OrangeBot.AI Digest — 2026-07-17

Your teachable AI intern for everyday browser work

A local-first whiteboard studio for teachers and creators

See the biggest startup yappers on X/Twitter

Shared memory across all apps and LLMs. In Claude

Fast & precise 3D CAD built natively for iPad and Mac

AI sales platform for modern B2B teams

Your AI data analyst, now with ideas of its own.

The world's first open 3T-class model

Pull your calendar availability in 3 seconds

Keep any Mac window visible in a floating mini screen

The memory layer for AI agents

run virtual wind tunnel aerodynamic analysis in your browser

Private inference for coding agents

AI account executives that demo and close B2B deals

Email for AI agents. They sign up, send and reply.

Recent advances in video understanding have spanned motion, long video, and streaming interaction, driving this field toward real-world applications. Despite this progress, current open-source models remain limited in several ways. They often struggle to generalize across diverse video types, making them effective only in specific domains. High computational demands further restrict their efficiency and scalability. Moreover, most models are only partially open, with key components such as training code, strategy, or datasets unavailable, which hinders reproducibility and slows community-driven development. To address these issues, we introduce VideoChat3, a fully open, efficient, and generalist video-centric MLLM. VideoChat3 advances video understanding through two complementary designs. For efficiency, we introduce Inflated 3D Vision Transformer (I3D-ViT) and Adaptive Frame Resolution for Streaming Video Perception, which enables efficient spatiotemporal representation and reduces the cost of processing video inputs during training and inference. For effectiveness, we develop a scalable video data synthesis pipeline that curates three diverse, high-quality training datasets: VideoChat3-Academic2M, VideoChat3-LV116K, and VideoChat3-OL617K, covering general, long-form, and streaming video scenarios, improving the model's generalization across domains. By integrating these designs, VideoChat3 achieves a rare balance of broad generalization and computational efficiency. Experiments across general, long-form, and streaming benchmarks demonstrate that VideoChat3 surpasses prior open-source models with equal or larger parameter counts with only 4B parameters and higher efficiency.

Large language models are increasingly trained as interactive agents for long-horizon tasks involving multi-turn interaction, tool use, and environment feedback. Outcome-based reinforcement learning (RL) provides a practical optimization paradigm, but its sparse trajectory-level rewards offer limited guidance on intermediate decisions, leaving a supervision gap between episode-level outcomes and token-level policy learning. We propose SEED (SElf-Evolving On-Policy Distillation), a self-evolving framework that converts completed on-policy trajectories into training-time hindsight skills and distills their behavioral effect back into the policy model. SEED first fine-tunes the policy to analyze completed trajectories and generate natural-language skills that capture reusable workflows, decisive observations, or failure-avoidance rules. During RL, the current policy both collects trajectories and serves as the analyzer that extracts hindsight skills from them. Policy updates therefore improve subsequent decision making and skill analysis together, allowing hindsight supervision to evolve with the policy. SEED then re-scores the sampled actions under ordinary and skill-augmented contexts, converting the skill-induced probability shift into a dense token-level on-policy distillation signal. This signal is jointly optimized with outcome-based RL, keeping the auxiliary supervision aligned with the current trajectory distribution. Extensive experiments on text-based and vision-based agentic tasks show that SEED consistently improves performance and sample efficiency, exhibiting robust generalization to unseen scenarios. Our code is available at https://github.com/jinyangwu/SEED.

Recent advances in Tool-Integrated Large Language Models have made web search a core capability of information-seeking agents. However, as interaction histories grow, agents increasingly struggle to track task progress. When search attempts fail to yield useful evidence, current single- and multi-agent systems can become trapped in repetitive loops, wasting search budgets and ultimately compromising the quality and completeness of the final output. We introduce SearchOS, a system-level multi-agent framework that turns fragile, implicit search progress into explicit, persistent, and shared state. First, we formulate open-domain information seeking as relational schema completion with grounded citations, where agents discover entities, populate attributes across linked tables, and anchor each value to source evidence. Then we design Search-Oriented Context Management (SOCM), which externalizes the evolving state into Frontier Task, an Evidence Graph, a Coverage Map, and Failure Memory. Built on SOCM, SearchOS applies a pipeline-parallel scheduling mechanism that overlaps the execution of sub-agents and continuously refills freed slots with tasks targeting unresolved coverage gaps to improve utilization and throughput. To schedule and control the execution of search agents, SearchOS introduces a Search Tool Middleware Harness that intercepts model and tool interactions to record grounded evidence and react to stalls or budget exhaustion, and provides a reusable hierarchical skill system comprising strategy and access skills to augment the agents' search process and avoid repeating failed search patterns across runs. On WideSearch and GISA, SearchOS leads all metrics among the evaluated single- and multi-agent baselines, paving the way toward robust information-seeking collaboration.

