OrangeBot.AI Digest — 2026-09-22

Local AI agent memory served via MCP

The deterministic memory layer for AI

Assign work to AI agents, like any teammate

Open-source email marketing with managed SMTP

Open omnimodal intelligence, trained in public

The laptop your Android phone has been waiting for

A standalone desktop editor for Markdown and MDX

Stop waiting for tokens, and start shipping

Clipboard that adapts to wherever you paste it

Your AI agent builds the app. You stay in control.

Video Courses on Anything

Shared memory for coding agents.

Dictation + meetings + read aloud in one app on-device

Push-to-talk dictation on the AssemblyAI Dictation API

AI agents that work as a team and make decisions together

An LLM agent's capability is largely magnified by its harness, namely the prompts, control flow, tooling, memory, and context management surrounding the frozen backbone model. Recent methods increasingly automate this process by iteratively proposing and selecting component-wise edits of an agent harness, practically establishing a form of recursive self-improvement (RSI) at the agent-system level. However, such recursive evolution may overfit by memorizing the training tasks, showing large in-distribution gains that shrink or even vanish on out-of-distribution benchmarks. We introduce Regularized Recursive Self-Improvement of Agent Harnesses (RRSI), which incorporates the principles of regularizations into harness self-improvement by constraining the evolution candidate proposal and selection. The proposer operates with a temporally annealed budget, limiting how many edits a candidate can bundle, and it encourages unexplored trajectories based on evolution history. The selector is equipped with a critic and a pruner: the critic screens benchmark-specific proposals, while the pruner, removes changes that are too small, too expensive, or no longer useful. Together these constraints favor reusable agent mechanisms over benchmark-specific ones or even noises. Across eight benchmarks spanning coding, agentic workspace and engineering design tasks, RRSI gains up to 14.1 points on the split it evolves against and up to 4.7 points on the five out-of-distribution benchmarks, while producing a harness that runs on 30% fewer policy tokens than the unregularized evolution. Code is available at https://github.com/google-research/rrsi and project page is https://regularized-rsi.com/.

Video world models enable interactive exploration of dynamic environments, yet struggle to respect prior observations over long horizons and across viewpoints. We present WorldCrafter, a video world model that learns a camera-queryable implicit 3D-aware memory for this purpose. The key insight is to let the requested viewpoint shape how multi-view evidence is compressed into the video generator's limited token budget. Trained jointly with the video generator, a memory encoder and pose-conditioned readout module integrate historical observations into a fixed set of target view-specific tokens before denoising, without explicit depth-based correspondences. By combining this memory with recent temporal context and few-step distillation, WorldCrafter enables streaming scene exploration from a single input image or text prompt. Experiments across static and dynamic scenes show substantial gains in long-horizon consistency and camera-control accuracy while preserving visual quality during minute-scale exploration.

Modern video games provide a measurable testbed for AI models, combining abilities of visual understanding, instruction decomposition, goal planning, and precise action control over multiple temporal horizons. Existing datasets and benchmarks, however, either cover a narrow range of games, lack language instructions, or rely on high-variance online rollouts. To address these challenges, we introduce GameHorizon, a unified data and evaluation suite that measures gameplay capabilities at different horizons for diverse model families. GameHorizon Suite consists of three components. First, GameHorizon-Annotator is a scalable and automated annotation pipeline for multi-horizon instructions. Second, utilizing the pipeline, we construct GameHorizon-Data, the first large-scale AAA gameplay dataset with temporally aligned videos, player actions, and multi-horizon instructions. It comprises 5,000 hours of recordings from 21 games, collected by 100 human expert players. Third, we build GameHorizon-Bench with reproducible offline and stepwise online testing. The offline track enables reproducible evaluation using thousands of standardized questions organized into three primary tasks and a series of diagnostic variants, while the online track tests whether offline scores reflect actual gameplay capabilities and localizes failures to specific steps within long-horizon gameplay. Based on our GameHorizon Suite, we evaluate 47 models through more than one million model invocations, revealing a meaningful hierarchy of task difficulty and pronounced differences in model capabilities. Our work can provide a standardized yardstick for evaluating gameplay capabilities across horizons and model families. We will release our dataset, annotator, and benchmark to facilitate future research.

