OrangeBot.AI Digest — 2026-09-09

Google's AI map of every possible human DNA mutation

A security layer for AI coding agents

From issue to production, run by agents.

Brands outbid each other to put their logo on my ass

Train your MicroDuck and win the Golden Beak Belt

Give your AI agent guides that show users where to click

Sharper visuals, faster flow, better creative control

Work with your data in your language

Understand your ad performance in seconds

Your personal AI agent that gets things done

Open-source OpenRouter

Turn any picture into a playable AI adventure

Where verified talent meets opportunity. Hire with trust!

Your product in AI and AI in your product

2D IDE for running agents across projects without fatigue

Recursive self-improvement (RSI) requires a concrete mechanism through which an AI system observes its capabilities and converts that evidence into the next round of learning. We present NeoHorse-1, a family of agent-native models developed to explore this path through agentic post-training. Our system combines a heterogeneous model pool with intelligent routing, recording the predicted capability demand, selected service tier, and subsequent interaction for each user turn. These records are converted into training examples that preserve interleaved reasoning, tool calls, and harness context, and are admitted through structural validation, six-dimensional semantic evaluation, and subscene-level labeling. Routing signals organize supervised fine-tuning into a three-stage curriculum and extend to routing-guided on-policy distillation, where a teacher supervises student-generated responses under the same progression. Capability-guided allocation then converts evaluation feedback into the next training mixture, closing an evaluation-selection-update loop in which what the system learns to do shapes what it learns from next. Across eleven benchmarks covering harness-based agents, tool use, coding, and instruction following, post-training raises the macro-average from 58.94 to 64.87 at 4B and from 65.60 to 69.04 at 9B, substantially narrowing the aggregate gap between the post-trained 4B model and the 9B base model. NeoHorse-1 provides an initial prototype of this feedback-driven process and a path toward harness-mediated RSI across successive iterations.

We introduce AuK, an open-source foundational model that unifies speech generation and editing through a common interface of natural-language instructions and audio context. To support this broad capability set, we construct approximately 3.03 billion instruction--audio instances and 1.95 million hours of effective supervision across five task families: speech generation, content editing, enhancement and separation, paralinguistic editing, and acoustic editing. AuK combines a multimodal large language model for semantic conditioning, an VAE jointly trained on speech, general audio, and music for acoustic conditioning, and a hybrid rectified-flow Transformer that performs dual-stream MMDiT blocks followed by unified single-stream DiT blocks for generation. Training begins with generation-only warm-up and proceeds to joint generation--editing pre-training. We then apply complementary post-training strategies: human-feedback preference optimization for open-ended editing and reward-based reinforcement learning for speech generation. To reduce inference cost, we further distill the model with consistency initialization and task-routed Decoupled DMD. The resulting AuK-Flash performs 4-step inference without classifier-free guidance and achieves a 4.5 wall-clock speedup over the full model under matched conditions. Experiments demonstrate leading performance on zero-shot and instruction-controlled speech generation and general instruction-guided editing, while remaining competitive on signal-level restoration tasks. We release both the source code and model weights to support reproducibility and further research.

In this work, we present Gander, an end-to-end model that unifies omni perception, realtime interaction, and agentic capabilities within a single framework. In contrast to turn-based conventional paradigms, Gander continuously receives streaming inputs across multiple modalities, including video, speech, and text, enabling natural full-duplex interaction in both everyday conversations and complex workflow-oriented agent scenarios. Users can interrupt the model at any time, while the model can also proactively provide intermediate feedback or ask follow up questions. To natively support these capabilities, Gander adopts two key architectural designs: 1) It employs a Cerebellum-Brain collaborative framework, in which the Cerebellum is responsible for realtime interaction and omni conversational capabilities, while the Brain handles complex reasoning and higher-level agentic tasks. The two components interact continuously through tool calling and the agent orchestration runtime. 2) The Cerebellum is built upon a streaming Thinker-Talker architecture, user inputs and model outputs are further flattened into an ordered token stream at the chunk level, providing a unified representation for low latency, continuous interaction. We conduct comprehensive evaluations of Gander across four dimensions: conversational ability, omni understanding, interactive capability, and agentic intelligence. Internal human evaluations demonstrate that Gander maintains the natural and expressive spoken dialogue capabilities of SOTA open source models while achieving competitive performance in omni interaction. Gander also demonstrates robustness in challenging real-world scenarios, including background noise interference, multi-party interactions, and backchannel communication. We release Gander together with its models, code, and data to facilitate further research and development in the community.