A growing gap separates inference context lengths from RL post-training: inference systems are approaching million-token contexts, while post-training workloads often remain at 256K tokens or below and rely on length generalization at deployment. The gap is especially important for AI agents, whose observations, tool outputs, documents, and prior decisions accumulate over long trajectories. LongStraw is an architecture-aware execution stack for million-token RL post-training under a fixed GPU budget, instantiated with Group Relative Policy Optimization (GRPO). It evaluates the shared prompt without autograd, retains only model-specific state needed by later tokens, and replays short response branches one at a time, reducing the live training graph at the cost of additional replay time. We implement it for the hybrid recurrent and full-attention Qwen3.6-27B and the compressed-attention mixture-of-experts GLM-5.2. On eight H20 GPUs, LongStraw completes grouped Qwen scoring and response backward at 2.1M positions for groups of 2 and 8; increasing the group size adds only 0.21 GB of peak allocated memory, while a separate stress test reaches 4.46M positions. On 32 H20 GPUs, we validate the end-to-end LongStraw execution path for a 2.1M-token prompt across all 78 layers of GLM-5.2. These experiments establish execution capacity rather than complete training correctness because the captured prompt state is detached and some distributed forward and gradient composition paths remain incomplete.

World-action models (WAMs) are emerging as a promising foundation for embodied control: rather than predicting actions alone, they learn representations that couple action generation with future world prediction. This coupling is often viewed as a source of robustness, interpretability, and safety, as a robot's action can in principle be checked against its imagined future. In this paper, we show that this assumption is fragile. We introduce BadWAM, a unified framework for modeling and evaluating World-Action Drift Attacks: a new class of WAM-specific adversarial attacks that use small visual perturbations to break the alignment between what a WAM imagines and what it executes. BadWAM characterizes this attack surface along two natural criteria: attack strength and stealthiness. When the adversary prioritizes disruption, BadWAM instantiates an action-only adversarial attack, which directly drives the model toward task-failing actions. When the adversary additionally prioritizes stealth, BadWAM instantiates an imagination-preserving adversarial attack, which seeks to induce harmful action shifts while keeping the model's predicted future close to its clean imagination. Together, these two attacks capture a spectrum of WAM-specific failures: from overt action hijacking to stealthier cases where the model appears to imagine a plausible future but executes a desynchronized action. We evaluate BadWAM across different variants of WAMs. Results show that our attacks substantially reduce task success rates under closed-loop execution. For example, our action-only attack reduces the model performance from 96.5% to 43.1% success. The results of our imagination-preserving attack further exposes a WAM-specific vulnerability: moderate future-preserving regularization can maintain strong attack performance while reducing future imagination drift.

Video generation increasingly relies on keyframe-based workflows, where creators specify a sequence of reference images to guide generation. Although recent models support multi-keyframe conditioning, it remains unclear whether they can faithfully reproduce the prescribed keyframes while maintaining overall video quality. We present KeyFrame-Compass, the first comprehensive benchmark for evaluating keyframe-conditioned video generation. The benchmark contains 386 carefully curated samples spanning three application domains, two video structures, two prompt granularities, two conditioning formats, and four keyframe densities, enabling controlled analysis under diverse generation settings. We further introduce an automated evaluation framework that jointly measures keyframe execution and overall video quality. Specifically, we decompose keyframe execution into six complementary metrics covering presence, fidelity, temporal ordering, localization, persistence, and uniqueness, while assessing overall video quality through evidence-grounded MLLM judgments augmented with specialized perception models. Experiments on nine representative video generation systems reveal several fundamental limitations. Current models exhibit a clear trade-off between faithful keyframe execution and natural video synthesis. Their performance further degrades as keyframe constraints become denser and most open-source models also fail to interpret storyboard-grid inputs as temporally ordered keyframe sequences.