Humans can seamlessly adapt to both physical and digital worlds, suggesting that while a digital-to-real gap exists in embodiment, environment and task, human intelligence itself may transfer across this gap. This naturally raises a fundamental question: can the intelligence of vision-language models (VLMs) similarly generalize from the digital world to the physical world for robotic control? We investigate this question through RoboDawn, a human-intuitive interface that exposes robotic control to an agentic VLM through a compact set of discrete translation, rotation, and gripper commands. Using this interface, the VLM controls a robot in a closed loop: it observes the current visual state, reasons about the next action, executes it, and adapts subsequent decisions to the resulting state. Furthermore, we introduce an in-context learning (ICL) scheme that uses a few demonstrations to ground the VLM in both interface usage and task-solving strategies. Experiments on RoboTwin 2.0 C2R and RoboDojo demonstrate that RoboDawn achieves strong performance without task-specific robot training. In the zero-shot setting, RoboDawn outperforms several strong policies trained on benchmarkspecific robot data, while a single in-context demonstration further yields substantial performance gains and establishes state-of-the-art (SOTA) results. On RoboTwin 2.0 C2R, the success rate increases from 53.2% zero-shot to 73.6% one-shot, exceeding the solid baseline π0.5 (46.0%). Similar gains are observed on RoboDojo, where success rate improves from 35.67% zero-shot to 47.17% one-shot. The same framework also transfers to real-world robots, performing block-in-basket and block stacking on Franka.

Educational foundation models must solve problems, understand curriculum structure, diagnose learner difficulties, and provide appropriate instructional support. Existing educational language models often focus on either problem solving or tutoring, with training mixtures organized by source or task rather than capability. We present OmniEdu, an open family of foundation models for K-12 learning and teaching. Its instruction-tuning corpus combines over 100 educational resources and general instruction sources, organized around four capabilities: subject competence, curriculum grounding, diagnostic reasoning, and pedagogical action and scaffolding. Our pipeline integrates deterministic cleaning, semantic auditing and rewriting, task-specific quality scoring, token-budgeted diversity selection, and pedagogical instruction assignment. It yields 69,999 examples and 15.96M supervised response tokens, including 60,951 education-specific examples. We fine-tune 4B, 9B, and 27B models and evaluate curriculum grounding, K-12 problem solving, and pedagogical tutoring, alongside general capability. Education-oriented tuning consistently improves all three educational benchmark groups across model scales. OmniEdu-27B achieves 63.12% EM and 76.69% F1 on K12-Bench, 85.89% on MathFish, 86.95% on EDUMATH, and 78.74% in MathTutorBench's Scaffold setting. It also achieves the highest Teaching average on LongTutor among the evaluated models, at 3.02. These results demonstrate the value of curated, capability-balanced supervision for adapting general language models to educational tasks spanning problem solving, curriculum understanding, and instructional support.

Retrieval-Augmented Generation (RAG) systems over enterprise knowledge bases must ingest heterogeneous document formats -- PDFs, Word documents, presentations, and scans -- whose content is locked inside complex visual layouts, multi-column pages, and dense tables. Rule-based extraction and OCR destroy reading order, flatten tables, and lose heading hierarchy, while fully agentic chunking over extracted text incurs high token costs and hallucination risk. We present Document Retrieval-Aware Chunking (D-RAC), an extension of our Web Retrieval-Aware Chunking (W-RAC) framework to arbitrary document formats. D-RAC first normalizes any input document into PDF, exploiting the fact that virtually every format has a faithful, deterministic PDF rendering. A single multimodal LLM pass then converts rendered pages into retrieval-optimized Markdown -- rewriting tables as self-contained prose statements and preserving heading hierarchy -- after which chunking proceeds exactly as in W-RAC: deterministic parsing into ID-addressable units followed by lightweight LLM-based chunk planning over identifiers rather than text. Source text is never regenerated during chunking, preserving W-RAC's cost, determinism, and observability benefits while unlocking every renderable format as a first-class input. On the 236-document, 795-page PDF subset of the RAG-Multi-Corpus benchmark spanning five enterprise domains, D-RAC converts and chunks the entire corpus in 72 minutes with zero errors, producing 1,748 retrieval-ready chunks. Compared to agentic chunking with frontier LLMs, D-RAC reduces chunking-stage output tokens by 95.7%, cutting chunking cost by 77.8% (GPT-4.1 pricing) to 85.6% (Gemini 2.5 Pro pricing) and chunking time by 75%. D-RAC scales linearly to documents of 500+ pages.

