OrangeBot.AI Digest — 2026-06-20

Control exactly who can access your company data

Work across 20+ apps in Slack with multiplayer collaboration

a small hosted memory layer for AI agents

Collaborative, proactive AI coworkers who work in Slack

Visual GitHub management directly from a graphical interface

AI video studio: text & image to video, with real control

AI support agent, help center & changelogs in a single inbox

Find out if you live forever in the brain of the LLMs

Build a room of ambient sound that lives in your notch

Surf like it's 1999

Fast LLMs for low-latency and high-performance workflows

60 second ultrasound-based full-body scanner that beats MRI

One API for WhatsApp: messaging, calling, and AI agents

An index for agents pushing the frontier of AI/ML research

Automate your entire backlink generation

While 10B-level industrial foundation models have pushed the boundaries of image inpainting, their prohibitive computational costs severely hinder practical deployment. Constructing a highly optimized task-specific specialist offers a promising solution; however, extreme structural compression inevitably triggers a severe representation bottleneck. To conquer this, we propose Moebius, a highly efficient lightweight inpainting framework. We systematically reconstruct the diffusion backbone by introducing the Local-λ Mix Interaction (LλMI) block. Comprising Local-λ and Interactive-λ modules, it elegantly summarizes spatial contexts and global semantic priors into fixed-size linear matrices, preserving complex latent interactions while drastically shedding parameters. Furthermore, to unlock the full representational capacity of this highly compact architecture, we synergistically pair it with an adaptive multi-granularity distillation strategy. Operating strictly within the latent space to avoid expensive pixel-space decoding, this strategy dynamically balances multiple gradient-based losses to achieve high-fidelity alignment. Extensive experiments across natural and portrait benchmarks demonstrate that this optimal synergy enables Moebius to rival or even surpass the generation quality of the 10B-level industrial generalist FLUX.1-Fill-Dev. Remarkably, Moebius achieves this using less than 2\% of the parameters (0.22B vs. 11.9B) while delivering a >15times acceleration in total inference time, setting a new efficiency standard for high-fidelity inpainting. Project page at https://hustvl.github.io/Moebius.

Dexterous interaction with articulated objects is important for household, assistive, and humanoid manipulation, where multi-finger hands can provide compliant contact patterns beyond parallel-jaw grasping. However, articulated-object manipulation differs from static-object manipulation: the target part cannot be directly actuated, and its motion must emerge through sustained physical hand--handle contact. This makes the transition from object-centric articulated generation to hand-driven dexterous hand--object interaction non-trivial, since geometric trajectory replay or open-loop execution does not model the contact dynamics required to move the articulated part. Moreover, policies trained only for task completion under fixed dynamics can overfit nominal contact loads, especially without tactile or force feedback, and may degrade when the contact load changes. To address these challenges, we present DragMesh-2, a contact-driven framework for dexterous interaction with articulated objects that extends articulated interaction from object-centric generation to hand-driven dexterous hand--object interaction, where articulated motion must arise through physical contact. We further propose PICA, a physically informed contact-aware training mechanism that injects physical signals into policy learning without tactile or force feedback, improving robustness and task success under changing contact loads. Finally, we conduct systematic evaluation across multiple damping conditions and articulated-object categories to study robustness under contact-load variation, and provide a pure-geometry dexterous interaction resource to support future loco-manipulation and humanoid hand--object interaction research. Across seven GAPartNet objects, DragMesh-2 achieves stronger robustness under contact-load variation than the compared methods while maintaining high task success across damping conditions.

Current agentic robot systems can write executable Code-as-Policy programs, observe feedback, and revise behavior across multiple attempts, but they remain largely task-driven: reusable skills are acquired only after explicit instructions. We study Playful Agentic Robot Learning, where an embodied coding agent uses self-directed play as a continual skill-learning stage before downstream tasks arrive. We introduce RATs, Robotics Agent Teams designed for play-time skill acquisition. During play, RATs proposes novel yet learnable exploratory tasks, plans and executes robot-code policies, verifies intermediate progress, diagnoses failures, retries with dense, step-level feedback, and distills successful executions into a persistent code skill library. At test time, the agent reuses relevant skills from this frozen library to help solve new tasks. Experiments in LIBERO-PRO and MolmoSpaces show that play-learned skills improve held-out downstream tasks over no-play and random-play baselines, with 20.6 and 17.0 percentage-point gains over CaP-Agent0 on LIBERO-PRO and MolmoSpaces, respectively. Moreover, the learned skills can be plugged into other inference-time Code-as-Policy agents by simply retrieving them into the context, improving RoboSuite and real-world transfer by 8.9 and 8.8 points, respectively, without finetuning the underlying model.