Multi-reference-to-audio-video (MR2AV) generation aims to generate coherent audio-video content conditioned on multiple references and textual instructions. Existing benchmarks mainly focus on text-driven generation, single-reference subject preservation, or isolated audio-video alignment, leaving the emerging MR2AV setting largely unexplored. Compared with these settings, MR2AV requires models to jointly reason over multiple references while generating synchronized visual and audio content. Models must not only preserve each reference faithfully but also correctly bind and compose multiple referenced entities into coherent audio-visual events. To address this gap, we introduce MultiRef-Compass, a unified benchmark for MR2AV generation. It comprises 350 carefully curated samples constructed through a scalable and controllable asset-composition pipeline, covering multi-view subject preservation, multi-entity binding, and human-object-scene composition. To provide interpretable assessment, MultiRef-Compass defines an evaluation protocol with four dimensions: Basic Quality, Reference Consistency, Audio-Visual Consistency, and Instruction Following, using 14 sub-metrics. MultiRef-Compass integrates automatic metrics with a rejudging-enhanced MLLM-as-a-Judge framework, enabling scalable and auditable evaluation of both perceptual fidelity and reference-conditioned composition. Extensive experiments on eight representative MR2AV systems reveal substantial room for improvement across multiple evaluation dimensions, underscoring the need for a comprehensive benchmark and positioning MultiRef-Compass as a foundation for future MR2AV research.

Building interactive worlds that respond coherently to player actions has long been a shared goal of computer graphics, games, and artificial intelligence. Recent video generative models provide a data-driven route toward this goal by predicting future observations conditioned on user actions, and are increasingly regarded as potential next-generation game engines. Realizing a genuinely interactive game world, however, requires interaction outcomes that follow rules over evolving game conditions, consequences that persist over long horizons, and a generation loop that operates in real time. Conventional game engines realize these properties through a recurrent action-state-observation loop, in which player actions update an explicit game state according to predefined rules and observations are rendered from the resulting state. Taking this loop as an organizing lens, this paper examines interactive game world modeling along four dimensions: player action control, game state dynamics, state-observation persistence, and real-time interactive generation. For each dimension, we start from the capabilities required by an interactive game world, group existing approaches into representative families, and discuss the strengths and trade-offs of each family. Complementing this analysis, we present a scalable data engine for Black Myth: Wukong that collects over 90 hours of gameplay with frame-aligned player actions, ground-truth game states, and visual observations, together with structured and semantic annotations, as a resource for state-aware game world modeling. We hope this paper offers a clear picture of where the field stands and fosters progress toward interactive game worlds.

Human cognition does not separate understanding and generation. A teacher at a whiteboard speaks and draws together, each modality reshapes the other. In this paper, we bring this coupled loop to artificial systems. Masked Diffusion Models (MDMs) are ideally suited to this task, yet existing samplers either decode text and image interleavedly or independently update them in parallel branches that share only previous-step history, but not the other modality's latest decisions within the same step; combined with MDMs' inability to remask, cross-modal contradictions are neither detected nor repaired. We introduce Self-Correcting Coupled Markov Jump Processes (SC-CMJP), a framework in which one modality's transition rates are functionals of the other modality's confidence score, as weighted by cross-modal attention. Furthermore, a remasking jump retracts commitments the moment cross-modal evidence turns against them. In conjunction with SC-CMJP, we introduce CO_2Jump (Self-text{CO}rrecting text{CO}upled text{Jump}), a novel training-free single-pass sampler for joint multimodal geneneration. For training and evaluation purposes, we have created and will release three large-scale joint multimodal generation corpora: JEdit-1M, JMaze-200K, JNono-200K, with matching in- and out-of-distribution benchmarks. CO_2Jump achieves best joint performance for image understanding and editing as well as visual reasoning (maze and nonogram solving). The performance of the sampler scales monotonically with the number of denoising steps, evidence that the benefits of cross-modal coupling compound across the trajectory. Project page: https://coupled-jump.github.io

Learning broad world knowledge directly from raw visual data is a fundamental capability of intelligence. We introduce UniVR, the first investigation into simultaneously learning complex reasoning, fine-grained physical dynamics, and long-term planning from pure visual demonstrations. At its core, UniVR features VR-GRPO, a reinforcement learning paradigm with complementary global and step-level rewards. This approach enforces logical coherence and physical consistency throughout the reasoning process without requiring task-specific heuristics or image-text pairs. To train and evaluate UniVR, we construct VR-X, a large-scale benchmark curated from 16 diverse sources spanning long-horizon manipulation, spatial puzzles, and physical reasoning. It is the first comprehensive suite to assess these heterogeneous capabilities under a purely visual protocol. Remarkably, UniVR achieves up to a 25% improvement on VR-X, and its superior visual reasoning also boosts performance on various multimodal understanding benchmarks. These findings underscore the vast potential of reasoning within visual spaces, with all code, data, and models are open-sourced for further research.