Recent advances in video generative models have enabled high-fidelity, temporally coherent video generation. However, these models often struggle to satisfy prompts requiring specialized knowledge, specific identities, physical consistency, or ordered events. In this paper, we present VideoGen-Agent, a multimodal agent trained through multitask agentic reinforcement learning to use external tools for video generation. The agent coordinates augmentation, generation, and verification tools through multi-turn interactions, using the prompt and intermediate observations to guide its decisions. We train a shared policy on a category-balanced dataset spanning six tasks. Supervised fine-tuning on teacher-generated trajectories establishes tool-use behavior, which is then refined through reinforcement learning. A category-aware hybrid reward evaluates tool-call validity, task-appropriate tool use, and generated video quality. We further introduce VABench, a held-out benchmark of 600 prompts covering procedural knowledge, single- and multi-entity identity preservation, physical consistency, scene composition, and multi-shot temporal structure. On VABench, VideoGen-Agent improves over its base text-to-video generator by 19.1 points, from 56.5 to 75.6. Upgrading the generation tools further raises the score to 86.1 without additional agent training. Human raters prefer the upgraded configuration over the strongest standalone baseline in 84.3% of comparisons. These results support learning tool use across video-generation tasks and show that the trained agent can benefit from subsequent advances in generation tools.

We present onPanda, an interactive tool for efficiently annotating LLM alignment data and agent trajectories. onPanda adopts token-level correction as its core interaction: while reading a model response, the annotator locates the first inappropriate token and either picks a substitute from the model's candidate tokens or types the correct text via free-form editing. The system then truncates everything after that position and continues generation from the corrected prefix, repeating this locate-correct-continue loop until a satisfactory response is obtained. This mechanism lets annotators precisely steer model outputs at low cost: a small controlled study suggests that onPanda reduces median annotation time by 52% over manual post-editing. Since the vast majority of tokens in the final response are generated by the model itself, the resulting data largely preserves the model's sampling distribution and is well suited for constructing on-policy SFT and preference data. Furthermore, the token-level corrections recorded during annotation provide fine-grained supervision with precise positions and naturally paired positive--negative samples. onPanda also connects to external tools and harnesses, enabling interactive trajectory annotation in realistic environments. In addition, we release Panda-CVL, a dataset annotated with onPanda, together with a benchmark for token-level correction.

Manipulation policies must know which objects matter and where they are, yet the pretrained backbones that current robot foundation models build on, from language in vision-language-action models (VLAs) to video generation in world-action models (WAMs), do not directly require this metric grounding, leaving it to be learned implicitly from robot demonstrations. We propose Grounded Action Models (GAMs), a new paradigm of robot foundation models built with 3D grounding. GAM can be conditioned using language, points, or box prompts, which are first transformed into a shared object-centric representation of the selected objects. This representation captures target-focused visual features and metric object geometry, which is mixed with robot state history through a multi-stream transformer to predict action chunks. Although GAMs can be run autonomously, they can also serve as a low-level controller that a high-level planner controls using its various input modalities, allowing for long-horizon and memory-dependent manipulation. On RoboTwin 2.0, GAM achieves an average success rate of 55.3% across 50 tasks (vs. 52.0% for Spatial Forcing), including 47.6% under scene randomization (vs. 30.4% for Abot-M0), with its action policy trained only on clean-scene demonstrations. On LIBERO-PRO, it achieves a state-of-the-art average success rate of 61% (vs. 53% for π_{0.5}) across 16 perturbation settings, with the largest gains when targets are relocated or newly designated. On two real robots, GAM retains 17/20 successes under visual shift on a bimanual YAM versus 4/20 for π_{0.5}, while its composition with a Molmo2 planner on a Franka achieves 64.7% ID and 49.8% OOD step completion on long-horizon and memory-dependent tasks.

Human video datasets offer an abundant and diverse source of interaction data that can complement expensive real-robot data. To bridge the human-to-robot embodiment gap, existing approaches either robotize videos in task-matched settings or address observation and action alignment separately at scale. In this work, we systematically examine whether robotized human videos can serve as an effective and scalable source of supervision for VLA pretraining. To this end, we develop a robotization pipeline that converts heterogeneous human videos into robot-aligned observations and action trajectories while inferring missing intermediate signals across annotation levels. Using this pipeline, we construct the HuRo dataset, comprising about 630K robotized episodes and 142M processed frames from five human-video sources. Across four real-world manipulation tasks, increasing the amount of robotized pretraining data improves overall completion from 51.5% to 80.3% and OOD completion under spatial and visual shifts from 34.9% to 72.2%. Ablations further show that visual robotization improves OOD robustness and that end-to-end pretraining with retargeted actions outperforms visual-only transfer. Project website: https://3587jjh.github.io/HuRo.

