-weak distillation setting, by independently varying rollout policy, token-level KL direction, and learning rate across the Llama3 and Qwen2.5 model families and reasoning tasks spanning scientific, medical, and arithmetic domains. Our analysis reveals a nuanced picture of distillation dynamics in which rollout policy does not necessarily play a central role. Instead, token-level KL direction more clearly shapes task performance and output coverage, while learning rate governs forgetting and update sparsity. Analysis of KL gradients and experiments along a continuous student-teacher rollout-policy spectrum explain this pattern: forward KL is remarkably robust to rollout policy, with its performance stable and strong despite changes to the rollout policy, whereas reverse KL is substantially more sensitive and favours student-generated rollouts. On-policy data nevertheless improves generalisation to harder variants of the Countdown arithmetic task under both KL directions, although this advantage does not reliably persist after subsequent RLVR. Our broader conclusions remain robust to removing gradient clipping, using sampled KL estimators, and training on tasks requiring longer reasoning chains. Overall, our results challenge the view that on-policy rollouts are inherently preferable and show that their value depends critically on the objective, evaluation setting, and optimisation hyperparameters.

AI reviewers can assign different judgments to manuscripts that report the same science in different wording, potentially rewarding rhetorical optimization over scientific improvement. We formulate Rhetorical Robustness as the joint requirement of stability across content-preserving rewrites and discrimination across papers. We introduce RobustReview, a controlled full-manuscript benchmark with 1,260 manuscript versions, and evaluate 30 reviewer configurations. The benchmark reveals false robustness, where low rewrite sensitivity coincides with score collapse across papers, and shows that human alignment and rhetorical robustness rank reviewers differently. Moreover, the evaluated content-focused prompting protocol does not consistently improve robustness across backbones. Motivated by these findings, we introduce SciCore, a dual-branch reviewer that averages a full-manuscript judgment with a judgment based on an extracted, structured science core. This design combines manuscript-level assessment with a content-normalized view intended to reduce rhetorical sensitivity. In our primary GPT-5.5 comparison, SciCore achieves a leading joint stability-discrimination profile among the benchmarked reviewers while maintaining competitive human alignment. These results identify rhetorical robustness as a distinct evaluation target and demonstrate the potential of science-core review to improve it.

Masked diffusion models (MDMs) generate sequences by progressively unmasking several tokens per denoising step, but their reverse process is typically factorized over positions, limiting sample quality in the few-step regime where diffusion's speed advantage over autoregressive decoding matters most. A recent line of work introduces a continuous Gaussian latent, trained as a variational autoencoder, to capture correlations across positions, but such approaches are prone to posterior collapse, where the latent is silently ignored. We propose Enhanced Mixture-of-Experts (E-MoE), which builds the reverse process as a mixture of factorized distributions over a discrete shared latent given by the expert-routing decisions of a Mixture-of-Experts (MoE) backbone, without increasing active parameters over the factorized baseline. Across synthetic multi-modal benchmarks, binarized MNIST, and LM1B, E-MoE improves few-step generation over factorized baselines.

Multi-reward reinforcement learning trains large language models to satisfy multiple behavioral objectives simultaneously. Reward-wise normalization, as used in GDPO, preserves reward-specific relative information within rollout groups, but different objectives can still exhibit uneven learning progress. We study this behavior through advantage energy, the sum of a reward's squared advantages over a batch. Under idealized GDPO normalization, we show that this energy is proportional to active-group density: the fraction of rollout groups in which the reward provides nonzero relative advantages. This reveals a residual batch-level signal imbalance and provides a basis for calibrating reward contributions. Based on this relation, we propose Density-Aware Reward Aggregation (DARA). We derive an inverse-square-root density correction that gives greater weight to signals from less frequently active rewards. DARA computes its weights from each rollout batch, adapting to changes in reward activity throughout training without modifying the underlying policy optimization objective. Experiments on tool calling and mathematical reasoning show that DARA learns the targeted behaviors faster than GDPO, reaching high format compliance in up to 26% fewer training steps on tool calling and near-saturated length compliance in up to 65% fewer steps on mathematical reasoning, while remaining competitive in final performance. Our code is available at https://github.com/zhaihaotian/DARA.

