OrangeBot.AI Digest — 2026-06-28

Add WebMCP-native AI chat to any Frontend

Lossless media compression without context switching

Your local semantic search app

"One AI agent across your browser, tools, and messages "

100+ AI models, one interface. ECO friendly.

Compose music with just your voice

Compare & Sync Videos Fast

The professional book writing app built for serious authors

Turn your Mac into a time machine.

Claude for PowerPoint, on steroids

Simulate AWS, GCP & DigitalOcean without paying the bill

3x Mobile Automation. Same QE Team.

Turn any Excel file into a live dashboard

See who's on your site. Right now.

AI workspace that’s already connected to everything

Modern image generation demands a single model that unifies diverse capabilities, including text-to-image (T2I), local editing, and global editing. However, these capabilities are rarely naturally aligned and often conflict. For instance, editing tends to degrade T2I performance, while global and local editing interfere with each other. Consequently, effectively composing these capabilities has become a central challenge for image generation model training. To tackle this, we introduce DanceOPD, an on-policy generative field distillation framework for flow-matching models that routes each sample to one capability field, queries one low-noise student-induced state, and trains with a simple velocity MSE objective. With each capability source defined as a velocity field over the shared flow state space, the student learns from fields queried on its own rollout states to compose expert capabilities. This formulation also absorbs operator-defined fields such as classifier-free guidance. Comprehensive experiments on T2I, editing, realism-field absorption, and CFG absorption show that our approach improves multi-capability composition, strengthening target capabilities while preserving anchor generation quality. We believe this work establishes a practical route for generative field distillation in flow-matching models.

Modern Vision-Language-Action (VLA) models often fail to generalize to novel setups, such as altered camera viewpoints or robot morphologies, because they are typically conditioned only on current observations and language instructions. By ignoring the underlying system configuration as a variable, these models implicitly assume a fixed execution context encountered during training, necessitating data-intensive fine-tuning for any new environment. In this work, we introduce In-Context World Modeling (ICWM), a framework that treats system identification as an in-context adaptation problem. ICWM enables robot policies to autonomously infer essential system variables from a short history of self-generated, task-agnostic interactions. Unlike traditional In-Context Learning that uses demonstrations to specify what task to perform, ICWM leverages the context window to understand how the system operates. By processing these interactions before task execution, the model implicitly captures the world dynamics of the current system, enabling adaptation to novel configurations without parameter updates. Extensive experiments in simulation and on real-world robot platforms demonstrate that ICWM significantly outperforms standard VLA baselines on novel camera viewpoints.

Outcome-based reinforcement learning provides a stable optimization backbone for language agents, but its sparse trajectory-level rewards provide little guidance on which intermediate decisions should be reinforced or suppressed. On-policy self-distillation offers dense token-level supervision, yet existing skill-conditioned variants often rely on external skill memories or retrieved privileged context, which are costly to maintain and can be mismatched with the state distribution induced by the current policy in multi-turn interaction. We propose OPID (On-Policy Skill Distillation), a framework that extracts skill supervision directly from completed on-policy trajectories. OPID represents trajectory hindsight as hierarchical skills: episode-level skills capture global workflows or failure-avoidance rules, while step-level skills capture local decision knowledge at critical timesteps. A critical-first routing mechanism uses step-level skills when critical decisions are identified and falls back to episode-level skills as default guidance otherwise. The selected skill is injected into the interaction history, allowing the old policy to re-score the same sampled response under both original and skill-augmented contexts. The resulting log-probability shift yields a token-level self-distillation advantage, which is combined with the outcome advantage for policy optimization. OPID thus preserves RL as the primary training objective while introducing dense, distribution-matched hindsight supervision. Experiments on ALFWorld, WebShop and Search-based QA demonstrate that OPID generally improves agent performance, sample efficiency, and robustness over outcome-only RL and existing skill-distillation baselines. Our code is available at https://github.com/jinyangwu/OPID/tree/main.