LiveCodeBench (LCB) has recently become a widely adopted benchmark for evaluating large language models (LLMs) on code-generation tasks. By curating competitive programming problems, constantly adding fresh problems to the set, and filtering them by release dates, LCB provides contamination-aware evaluation and offers a holistic view of coding capability. However, LCB remains restricted to Python, leaving open the question of whether LLMs can generalize across the diverse programming languages required in real-world software engineering. We introduce Multi-LCB, a benchmark for evaluating LLMs across twelve programming languages, including Python. Multi-LCB transforms Python tasks from the LCB dataset into equivalent tasks in other languages while preserving LCB's contamination controls and evaluation protocol. Because it is fully compatible with the original LCB format, Multi-LCB will automatically track future LCB updates, enabling systematic assessment of cross-language code generation competence and requiring models to sustain performance well beyond Python. We evaluated 24 LLMs for instruction and reasoning on Multi-LCB, uncovering evidence of Python overfitting, language-specific contamination, and substantial disparities in multilingual performance. Our results establish Multi-LCB as a rigorous new benchmark for multi-programming-language code evaluation, directly addressing LCB's primary limitation and exposing critical gaps in current LLM capabilities.

Real-world spatial intelligence requires reasoning over a continuous and evolving 3D world, yet existing VLMs and tool-augmented agents largely remain tied to static, stateless inference from isolated visual observations. We introduce \textsc{S-Agent}, a spatial tool-use agentic paradigm for understanding and reasoning over continuous multi-view images and videos. By formulating spatial reasoning as spatio-temporal evidence accumulation rather than isolated frame-level prediction, S-Agent reshapes spatial perception into scene-centric understanding beyond frame-centric recognition. Specifically, S-Agent casts the VLM as a semantic planner that decides what evidence is needed, while a hierarchy of spatial tools and experts grounds objects in 2D, lifts them into 3D geometric evidence, and aggregates this evidence into high-level spatial knowledge (e.g., counting, measurement, orientation, and relative position). Additionally, a temporal memory mechanism, including Scene Memory for maintaining the evolving scene state and Agent Memory for accumulating reasoning context, enables evidence integration across frames and reasoning steps. Comprehensive experiments on multi-view and video spatial reasoning benchmarks show that S-Agent consistently improves both open-source and closed-source VLMs in a training-free manner. Beyond inference-time augmentation, supervised fine-tuning (SFT) on S-Agent-generated spatial trajectories S-300K yields S-Agent-8B, a compact spatial agent that significantly surpasses similar-scale baselines (e.g., Qwen3-VL-8B) and performs comparably to advanced closed-source models (e.g., GPT-5.4 and Gemini 3).

Advances in radiance fields have enabled photorealistic novel view synthesis. In several domains, large-scale real-world datasets have been developed to support comprehensive benchmarking and to facilitate progress beyond scene-specific reconstruction. However, for distractor-free radiance fields, a large-scale dataset with clean and cluttered images per scene remains lacking, limiting the development. To address this gap, we introduce DF3DV-1K, a large-scale real-world dataset comprising 1,048 scenes, each providing clean and cluttered image sets for benchmarking. In total, the dataset contains 89,924 images captured using consumer cameras to mimic casual capture, spanning 128 distractor types and 161 scene themes across indoor and outdoor environments. A curated subset of 41 scenes, DF3DV-41, is systematically designed to evaluate the robustness of distractor-free radiance field methods under challenging scenarios. Using DF3DV-1K, we benchmark nine recent distractor-free radiance field methods and 3D Gaussian Splatting, identifying the most robust methods and the most challenging scenarios. Beyond benchmarking, we demonstrate an application of DF3DV-1K by fine-tuning a diffusion-based 2D enhancer to improve radiance field methods, achieving average improvements of 0.96 dB PSNR and 0.057 LPIPS on the held-out set (e.g., DF3DV-41) and the On-the-go dataset. We hope DF3DV-1K facilitates the development of distractor-free vision and promotes progress beyond scene-specific approaches. The dataset and leaderboard are available at https://johnnylu305.github.io/df3dv1k_web/.

Agent benchmarks are growing fast, but no single benchmark touches more than four or five of the dimensions that deployment exposes. This paper aggregates the largest coordinated deep-dive of one MCP-based industrial-agent benchmark to date: fourteen parallel implementation studies covering new asset classes (including a multi-modal visual extension), alternative orchestrations, retrieval strategies, reasoning modes, infrastructure optimizations, and evaluation-methodology probes. Consolidating those studies with seven prior agent benchmarks, we argue that aggregate-score leaderboards systematically underspecify deployed-agent evaluation. Rankings derived from aggregate scores do not transfer to out-of-distribution settings; recent public-to-hidden competition retrospectives provide direct empirical evidence of this rank instability. We propose ranking configurations by predictive validity, the correlation between in-sample and out-of-sample rank, rather than in-sample mean, and report a twelve-tier measurement apparatus that exposes the deployment-relevant dimensions HELM and its agent-era successors collapse. The position is operationalized through three falsifiable out-of-distribution criteria with explicit thresholds; existing evidence partly supports it but is too thin to confirm. We close with a pre-registered pilot design and a field-level vision for what the next generation of agentic benchmarks should report.