Decentralized multi-agent path finding (MAPF) with communication requires agents to reach individual goals without collisions under partial observability. Learnable policies trained on expert data provide an effective approach to this problem. However, when several coordinated joint actions are valid in the same context, independently sampling from per-agent distributions can recombine locally valid choices into incompatible joint actions. This failure can arise from the final sampling mechanism even when the per-agent action distributions are learned correctly. DMM (Decentralized Master-Mind) addresses this by replacing one-shot action sampling with discrete, iterative refinement of action intents across communication rounds, inspired by denoising in diffusion models. Agents initialize random action intents and refine them through local communication, coupling their choices before commitment. DMM is pretrained with imitation learning on expert MAPF solutions and further optimized with MICPO, a critic-free group-relative reinforcement-learning method designed for multi-agent, multi-round action refinement. DMM generally achieves higher success rates and lower solution costs than the evaluated learnable baselines. On 1,600 MovingAI tasks, DMM fine-tuned with MICPO solves 1,598, the highest coverage among the evaluated methods, while achieving solution costs close to those of the strongest baselines. DMM also scales to over one million simultaneously acting agents in obstacle-rich environments. These results show that round-level intent refinement can improve joint-action coordination while preserving decentralized execution.

We study test-time evolution for humanoid loco-manipulation: solving tasks that a controller was never trained for by repurposing its existing skills, improving from its own attempts, and retaining what it learns, without retraining. Our key insight is that a broad controller already holds much of the competence a new task needs, and that this competence becomes accessible through an interface between planning and control that is expressive enough to specify contact-rich, multi-stage interactions, yet executable and measurable enough that execution feedback can guide planning from experience. InterEvolve realizes this interface with two components. First, we develop an object-aware forward-backward (FB) behavioral foundation model, whose object residuals on a frozen body prior turn a new reward about the body or objects into loco-manipulation behavior at test time. Second, we specify tasks as reward programs: staged rewards with completion conditions and tunable constants. A large language model (LLM) agent revises the program structure in context, drawing on execution feedback and a skill library of verified programs, while a numerical optimizer tunes its constants. With every candidate verified across parallel simulation scenarios, the program explores new ways to induce, repurpose, and compose the controller's existing motor competence for the task at hand, and thus improves over iterations. Experiments show that human-designed rewards leave much of the FB model's loco-manipulation competence untapped, whereas the programs InterEvolve evolves release it, sometimes through novel strategies. It further produces behaviors for diverse tasks, complex scenes, and long-horizon compositions in simulation, and evolved skills run autonomously on a physical Unitree G1 from egocentric onboard perception.

Real-world embodied tasks, from everyday activities to professional procedures, require agents to act under physical constraints while tracking evolving object and task states. Tool use sits at the heart of such tasks, as many everyday and professional activities are tool-mediated. Understanding them requires reasoning about affordances, hand-tool-object geometry, procedural progress, and causal effects on target objects. Yet despite strong performance on perception-oriented video tasks such as captioning and general video QA, current multimodal video models remain limited in this form of tool-centric embodied reasoning. Progress in this direction has been limited by the lack of real-world egocentric data and diagnostic benchmarks. To address this gap, we introduce EgoTools, the first comprehensive suite for egocentric tool-use understanding. It consists of two complementary components: EgoTools-Data, a large-scale corpus of 100 hours of tool-centric egocentric recordings with synchronized audio, dense captions, reasoning-heavy narrations, and supplementary 3D information; and EgoTools-Bench, a diagnostic benchmark of 1,000 QA pairs across four tracks that cover tool-use understanding from perception and geometry to procedure and causal reasoning. Experimental results show that current models still struggle to ground tool use in visual evidence: Gemini-3.1-Pro achieves 66.9% overall accuracy but only 51.7% on Perception & Grounding. Beyond evaluation, we validate EgoTools-Data as a training resource. On the full 1,000-question benchmark, full supervised fine-tuning improves Qwen3-VL-8B-Instruct from 50.0% to 60.9%, under strict source-video separation. Together, these results establish EgoTools as a unified resource for both training and diagnostic evaluation of real-world egocentric tool-use understanding.

Vision language model (VLM) agents can control robots through visual feedback and action primitives, but repeated model invocations and redundant observations incur substantial token overhead. We introduce PyRUA-Lean, an interactive code-execution framework that couples feedback-driven primitive composition with selective observation: the agent composes classical robot primitives and learned vision-language-action (VLA) policies into Python cells that perform conditional checks and local retries, returning only explicitly requested images and state feedback for replanning. Across 700 simulated task instances from LIBERO-PRO, RoboTwin 2.0, and RoboCasa365, we compare PyRUA-Lean with a tool-calling baseline using the same GPT-6 Astra planner and underlying robot primitives. Under equal LLM-call budgets, PyRUA-Lean increases overall success from 63.1% to 71.7%. On instances solved by both agents, it uses 49% fewer LLM calls and 65% fewer input tokens.