While text-to-image (T2I) models have achieved remarkable progress, they struggle with real-world requests that are often underspecified, implicit, or dependent on up-to-date knowledge. We identify this challenge as the Context Gap: the mismatch between the user context and the sufficient generation context for T2I models. To bridge this gap, we propose Qwen-Image-Agent, a unified agentic framework that integrates plan, reason, search, memory and feedback in a context-centric manner. Qwen-Image-Agent treats user input as partial context and progressively constructs the generation context through Context-Aware Planning and Context Grounding. Specifically, Context-Aware Planning identifies missing context and plans how it should be acquired and used, while Context Grounding gathers this context from reason, search, memory, and feedback. To evaluate agentic image generation, we further introduce Image Agent Bench (IA-Bench), a benchmark covering four core image agent capabilities: Plan, Reason, Search, and Memory. Experiments on IA-Bench, Mindbench and WISE-Verified show that Qwen-Image-Agent outperforms strong baselines and achieves state-of-the-art performance.

A classical intuition holds that verifying a solution is easier than producing one. For today's coding agents, this intuition is being inverted: as foundation models develop stronger reasoning capabilities and engineering harnesses grow more sophisticated, generating complex candidate solutions is no longer difficult -- reliably verifying them has become the harder problem. Every verifier we can build is only a proxy for human intent, never the intent itself. This makes verification subject to a twofold difficulty: first, intent is underspecified by nature, making it inherently hard to faithfully check whether it has been fulfilled; second, during model training, optimization widens the gap between proxy and intent -- manifesting as reward hacking or signal saturation. To address this, we characterize the quality of verification signals along three dimensions -- scalability, faithfulness, and robustness -- and argue that achieving all three simultaneously is the central challenge. We further study four reward constructions: a test verifier for general coding tasks, a rubric verifier for frontend tasks, the user as verifier for real-world agent tasks, and an automated agent verifier for long-horizon tasks. Across different task types and policy capability levels, we conduct in-depth analysis and experiments on the core challenges of reward design and how to more effectively leverage reward signals. Experiments show that targeted verification design can effectively suppress reward hacking, improve task completion quality, and achieve significant gains across multiple internal and public benchmarks. These experiences collectively point to a core observation: no fixed reward function can remain effective as policy capability continues to grow; and verification must co-evolve with the generator.

A unified representation for text and vision is a natural pursuit, as it enables simpler multimodal modeling and more efficient training. However, representing images as discrete signals in the same way as text inevitably introduces severe information loss. Existing work struggles to balance low-level details and high-level semantics in discrete representations: reconstruction-oriented representations often lack semantic information, whereas semantically stronger features typically suffer from severe loss of detail. We present ViQ, a Visual Quantized Representations framework, which is designed to balance semantics and details in discrete representations while supporting inputs at native resolutions, thereby enabling it to serve as a unified and general discrete representation for arbitrary visual inputs. Our approach structures quantization learning into two stages: text-aligned pre-training and feature discretization. With text-aligned pre-training, we enhance the visual encoder semantic-rich supervision from the pretrained language model and enable it to process native-resolution visual inputs. During discretization, we propose a proximal representation learning strategy to progressively compact the feature space, along with a position-aware head-wise quantization mechanism that enables flexible processing of arbitrary resolutions. Extensive experiments on multimodal tasks demonstrate that ViQ achieves competitive performance compared to state-of-the-art multimodal vision encoders with continuous and high-dimensional visual features, while maintaining high precision in low-level reconstruction. We also show that multimodal training with visual quantized representations largely improves efficiency, yielding up to 20\%-70\% acceleration with different base LLMs and training recipes.

Speculative decoding (SD) accelerates autoregressive Large Language Models (LLMs) by drafting multiple tokens and verifying them in parallel, but it faces a scaling limitation: increasing the draft budget improves speed only when acceptance remains high and drafting overhead stays low. This ceiling has been difficult to break because prior head-based SD methods face a causality-efficiency dilemma. Autoregressive drafters produce path-conditioned candidates that are effective for tree speculative decoding with higher acceptance length, but their drafting cost grows with tree depth. Bidirectional block-diffusion drafters generate all positions in one pass, but their branch-agnostic marginals can form individually plausible yet mutually inconsistent trees, wasting budget and reducing acceptance. We propose JetSpec, a head-based SD framework that combines one-forward drafting efficiency with branch-wise causal conditioning. JetSpec trains a causal parallel draft head over fused hidden states from the frozen target model, producing candidate trees whose scores align with the target model's autoregressive factorization. This enables JetSpec to convert larger draft budgets into longer accepted prefixes and higher end-to-end speedup. Across math, coding, and chat benchmarks on dense and MoE Qwen3 models, JetSpec consistently outperforms bidirectional-head and tree-based SD baselines. On H100 GPUs, JetSpec achieves up to 9.64x speedup on MATH-500 and 4.58x on open-ended conversational workloads, with further latency gains demonstrated through vLLM integration under realistic serving loads. Our code and models are available at https://github.com/hao-ai-lab/JetSpec.