Style-content dual-reference generation aims to synthesize an image that preserves the structure and semantics of a content reference while adopting the style of a separate style reference.Despite recent progress, this setting remains challenging because models must balance content fidelity, style alignment, and instruction following avoiding semantic leakage from the style reference.A key bottleneck is the lack of large-scale triplet data with clean content-style separation and broad long-tail style coverage.In this work, we propose FreeStyle, a scalable dual-reference generation framework based on community LoRA mining.We treat community LoRAs as compositional anchors for style and content, and design a rigorous generation and filtering pipeline to construct large-scale Style-Reference and Content-Reference triplets across multiple base models.To address content leakage, we adopt a two-stage curriculum with stage-specific disentanglement mechanisms: an attention-level enrichment constraint that suppresses style-reference leakage in the style-transfer stage, and a frequency-aware RoPE modulation strategy that targets positional-correspondence-based leakage in the harder dual-reference stage.We also introduce a benchmark covering both style-reference and dual-reference generation, with evaluations on style similarity, content preservation, aesthetics, instruction following, and leakage rejection. The benchmark incorporates a style-invariant Content Alignment Score (CAS) and introduces a calibrated VLM-based Rejection Score for evaluating generation reliability and leakage suppression.Extensive experiments show that our model achieves a strong balance among style alignment, content preservation, and leakage suppression.

Conditional diffusion and flow models routinely fail to satisfy the very constraints that define their task. For instance, a depth-conditioned model often produces images whose re-extracted depth disagrees with the input, even though the forward operator--the depth predictor defining the constraint--is available during both training and inference. Existing approaches generally fall into two categories: supervised models that treat the conditioning signal as a static cue and ignore alignment information at inference, and guidance-based methods that consult it through hand-tuned linear updates, typically trading fidelity to the condition against the plausibility of the generated sample. We argue that the fundamental gap in both paradigms is that the model is never trained to utilize its own alignment error. We introduce FlowBender, a closed-loop framework that treats this error as a first-class input, training the network to learn a correction policy conditioned on inference-time feedback. At each step, an unguided look-ahead pass estimates the clean signal, a task-specific deviation is computed via the forward operator, and a refinement pass consumes this signal to produce a corrected velocity. We propose several variants of FlowBender, including a gradient-based formulation for differentiable operators and a zero-order variant for non-differentiable settings such as JPEG compression. For efficient sampling, we introduce a prior-step shortcut that enables closed-loop correction at a minimal additional computational cost. Across image-to-image translation, restoration, and 3D mesh texturing, FlowBender consistently outperforms standard supervised baselines, alignment-loss-augmented training, and state-of-the-art inference-time guidance, improving fidelity and plausibility simultaneously rather than trading them against each other. Project page: https://flow-bender.github.io/

Creating 3D visual illusions, a single 3D mesh that reveals entirely different semantics from various viewing angles, is a fascinating but tough challenge. Existing optimization-based methods are slow and can produce oversaturated colors. In contrast, naive stitching approaches fail to produce geometrically coherent objects. This results in visible unnatural seams and semantic leaks. In this paper, we present a fast and training-free framework for generating text-driven 3D visual illusions. Our approach decouples the generation into two stages. First, we propose a cross-space dual-branch denoising process. This process dynamically decodes 3D latents into voxel space for CLIP-guided orientation alignment and Signed Distance Field (SDF) blending, which ensures seamless geometric fusion. Second, we introduce a view-conditioned texture synthesis module that projects and aggregates view-specific 2D diffusion priors onto the fused geometry. Extensive experiments demonstrate that our method generates highly realistic, dual-semantic 3D illusions in just 3-5 minutes. It significantly outperforms existing methods in geometric integrity, semantic recognizability, and efficiency. Project page: https://siang1105.github.io/JanusMesh.github.io/