Repository-level software engineering (SWE) comprises heterogeneous task categories, whose progress under pooled agentic reinforcement learning can be uneven: gains in some categories coincide with regressions in others, while aggregate resolution obscures these changes. Motivated by this category see-saw, we develop a category-aware expert-training and policy-integration framework. Executable task construction and SWE Labeler, an evidence-grounded multi-axis labeling system, organize the training pools. Initial category-specific RL improves average training success while leaving uneven instance-level progress, motivating explicit consolidation of successful behavior and policy-adaptive task selection. Same-origin category experts alternate long-horizon Agentic-miniRL with Refresh-Repair-Expand (RRE): the updated policy refreshes instance mastery, reuses its own verified successful trajectories for Repair SFT, and reselects tasks for further RL. Label-routed multi-teacher on-policy distillation (MOPD) consolidates the experts into one deployable student, with ReLU-gated reward extrapolation keeping only each teacher's improving direction over the reference. Expert training and policy integration require no external model to provide solution trajectories or action targets. We evaluate Pooled RL and Balanced RL, expert development, and single-model integration through aggregate and per-category resolution, the minimum category lift over each joint-RL baseline, and expert-gain recovery. The final MOPD policy achieves mean resolution of 58.04% on Pro-618 and 59.00% on SWE-bench Multilingual, improving over the base model by 5.39 and 2.78 percentage points, respectively.

Natural interaction in digital and physical environments requires continuous perception and timely responses. Spoken dialogue relies on acoustic and linguistic cues, while video interaction also requires grounding the conversation in evolving visual context. We present Realtime-Venus, a proactive full-duplex interaction system with two separately trained 9B models: Realtime-Venus-Omni for audio-visual interaction and Realtime-Venus-Audio for spoken interaction. Each model serves as a complete conversational frontend, integrating continuous perception, conversational control, and native speech generation through a shared causal timeline for user inputs, model outputs, and delegation events. A dual-loop runtime coordinates live interaction with background reasoning and tool execution. Foreground interaction continues while Realtime-Venus-Harness executes tasks asynchronously and returns results for integration into the ongoing dialogue. Both models follow a common post-training recipe combining offline understanding, proactive full-duplex trajectories, and delegation workflows. Among the evaluated online models, Realtime-Venus-Omni achieves the highest scores on six of eight video benchmarks, including StreamingBench (70.2%), OVO-Bench (64.7%), and Daily-Omni (81.3%). Across eight audio understanding and spoken question answering benchmarks, Realtime-Venus-Audio leads the compared models on MMAU (78.0%), MMAU-Pro (63.2%), Llama Questions (83.8%), and Speech CMMLU (67.8%), while matching the best VoiceBench AlpacaEval score of 4.81. On Full-Duplex-Bench v1.5, Realtime-Venus-Audio responds to 75% of user interruptions and achieves continuation rates of 97%, 88%, and 86% under backchannels, other-directed speech, and background speech, respectively, exceeding Gemini 3.1 Live and GPT-4o on all three continuation metrics.

Existing approaches to persona simulation with Large Language Models (LLMs) mostly rely on shallow character descriptions that fail to sustain coherent character behavior across extended interactions. We introduce Deep Persona, a psychologically grounded, three-layered architecture that organizes personas into hierarchical levels of observable expression, latent beliefs, and core motivational drives, for constructing highly convincing role-playing agents. Governed by the principles of scripted determinism and bounded agency, the architecture restricts the model to a reactive engine guided by a structured internal script. We further propose a reference-free evaluation framework that benchmarks dialogue naturalness against empirical human distributions using established psychological clinical instruments and adversarial stress-tests. Empirical evaluation reveals that while LLMs achieve high pragmatic fluency, they exhibit systematic limitations in emotional expression and joint attention. In addition, we present a case study of two Deep Personas and evaluate them using the proposed framework, demonstrating that structured personas can produce interactions that more closely align with human conversational behavior.