Reinforcement learning has become a standard post-training recipe for large language models, but dense full-parameter updates create two deployment-relevant bottlenecks: suppressed reasoning performance, often reflected by premature saturation of test-time scaling, and interference when consolidating multiple capabilities through multi-domain training or model merging. We show that the reasoning-effective component of these updates is largely concentrated in the base model's spectral space, motivating Subspace-Aligned Rewiring (SAR), a post-hoc editing method that retains this spectral core while removing orthogonal components. SAR therefore preserves reasoning gains and filters residual update directions that suppress performance or amplify cross-domain interference. Across several model families and scales, SAR extracts compact reasoning cores using as little as approximately 0.58% of total parameters: it preserves over 99% of post-training performance and improves high-k exploration in mathematical reasoning, and generalizes to agentic coding by improving six of seven open benchmarks on an in-house model. SAR also purifies mixed-domain training updates by releasing suppressed coding capability while maintaining math reasoning and instruction following. It further enables model merging across experts, yielding cross-domain generalization that surpasses previous merging baselines and even the best single-domain experts. Overall, SAR shows that extracting reasoning-effective updates from parameter geometry can serve as a training-free mechanism to improve reasoning and multi-domain performance.

We present Wan-Streamer v0.3, which reframes our native-streaming interaction model under a single organizing view: a video is a world plus an event stream. The world is the persistent context in which a video unfolds, including the environment, scene, subjects, ambient acoustic conditions, voice characteristics, and other relatively stable conditions. The event stream is everything that changes over time within that world, including scene or environmental changes, subject behavior, speech, and other sounds. This yields a general-purpose pretraining task over large amounts of real video: given a world and incoming input, predict how the world moves, changes, and responds in real time. The resulting competence can be specialized to a broad family of real-time downstream tasks. We instantiate it on real-time full-duplex audio-visual interaction, where the event stream is the agent's speech together with free-form behavior. Functionally, the model's multimodal understanding process is vision-language-action-like: it maps multimodal user input to language-form speech and behavior actions. Wan-Streamer v0.3 preserves the v0.2 operating point: 640x368 video at 25 FPS, a 160 ms streaming unit, approximately 200 ms model-side response latency, and approximately 550 ms total interaction latency under a 350 ms bidirectional network budget.

On-policy distillation (OPD) has become a key paradigm in LLM post-training, yet its training dynamics remain poorly understood. We present a systematic study examining the role, pathologies, and regulations of OPD. We first clarify the role of OPD as an exploration catalyst: it steers the student toward correct reasoning paths via dense token-level guidance, without expanding capability ceiling. We confirm this by showing that prompt diversity matters more than per-problem sampling numbers, and critically, that the effectiveness of OPD hinges entirely on the quality of its guiding signal. This dependency exposes two pathologies that derail exploration. The Student-Teacher Mismatch occurs when a large teacher-student distributional gap causes the guiding signal to misalign with task correctness, steering exploration in counterproductive directions. Length Exploitation arises when the aggregated token-level objective creates length-dependent shortcuts, allowing the student to game the reward landscape through response truncation or redundant padding, exploring degenerate length modes rather than reasoning strategies. To tame these pathologies, we investigate lightweight signal regulations: advantage clipping and log-scale compression, ensuring exploration is guided by faithful signals. Experiments across seven benchmarks demonstrate that these regulations alleviate length exploitation and enable effective distillation, stably surpassing OPD variants and RLVR baselines, thereby confirming that well-regulated signal quality, rather than mere teacher scale, governs successful exploration in OPD.

Recent robot foundation models operate with single-step or short-history visuomotor context. We introduce Test-Time-Training Robot Policies (RoboTTT), a robot model and training recipe that scale visuomotor context to 8K timesteps, three orders of magnitude beyond state-of-the-art policies, without growing inference latency. At this context length, we unlock new robot capabilities: one-shot in-context imitation from human video demonstrations, on-the-fly policy improvement, robustness to perturbations, and stronger performance on multi-stage, long-horizon tasks. We also observe, for the first time, steady gains in closed-loop performance as pretraining context length scales. At its core, RoboTTT integrates Test-Time Training into robot foundation models such as Vision-Language-Action policies, yielding a sequence model whose recurrent state consists of fast weights, parameters updated by gradient descent during both training and inference, compressing histories into weight space and retrieving contextual information for long-context conditioning. To scale training context length, the recipe combines sequence action forcing with truncated backpropagation through time. On challenging real-robot manipulation tasks, RoboTTT improves overall performance by 87% over the single-step context baseline and fully completes a five-minute, ten-stage assembly task, which no baseline ever does. RoboTTT trained with 8K-timestep context outperforms the same model pretrained with 1K timesteps by 62%, suggesting context length as a new scaling axis for robot foundation models. Videos are available at https://research.nvidia.com/labs/gear/robottt/