World Action Models (WAMs) commonly rely on video generation to bridge visual world modeling and robot control. However, video-based WAMs face three coupled limitations: dense multi-frame future tokens make inference costly, full video prediction spends capacity on action-irrelevant temporal and appearance details, and long-horizon future imagination may introduce errors that mislead action prediction. These issues raise a simple question: Does world action model really need video generation? We propose ImageWAM, a simple WAM framework that repurposes pretrained image editing models for robot action prediction. In contrast to video generation, image editing provides a better-matched prior: it only needs to model a target-frame transformation, focuses on action-relevant current-to-target visual differences, and grounds task instructions to localized visual changes through edit pretraining. In practice, ImageWAM does not decode the target frame at inference time; instead, it conditions a flow-matching action expert on the KV caches produced by image-editing denoising, using them as a compact world-action context. ImageWAM outperforms standard VLA baselines and matching competitive WAMs without additional policy pretraining across different simulator and real-world experiments. It also reduces FLOPs to 1/6 and latency to 1/4 of video-based WAMs. Attention analysis further shows that editing caches focus on task-relevant change regions, supporting image editing as an effective alternative to video-based world-action modeling.

World models are increasingly regarded as a decisive step toward artificial general intelligence, yet modeling the physical world demands more than rendering convincing frames on demand: it requires an internal world state that keeps evolving over time, decoupled from observation, so that objects endure and events run to their conclusions whether or not a camera is watching, much as the moon holds to its orbit when no one is looking. This requirement is a blind spot of existing benchmarks, which reward surface properties such as fidelity, motion, and camera controllability while never asking whether a generated world keeps evolving once it is unobserved. We introduce WRBench, the first systematic diagnostic benchmark that treats camera motion as an intervention on observability and resolves evaluation into a human-calibrated chain that asks whether the camera executes the requested interaction, whether the scene stays continuous and identifiable while in view, and whether a returning target remains consistent with the event that was set in motion. Across 9{,}600 videos from 23 models spanning four control paradigms, one finding proves stubborn: current systems maintain the observed world as a tracking shot, resuming a returning target in the state at which it was abandoned rather than advancing the event while it went unseen. Because this failure recurs across control paradigms, model families, and increments of scale, robust world-state evolution does not follow from cleaner imagery, tighter control, richer geometric priors, or sheer parameter count We therefore argue that the stability of the physical state kernel and the consistency of worldlines under viewpoint intervention should become first-class objectives of world-model design, so that a world model captures how the world will unfold rather than how the next frame appears.

Visual thinking should not only sound right; it should show its evidence. While recent vision-language models (VLMs) can produce natural-language reasoning traces, these traces often leave the supporting image regions implicit, making them hard to verify and difficult to supervise. We introduce visually grounded thinking, a reasoning process in which models interleave natural-language thoughts with explicit point or box groundings of the visual evidence used at each step. This lets the model express intermediate reasoning in language while grounding key objects in the image regions they refer to. To train this behavior, we construct a scalable synthesis pipeline that distills correct visual reasoning traces, extracts the visual objects required by the traces, grounds them with a SAM3-based agent, and derives aligned point and box supervision from the resulting masks. We further propose grounding-aware reinforcement learning, which combines answer correctness rewards with dense grounding rewards that score whether generated object references match the correct image evidence. Across two counting benchmarks and four spatial reasoning benchmarks, adding visually grounded thinking to Gemma3-4B-IT consistently improves performance over the original model and the non-grounded thinking baseline. On spatial reasoning, the visually grounded thinking 4B models match, and in some cases surpass, Gemma3-27B-IT from the same model family. Our analysis shows that point grounding is well suited to counting, while box grounding benefits most from explicit grounding rewards on spatial tasks. Overall, our results show that VLMs think better when their intermediate thoughts are tied to the image regions that make them true.

Large language models (LLMs) often fail when answering requires identifying a small but decisive piece of evidence within a long or complex context, such as a single line in a tool trace or a subtle detail in an image. We propose ContextRL, a context-aware reinforcement learning (RL) method that improves long-horizon reasoning and multimodal performance through an indirect auxiliary objective. Instead of supervising only the final answer, ContextRL presents the model with a query, an answer, and two highly similar contexts, and rewards it for selecting the context that supports the query--answer pair, thereby encouraging fine-grained grounding. We construct contrastive context data in two domains: for coding agents, trajectories serve as contexts, yielding 1k pairs built via condition filtering; for multimodal reasoning, images serve as contexts, yielding 7K pairs built via generative editing and similarity search. ContextRL achieves average gains of +2.2% over standard GRPO on 5 long-horizon benchmarks, and +1.8% across 12 diverse visual question answering benchmarks. To disentangle the effect of the proposed objective from that of additional data, we compare against data-augmentation baselines that repurpose the same contrastive contexts as standard query--context--answer examples. These baselines provide little to no improvement, showing that the gains arise from the proposed context-selection objective rather than from the contrastive data alone.