Computer-use agents can execute software tasks through either graphical interfaces or programmatic command interfaces, but existing evaluations confound interaction modality with differences in tasks, initial states, verifiers, and permitted actions. We introduce a matched execution-layer benchmark of 440 desktop tasks across 18 applications and 12 workflow categories, where screen-only GUI agents and skill-mediated CLI agents receive identical goals, states, and final-state verifiers while being restricted to modality-native actions. In this controlled setting, the strongest GUI agent reaches a 59.1% full pass rate, outperforming the strongest original-skill CLI agent at 48.2%; however, verifier-guided skill augmentation raises CLI success to 69.3%, showing that much of the CLI deficit comes from incomplete skill coverage rather than model capability alone. These results suggest that GUI and CLI expose different execution bottlenecks: GUI agents are limited by reliable grounded interaction over long-horizon workflows, whereas CLI agents are limited by the coverage and scalability of their skill interfaces.

Joint-Embedding Predictive Architectures (JEPAs), including recent LeWorldModel (LeWM), have become a promising foundation for reconstruction-free visual world models. For visual planning, however, LeWM evaluates candidate action sequences by repeatedly applying a local one-step latent transition model. This autoregressive rollout makes planning computationally expensive and exposes the predicted trajectory to accumulated latent errors as the horizon grows. We propose Fast LeWorldModel (Fast-LeWM), a fast latent world model that replaces repeated local rollout with action-prefix prediction. Given the current latent and a candidate action sequence, Fast-LeWM encodes its prefixes and predicts the future latents reached after executing those prefixes in parallel. By making action prefixes the basic prediction unit, Fast-LeWM directly models action effects accumulated to different extents over multiple horizons. This prefix-level supervision forces the model to learn how states continuously evolve under different action prefixes, rather than only fitting one-step state transitions. During planning, the predictor can use the last prefix token from the encoded action sequence to evaluate the corresponding future latent without explicitly rolling through each intermediate imagined state. Across multiple tasks, Fast-LeWM improves average success over LeWM while substantially reducing planning time, achieving lower open-loop latent loss whose growth becomes significantly slower as the rollout horizon increases.

As agentic systems continue to evolve and are widely deployed in real-world scenarios, there is a growing demand to faithfully evaluate their capabilities. However, current benchmarks are typically built on popular applications with relatively simple tasks and focus on a narrow set of capabilities while overlooking broader dimensions, resulting in saturated performance on modern agents and failing to probe their limitations. To this end, we introduce GauntletBench, a web-based benchmark for evaluating agent generalisation in challenging scenarios, focusing on three underexplored capabilities (temporal perception, graphical understanding, and 3D reasoning), across five less-covered professional applications (Video Editor, Workflow Builder, 3D Modeller, Flight Analyser, and Circuit Designer), each with 20 vision-intensive tasks (100 in total). Our benchmark provides a modular pipeline that comprises an environment compatible with both open- and closed-source agent frameworks, a controlled web-based application, a well-structured task suite, and an automated evaluation engine with diverse metrics. Contrary to widespread expectations, our empirical results reveal that frontier agentic systems remain far from achieving human-level performance. Even the state-of-the-art agent achieves only a 19.1% success rate on our GauntletBench, highlighting the limitations in these overlooked capabilities and generalisation. By comparison, non-expert human annotators achieve over 80% success on our challenging yet feasible tasks, revealing the substantial gap between current agent capabilities and those required for complex real-world scenarios.