Weak-to-strong generalization asks whether stronger models can learn from weaker supervisors and surpass them. This question is particularly important for successive model generations and multi-domain consolidation, where repeating frontier-scale post-training from scratch can be prohibitively expensive. Yet conventional distillation treats the weak teacher as an optimization target, potentially imposing its capacity ceiling on the student. We introduce On-Policy Reverse Distillation (OPRD), which evaluates the teacher's policy shift relative to its reference policy on student rollouts and amplifies the component of the student's verifier-driven policy gradient along that direction. By rescaling only verifier-supported updates, OPRD preserves the stationary points of policy optimization while accelerating learning beyond the teacher. In both successive model transfer and multi-teacher distillation, OPRD achieves higher performance with fewer student updates than existing RL and distillation approaches. Response-style analysis shows that OPRD students remain closer to models trained with verifier-based RL alone than to their weak teachers, suggesting that teacher guidance accelerates rather than redirects the student's own optimization. Results in conventional strong-to-weak distillation further demonstrate that OPRD effectively combines verifier-driven policy optimization with teacher guidance regardless of capacity ordering.

Most end-to-end autonomous-driving systems learn by imitating human driving logs, leaving their learned behavior constrained by the quality and behavioral coverage of the recorded trajectories. This report presents DriveZero, an end-to-end system that learns driving behavior beyond human demonstrations. It decomposes driving into a perception model and an action model, pretrains each in the regime best suited to it, and combines them into one end-to-end planner. The two models call for different learning recipes: perception must understand the world, and benefits from massive and diverse visual data; action must interact with it, and requires closed-loop feedback. On the action side, we introduce DriveRL, a mixed-agent closed-loop reinforcement-learning framework. It converts real driving logs into interactive worlds, where a privileged teacher policy is trained with PPO through closed-loop rollouts. For the perception model, DriveVFM consolidates multiple frozen vision foundation models, including DINOv3, SigLIP2, SAM and Depth Anything V2, into a single backbone from raw images alone, requiring no task-specific annotations. DriveZero then unifies the two: a camera-only planner that distills the frozen DriveRL teacher through its rolled-out trajectories. The goal-conditioned teacher can moreover be queried under augmented driving intents, yielding diverse, goal-consistent supervision that logged data cannot provide. On nuPlan, DriveRL with value-guided test-time action search achieves a mean score of 93.57 across the Val14, Test14-hard, and Test14-random community splits in both non-reactive and reactive modes, exceeding the Log-Replay expert on all three splits. DriveZero achieves state-of-the-art performance on NAVSIMv1, NAVSIMv2 and the closed-loop HUGSIM benchmark without any human trajectory supervision.

World-Action Models inherit world knowledge from video-generative priors, and channel it into executable control signals through embodied experience. Existing systems, however, are monolithic: the generative backbone, visual representation, architecture, information flow, inference procedure, and training data are tightly coupled, obscuring which design choices matter and why. We introduce OpenWAM, an open research stack that turns world-action pretraining into a controlled experimental program. OpenWAM-Infra factorizes the WAM design space into composable modules with unified training, inference, deployment, and evaluation. On this substrate, OpenWAM-Study examines three questions through controlled experiments: what to inherit, how world and action learning interact, and how their synergy scales; and distills three principles: upstream knowledge transfers through a sufficiently capable generative backbone and a compact, information-rich latent space; world-action synergy requires dedicated action capacity, explicit world-to-action information flow, and synchronized joint denoising; and embodied pretraining principally improves out-of-domain generalization, with one-stage co-training over egocentric and robot data integrating world coverage and action grounding. Composing these principles, we build OpenWAM-α, an open WAM pretrained on roughly 6,400 hours of egocentric human and robot data and evaluated across simulation and real-world benchmarks. Across the eight simulation benchmarks and the real-robot experiments, which together span embodiments from single-arm and bimanual manipulation to dexterous hands, OpenWAM-α delivers consistently excellent performance, sustaining its top-tier standing from simulation to the physical world. We release the full stack, including infrastructure, evaluation protocols, pretrained models, and data recipes, to facilitate future research.