Multi-step LLM pipelines fail through interactions among retrieval, reasoning, and formatting steps, so prompt-only optimization can miss bottlenecks in the chain. We present FAPO (Fully Autonomous Prompt Optimization), a framework that lets Claude Code optimize an LLM pipeline inside a standardized codebase. FAPO evaluates a pipeline, inspects intermediate steps, diagnoses failures, proposes scoped changes, and validates variants repeatedly to optimize against a score function. It first tries prompt edits and, only when prompt optimization appears insufficient, changes chain structure within the permitted scope when attribution identifies a structural bottleneck. Across six benchmarks and three task models, FAPO beats the baseline GEPA in 15 of 18 model-benchmark comparisons. In 11 model-benchmark comparisons, FAPO wins with non-overlapping mean pm trial-standard-deviation ranges, and the mean FAPO-GEPA gain is +14.1 pp. In the six HoVer and IFBench comparisons where prompt-first search escalated to structural changes, FAPO wins all six with a mean gain of +33.8 pp. FAPO also improves performance on security tasks: on CTIBench-RCM, a security CVE-to-CWE task, prompt-only FAPO lifts test accuracy by +4.0 pp on GPT-5, +7.1 pp on Foundation-Sec-8B-Instruct, and +2.0 pp on Foundation-Sec-8B-Reasoning. These results position FAPO as a state-of-the-art pipeline optimization technique for both general-purpose and security-focused tasks.

Aisha Down / The Guardian : A speculative scenario titled “Europe 2031” projects economic and political instability in the EU if it fails to keep pace with the US and China in the AI race —  A speculative thought-experiment warns the continent could pay a heavy price for lagging behind the US

Wall Street Journal : How success of AI-related companies in South Korea, Taiwan, and Japan is driving stock gains, bigger bonuses, and a retail investing frenzy in Asian markets —  Global success of AI-related companies in South Korea, Taiwan and Japan stokes market fever  —  Na Se-bin has lost all sense of the value of money.

Ella Creamer / The Guardian : Granta says it will stop publishing short story contest winners or join publishing partnerships it doesn't control after AI use allegations against a winner —  Literary magazine will no longer engage in ‘external publishing partnerships’ after Commonwealth prize furore

Angela Cullen / Bloomberg : Claude Guillemot, co-founder of Ubisoft and chairman of gaming hardware company Guillemot Corporation, died at 69 after a plane crash in France —  Claude Guillemot, who co-founded French video-game publisher Ubisoft Entertainment SA with his brothers in 1986, has died, according to the company.

Tim Fernholz / TechCrunch : An interview with Smartbird CEO Nadia Carlsten about the shoe company Allbirds becoming an AI infrastructure company, plans to deploy compute clusters, and more —  When Allbirds pivoted to AI in April, it felt like a joke from “Silicon Valley” breaking free of the TV: The direct …

Nikita Ostrovsky / Time : A look at Russia's push to develop homegrown AI talent, as the country is hampered by scarce access to AI hardware and a brain drain of top technical talent —  Nikita Ostrovsky … In early April, on a stage in the southwestern outskirts of Moscow, a moderator at Russia's annual Data …

Mishal Husain / Bloomberg : Q&A with Signal's Meredith Whittaker on why online child safety efforts risk mass surveillance, leaving the markets that demand weakening of encryption, more —  Meredith Whittaker has spent years arguing that privacy is a prerequisite for a free society.

Financial Times : Sources: the UK government is expected to consult as early as this month on rules to make public service news more prominent on social media and video platforms —  Move expected in British government green paper would set stage for fresh battle with Big Tech over online misinformation

Wall Street Journal : Sources: PC makers, including HP, are in talks with their supply-chain partners about using CXMT's memory chips in products bound for Asia as DRAM prices soar —  Prices soar because capacity isn't growing quickly, while China option is limited by national-security concerns

Anna Heim / TechCrunch : Paris-based Kyber, which develops a low-latency remote device control SDK and is founded by VLC lead developer Jean-Baptiste Kempf, raised $5M led by Lightspeed —  You've probably used VLC Media Player, the free video player with the orange traffic-cone icon — it's been downloaded more than 6 billion times.

Financial Times : Sources: Bain Capital stands to make $15B+ in profits on its 2018 Kioxia buyout, a ~20x return, as Kioxia's stock has surged 5,000%+ since its December 2024 IPO —  Bain Capital stands to pocket profits of $15bn on 2018 buyout of Kioxia, the former Toshiba Memory

Tyler Cowen / Conversations with Tyler : An interview with Roblox CEO Dave Baszucki on his early decision not to prioritize ad revenue, whether every mega platform becomes an everything app, and more —  Dave Baszucki is co-founder and CEO of Roblox, the user-generated gaming platform where all the games are built by the community itself.