MeanFlow generators achieve fast few-step sampling by predicting average velocities over time intervals, making them attractive for efficient generation. Reinforcement learning (RL) has become a powerful way to align diffusion and flow models with human preferences and task-specific objectives. In particular, DiffusionNFT offers an efficient forward-process RL framework that does not require reverse-process trajectories or likelihood estimation. However, applying such RL methods to MeanFlow remains underexplored. DiffusionNFT optimizes instantaneous velocities, whereas MeanFlow samples with average velocities. To bridge this gap, we introduce MeanFlowNFT. Inspired by the MeanFlow identity, which bridges average and instantaneous velocities, we construct an induced instantaneous-velocity predictor. We apply the DiffusionNFT objective to this predictor, making reward optimization well-defined for MeanFlow. Sampling remains based on the average velocity, preserving MeanFlow's fast few-step generation. We further prove that MeanFlowNFT inherits DiffusionNFT's strict policy-improvement guarantee. Experiments on image and video generation show that MeanFlowNFT consistently improves baselines. Moreover, it outperforms prior state-of-the-art RL-tuned few-step generators on most metrics (6 of 8 on SD3.5-M), and can even surpass multi-step RL-tuned diffusion while using only a few sampling steps. For instance, on Wan 2.1, 4-step MeanFlowNFT reaches a VBench score of 84.33, surpassing 50-step LongCat-Video RL (82.57).

Reuters : The Philadelphia Semiconductor Index sank ~10% for the week, its largest weekly fall in over a year, and is now down ~20% from its late-June all-time high —  A brutal week for chip stocks — the same names that fueled this year's blistering market rally — has left investors from Seoul …

The Information : Sources: Valar Atomics, which makes small nuclear reactors intended to power data centers, is in talks to raise $1B at a ~$5B pre-money valuation —  Valar Atomics, a three-year-old startup that makes small nuclear reactors intended to power data centers and other industrial facilities …

Joe Rossignol / MacRumors : Apple raises iPhone prices in Japan by up to 11%, likely due to the Japanese yen's depreciation against the US dollar over the past year —  After raising prices on all Macs and iPads worldwide last month, Apple has now increased the price of iPhone models in Japan.

Robert Booth / The Guardian : Amazon apologizes after some AWS users received bills as high as $1.5T due to “an issue with unit pricing within the estimated billing computation subsystem” —  One UK man whose bill is usually less than £1 says he ‘almost had a heart attack’ when he saw £5.8bn invoice

The Information : Sources: OpenRouter has discussed a potential sale to a bigger tech company that could value it at billions of dollars, a premium to its $1.3B valuation in May —  OpenRouter, which helps app developers access hundreds of AI models, has discussed a potential sale to a bigger tech company …

Wall Street Journal : Sources: SpaceX is in talks with the DOD about providing the agency with access to data-center capacity worth billions of dollars for running AI models —  Elon Musk's company would sell military billions of dollars' worth of data-center capacity for running AI models

Chance Miller / 9to5Mac : Apple raises the price of Apple Music, with the individual plan up by $1 to $11.99, and some Apple One bundles, citing rising licensing costs —  Apple has raised the price of Apple Music today in the United States and other countries.  Starting today, Apple Music's Individual plan now costs $11.99 per month, up from $10.99.

Bloomberg : Sources: Apple and the DOJ are in early discussions about settling a 2024 antitrust lawsuit alleging that Apple violated antitrust laws —  Apple Inc. and the US Justice Department are in early discussions about settling a 2024 lawsuit that alleges the iPhone maker violated antitrust laws.

New York Times : Sources: Meta is in talks to rent computing power from its data centers to Anthropic in a deal that could be worth ~$10B over two years —  A deal would underline how scarce computing power is for artificial intelligence development, and could create a new business for Meta.