World-action models (WAM) predict future states to guide robot actions, enabling learning from both action-free video and action-labeled interaction. Most inherit pretrained video generators, leaving WAM pretraining and scaling underexplored. We introduce Genie Envisioner Act 2.0 (GE-Act 2.0), a world-action model whose trainable generative and action components are all initialized from scratch on manipulation data. It combines a control-oriented autoencoder (CoAE), a single-step visual planner (SVP), and an inverse dynamics model (IDM). CoAE retains action- and instruction-relevant information under aggressive compression, while SVP produces a complete future state in one differentiable pass, so visual planning and inverse dynamics can be pretrained separately on complementary data. The components are then jointly trained with knowledge-aligned selective optimization (KASO), which reduces mismatched supervision by selecting only predicted futures judged behaviorally compatible with the recorded action. We evaluate pretrained checkpoints directly, without per-task fine-tuning, on 100 tasks across 20 manipulation skill groups with held-out scenes, backgrounds, lighting, and object instances. Scaling co-training data from 300 to 30,000 hours raises success from 17.1% to 44.1% on G1-OP and from 13.4% to 31.1% on G2-90D; despite comprising less than 2% of the co-training data, G2-90D improves by 17.7 points, suggesting cross-embodiment transfer. Gains span 19/20 and 18/20 skill groups, and skill-specific coverage strongly correlates with zero-shot out-of-distribution (OOD) success (Pearson r=0.80; Spearman rho=0.85). Under the same protocol, the model grounds object, color, shape, and position references in at least 90% of trials and follows explicit instructions even when they conflict with an already-committed behavior or a conventional scene association.

Monocular depth estimation is a ubiquitous yet highly ill-posed computer vision task, with downstream applications in scene reconstruction, computational photography, and robotics, among others. Despite the field's maturity, recent models still struggle to generalize to out-of-distribution inputs and to produce sharp and detailed depth maps. In this paper, we revisit Marigold, a set of techniques for repurposing modern image generation and editing models, powered by the diffusion transformer (DiT) architecture, into state-of-the-art monocular depth estimators. Our recipes target single-step inference from pretrained multi-step flow-matching models, with quantization where needed, preserving model capacity while remaining cheap to run. We analyze the artifacts of naive training and identify two effective remedies: aligning the model's internal representations with semantic features extracted from ground-truth, and adopting a 2-stage fine-tuning protocol built around a novel Sinkhorn-based loss. The results are crisper, cleaner depth maps that generalize well out-of-distribution, with 16-26% improvement in AbsRel over the previous best on KITTI and ETH3D. Qualitatively, our model resolves fur, foliage, and hair-thin edges that have eluded prior models. Furthermore, Marigold V2 achieves state-of-the-art results when applied to other dense regression tasks, such as surface normals estimation and intrinsic image decomposition. Project website: https://hf.co/spaces/huawei-bayerlab/marigold-v2-web

We present Miles v0.1, a full-stack, production-ready system for frontier post-training. Building upon the clean design of slime, Miles designs each stage of the reinforcement-learning (RL) training loop around a single principle: components should be verified, clean, and customizable. With accuracy, efficiency, reliability, and scalability as first-class goals, Miles aims to make frontier-scale RL accessible to researchers and enterprises alike. This report walks through the system end to end: rollout engines built on SGLang, a trainer with a choice of two backends (NVIDIA Megatron-LM and PyTorch FSDP), and three weight-synchronization transports for different deployment topologies. Beyond full-parameter RL, Miles also supports LoRA RL, on-policy distillation, supervised fine-tuning, and true-on-policy rollout-training alignment, and extends the same architecture to diffusion models. We close with an end-to-end case study: fully asynchronous agentic RL on a GLM-5.2 744B-A40B model over terminal-use coding tasks, running on 64 NVIDIA GB300 GPUs with a median step time of 263 seconds over the first 30 measured steps. Miles is open-sourced at https://github.com/radixark/miles, with the project website at https://miles.radixark.com.

Autoregressive (AR) video diffusion models have shown great potential in real-time video generation. Recent methods distill pretrained bidirectional video diffusion models into causal AR students through Distribution Matching Distillation (DMD), but the generated videos often suffer from over-saturation and over-smoothing issues, resulting in limited visual quality and realism. The key contributing factor is the mode-seeking behavior of the reverse KL objective in DMD, which can cause the student distribution to collapse onto only a few modes of the teacher distribution. To address this, we propose Mask Forcing, a Dual-Noise Masking Rollout strategy that perturbs the AR student self-rollout to mitigate mode collapse induced by reverse-KL mode seeking. The core idea is to inject cleaner signals into noisy rollout inputs via random masks along spatial and temporal axes during the self-rollout process of AR diffusion distillation. Such perturbations encourage the student rollouts to explore more regions of the teacher distribution, allowing DMD to provide learning signals beyond the modes already covered by the student. Moreover, the cleaner tokens act as denoising guidance for other noisier tokens, improving the intermediate rollout predictions and reducing error accumulation. Extensive experiments demonstrate that our method improves multiple AR video diffusion distillation methods with higher visual quality efficiently, without incorporating real video data or additional post-training stages.