Ben Schoon / 9to5Google : Nothing co-founder Akis Evangelidis says the phonemaker won't launch a new phone this year in its budget-focused CMF Phone series due to surging memory prices —  Nothing's CMF Phone series has released two models thus far and, at least for now, won't be getting a third …

Naga Avan-Nomayo / The Block : DOJ says two brothers pleaded guilty to robbing a Minnesota family of $8M+ in cryptocurrency after holding them at gunpoint for over eight hours in 2025 —  Quick Take  — Two brothers pleaded guilty to robbing a Minnesota family of more than $8 million in cryptocurrency after holding …

Reuters : Sources: Abu Dhabi's MGX is exploring buying Singapore-based data center operator DayOne; last month, sources said DayOne planned a US IPO at a $20B valuation —  Abu Dhabi-backed artificial intelligence investor MGX has been exploring buying Singapore-based data centre operator DayOne …

2020 年 12 月现代汽车以 8.8 亿美元从软银手中收购波士顿动力 80% 的控股权，该交易对波士顿动力的估值为 11 亿美元。收购协议包含了一项卖出期权，允许软银在未来某个时候将剩余股份出售给现代汽车。现在软银行使了卖出期权，以 3.25 亿美元将其持有的剩余股份出售给现代汽车。波士顿动力已有逾三十年历史，但其机器人的商业化一直进展缓慢。四足机器人 Spot 是首个取得商业成功的产品，而其人形机器人 Atlas 要证明其商业价值仍然面临重重困难。

随着越来越多的专业人士开始在工作中依赖 AI 工具，他们来之不易的技能会不会逐渐退化？此种担忧不无道理。有证据表明，AI 驱动的技能退化现象正在医学、计算机科学等领域发生。研究人员正探讨如何在 AI 时代保留重要的人类专业知识。一项针对波兰内窥镜医生的研究表明，AI 工具会迅速削弱人类的能力。这些医生在其职业生涯至少进行 2000 例结肠镜检查。研究人员让他们使用一套 AI 系统，能实时分析结肠镜图像，标记名为腺瘤的肠道癌前期病变。一旦医生开始使用该系统，每当系统不可用时，他们的诊断能力会显著下降。在引入 AI 工具前三个月里医生在 28.4% 的结肠镜检查中发现了至少一个腺瘤。而在引入 AI 工具后三个月里，在没有 AI 辅助的情况下腺瘤检出率下降至 22.4%。

挪威首相 Jonas Gahr Stoere 周五表示，挪威将限制小学生使用生成式 AI 工具。他表示，使用 AI 会增加儿童跳过重要教育阶段的风险。儿童在学校里最重要的是学会阅读、写作和数学。挪威政府表示，6-13 岁或一至七年级学生不应使用 AI；14-16 岁初中生可在教师指导下谨慎使用 AI 工具，17-19 岁高中学生应学习如何正确使用 AI 以便为未来的学习和工作做好准备。挪威政府还表示将立法为课堂使用更多书籍提供资金以扭转平板电脑普及的趋势。新标准将从 8 月下旬的新学年开始实施。

加州亿万富翁税提案获得足够签名达到在 11 月公投的资格。亿万富翁税（California Billionaire Tax Act）将对任何身价逾 10 亿美元的加州居民一次性征收 5% 的税，该税将用于医保和教育等项目。加州是全美亿万富翁人数最多的州，总数超过 200 人，很多人在 AI 热下积累了巨额财富。该税遭到了加州亿万富翁们的强烈反对，Google 联合创始人 Sergey Brin 一人至少投入了 8200 万美元反对该税，并搬到了靠近加州的内华达州居住。Palantir 联合创始人 Peter Thiel、Google 前 CEO Eric Sc​​hmidt、加密货币亿万富翁 Chris Larsen 以及DoorDash CEO Tony Xu 等也捐出了数百万美元反对该税。英伟达 CEO 黄仁勋则对该税没有异议，称会继续居住在加州。尽管提案已获得足够的签名，但相关组织必须在 6 月 25 日之前决定是否继续推进，或者与州政府达成协议。

在美国国会参议院周三通过一项法案阻止政府拆除 3.68 亿美元的海洋观测计划（Ocean Observatories Initiative）之后，特朗普政府表示将撤销拆除决定。海洋观测计划由逾 900 台深海仪器构成，用于监测洋流、海洋生态系统、碳吸收、热浪、渔业、沿海洪水和气候变化。每个观测站由多个锚定装置组成。这些设备测量从水面到数千英尺深处的洋流以及化学生物状况。仪器经过加固能承受深海的压力、腐蚀性海水以及可能损坏电子设备的海洋动植物。锚定装置周围的遥控机器人和滑翔机负责收集数据并将其传输到研究实验室。它于 2016 年投入运作，原计划运行 25 年，每年的运行成本为 4800 万美元。管理该项目的国家科学基金会已经宣布即日起停止移除设备，将继续运行和维护现有设备。