Gavin Baker / @gavinsbaker : Models like Kimi K3, Grok 4.5, and Muse 1.1 may prevent the dominance of 2-3 frontier labs with 90% inference margins from hurting other AI ecosystem layers —  Kimi K3 may be an important inflection point for AI. Potentially negative for Anthropic and OpenAI while being net positive for essentially every other company in the world. I mean that very literally. Although the real “Sputnik moment” would be an open-source frontier model that [image]

Transformer : Kimi K3 hype shouldn't alarm the US about “losing the AI race” to China; K3 is a good model but not frontier-level and likely lacks dangerous cyber capabilities —  Transformer Weekly: NY data center moratorium, NDAA export controls and Amodei's $1m to safety super PAC

Sarah Jacob / Bloomberg : ASML plans to give its ~45K employees globally a €20K bonus, issued as a share award that vests in 2030, joining other chip industry companies offering payouts —  ASML Holding NV will give employees globally a one-time €20,000 ($22,862) bonus, joining other firms in the chip industry …

Tim Fernholz / TechCrunch : AI inference startup General Compute gets a $400M loan from tech investment firm Upper90, seemingly the first deal to use inference-specific chips as collateral —  General Compute, an AI inference cloud startup, has landed a $400 million loan from Upper90, a tech investment firm.

Foo Yun Chee / Reuters : Sources: the EU is set to approve the $55B acquisition of EA under its subsidy rules on July 30; the deal includes Saudi Arabia's PIF, Silver Lake, and Affinity —  A group of investors including Saudi Arabia's Public Investment Fund is set to secure European Union approval for its $55 billion acquisition …

Matt Burgess / Wired : San Francisco sends legal notices to Apple and Google, demanding they take down 13 AI apps used to make deepfake nude images; Google says it deleted five apps —  The City Attorney's Office sent the tech giants cease-and-desist letters this week telling them to stop profiting from 13 …

危地马拉舒尔通玛雅遗址一面墙壁上镌刻的数学公式，让学界首次确认了一位重要的玛雅数学家兼天文学家的真实姓名。研究团队指出，这位名为 Sak Tahn Waax（意为“白胸狐狸”）的学者足以比肩人类历史上的数学巨匠。 墙壁绘有人物画像与象形文字，其中包含基于天文历法的精密数学运算。玛雅人曾依靠这些运算确定国王登基等重大仪式的举办时间。研究团队重点分析了编号为“19号文本”的一组象形文字。“19号文本”是一组呈L形排布的11个象形文字，整体高度约10厘米。研究发现，前 9 个象形文字完整记录了玛雅历法与天文周期的换算逻辑。这套公式破解了一个2920 天周期的拆分规律，可适配玛雅各类历法单位。2920 天是玛雅文明的核心天文周期，完美契合5个金星周期（每周期584天）与8个太阳年（每周期365天）。不仅如此，该运算还将2920天与乌伊纳尔（每月20日）、卓尔金历（260天神圣历法）、通年（每年360天）以及780天火星周期建立了精准数理关联。 团队在“19号文本”倒数第二个符号中破译出“如是说”的句式，其后紧跟的最后一个象形文字便是署名 Sak Tahn Waax，代表该学者是这套运算公式的创作者。从铭文缺失女性专属前缀可判定，这是一位男性学者。

天文学家直接探测到系外行星 LHS 1140 b 大气层中的氦气。LHS 1140 b 是一颗岩石行星，距离地球 48 光年，位于其母星的宜居带。这是首次在宜居带内的类地球岩石行星确认探测到大气层。LHS 1140 b 围绕着一颗比太阳更小更冷的红矮星运行，其轨道比日地距离更近，温度适宜，其表面可能存在液体水，可能有铁质内核。就对地球的生命研究而言，液态水是生命存在的必要条件。论文第一作者 Collin Cherubim 表示需要进一步研究去确认行星是否存在水。

在今天的意大利漫步，我们仍然能看到有近两千年历史的混凝土建筑。相比下，现代混凝土建筑在百年内就会坍塌。为什么罗马时代的混凝土耐用性如此好？科学家认为要归功于名为“火山灰反应”的关键化学过程——即火山灰与石灰和水发生反应。根据发表在《Science Advances》上的一项研究，名为碳化的反应也有助于增强混凝土的耐久性。研究人员从有 1900 年历史的哈德良(Hadrian)皇帝庄园的马桶座圈上收集了混凝土样本，用高倍显微镜观察，用 X 射线扫描分析其化学成分。研究发现，样本含有火山灰、石灰和水等材料的证据，但对混凝土孔隙和裂缝的观察显示，方解石是其主要的粘结剂。当大气中的二氧化碳与混凝土中的钙化合物发生反应时，会形成坚硬的方解石矿物，它含有大量的碳酸钙。方解石填充了混凝土中的细小裂缝和孔隙，使古老的建筑结构能随着时间的推移而加固和修复。