Diverse and simulation-ready indoor scenes are essential for interactive entertainment and embodied AI, yet their scalable generation remains challenging. Recent agentic text-to-3D scene pipelines that rely on vision-language models (VLMs) can generate scenes of high fidelity but require costly iterative object placement and refinement. Another mainstream paradigm, parametric image-to-3D scene models, produces scenes efficiently from strong priors learned from 2D images but often leads to imprecise and physically invalid scenes. More importantly, both paradigms struggle to output diverse scenes for a single input, making it hard for them to reflect the dynamically changing nature of real scenes. In this paper we propose SceneMosaic, a framework that combines the merits of both paradigms. It obtains the initial candidate from the learned image-based prior, and subsequently evolves the result through VLM agents, ensuring both efficiency and physical validity. Within the evolution process, SceneMosaic exploits the locality of natural scenes and decomposes a scene into independent local units, allowing separate evolution within each unit before composing the global scene via Cartesian product. On SceneEval-100, SceneMosaic matches the strongest agentic baseline in semantic layout quality with a 24x speedup, substantially reduces physical violations, and receives the highest human ratings. Our code is publicly available at https://github.com/rxjfighting/SceneMosaic.

Large Reasoning Models (LRMs) achieve superior problem-solving through extended Chain-of-Thought (CoT) generation, but the resulting key-value (KV) cache grows linearly with sequence length and creates severe memory bottlenecks, often exceeding GPU capacity for long reasoning traces. Existing KV cache compression methods rely on recent queries to estimate future token importance, implicitly assuming these serve as reliable proxies for future attention patterns. We demonstrate that this assumption fails in long-horizon reasoning: certain decoding steps generate Thought Revisiting Tokens (TRT) that re-attend to distant previous context, such as task-solving plans formulated early in the trace. Through systematic analysis, we discover that queries corresponding to the TRT cluster into a small number of similarity groups in the embedding space. Based on this insight, we propose BeaconKV, a training-free KV cache compression method that maintains beacon queries, compact representatives for each global query cluster, to anticipate which KV pairs will be revisited without storing the entire query history. Across four open-source LRMs and diverse reasoning benchmarks, BeaconKV generally outperforms existing compression methods, achieving up to 5.8times memory reduction while nearly preserving full cache accuracy and improving throughput by over 4.3times.

Latent visual reasoning aims to perform multimodal reasoning through hidden-state computation rather than explicit textual chains of thought. However, visual information being present in a latent state does not imply that the model actually relies on that state when producing its answer, especially when alternative image-conditioned paths remain available. We introduce Causal Visual Recurrent Reasoning (CVRR), which preserves pretrained visual competence while making recurrent computation the required image-conditioned path to prediction. CVRR initializes recurrence from the question hidden state after the pretrained vision-language model has incorporated the image, then repeatedly updates this state while re-reading the same fixed visual evidence. Before decoding, visual states and the original multimodal KV cache are removed so that only the final recurrent state carries image-conditioned information to the answer. Across the V^*, MMVP, BLINK, and MME-RealWorld-Lite benchmarks, CVRR retains strong performance under this strict interface, while compatible latent reasoners fail to recover comparable visual competence even when retrained under the same constraint. Causal interventions further show that predictions remain sensitive to recurrent content when the question is held fixed, and that persistent visual evidence causally revises the recurrent trajectory. These results distinguish latent informativeness from latent computation that is actually used for prediction.

As large language models (LLMs) are increasingly deployed in alignment-sensitive contexts, activation steering has emerged as a lightweight, inference-time alternative to fine-tuning methods (e.g., RLHF, DPO) for behavioral control. However, existing work typically validates steering on isolated behaviors, leaving it unclear whether steering vectors encode coherent semantic structure or merely exploit behavior-specific shortcuts. We investigate whether the latent geometry of LLM steering vectors reflects theory-specified structure in human values and morality. Using Schwartz's Theory of Basic Human Values as our primary fine-grained framework, we introduce a 26K-sample benchmark covering 20 human values and analyze distribution-driven methods (e.g., CAA, SphericalSteer, ODESteer) and behavior-centric approaches (e.g., COLD-Steer, BiPO) across diverse model families and sizes. We find that distribution-driven methods recover human value topologies aligned with theoretical predictions (Spearman ρ up to 0.51, p < 10^{-13}). In contrast, behavior-centric methods achieve comparable steering performance but show little correlation with the expected value geometry. Geometric fidelity improves with model scale but drops after instruction tuning. Finally, better geometric alignment also leads to more human-consistent transfer across values: steering one value correctly lifts compatible values and suppresses opposing ones. Code and data are available at: https://github.com/DeepRCL/Steering_Geometry.