在正午高光、高温双重胁迫下，农作物光合作用会被大幅抑制，光能利用效率大幅下降，造成作物平均减产 30% 左右。这种作物“午睡”现象长期制约着光能利用效率与产量提升，自 1910 年被发现至今，困扰科学界长达百年。根据发表在《细胞》期刊上的一项研究，中科院等研究团队通过 10 年跨学科联合攻关，发现植物体内一种名为 MBS1 的超小蛋白可响应强光，形成凝聚体小液滴，通过类似“防护服”的保护作用过滤强光实现“防晒”。研究团队随后在海南、北京、吉林、黑龙江等地开展大田试验。测产结果显示，与底盘对照品种稻花香相比，增强表达 MBS1 的材料在光热胁迫较轻的吉林等地，增产约 10%~15%；在北京等中等光热地区，增产达 20%~30%；而在强光高热的海南，增产幅度高达 40%。这为应对全球气候变暖、实现粮食增产提供了新的基础理论与技术路径。

在内存和 SSD 之后，又一个计算机核心元件面临价格快速上涨。这一次是因为中东冲突导致生产高纯度树脂的沙特工厂停产。树脂是制造电路板（PCB）的基本原材料，位于沙特阿拉伯 Jubail 的一家工厂为陶氏化学与沙特阿美的合资企业，它供应了全球七成的高纯度 PPE 树脂。该公司于三月下旬因中东冲突关闭，4 月 6 日和 7 日还遭到导弹袭击。陶氏化学 CEO Jim Fitterling 此前表示公司预计物流和供应链恢复正常至少需要 275 天。高盛报告称 3 月到 4 月间 PCB 价格上涨了 40%。全球最大 PCB 制造商之一的胜宏科技警告中东冲突可能会推高铜和树脂的价格。替代树脂需要重新设计电路板、重新进行可靠性和性能测试，以及重新获得认证。环氧树脂原料的交货期已经从三周延长到了十五周。

在成功推出开源电子纸显示屏 DIY 工具包 Paper Monitor 和 Dev Kit 后，创业公司 Modos 准备推出一款完整的量产版显示屏。它在 Crowd Supply 上发起了众筹活动，筹款目标是 17.5 万美元，该目标已经达成，目前的金额达到了 45.6 万美元。Crowd Supply 计划推出的是 13.3 英寸、分辨率 3,200 x 2,400，支持触控，刷新率达到 60-Hz 的电子纸显示屏，其中黑白版本的众筹价格是 619 美元，彩色版本价格 719 美元。公司的两位联合创始人 Alexander Soto 和 Wenting Zhang 接受了 IEEE Spectrum 的采访。

根据发表在《科学》期刊上的一项研究，乌克兰研究人员利用相机陷阱调查了武装冲突对野生动物的影响，将 2022 年观察到的情况与 2021 年同期进行了对比。他们发现哺乳动物会通过行为调整应对武装冲突，其中包括减少夜间活动。武装冲突对卷入其中的人类是可怕的；野生动物也同样会被殃及。然而由于研究人员难以进入武装冲突地区且会面临危险，因此要理解此类冲突的影响会充满挑战。研究人员利用已运作中的相机陷阱来了解战争对野生动物的影响。他们发现冲突对该地区的哺乳动物产生了明显影响，其中包括这些动物的活动减少，尤其是在激烈冲突期间。此类影响证实，政治动荡所伤害的不仅是直接卷入其中的人类。

地球之水来自何处？科学家其实并不真正了解。水的来源有多种假说，其中最主流的是彗星说——撞击地球的彗星将水带到了地球；此外还有小行星说——撞击地球的小行星将水带到了地球，以及水由地球自身创造说。1986 年 Giotto 探测器对哈雷彗星的观测数据基本上否定了彗星假说，因为地球水的化学特性与彗星水完全不同。后续对 Hale-Bopp 彗星以及 Rosetta 探测器对 Churyumov-Gerasimenko 彗星的观测也都证实彗星之水与地球之水截然不同。那么地球之水是否可能来自小行星？科学家发现小行星上的惰性元素比例与地球也存在差异。那么地球上的海洋是否主要是由它自身形成的？早期地球的岩浆海洋富含氧气，而大气富含氢气，但氢气和氧气并不会自然结合。过去几年科学家做了一系列实验探索早期地球环境氢气和氧气是否能发生反应形成水。实验证实，地球上至少有一部分水能靠自身形成，但是否能形成今天覆盖整个地球的海洋，还无法下定论。