黑胶唱片过去十年一直是实体音乐复兴的代表，但今年上半年美国 CD 唱片销量涨幅远超黑胶唱片。CD 销量飙升 16% 达到 1630 万张，相比下黑胶唱片销量上涨 2.4%。韩国 K-pop BTS 热门专辑《ARIRANG》在 CD 销量增长中起到了重要作用，但排除 K-pop 之后，CD 销量仍然同比增长了 6.7%。美国包括 LP、CD 和磁带的实体专辑总销量上半年增长了 7.8% 达到 3820 万。实体音乐复兴的一大原因是年轻一代听歌习惯发生了变化，六成 Z 世代听众表示最常听的是音乐是 1990 年代或更早期的，相比下 2021 年这一比例仅为 18%。无论是怀旧、通过流媒体发现新音乐，还是想要拥有艺术家作品的实体唱片，年轻听众在拥抱实体唱片。

研究团队结合了地面大型巡天计划 LAMOST DR7 的光谱观测，与 ESA 盖亚太空望远镜的高精度视差资料，建立起包含 155,600 颗邻近太阳次巨星（Subgiant stars）的庞大基准样本。由于这类恒星形成于银河系早期，其化学组成几乎保留了最原始的样貌，是推算时间极佳的恒星化石。团队利用马可夫链蒙地卡罗（MCMC）算法重建真实年龄分布，推导出样本中最古老恒星的真实年龄为 137.3（+1.8 / -1.5）亿年，若考虑大爆炸后约 2 亿年才形成首批长寿命恒星，这与宇宙微波背景辐射预测的 138 亿年宇宙年龄完全相符。最新结果显示，目前并没有可信的观测证据支持宇宙年龄超过 138 亿年。

微软宣布开源 Comic Chat，源代码采用 MIT 许可证托管在 GitHub 上。Comic Chat 是一款能自动将 IRC 中的对话转换为漫画格形式的聊天客户端，其中包含插图人物、对话气泡和表情。它的一个目的是帮助世界认识 Comic Sans 字体。Comic Sans 字体最早由微软字体设计师 Vincent Connare 于 1994 年设计，在 Comic Chat 中找到了它真正的家。它非正式的手写风格与软件的对话气泡完美匹配。Comic Chat 最早于 1995 年开发，1996 年随 Internet Explorer 3 推出。开源 Comic Chat 的一个意图是保存软件历史，以及让开源社区在此基础上探索、学习和二次开发。

月之暗面宣布了 2.8 万亿参数开放权重模型 Kimi K3，完整的权重将于 7 月 27 日发布。月之暗面称，Kimi K3基于 Kimi Delta Attention 和 Attention Residuals 构建，参数规模 2.8 万亿，具备原生视觉功能和 100 万个词元上下文窗口。它是全球首个 3 万亿参数级别的开放权重模型，专为长程编码、知识工作和推理等前沿智能领域而设计。在基准测试中，Kimi K3 整体性能仅落后于 Claude Fable 5 和 GPT-5.6 Sol。Kimi K3 现已在 Kimi.com、Kimi Work、Kimi Code 和 Kimi API 上线。发布初期 Kimi K3 将默认采用最大思考强度模式，低强度和高强度模式将在后续更新中推出。

从 7 月 22 日起，美国第三方应用商店将可以通过 Google 的应用商店发行其客户端。Google 公布了“加入 Google Play 目录访问权限计划”的文档，美国第三方应用商店可以访问 Google Play 商店的应用目录。届时，美国第三方 Android 应用商店将能向用户提供这些应用，而用户仍将通过 Google Play 完成下载，其下载条款与直接在 Google Play 商店下载完全一致。通过此类方式下载的应用将需要继续支付 Google Play 服务费。Google 还将要求上架的应用商店屏蔽恶意软件、尊重知识产权，提供应用更新和卸载机制。如果超过 1% 的应用安装尝试疑似恶意软件或不需要的软件，则该应用商店可能会被移除。

根据发表在《European Heart Journal》上的一项研究，相比心脏供血正常的人，心脏病发作患者血液内的微塑料和纳米塑料含量更高。研究还发现，吸烟者和暴露于较高空气污染水平人群血液中微塑料和纳米塑料含量也更高。研究人员分析了 61 名意大利患者，采集了其血液样本，调查了他们是否吸烟以及空气污染情况。结果显示，84% 的心脏病发作患者血液中检测到微塑料和纳米塑料，慢性缺血性心脏病患者为 40%，冠状动脉正常的患者为 32%。心脏病发作患者血液中塑料种类更多。最常见的塑料类型是聚乙烯，这种塑料常用于包装和消费品。长期暴露于较高 PM2.5 水平的患者血液中更容易检测到微塑料，吸烟者血液中微塑料的检出率是不吸烟者的六倍。