Transferring human hand demonstrations to robotic grippers has recently emerged as a cost-effective solution for robot learning. However, existing methods are largely confined to simple, planar tasks and fail to handle complex spatial movements (e.g., intricate trajectories involving rotations or flips) that are essential for robot manipulation. Motivated by this gap, we adopt an implicit, data-driven approach guided by fine-grained hand-pose motions. To this end, we introduce a scalable acquisition pipeline to collect hand-gripper paired demonstrations, governed by a rigorous protocol that prioritizes motion complexity and leverages a handheld gripper for seamless action mimicry. This yields a large-scale paired dataset comprising 6,189 episodes across 1,254 unique objects, exhibiting significantly higher spatial complexity than existing benchmarks. However, learning such complex mappings remains challenging. We observe that naive end-to-end generation of full gripper pose sequences is insufficient, as minor trajectory deviations compound rapidly under intricate dynamics. To address this, we propose a two-stage framework: Stage I predicts sparse gripper keyframes (initial and terminal) to simplify the mapping objective, while Stage II generates the full continuous action sequence conditioned on these keyframes. Furthermore, to mitigate cumulative drift, we keep the gripper's orientation being learned while post-optimizing its translation based on the grasping heuristic and kinematic consistency. In both simulation and real-robot experiments, our framework enables stable and precise hand-to-gripper transfer of complex spatial manipulations, significantly outperforming traditional baselines. Project page: https://cosmoh2g.github.io.

Samantha Cole / 404 Media : Slack message: Automattic CEO Matt Mullenweg says CFO Mark Davies “conspired” with board members to put him on paid leave and Davies was voted to be interim CEO —  Mullenweg, co-founder of Wordpress, wrote in a company-wide Slack channel on Wednesday that board members “conspired” …

Reed Albergotti / Semafor : Source: Meta AI researcher Andrew Tulloch, who came over from Thinking Machine Lab as one of the highest paid employees last year, is leaving the company —  Meta AI researcher Andrew Tulloch, who gained attention for being one of the highest paid employees in tech industry, is leaving the company …

Lily Hay Newman / Wired : New opt-in Apple Watch Siri audio features can listen to and transcribe conversations all day; Apple says raw audio is inaccessible and deleted after processing —  The new Apple Watch includes several “intelligent” listening features that have privacy and security baked in.

Supreeth Koundinya / AIM : Tailwind Labs, the company behind Tailwind CSS, is joining Shopify; in January, it laid off three of four engineers, citing AI's impact on documentation traffic —  Tailwind Labs, the company behind the Tailwind CSS web-development framework, is joining Shopify, with the team moving under …

Cameron Faulkner / The Verge : Apple raises iPhone prices by $100 on all of the older models it is offering, including the iPhone 16, the iPhone 17 lineup, as well as the iPhone Air —  The price increase impacts the iPhone 16, the full iPhone 17 lineup, as well as the iPhone Air.

Anthropic : Anthropic details four incidents where Claude gained unauthorized access to third-party systems, including a new Opus 4.6 case; METR will investigate them —  We present an alignment assessment of four incidents in which Claude models gained unauthorized access to real third-party systems.

Business Insider : Sources: Salesforce has been in talks to buy Listen Labs, an AI-powered customer research platform for $2B; Listen Labs was valued at $500M earlier this year —  - Salesforce has been in talks to acquire AI startup Listen Labs for around $2 billion.  — Listen Labs, which was last valued …

Terrence O'Brien / The Verge : iPhone Duo supports Apple Pencil and comes with iOS redesigned for multitasking on different screen sizes/orientations, with auto-adjusting controls and apps —  iPhone Duo is also the first iPhone to support Apple Pencil. … Obviously, part of the appeal of a foldable phone is all that additional screen real estate.

Igor Bonifacic / Engadget : Apple says iOS 27, iPadOS 27, watchOS 27, and macOS 27 Golden Gate, all with Siri AI, will be released on September 14 —  It's almost time for Siri AI.  —  The next version of iOS will arrive on Monday, September 14, Apple announced today at its fall iPhone event.

The Verge : Hands-on with the iPhone Duo: the folding mechanism is smooth, the crease is almost impossible to find with a finger, and the inner screen has a matte finish —  We've tried the first ever foldable iPhone at Apple's launch event. … The first foldable iPhone is finally here …

Dominic Preston / The Verge : The iPhone Duo starts at $1,999 for 256GB, with a 2TB model costing $3,199, in star white and night sky colors; preorders start Oct. 16, and sales begin Oct. 23 —  Seven years after foldable phones arrived, Apple thinks the tech is ready for an iPhone.