Agent harnesses, the external systems that mediate model-environment interaction, can substantially improve agent performance, but their gains remain tied to the harness at deployment. Because the best harness varies across domains, instances, and models, a general-purpose agent must either settle for a suboptimal shared harness or route among an ever-growing set of specialized ones. We therefore study agent harness distillation: using a domain- or instance-optimized harness as training-time guidance and transferring the behaviors it induces into model weights, so that its gains survive under a single fixed target harness. The challenge is that the two harnesses differ in action space and available information, so guidance from the optimized harness cannot serve directly as supervision for the target one. We introduce Harness-Zero, which enables harness distillation through agent-as-harness. Guided by the optimized harness, a harnessing agent corrects student responses before execution in the target harness's action space, turning harness guidance into training demonstrations. Fine-tuning on the resulting trajectories internalizes harness-induced behavior into the model, so the specialized harness can be removed at deployment. Our experiments spanning knowledge work, tool use, and science domains show that: (1) For frontier LLMs using the same evolved harness, agent-as-harness outperforms code-as-harness. (2) With the specialized harness removed at deployment, Harness-Zero improves the base model's macro-average task success from 23.3% to 44.3%, even exceeding the 41.7% it reaches with that harness still attached. (3) Harness-Zero recovers harness-induced behaviors absent from the base model, with 82.3% average recovery across 28 patterns in the three domains.

Vision-Language-Action (VLA) policies achieve strong performance in robotic manipulation but remain brittle once execution deviates from nominal trajectories. We propose CARE (Corrective Atomic Robotic Execution), a framework that improves recovery by learning from failures encountered during execution. Instead of generating corrective data from manually designed or random perturbations, CARE collects failed rollouts, models stage-conditioned post-failure deviations, and uses the resulting empirical distributions to synthesize representative failure states and corrective demonstrations. At inference time, CARE combines stage-wise planning with physically grounded 3D monitoring to trigger atomic adjustments or re-operations while preserving task progress. We further introduce the Failure State Recovery Benchmark (FSR-Bench), which evaluates recovery from intermediate failure states under local deviations and structural anomalies. Experiments across multiple VLA backbones, simulation benchmarks, and real-world dual-arm tasks show consistent improvements, with average task-success gains of 14.5 points in simulation and 15.9 points in the real world. Code, models, and data are available at https://github.com/xiaojunlan/care

Maria Deutscher / SiliconANGLE : Firecrawl, which provides web scraping tools for AI agents, raised a $75M Series B led by Smash Capital —  Firecrawl Inc., a provider of web scraping tools for artificial intelligence agents, today announced that it has raised $75 million in funding.  —  Los Angeles-based venture capital firm Smash Ventures led the Series B round.

Julian Chokkattu / Wired : Rabbit launches OS3, a cloud AI agent that connects to local apps and files on Windows, Mac, and Linux and works via web, Telegram, iMessage, or its R1 device —  Two years after trying to sidestep mobile apps with dedicated AI hardware, Rabbit is launching OS3, a cross-platform agent that lives on the screens you already use.

James Rundle / Wall Street Journal : Data security startup Cyera raised $400M from Goldman Sachs in a Series G extension, bringing its total funding since June 2025 to $1.94B —  Data-security startup raises more than $1.9 billion in 15 months as Wall Street AI anxiety leads to an investor surge  —  Wall Street is getting nervous about artificial intelligence.

Stephen Nellis / Reuters : Microsoft says it will give DARPA on-site access to its latest quantum computing hardware, including the Majorana 2 chip, at its new Maryland research center —  Microsoft (MSFT.O) on Tuesday said it will give the US Defense Advanced Research Projects Agency direct, on-site access …

Semafor : Sources: Scott Bessent is emerging as a frontrunner for the AI czar position; other names include WH OSTP Director Michael Kratsios and OPM Director Scott Kupor —  THE SCOOP  —  Treasury Secretary Scott Bessent is emerging as a frontrunner for President Donald Trump's new “AI czar” …

Amanda Silberling / TechCrunch : TikTok US joins child safety initiative Lantern, which says members like Meta and X shared 2M+ signals of potentially violating behavior from 2023 through 2025 —  Eight months after it was spun out of TikTok, the social media platform's U.S. unit, known as TikTok USDS Joint Venture …

Axel Metz / TechRadar : Qualcomm unveils the Snapdragon 8 Elite Gen 6 and 8 Elite Extreme Gen 6, both on TSMC's 2nm process, with two 5GHz prime cores and six 4GHz performance cores —  Extreme by name, extreme by nature  —  Qualcomm has announced not one, but two new flagship mobile chipsets at its annual Snapdragon Summit …

Heather Knight / New York Times : San Francisco sues Trump Media & Technology, claiming its Truth API service violates California's Unfair Competition Law by creating an insider trading market —  The city attorney argued that the service, available for up to $100,000 a month, could allow wealthy traders to manipulate the market …

Sarah Perez / TechCrunch : Nat Friedman says Meta built Muse “from scratch” but was “heavily inspired” by OpenClaw, as users compare the two, including calling Muse “OpenClaw for normies” —  Early adopters of Meta's Muse have been speculating that the reason the AI works …

Stephanie Palazzolo / The Information : New York-based Ande, whose AI agents help businesses arrange corporate events for employees or customers, emerges from stealth with $52M in seed and Series A —  Personal AI agents like Instinct and Muse have promised to take away the hassle of booking hard-to-get restaurant reservations, Broadway tickets and sporting events.