一加确认退出美国和欧洲市场，表示将会继续为现有用户提供软件更新，但一加手机搭载的 OxygenOS 系统将被母公司 OPPO 的 ColorOS 系统所取代。作为关闭全球业务计划的一部分，一加手机的 OxygenOS 也将随之关闭。所有在售一加手机运行的操作系统从 Android 17 更新开始将逐步迁移到 ColorOS，一加称此举有助于简化软件开发流程，加快更新推送速度，提升软件质量，更好地利用共享的工程和研发资源。对于不会更新到 Android 17 的旧型号设备，一加将会提供操作系统的维护支持，但新设备需要更新到 ColorOS 才会获得所有形式的支持。如果客户想要旧的 OxygenOS 使用体验，在更新到 ColorOS 之后可选择回滚。

xAI 在 GitHub 上公开了其辅助编程智能体 Grok Build 的源代码，此举可能是某种重新赢得用户信任的补救措施。因为在这之前它被发现存在严重的隐私安全问题，会上传用户的完整代码库。Grok Build 被发现在读取或处理文件时，该文件的内容未经任何编辑就被传输到 xAI 使用的 Google Cloud Storage 中。Grok Build 的数据保留远超 Claude Code、Gemini 和 Codex 等类似工具。有 Grok Build 用户报告，包含 SSH 密钥、密码管理器数据库等的完整用户目录都被上传了。Elon Musk 表示该公司将彻底删除此前上传到服务器上的用户数据。

发表在 PLOS One 上的一项新研究发现，马不依赖声音、气味或过往经验就能识别屏幕上的捕食者。传感器显示，马看到屏幕上的狼后心跳会加速，但面无表情十分镇定。研究显示，马没有摇头，没有摇尾巴，其目光也没有锁定屏幕以表明大脑在处理威胁信息。论文第一作者 Zeynep Benderlioglu 称，马在评估潜在威胁时展现出惊人的认知克制而不是惊吓。在实验中，马看到屏幕上的袋熊时心率平稳，但看到屏幕上的狼——无论狼是在攻击还是梳理——时其心率会显著加快，其中雄性马的反应更强烈。Benderlioglu 说，马的认知处理能力出乎意料的高。它们高度警惕，但威胁并没有出现，所以它们没有表现出任何异常行为，它们正在进行认知评估。”

大型图书出版商 Hachette、Cengage 和 Elsevier 以及作家 Scott Turow 指控 Google 在训练 Gemini 模型过程中未经许可使用了数百万受版权保护图书，声称这是历史上最严重的版权侵犯事件之一。出版商称，Google 挪用了用于 Google Books、Google Play Books 和 Google Scholar 等服务的图书，这些服务允许 Google 以特定方式使用相关图书——如显示可搜索的片段或销售电子书，但无权将这些图书用于训练商业 AI 产品。诉讼书称：“为了维持其在互联网领域的统治地位，Google 放弃了其早期‘不作恶’的座右铭，犯下了历史上最严重的版权侵权事件之一。”诉讼书称，Google 内部早已认识到该问题，它可能会面临“100-1000 亿美元的潜在罚款”，但仍然未经许可将这些图书用于训练 Gemini。

深度求索（DeepSeek）计划启动新一轮融资，目标融资高达 500 亿元人民币，估值高达 4800 亿人民币。该公司也已开始就可能在上交所科创板上市进行初步探讨。该公司的内部目标是计划在今年提交 IPO 申请。今年 6 月，DeepSeek 完成了成立以来的首轮外部融资，募资总额逾 500 亿元人民币，融资后估值约为 3380 亿人民币。首轮融资中，梁文锋个人出资 200 亿元人民币，腾讯控股和电池巨头宁德时代分别出资 100 亿元人民币和 50 亿元人民币，成为最大的外部股东。京东、网易及 IDG 资本各出资 30 亿元，国家人工智能产业投资基金出资 10 亿元。DeepSeek 如此迅速的融资节奏，源于其对资本支出的预期增加。该公司计划建设自有数据中心并采购更多 AI 芯片。

对部分戴尔电脑用户而言，Windows 的周二例行安全更新之后紧跟着的是周三的电脑出问题了。微软确认，由于意外关机、性能下降、发热增加和电池耗电过快，它停止向部分配备英特尔处理器的戴尔电脑推送更新。但微软和戴尔尚未披露受影响设备的型号。微软只是表示正与戴尔合作，防止受影响型号出现问题，计划在未来几天内发布针对受影响设备的解决方案。