Looped Transformers achieve parameter efficiency by repeatedly executing a shared block across recurrent loops. Each loop yields an intermediate representation decodable for the same next token, yet standard decoding discards earlier states. Because earlier loops embody less computation, recurrence inherently supplies aligned weak-and-strong prediction pairs without auxiliary models or external training. We introduce LoopCD, a training-free contrastive decoding framework that guides token selection by contrasting the final prediction with an earlier recurrent pass, operating either in logit space with one extra output pass (LoopCD-Logits) or in hidden-state space with zero output overhead (LoopCD-Hidden). Across four looped Transformer families, LoopCD delivers substantial, consistent gains at full recurrent depth: LoopCD-Logits raises Ouro-2.6B-Thinking's AIME 2024 pass@1 from 61.88% to 73.33%, while LoopCD-Hidden lifts Huginn's HumanEval pass@1 from 22.56% to 31.71%. Crucially, these performance gains enable halving the number of recurrent loops while still matching or exceeding full-depth unguided baselines, reducing forward FLOPs by 22.5% to 48.2%. By transforming intermediate recurrent states into effective guidance signals, LoopCD achieves superior decoding quality while substantially reducing inference compute.

We present Multimodal Flow, a fully continuous generative model of language and vision. Most unified multimodal models either model both language and quantized images as discrete tokens or combine discrete language prediction with continuous image generation. The former introduces a visual quantization bottleneck. The latter requires modality-dependent objectives and sampling procedures. Fully continuous modeling avoids these trade-offs and enables a shared generative process, but remains underexplored for multimodal pretraining. Multimodal Flow introduces a unified continuous architecture that integrates multimodal continuous representations with a shared chunk-causal flow backbone. It organizes text blocks and images as ordered continuous hyperchunks, preserving textual token order and visual spatial structure. The backbone learns a single vector field over these hyperchunks through Flow Matching. Joint attention enables cross-modal interaction, while modality-specific feed-forward networks process each modality. The model predicts multiple target chunks in parallel during training and generates hyperchunks sequentially at inference. We instantiate MF-1 and pretrain it on multimodal data. Across 0.6B, 1.2B, and 1.6B scales, continued pretraining consistently improves multimodal modeling. With only 150B pretraining tokens, MF-1 achieves an average score of 82.8 across GenEval and DPG-Bench and 75.3 across VQAv2, MMBench, and POPE, remaining competitive with unified models trained on substantially more data. Under matched data, optimization, and parameter budgets, Multimodal Flow further outperforms representative hybrid and discrete models. These results establish continuous chunk-based embedding flow modeling as a new fully continuous paradigm for unified multimodal modeling. The related code and model are publicly released at https://github.com/hustvl/Multimodal-Flow.

High-resolution 3D generation increasingly relies on voxel latents and multi-stage pipelines that first predict active structure and then synthesize local geometry. While effective, this design fragments continuous surfaces into many local tokens, inflates generation cost, and often weakens topological consistency for thin or highly connected shapes. We introduce SILSA, a topology-aware 3D generation framework that represents shapes with compact sliding-window slice latents. Instead of generating expensive voxel tokens, SILSA uses a fixed set of overlapping slices along the three canonical axes, where each token summarizes a local depth window to preserve cross-sectional continuity and support single-stage rectified-flow generation. A Slice VAE encodes oriented surface samples into multi-axis slice latents and reconstructs them with a sparse volumetric decoder, while a Volumetric Anchor Lattice coordinates directional slice streams through a shared 3D workspace. To preserve structural correctness, we introduce slice-level topology supervision that matches persistence diagrams and aligns Betti transitions across neighboring slices. Experiments show that SILSA improves structural fidelity while substantially reducing generation cost. SILSA improves PSNR by 8.7%, coverage by 5.96 absolute points, and Betti error by 9.2% over the strongest baseline, while using 70.0% fewer tokens than the next-most compact baseline and over 98% fewer tokens than sparse or hierarchical tokenizers, effectively reducing training memory by 40.4% and inference time by 58.5%. Qualitative results further show improved preservation of thin structures, repeated components, and long-range connectivity.