Chance Miller / 9to5Mac : Apple announces its first foldable, the iPhone Duo, with a 5.4-inch outer display, a 7.6-inch inner display, A20 Pro with a new display engine, and Touch ID —  Apple has officially unveiled the highly anticipated foldable iPhone, and it's called iPhone Duo.  Here's everything you need to know.

Jeremy Gray / PetaPixel : iPhone 18 Pro has a faster 48MP Fusion Main camera with variable aperture that lets in ~50% more light in low-light, along with enhanced camera app controls —  Apple has unveiled the iPhone 18 Pro and iPhone 18 Pro Max, its latest flagship slab smartphones.

Zac Hall / 9to5Mac : Apple unveils the Apple Watch Series 12 and Ultra 4, with a new S11 chip, upgraded health tracking, an Audio Intelligence feature, a readiness feature, and more —  Apple Watch Series 12 and Apple Watch Ultra 4 are official.  Apple has announced its two new Apple Watch models as part of its product event today.

Ryan Christoffel / 9to5Mac : Apple announces two AirPods 5 models with ANC that removes up to 50% more noise than AirPods 4, for $129 or $149 with a wireless charging case —  Today during its iPhone 18 Pro event, Apple announced the next generation AirPods 5 with improved active noise cancellation.  Here are the details.

青少年的阅读、数学和科学成绩降至 2000 年国际 PISA 测试启动以来的最低水平，15 岁学生的阅读能力相当于过去低一岁学生的水平。经合组织（OECD）将成绩下滑归因于屏幕使用时长增加、出于兴趣的阅读减少以及数字设备带来的干扰。OECD 还发现，经常将 AI 聊天机器人用于起草论文或总结文本等任务的学生，其科学成绩平均低了 20分，相当于一年的学习差距。PISA 阅读成绩指数于 2012 年达到 501 分的最高值，到去年降至了 466 分。OECD 秘书长 Mathias Cormann 表示：“在阅读上，屏幕使用时长的增加和兴趣阅读的减少，与成绩下滑有着密切联系。我们还观察到更多‘仓促阅读’的现象，即学生匆忙浏览文本后给出草率且错误的答案。”研究还发现，富裕家庭学生阅读成绩下滑幅度更明显。由于阅读水平下降，青少年在科学和数学上也处于本世纪以来的最低水平。科学成绩从 2009 年 506 分的最高值降至 486 分，数学成绩则从 502 分降至 469 分。东亚地区表现最出色。参与测试的中国城市、日本、韩国、新加坡和台湾等拥有最优秀的教育体系，其它地区仅有英国和爱沙尼亚跻身阅读、数学和科学的前十名。

为遵守欧洲的数字市场法律《Digital Markets Act（DMA）》，Google 宣布将调整欧洲的搜索服务，提升 Expedia 和 Hotels.com 等竞争对手比价服务的权重，移除酒店、航空公司和餐厅搜索结果中的部分实时信息功能。Google 高级副总裁 Nick Fox 表示，“这些改变降低（degrade）了欧洲用户的体验——以牺牲本地企业为代价提升了在线中介平台的地位，移除了用户每天依赖的实用功能。”DMA 禁止科技巨头在排名中给予自家服务高于第三方服务的待遇。搜索排名仍然由 Google 的算法决定，这意味着搜索巨人在搜索结果的展示方式上仍掌握着很大程度的控制权。Google 尚未披露欧洲用户何时会看到这些改变。

1966 年 9 月 8 日，《星际迷航：原初系列》上映。60 年之后，《星际迷航》系列包含了 967 集电视剧、14 部电影，以及大量小说等其它媒介，形成了庞大的《星际迷航》宇宙，是史上最成功的科幻作品系列之一。《星际迷航》最著名的开头旁白是：“宇宙，终极的边疆。这里叙述的是星舰进取号的旅程，它的五年任务，是为了要探索这全然未知的新世界戏中，寻找新生命和新文明，勇踏前人未至之境。”以前被称为 Trekkies 的星际迷航粉丝可能只记住 James T. Kirk 这位进取号船长的名字，如今 Picard、Janeway 或 Sisko 等船长的名字对大部分粉丝而言都如雷贯耳。