三个微软在 2011 年颁发的安全启动 (Secure Boot) 证书将于 6 月 24 日过期。安全启动检查系统启动期间加载的所有固件的数字签名，确保其来自可信提供商。安全启动旨在设计阻止会纂改 UEFI 的恶意程序 UEFI bootkits，一旦安装此类恶意程序很难检测到，即使重装系统也没用。安全启动使用加密签名确保启动过程中加载的每个固件都受到计算机制造商的信任，它旨在建立信任链，防止攻击者用恶意固件替换预期的启动固件。但在 2023 年研究人员发现了存在于几乎所有 Windows 和 Linux 系统 UEFI 启动过程中的严重漏洞 LogoFail。该漏洞存在于启动时显示硬件制造商徽标的软件中，攻击者能利用其图像解析 bug 绕过安全启动，用恶意固件感染 UEFI。微软因此移除了三个在 2011 年颁发的旧证书，用 2023 年颁发的新证书取代。Windows 用户可通过 Windows 安全设置 > 设备安全性 > 安全启动 去检查证书是否已经更新。Linux 用户可关注名叫 shim 的程序更新。

美国投行摩根大通已禁止香港员工访问 Anthropic 的模型，显示这一技术在美国境外的应用正面临极其严格的审查。由于 Anthropic 与摩根大通的许可协议中有关“使用条款”的特定措辞，摩根大通已将 Claude 模型从其驻港员工获批使用的大型语言模型（LLM）内部名单中移除。在此之前，高盛也做出了类似决定，于 4 月将 Claude 从其香港员工的获准使用工具名单中剔除。今年 4 月 Anthropic 首次向少数企业和机构开放 Mythos 模型测试，并警告该模型具备发现网络安全漏洞的能力，不宜广泛推广。6 月初 Anthropic 发布了 Mythos 级模型的首个公开版本 Fable 5，但为管控其突破网络漏洞的能力，同步设置了许多限制措施。然而华盛顿仍以国家安全为由下达紧急出口管制令，迫使 Anthropic 在全球范围内关停 Mythos 5 和 Fable 5 模型。

勒索组织 FulcrumSec 宣称入侵了制药巨头诺和诺德（Novo Nordisk）的网络，窃取了约 1.3 TB 的数据，包括源代码、药物研究、临床试验记录、员工和医生信息、生产系统信息以及内部 AI 模型数据。它向诺和诺德勒索 2500 万美元赎金，但未获成功，因此考虑出售部分数据。FulcrumSec 称诺和诺德的代表于 6 月 3 日联系了他们。FulcrumSec 表示考虑通过开源来遏制企业不想支付赎金的情况。诺和诺德发言人表示它正与相关机构保持联系。

科学家发现了已知最古老的鼠疫证据，将其出现的时间追溯到约 5500 年前——比之前认为的早了约 200 年。研究人员在西伯利亚贝加尔湖附近的四个墓地寻找鼠疫杆菌的痕迹。他们在 18 位古代狩猎采集者的牙齿中发现了鼠疫 DNA 残留。对骨骼碳年代测定显示，发现这场瘟疫引发了两波疫情，第一波出现在 5500 年前。病菌可能是通过土拨鼠传播的，当地人可能是通过食用生内脏或屠宰过程中接触携带病菌的兽皮而感染鼠疫。死者中有很多是 8-11 岁幼童。早期的鼠疫和中世纪的黑死病同样致命，不仅摧毁人口稠密的城市，也摧毁小型游牧狩猎采集群体。

山西大学研究人员在 PLOS One 上发表了一篇论文，指出青少年的心理健康、体重指数以及屏幕时间与睡眠质量有显著联系，且女孩和生活在农村地区的青少年睡眠质量往往较差。研究人员调查了中国六个城市的 5,713 名 13-18 岁青少年，这六个城市分别是：上海、苏州、太原、婺源、兴义和乌鲁木齐。他们使用匹兹堡睡眠质量指数（PSQI）收集了睡眠质量数据，同时还收集了 BMI、体质健康、静坐时间、屏幕使用时间及心理健康等数据。此外还获得了每位参与者的居住地（城市或农村）和性别信息。总体上有 33.71% 的受访者睡眠质量不佳。他们发现不同居住地点和性别之间存在显著差异。农村青少年睡眠质量不佳的比率高于城市青少年（分别为 35.78% 和 31.90%），在入睡时间、睡眠时长和睡眠干扰几个方面的表现均较差。女孩在几乎所有睡眠衡量指标方面上的表现均不及男孩，女孩睡眠质量较差者的比率为 38.40%，而男孩为 29.20%。较高的体重指数对女孩的睡眠有更显著的不利影响。