Real-world enterprise data science and analytics workflows require reasoning across dozens of tables, performing statistical analyses, and acting on the results. Established text-to-SQL benchmarks evaluate query generation alone, and audits have found their answer keys frequently wrong. Because real enterprise warehouses are too sensitive to release, these benchmarks are built on public datasets where a business event fits in a single table. We introduce Argo-Bench, an evaluation framework comprising 210 data science and analytics tasks. Drawing on public data, peer-reviewed industry literature, and regulatory filings, we simulate a food delivery platform in New York City at true scale, with 81 million orders in 2024, grounded economics, fraud patterns, and marketplace incentives. We export this world to an ERP warehouse of 235 tables and 7.5 billion rows, modeled on the Oracle E-Business Suite schema. The simulator's ground-truth state is withheld from the warehouse the agent sees, so tasks require reconstructing facts by navigating the warehouse before acting on them. Argo-Bench goes beyond text-to-SQL: the agent files actions such as banning fraudulent accounts, allocating courier incentive budgets, or issuing back pay, and the grader scores each by its consequences in the simulator. Every task has an executable reference solution that demonstrates solvability using only the warehouse. The strongest of 14 frontier and open-weight models scores 95 or higher on only 34.8% of tasks and averages 59.5 points. We hope Argo-Bench drives progress toward agents that understand, navigate, and act within real data environments.

Group Relative Policy Optimization (GRPO) is widely used to train reasoning language models, where it computes advantages by centering and normalizing rewards across rollouts of the same prompt. For multiple rewards, GRPO sums the reward components and normalizes the total reward by its within-group standard deviation. The corresponding variance equals the sum of all pairwise reward covariances. For a fixed centered reward, larger aggregate covariance produces smaller advantages, and vice versa, allowing update magnitudes to adapt to reward dependence. However, correlated rewards with large scales can dominate this normalization and suppress signals from smaller-scale rewards. We propose Correlation-Normalized GRPO (CorrGRPO), which normalizes pairwise covariances into Pearson correlation coefficients. CorrGRPO keeps the centered total reward unchanged while balancing the influence of differently scaled rewards on the correlation-based normalization. This allows advantage magnitudes to adapt to reward correlations without the normalization being dominated by large-scale reward components. We compare CorrGRPO with GRPO and other variants on code generation, tool calling, and agent security, using models ranging from 0.5B to 8B parameters. These tasks all involve multiple rewards that can improve together or present tradeoffs. Results show improvements across three domains, including code generation, tool calling, and agent security. Our code is available at https://github.com/HKUST-KnowComp/CorrGRPO.

Vision-Language Models (VLMs) have shown strong multimodal reasoning capabilities, yet whether they truly capture the physical consistency underlying real-world dynamics remains unclear. Existing benchmark paradigms often suffer from fragmented evaluation, focusing on isolated cognitive stages while overlooking the inherent synergy between perception, reasoning, and physical judgment. The lack of a holistic perspective limits the ability to diagnose whether VLMs can reliably evaluate the physical authenticity of emerging generative models. To address these issues, we introduce PhysVista, a benchmark designed to evaluate physical intelligence in VLMs through a closed cognitive loop framework inspired by the human seeing-reasoning-assessment process. PhysVista restores this loop by jointly evaluating physical state perception, physical dynamics reasoning, and physical plausibility assessment. It further distinguishes event-level reasoning and scale-level reasoning to enable fine-grained analysis of physical understanding. In addition, PhysVista incorporates both real-world and AI-generated videos, allowing evaluation across diverse domains and emerging generative scenarios. Extensive experiments across a diverse set of VLMs reveal substantial limitations in physical reasoning and plausibility assessment, highlighting a persistent gap between visual recognition and genuine physical understanding, and pointing toward more principled designs for physically grounded multimodal intelligence.

Chance Miller / 9to5Mac : Apple releases an update for iPhone 18 Pro Max devices on AT&T to address cellular failures; units that already lost service require hardware replacement —  Apple has confirmed an issue causing a “small number” of iPhone 18 Pro Max devices on AT&T to lose cellular service.