NVIDIA 2026 年度创业企业展示活动的半程展示将于 10 月 15 日在苏州金鸡湖国际会议中心 A 馆举办。 半程展示将聚焦 AI 智能体、物理 AI、开放模型及 AI 应用等前沿方向。届时，30 余家 NVIDIA 初创加速计划会员企业将带来路演，NVIDIA 技术专家联合生态伙伴共同把脉 AI 产业趋势，现场设专属展区供近距离体验。 活动面向科创企业、投资人、大企业、云厂商、创业加速器等生态从业者开放报名，审核制确认，报名从速！ https://scrm.nvidia.cn/events/semi-final-20261015?pushId=8kJQ2pgiiMFxtKL7Gapgdg1

非营利组织费米探索者任务（Fermi Explorer Mission）公布了一项低成本星际方案，计划 2029 年底前发射小型探测器，飞向约 4.4 光年外的半人马座α星系统。路线由 AI 规划，全程耗时约 8 万年，总预算低于 1500 万美元。项目联合创始人兼主席菲利普·约翰斯顿在声明中称，旅程结束时无人能亲见结果，而这正是意义所在：“这是在迈出人类能够做到的第一步，也是为了激励后代继续突破我们所继承下来的极限”。任务设四项硬指标：探测器应朝半人马座α星系统飞行，8 万年内至少完成整个旅程的 99%；有效载荷至少 1 公斤，封装在 10×10×10 厘米空间内；2029 年底前发射；全周期成本低于 1500 万美元。半人马座α星系统由半人马座αA、αB与比邻星组成，是太阳系最近的恒星邻居。

澳大利亚玩家通过社媒报告，Valve 要求澳大利亚玩家用信用卡验证年龄，之后才允许访问 Steam 商店的 R18+ 类游戏。玩家抱怨，他们的账号历史都超过 20 年了，如今却还被迫要通过信用卡去验证年龄。而在澳大利亚，信用卡并不像美国那样普及，大部分成年人都只有借记卡，为了访问 R18+ 游戏页面去申请信用卡有点小题大作。此前 Valve 也对访问 R18+ 游戏的英国玩家推行了信用卡年龄验证要求。

《南方公园（South Park）》创作者 Matt Stone 和 Trey Parker 宣布，“受苹果和 Google 的勇敢和爱国精神的启发，我们将把《南方公园》更名为《南方美国（South America）》。”美国总统特朗普上个月签署了行政令，将安大略湖（Lake Ontario）更名为美国湖（Lake America），去年他下令将墨西哥湾（Gulf of Mexico）更名为“美国湾”（Gulf of America）。《南方公园/南方美国》将于 9 月 16 日播出第 29 季第一集，它曾在 27 季刻画了特朗普睡了撒旦并让其怀孕的情节，被白宫抨击是“如垂死蛛丝，用枯竭的灵感，不择手段博取关注”。

许多女性在孕期或使用口服避孕药时，经常会有忘记物品摆放、难以集中注意力的经历。这种常被戏称为“孕傻” （Mom Brain）的现象，长期以来缺乏明确的生物学解释。发表在《Science Bulletin》上的一项研究揭示了持续水平高雌激素并非直接作用于记忆中枢，而是通过干扰一条特定的“下丘脑-海马”神经环路，从而暂时性损害记忆功能，从机制层面解答了高水平雌激素影响认知的生物学基础。研究人员首先构建了模拟孕期持续高雌激素水平的小鼠模型，发现高雌激素会引起可逆的记忆损伤，但并不影响情绪或探索动机，表明这是一种特异性的认知效应。在机制层面，研究团队发现，外侧下丘脑中占主导地位的雌激素受体集中在GABA能神经元上。这类神经元通常向大脑其他区域发送抑制性信号。单核RNA测序与电生理实验表明，高雌激素水平会抑制这些神经元中的GABA-A受体信号，导致其自发放电频率显著增加。当特异性敲除这些下丘脑神经元中的雌激素受体后，小鼠中由高雌激素及孕期本身引起的记忆损伤均得到恢复。进一步神经环路追踪显示，这些过度活跃的下丘脑神经元直接投射到海马CA3区（记忆形成的关键枢纽）。利用化学遗传学技术，研究人员证明：沉默这一“下丘脑-海马”通路可保护小鼠免受高雌激素诱导的记忆损伤；而在无雌激素升高的情况下，人工激活同一通路本身已足以损害记忆。研究表明，低水平、周期性的雌激素暴露主要作用于海马，维持认知功能；但持续高水平的雌激素暴露则主要激活了以下丘脑为中心的神经通路。这一不同作用靶区的发现，成功调和了该领域长期存在的争议。