Joseph Cox / 404 Media : ShinyHunters claims it used an Oracle PeopleSoft zero-day to hack FBI-related services and steal employee and applicant data; it also defaced the FBI jobs site —  A sample of 5,000 alleged agents seen by 404 Media includes names, addresses, phone numbers, and details on FBI employees' spouses.

Financial Times : In his UN General Assembly speech, Trump rejects a “globalist scheme to control” AI, saying the US is “leading now over China by a lot and everyone else” —  President says US is 'leading over China by a lot and everyone else, and we're going to keep it that way'

Bloomberg : Sources: Mirendil, founded by former Anthropic researchers to build self-improving AI, is in talks to raise ~$1B led by Kleiner Perkins at a $5B valuation —  Mirendil, an artificial intelligence startup launched by former Anthropic PBC researchers, is in talks to raise a new round of funding …

Nat Rubio-Licht / The Deep View : GPT-6 Sol costs $2/1M input and $10/1M output tokens, and GPT-6 Luna costs $0.10/1M input and $0.50/1M output tokens, both about 50% cheaper than GPT-5.6 —  OpenAI is releasing a more budget-friendly version of its most powerful model.  —  On Tuesday, the company unveiled GPT-6 Sol and GPT-6 Luna …

David Gewirtz / ZDNET : OpenAI launches GPT-6 Sol and Luna, saying Sol makes about half as many mistakes as GPT-5.6 Sol and Luna matches GPT-5.6 Sol's performance at ~1% of the cost —  GPT-6 Sol and Luna arrive less than three months after GPT-5.6, with OpenAI claiming major accuracy gains, dramatic cost reductions …

微软 Xbox 游戏业务宣布旗下第一方工作室 Halo Studios 等裁员 268 人，动视将负责下一代 Halo 游戏的开发，而原来负责开发 Halo 的 Halo Studios 则转变成辅助工作室角色。Obsidian 工作室将成为 Bethesda 的一部分，将继续开发 Grounded 以及新 Fallout 游戏。King 工作室将合并微软的休闲游戏部门 Microsoft Casual Games。开发 Forza 系列的 Playground 和开发新 Fable 游戏的 Turn 10 将合并为一家工作室。 Ninja Theory 工作室预计将会关闭。Arkane 工作室仍然在磋商中。

盖洛普 8 月民调显示，仅有 54% 的美国人饮酒，而 2010 年这一数字是 67%。千禧一代和 Z 世代推动了减少饮酒的趋势，而 50-64 岁的中老年人的酒精消费则在上升。发表在《Annals of Internal Medicine》期刊上的一项研究分析了逾 11.4 万名美国成年人的调查数据，受访者在 2018-2024 年间参加了 CDC 的年度健康调查 National Health Interview Survey，其中包括了饮酒的情况。结果显示，2022-2024 年期间，美国人的总体饮酒率下降了 2%，重度饮酒率下降了 8%。Z 世代的总体饮酒率降幅最为显著下降了近 6%，千禧一代降幅约 2%。与此同时，2018-2024 年间 50-64 岁中老年人总体饮酒量率加了 4%，重度饮酒率激增了 36%。相比中老年人，年轻人更了解酒精的负面影响。老年人也可能更富裕能承担更低酒精消费。

天文学家发现了已知最年轻的行星——Elias 2-24 b。这颗不到 100 万年的木星大小行星仍被其形成时的气体和尘埃包围。其令人惊讶的形成速度挑战了关于巨行星如何形成的主流理论。现有的行星形成理论认为，一颗大质量行星不可能形成地如此迅速，尤其是在距离恒星如此遥远的地方。当前模型表明，在太阳系中木星的位置形成一颗木星大小的行星大约需要 500 万年，那么距恒星更远的巨行星形成时间应该更长。然而 Elias 2-24 系统中这个微弱天体到恒星的距离约为地球到太阳距离的 55 倍，而且已经显示出行星形成的迹象。Elias 2-24 b的质量与木星相当，围绕一颗距离地球约 450 光年的恒星运行。由于该系统还很年轻，天文学家通过研究它可以一瞥数十亿年前太阳系的样子。