Karissa Bell / Engadget : Meta announces Muse Gadgets, providing an open source ESP32 microchip firmware and a Linux SDK to let users bring Muse to their hardware, like Raspberry Pi 5 —  The company created its own smart home device, the Muse Home Link  —  We still don't know too much about Meta's tamagotchi-like AI device …

Ashley Gold / Semafor : Meta says it is letting go of employees it hired from AI safety startup Virtue AI four months after they joined the company, citing clashing work styles —  Meta said it's letting go of employees it hired from AI safety startup Virtue AI four months after they joined the company in June …

Colin Demarest / Axios : Memo: the US Army is creating an autonomous systems command, after Defense Secretary Pete Hegseth announced the Meridian and Agincourt robotic warfare projects —  The U.S. Army is creating a new autonomous systems command and appointing an acquisition executive to prioritize buying …

Elias Schisgall / Wall Street Journal : TypeSafe CEO Diogo Almeida says Jev is in use by ~25% of Fortune 500 companies and “we were at a trillion tokens per day about a week ago” —  The Jev model, which TypeSafe's CEO considers at the forefront of a ‘new class of AI,’ has set Silicon Valley circles abuzz

@tavus : Tavus unveils Griffin, the “first Human Interaction Model”, which it says passed the “video Turing test”, with 48% of users thinking it was human in live chats —  Introducing Griffin, the first model to pass the video Turing test. 48% of people who talked to it live thought it was a real human. Previous systems have had a pass rate <3%. It is #1 on NVIDIA's benchmark for full-duplex AI video. It's the first Human Interaction Model (HIM).

Sarah Perez / TechCrunch : Apple says it is adding additional controls around “Full Disk Access” on macOS as AI agents have increased “the risks associated with this level of access” —  Days after a journalist claimed that Meta's Muse app on Mac read their private messages — a claim that Meta disputed …

Anthropic : Anthropic launches the Claude Frontier Academy with a $100M commitment to train 10K “Frontier Deployed Engineers” by 2028, starting with Accenture, Bain, others —  The first-of-its-kind Academy trains Frontier Deployed Engineers using the same standard of skills as Anthropic's own engineers …

Grace Kay / The Information : Sources: SpaceX's AI unit held talks about leasing computing capacity to Microsoft, as it fully embraces the neocloud business despite Musk's initial reluctance —  SpaceX's AI unit held talks over the summer about leasing computing capacity to Microsoft, say people familiar with the matter …

Rya Jetha / Business Insider : Source: FieldAI, which is developing a “universal general-purpose brain” for robots, is raising $700M at a $10B valuation, up from $2B in 2025 —  - FieldAI is raising $700 million at a $10 billion valuation, up from $2 billion last year.  — FieldAI is building what it says is a …

Toby Sterling / Reuters : Hans Anders, one of the largest Dutch eyewear retail chains, suspends sales of Ray-Ban Meta Glasses in the Netherlands and Belgium amid growing privacy concerns —  One of the largest Dutch eyewear retail chains said on Friday it had suspended sales of Meta's (META.O) Ray-Ban smart glasses …

Laura Cress / BBC : Slovenia says 44,000 domain names with the .si suffix were registered in September, up from ~2,000 in August, as Trump renamed AI “super intelligence” —  Donald Trump's push to rename AI may be behind an “unprecedented” rise in registrations for domain names in Slovenia, according to its official national registry.

Bloomberg : Sources: Broadcom's Wall Street syndicate is amassing $60B in AI chip financing to help Anthropic and others access chips and other key AI infrastructure —  Broadcom Inc.'s Wall Street syndicate is starting to gather $60 billion of fresh AI chip financing to benefit Anthropic PBC and other companies …

Bloomberg : Sources: an EU investigation is set to find that Azure and AWS meet DMA thresholds, as the EU strengthens its regulatory scrutiny of the cloud sector —  Microsoft Corp.'s Azure and Amazon Web Services are set to be designated under the European Union's strict rulebook for Big Tech.

Jess Weatherbed / The Verge : Suno debuts Speech, which generates spoken voices from scripts or prompts, with optional background music, in public beta, as its music generator faces lawsuits —  The new Speech feature provides synthetic voiceovers embellished with AI background music.