基于 Chromium 的浏览器 Brave 公布了一份测试结果，称其桌面版比竞争对手 Chrome、Microsoft Edge 和 Firefox 占用更少的系统资源，页面加载速度更快。测试使用的 Firefox 不是最新版本 v155，而是旧版本 v146.0.1。结果显示，Brave 的平均 CPU 占用率为 33%，Chrome 为 47%，Edge 为 53%，Firefox 为 78%；Brave 使用了约 1.2 GB 内存，而 Chrome 为 1.75 GB，Edge 为 1.62 GB，Firefox 为 1.65 GB；Brave 完成网页加载约需要 4.4 秒，Chrome 需要 5.1 秒，Firefox 需要 5.3 秒，Edge 需要 6 秒。Brave 内置了广告屏蔽功能，其网页加载速度受益了这一功能。

Flinders 大学的研究人员发现，冲马桶会向周围空气释放气溶胶和生物气溶胶，气溶胶颗粒甚至会进入到成年人的呼吸区，而冲水后气溶胶会在空气中悬浮至少 20 秒。这些发现是基于对 22 项马桶气溶胶研究的分析。结果表明，保持良好的厕所卫生，包括定期清洁马桶及其周围表面，以及使用后洗手，有助于最大限度减少微生物污染和潜在的微生物疾病风险。使用马桶的低冲水模式也有助于最大限度减少气溶胶的产生。充足的通风有助于扩散和清除悬浮的空气颗粒，关闭马桶盖会改变气溶胶的扩散方向，气溶胶会从马桶盖和马桶座之间的缝隙逸出，而不是向上扩散。研究人员建议保持卫生间通风良好，在冲水前盖上马桶盖。

焦虑是人类最常见的精神疾病，全球约有 3.59 亿人受到影响。控制呼吸被认为有助于控制焦虑，根据发表在 PNAS 期刊上的一项研究，科学家基于小鼠研究揭示了这一现象背后的鼻脑回路（nose-to-brain circuit）机制。鼻脑回路始于嗅觉感觉神经元（OSN），OSN 感知鼻腔吸气并将信号传递给嗅球的僧帽细胞，信息随后从嗅球传输到海马旁皮层中的长投射中间神经元，最终到达杏仁体基底外侧核的谷氨酸能神经元。研究发现，鼻腔气流通过这条通路以频率依赖的方式调节类焦虑行为，这种效应可能取决于通过鼻腔吸气的频率。这种效应是双向的，意味着加快呼吸可能会加剧焦虑，而减慢呼吸则能缓解焦虑。

开源媒体服务器 Jellyfin 项目释出了 v12.0。Jellyfin 采用了新版本号，在旧版本下 Jellyfin 12.0 其实就是 10.12.0，10.11.x 将是最后一个使用旧方案的分支。Jellyfin 12.0 主要变化包括：重写了媒体库数据库，显著改进了性能，但数据库重构尚未完全完成；电视剧集支持多版本，可同时包含电视版和加长版，或者 1080p 和 4K 等不同分辨率版本；支持图书和漫画；等等。

在禁止儿童使用社媒平台之后，澳大利亚工党政府提出了一项新的法律草案，要求社媒平台允许用户退出算法驱动的信息流。用户可选择算法推荐的个性化内容作为默认信息流，或者拒绝接收算法推荐的内容，只浏览用户关注的朋友和创作者的内容。被称为 My Feed, My Way 的法案旨在给予用户选择权，违反者将面临最高 1.092 亿澳元的罚款。澳大利亚总理 Anthony Albanese 表示，“它赋予用户选择权，如果大型科技公司不遵守我们的法律，我们将追究它们的责任。”

旨在将 Linux 移植到运行 Apple Silicon 芯片的 Mac 电脑的发行版 Asahi Linux 宣布支持 M3 系列 Mac。开发者表示，Linux 对 M3 系列 SoC 及其相关设备支持已达到几乎与 M1 和 M2 系列设备相当的水平，绝大多数功能都能正常工作。其中包括：网络摄像头、内置麦克风、USB（最高支持 USB 3.0 的 10 Gb/s）、硬件加速视频解码（包括 AV1 解码）、WiFi 和蓝牙等，完整的 DCP 支持和 GPU 功能尚未完成，用户暂时不要期待高性能或高能效的 3D 加速。

美国军方正在禁用设备上的广告追踪功能，防止敌人借助于购买的公开追踪数据去锁定美国士兵的位置。此前有报道称，商业追踪数据被用于锁定驻扎在中东的美军。美国陆军在一份声明中表示，Windows PC 上的广告 ID 功能早在 2021 年之前就被禁用，但 Android 和苹果移动设备上的广告 ID 则“至少从 2026 年 2 月起”才默认禁用。美国军方还在考虑对手机使用实施更严格的限制。