对加州亿万富翁征收一次性 5% 税的提案 Initiative Number 25-0024 将在 11 月 3 日进行公投。胡佛研究所的研究显示，面临征税的亿万富翁们已有近三成迁出加州。Larry Page 在迈阿密 Coconut Grove 购买了两栋临水豪宅，总价 1.73 亿美元，同时将家族办公室 Koop 从加州转到注册地特拉华州、办公地址佛罗里达的公司。Sergey Brin 在迈阿密 Allison Island 购买了一栋价值 5100 万美元的临水豪宅，将内华达州登记为正式居住地，他资助了反对征税的政治行动委员会 Building a Better California。Peter Thiel 在 2025 年 12 月将其家族投资公司从加州迁至迈阿密。英伟达 CEO 黄仁宇则是少数公开表示会纳税的亿万富翁，他预计将缴纳 80 亿美元的税。

美国国会的预算法案取消了对 NASA 火星样本采集送回任务 Mars Sample Return(MSR)的资助，虽然该法案还需要通过国会两院的批准以及总统的签署才会生效，但实际上代表着 MSR 计划的终止。MSR 计划因为不断膨胀的成本而备受争议，2024 年其成本膨胀至 110 亿美元，如果推行将占用大部分 NASA 科学预算。2025 年 NASA 设法将项目成本降至 70 亿美元，但费用仍然过高，而 NASA 同时正面临特朗普政府削减科学预算的挑战。MSR 项目的终止意味着火星漫游车毅力号收集的样本无法送回样地球实验室进行分析。

成都六岁女童连允之在世界魔方协会（World Cube Association）在两场赛事中，三天内两次打破了女子三阶魔方世界纪录，成为全球唯一一位平均还原时间低于 4.5 秒的女子魔方选手。她分别在武汉和广州举行的比赛中以 4.52 秒和 4.27 秒的平均成绩两次刷新了纪录。她过去一年进行了高强度训练，每天投入两到三个小时练习魔方，目前已掌握逾 1300 种魔方还原算法。

阿里巴巴 CEO 吴泳铭在阿里云年度云栖大会上透露，该公司的千问（Qwen）团队正持续研究模型架构与数据优化，目标是完成更复杂、长周期任务，向人工智能超级智能（artificial superintelligence）前进。阿里巴巴的旗舰模型 Qwen 3.8 Max 有 2.4 万亿参数，正在训练中的 Qwen 4 将会继续扩大参数规模，未来的 Qwen 4.5 和 Qwen 5 系列将扩大至 5-10 万亿参数规模。吴泳铭表示，阿里巴巴自研的 M890 AI 超级节点能处理参数规模逾 2 万亿模型的推理任务，新一代的真武 V900 性能三倍于上一代的 M890，预计于 2027 年第一季度量产。

AMD 周一股价上涨 9.6% 至 613.31 美元，市值突破一万亿美元，成为英伟达、博通和美光之后第四家市值突破万亿美元的美国芯片公司。英伟达在 2023 年市值突破万亿美元，如今市值逾五万亿美元，是全世界市值最高的公司。AMD 被认为是英伟达在 GPU 芯片领域最强大的竞争对手。AMD 股价在 2026 年上涨 185%，远超聚集众多科技股的纳斯达克指数 15.8% 的涨幅，是标普 500 指数中表现最佳的股票之一。

Google 因地理位置数据处理被爱尔兰数据保护委员会（DPC）罚款 4.03 亿欧元。DPC 对 Google 的调查持续了六年，涉及 Google 在 2018 年 5 月 25 日至 2020 年 2 月 4 日间 Web & App Activity、Location History 和 Location Accuracy 三项功能的位置数据处理。DPC 的报告认为 Google 的位置数据处理违反了 2018 年生效的数据保护法律 GDPR，可能导致用户未意识到自己的位置信息正被用于投放定向广告或推断其兴趣偏好，丧失对自己个人数据的控制权。Google 发表声明，表示它从 2019 年起就调整了位置数据管理。引入了位置数据自动删除功能。

深度集成 Gemini、运行 Android 的笔记本电脑 Googlebooks 将于 10 月 4 日上市。Google 硬件合作伙伴中除了宏碁推出一款起售价 899 美元的型号外，其余厂商的产品都超过 1000 美元。Googlebook 不同于 Chromebook 面向低端市场，它面向的是中端笔记本电脑市场。Googlebooks 的 Continue On 功能允许用户在手机或 Googlebook 之间无缝切换，但需要应用开发者支持；Cast My Apps 可以直接在 Googlebook 上使用 Android 手机已安装应用；Play Store 是 Googlebook 获取应用的主要渠道，侧载受到了限制，只能安装运行已通过 Google 验证身份的开发者的应用；通过深度集成 Gemini Intelligence，用户仅仅移动光标就能激活被称为“Magic Pointer”的 AI 功能，AI 会分析屏幕上的内容，根据上下文提供建议，能从多个应用中提取数据。比如将光标指向电邮中的日期即可创建日历预约。