当前一代游戏机的模拟器通常需要较长时间才能成熟，但 PS5 的模拟器仅仅几个月时间就让许多 PS5 游戏能在 PC 平台上可玩。SharpEmu 从 5 月的极早 Alpha 阶段到现在具备加载真实游戏 eboot.bin 文件、执行原生 CPU 指令以及部分处理 GPU 相关功能的能力。已有 10 款游戏被标记为可玩，其中包括简单 2D 游戏如 Tetris Forever，也有复杂 3D 大作如 Astro Bot 和 Demon’s Souls 重制版。另一款 PS5 模拟器 KytyPS5 也于上周发布了首个公开版本，它是 PS4 模拟器 Kyty 的扩展版，能启动 2D 游戏及部分 3D 游戏，包括使用虚幻引擎 4/5、Unity 以及自研引擎开发的作品。有 134 款游戏标记为可玩，但很多存在严重 bug。

一名玩家购买了一个二手 CPU Ryzen 7 5800X3D，结果发现无法启动 Riot 工作室旗下的多款游戏，每次启动游戏就被踢出，在联络了 Riot 的客服之后才知道该 CPU 被列入了封禁黑名单，因为其前任主人有作弊行为。Riot 工作室旗下的所有游戏都受到影响，其中包括了 Valorant、League of Legends、Teamfight Tactics、Legends of Runeterra、2XKO 等。Riot 使用了内核级反作弊系统 Vanguard，它深度嵌入在 Windows 内核模式中。暂时不清楚 Riot 是如何唯一标识 CPU 的，CPU 会向操作系统报告硬件 ID，但该硬件 ID 不是唯一标识符，只是告诉操作系统其 CPU 型号。它可能是通过 CPU 运行产生的独特指纹标记 CPU，每个 CPU 在运行时候都会有微小的差异。

为了提高生育率，新加坡试点推出了为公务员牵线搭桥的约会应用 FirstDate。新加坡的总和生育率已降至每名女性生育 0.87 个孩子，而十年前这一数字为 1.24。政府最近成立了一个专门研究生育率下降问题的工作组，预计该作组将在 2027 年初发表研究结果。FirstDate 面向 21-35 岁的单身人士，目前仅向公务员开放，申请截止日期为 10 月 5 日。用户无需浏览海量的个人资料，而是填写一份关于兴趣、习惯、价值观和偏好的问卷。FirstDate 会在每个周期内（即双方互相接受并预计见面所需的时间）为用户推送一个匹配对象。每项匹配结果都包含匹配度评分、对方的简介和一段个人留言。该服务使用了 Gale-Shapley 稳定婚姻算法。有公务员认为该应用的一大优势可能是被诈骗的可能性较低。

根据发表在 PLOS One 期刊上的一项研究，猫的数量与幸福感正相关，一个国家的猫越多，其幸福感通常越高。但主要通过猫传播的弓形虫感染率与幸福感负相关，弓形虫感染率较高的国家的幸福指数通常较低。两个关系似乎矛盾，但也可能与卫生、经济条件相关联。研究分析了 93 个国家和地区的数据，幸福指数平均值为 5.72，每万人拥有猫数量的平均值为 989 只，其中中国的幸福指数为 5.97，略高于平均水平；每万人拥有猫的数量为 376 只，显著低于平均值，中国的弓形虫感染率与平均水平相当。研究作者提出，猫对幸福感的影响可能与经济因素有关。也就是随着收入水平的提高，公共卫生条件会得到改善，虽然猫的数量增加，但感染风险不会有相应的增加。

由于索尼频繁更新 PS5 的固件，而大部分 PS5 越狱方法只针对特定固件版本的漏洞，因而这些越狱方法实用性相当有限。但情况在本周二发生了变化，名为 Relapse 的漏洞利用方法适用于最高固件版本 v13.6 的 PS5 游戏机，而 v13.6 是在今年 7 月释出的，意味着 PS5 只要不更新最新固件，就能成功越狱。Relapse 利用了 PS5 浏览器的一个已知的 WebKit 漏洞，提权获取内核的写入访问权限，安装 ELF 加载器去简化任意代码的运行。越狱后的 PS5 除了能备份游戏外还能运行模拟器以及 PS4 游戏的非官方 60 帧 MOD。

中国互联网络信息中心（CNNIC）发布了《生成式人工智能应用发展报告（2026）》，截至 2026 年上半年，我国生成式人工智能用户规模突破 7亿 人，普及率超 50%。76.0%的 用户表示自己会让生成式人工智能回答问题；使用生成式人工智能处理图片/视频、文本、工作总结/会议纪要/PPT的用户占比分别为 47.8%、37.6% 和 32.5%。数据显示，38.7%的网民近半年在网上购买过智能硬件设备。其中可穿戴设备和3C数码产品是我国网民接触智能硬件设备的首要入口。购买过智能可穿戴设备的网民比例为 20.2%；购买智能手机、平板电脑等3C数码产品的网民比例为18.2%。报告称，深度求索、月之暗面等本土企业先后发布多个万亿级参数开源大模型，全球主流大模型调用榜单上排名前六的模型全部来自中国团队。

新奥集团发表新闻稿，称其“玄龙-50U”装置实现氢硼聚变反应。新闻稿称：氢硼聚变具有无中子、燃料丰富易得、低成本等商业化优势，产物是氦（α粒子），但相对于氘氚聚变，反应温度及三乘积要求更高，反应条件更苛刻。本次新奥聚变团队通过高能中性束注入与射频波的协同，大幅提高了氢硼反应第一共振峰的非热平衡快质子份额，实现了大于 10^8/秒的氢硼聚变反应率，表明新奥氢硼聚变迈入燃烧等离子体相关实验阶段，是中国多路径聚变能发展的重大突破。来自全球多个国家科研院所与知名高校的十余位聚变权威专家就本次实验成果召开专题论证会，一致认为：本次实验实现了球形环装置中质子能谱及氢硼聚变反应产物α粒子的有效、可重复测量，探测方法可靠，可支撑氢硼聚变反应验证，是球形环氢硼聚变创新探索实践的里程碑突破，对全球磁约束氢硼反应的科学研究具有重要价值贡献。

过去几年极端气候频发，美国佛蒙特州每年都会因此发生十几次持续数小时的断电事故。当地电力公司 Green Mountain Power(GMP)记录到的 10 场最具有破坏性的飓风有 7 场发生在过去十年，造成了逾 2.25 亿美元的损失。为了应对气候变化导致的断电，该公司推出了分布式电池储能网络，向参与该网络的家庭出租两块电池，租期十年，每月费用为 55 美元。该州有超过 5,500 人参与了该家庭电池网络，半数家庭还安装了太阳能电池板。该项目目前提供约 110 MW 的电力，相当于一座中小型天然气发电厂的装机容量。美国其他州也有类似的电池储能网络。

文学家团队借助韦伯望远镜在距地球 500 光年的巨行星 HATS-6 b 大气中探测到水、甲烷、氨，同时发现这颗行星温度可能比标准推算温度低得多。这是透射光谱技术第二次在系外行星大气中检出水汽之外的氨信号。由于氨这类含氮分子在较冷的巨行星中本应比在炽热类木星行星中更为丰富，这一发现支持了一个判断：围绕 M 型矮星运行的行星可能在化学上构成独特群体。HATS-6 b 体积大致相当于木星，每 3 天绕一颗体积小的低温红矮星公转一周。按现有认识，小恒星周围的气体尘埃盘既缺乏足够物质，也缺乏足够时间积聚出木星、土星级别的行星。目前人类已在太阳系外发现 6000 多颗行星，多数与太阳系内行星毫无相似之处。团队表示，弄清它们由什么构成、怎样形成，是判断其他星系是否与太阳系有共同起源的前提。

微软宣布 WSL Containers GA，该工具为 Windows 11 开发者提供了一种通过 Windows Subsystem for Linux 构建、运行和部署 Linux 容器的内置方案。微软同时提供了容器管理工具 wslc.exe，GPU 支持、网络改进、健康检查、存储挂载、与 Microsoft Defender for Endpoint 和 Intune 的集成。微软还声称，当 Linux 环境访问存储在 Windows 中的文件时，性能最多可提升一倍。

日本人的身高在 1896～1996 年的 100 年间，男性增长约 14.6 厘米，女性增长了约 16 厘米。但 1990 年代之后升高增长日益乏力，相比下中韩平均身高则在快速增长。2019 年日本 19 岁男性平均身高约 172.1 厘米，女性约 158.5 厘米。韩国男性为 175.5 厘米，女性为 163.2 厘米，中国男性为 175.7 厘米，女性为 163.5 厘米。日本人口身高增长停滞有三种解释：其一是能量摄入减少，日本人均每日能量摄入量在 1995 年至 2023 年间缓慢下降并趋于停滞，中国和韩国的能量摄入则比日本高出约 760-800 千卡。其二是能量摄入问题可能对正在怀孕的女性产生影响，日本每 10 名新生儿中就有 1 人出生体重不足 2500 克，高于中韩。其三可能与婴幼儿时期的睡眠时长相关，日本婴儿（11.62 小时）显著低于中国（12.49小时）和韩国（11.9小时）。