Saturn 的一项研究显示，ChatGPT、Claude、Copilot、Grok 和 Gemini 等主流 AI 模型在回答财务相关问题时，平均有 57% 会给出错误答案。研究使用了逾百个财务相关问题，分别测试了 ChatGPT、Gemini、Claude 和 Copilot 提供的免费及付费 AI 模型。每个问题最多重复提问五次，向 18 种 AI 模型共提出了逾 10,000 个问题。研究发现，AI 模型的回答中包含计算错误、遗漏即将实施的税收政策变更，或是凭空捏造规则（即幻觉）。在最严重的情况下，依赖 AI 对税务问题的回答可能导致严重的经济损失。研究发现，付费模型的回答比免费模型更准确，较新的模型表现优于较旧的模型。表现最好的是推理模式的 Claude Opus 5，但仍然有 39% 的答案存在错误。

青少年有社交网络成瘾问题，年龄超过 60 岁的婴儿潮一代也存在类似问题。YouGov 在 2026 年进行的一项民调发现，55 岁以上人群有超过五成每周至少使用一次社交媒体。62 岁的 Tracie 女士说，拿起手机就忘记放下了，称很多同龄人都沉迷于不断向下滑动手机屏幕浏览信息。Surrey 大学副教授 Brigitte Stangl 认为，社交媒体在缓解老年人孤独感方面非常有效，“上网可消除行动不便或交通出行困难带来的问题”——这些问题对老年群体影响最为显著。61 岁的演员 Kerri Hedley-Cheney 称，社交媒体让她能加入表演组织，建立人脉和互相支持。她同时也承认可能会在社媒上浪费几小时时间。她指出，使用社交媒体也缩短了她的注意力持续时间，类似年轻人，她现在的注意力间隔只有大约 10 秒钟。

位于埃及帝王谷的图坦卡蒙陵墓一直令考古学家困惑不已，怀疑墙壁后面有隐藏密室。借助于雷达，研究人员认为他们找到了密室存在的证据。但是否发掘密室还需要等待埃及最高文物委员会的批准。如果批准，发掘工作最快将于 11 月启动。图坦卡蒙陵墓建于公元前 14 世纪，于 1922 年被发现。英国埃及学家 Nicholas Reeves 认为，隐藏密室可能是图坦卡蒙前任统治者娜芙蒂蒂（Nefertiti）的安息地—— 娜芙蒂蒂是法老阿肯纳顿的王后，她的女儿安克姗海娜曼嫁给了同父异母的弟弟图坦卡蒙，她在阿肯纳顿死后曾短暂统治埃及。Reeves 认为图坦卡蒙陵墓最初是一座规模庞大的皇家陵墓，当年轻的图坦卡蒙意外去世时，陵墓入口被扩建并改造以供其安葬，更深处的区域则被封闭起来。探地雷达显示陵墓后有一条宽两米、填满碎石的通道。

根据发表在《Sociology of Religion: A Quarterly Review》上的一项研究，互联网普及与宗教信仰下降相关。研究人员汇集了一个 81 个国家的数据集，时间跨度从 1990-2022 年。研究团队构建了一个包含六项指标的指数去衡量各国的平均宗教水平。这些指标涵盖了宗教仪式参与率、神（God）在个人生活中的重要性、宗教的总体重要性、自我认定的宗教身份、对教会的信任度以及是否正式隶属于某个教派。结果显示，一个国家的互联网普及度越高，其宗教信仰程度就会呈下降趋势。在考虑了其它因素之后这种负关联仍然存在，且独立于整体经济或教育水平的提升。

日本总务省公布的人口推算数据显示，截至 9 月 15 日，日本 65 岁以上老年人口较上年增加 2 万人，达到 3624 万人。老年人口时隔两年再次增加，在总人口中的占比为 29.6%，刷新历史最高纪录。截至 2025 年，日本 65 岁以上就业人数达到 943 万人，连续 22 年增加，创历史新高。按性别来看，日本的男性老年人口为 1571 万人，女性为 2054 万人。2025 年死亡老年人口多于新满 65 岁的人数，日本的老年人口总数有所减少，但 2026 年再次转为增加。