日本总务省公布的 2025 年人口普查确定值显示，截至 2025 年 10 月 1 日，包含外国人在内的日本总人口为 122,972,528 人，较 2020 年的上次调查减少 3,173,571 人，降幅为 2.5%。自 2015 年的调查起连续3次负增长。65 岁以上人口在总人口中的占比（老龄化率）为 29.4%，创历史新高。较上次调查上升 0.9 个百分点。总人口中的日本人减少 3.5% 至 119,131,935 人，自 2010 年的调查起连续 4 次减少。居住在日本国内的外国人增加 39.8% 至 3,840,039 人，创历史新高。从国籍来看，越南和尼泊尔增幅显著。日本整体的家庭数量比上一次 2020 年调查增加了 149 万 4465 户，达到 5732 万 4619 户，创出有可比数据以来的新高。每户平均人数降至 2.10 人，降至最低水平。一人家庭达到 2368 万 9024 户，增加了 253 万 7982 户。其中增长最多的是 65 岁以上的老年人，占一人家庭总数的 34.4%。一人家庭中，男性有 24.6% 为 65 岁以上，女性有 44.7% 为 65 岁以上。一人家庭的比例达到 41.4%，首次超过4成。

世界正日益城市化，夜空的亮度每年都在增加10%。全球八成的人口生活在受光污染影响的夜空之下，真正的黑暗日益稀缺，对大部分人而言正变得遥不可及。享受黑暗的意义不止于观星。黑夜本身就是一个独特的生态环境。人造光会干扰依靠月光导航的动物：无论是误将停车场当成大海而迷途的幼龟，还是绕着灯泡团团转的飞蛾，都深受其害。萤火虫和青​​蛙需要黑暗环境完成求偶仪式；候鸟常因大城市的强光照射而偏离迁徙路线。研究表明，光污染正在破坏植物与授粉昆虫之间的关系，改变树木的开花时节，扰乱包括人类在内的所有生物的昼夜节律。无论是室内的灯光，还是夜间透过窗户射入的光线，研究都证实我们需要黑暗环境休息、恢复体力和保持健康。大多数光污染源自彻夜长明的路灯、建筑物、停车场和运动场，但来自天空本身的光污染威胁也日益增加。地球轨道上的卫星越来越多。

FBI 与荷兰合作逮捕了疑似 ShinyHunters 组织领导成员、24 岁的 Pepijn van der Stap。他是在 9 月 16 日左右被捕的，发生在 ShinyHunters 入侵 FBI 招聘网站 apply.fbijobs.gov 纂改网页窃取逾 2TB 雇员数据之前——有一种解释是 ShinyHunters 组织的领导者换人了，该组织在新领导人的管理下变得更激进，并试图将入侵 FBI 的行动嫁祸给被捕的 van der Stap。van der Stap 曾使用化名 Umbreon，其头像就是宝可梦 Umbreon。入侵 FBI 的 ShinyHunters 黑客在纂改网页时植入了宝可梦 Umbreon 的 ASCII 艺术图。接管 ShinyHunters 的据称是约旦的少年黑客 Rey。

据清华大学经济管理学院官方微信公号称，今年清华经管学院顾问委员会新增委员三人，新增接任委员五人。三位新增委员分别是黄仁勋、苏姿丰，以及瑞士百达集团高级管理合伙人百达铭（Marc Pictet）。五位新增接任委员分别是宝马集团董事长聂科维（Milan Nedeljković）、沃尔玛公司总裁兼首席执行官方威翰（John Furner）、可口可乐公司首席执行官柏瑞凯（Henrique Braun）、泛大西洋资本集团联席总裁、全球成长型股权投资负责人马丁·埃斯科瓦里（Martín Escobari），以及bp集团首席执行官梅格·奥尼尔（Meg O’Neill）。清华经管学院称，这五位接任委员所在公司的前任都曾是学院顾问委员会委员，因工作变动不再担任。清华经管学院现任顾问委员会主席为苹果董事会主席库克。马斯克（Elon Musk）、扎克伯格（Mark Zuckerberg）以及微软 CEO 纳德拉(Satya Nadella)都是委员。