Tool use enables large language models (LLMs) to perform complex tasks, and recent agentic reinforcement learning (RL) methods show promise for enhancing model capabilities. However, RL alone often leads to instability or limited gains in tool-use tasks. In our experiments, some models exhibit catastrophic collapse, where performance abruptly drops and tool-invocation structures fail. The analysis reveals that these failures stem from unexpected probability spikes in specific control tokens, disrupting structured execution, yet the underlying tool-use capability remains intact, merely obscured by specific formats. To address this, we systematically investigate a diverse set of supervisory signals, including off-policy supervision, hint-based guidance, erroneous example supervision, and others, applied under both synchronous and interleaved training schemes. We find that interleaving supervised fine-tuning (SFT) with RL substantially improves stability, but exhibits degraded performance under format and content out-of-distribution (OOD) evaluation. We also analyze the impact of learning rates and generalization across settings. These results highlight the importance of understanding RL failures and demonstrate how diverse supervisory signals can guide exploratory learning, enabling robust training of LLMs for complex, multi-step tool-use tasks. Our Code is available at https://github.com/hypasd-art/Tool-RL-Box.

The prevalent dual-branch paradigm, i.e., training a side network to encode visual conditions and fusing its intermediate-layer features to a frozen pretrained main network, has shown remarkable success in visual-condition controllable generation. Despite its widespread adoption, the role of the side branch and its training efficiency remain underexplored. In this paper, we first revisit this mainstream paradigm through the lens of score-based generative modeling: 1) The main network preserves visual perceptual quality by providing a prior unconditional score. 2) The side network steers conditional control by implicitly contributing a likelihood score. Guided by this perspective, we propose LIkelihood Score Alignment (LISA), an effective regularization method that explicitly aligns the intermediate feature of the side network with an approximated likelihood score. Specifically, we first hook features from a designated layer of the side network and project them into the score latent space by a lightweight decoder. Then, we construct an approximated likelihood score target and calculate the distance between the decoder's output and this target as an additional regularization loss. Finally, we jointly optimize the side network and decoder with both standard diffusion loss and our regularization loss. Experiments across various image/video tasks, architectures, and diffusion/flow models demonstrated that LISA can not only consistently accelerate the training convergence and improve final synthetic results, but also encourage the side network's features to be more disentangled for conditional modeling with negligible additional training cost and zero extra inference cost.

Despite their widespread use, the role of reward models in shaping reinforcement learning is poorly understood. Reward models offer a tempting promise: they automatically estimate response quality in the absence of verifiers or human judges. Unlike "verifiable rewards" which typically produce binary scores, reward models typically produce continuous scores, allowing them to be sensitive to fine-grained differences in responses. However, we show this apparent strength is a serious weakness: many popular reward models are oversensitive, assigning different scores to equally good responses. Theoretically, we show that seemingly perfect reward models can be highly oversensitive; empirically, this oversensitivity can lead to bad policies. In place of existing notions of "reward model accuracy," we propose evaluating reward models using distinct measures of "discriminative ability" and "specificity" (the complement of oversensitivity). As a solution, we describe a training-free algorithm that uses Monte Carlo dropout on any neural reward model to produce discrete reward clusters. Theoretically, we prove there exist discretizations that reduce oversensitivity at minimal expense of discriminative ability; empirically we show, in both controlled and natural RL settings, that discretizing rewards leads to less reward hacking and better policies than training on the original rewards.

Reasoning capability has advanced rapidly in large language models (LLMs), leading to an increasing size of key-value (KV) cache in both prefilling and decoding stages. Existing KV cache compression methods mainly rely on attention weights to estimate token importance. While attention effectively captures contextual relevance, it overlooks complementary information-theoretic signals related to predictive uncertainty and token informativeness. In this paper, we revisit token importance from a forward-looking perspective and introduce Forward Influence, a metric that measures how compressed tokens affect future contexts. Our analysis reveals that tokens selected by attention scores mainly influence nearby contexts, whereas tokens associated with high predictive uncertainty exhibit substantially stronger influence on distant future contexts. Based on the observation, we propose InfoKV, an entropy-aware KV cache compression framework that incorporates information-theoretic signals. It combines token-level predictive uncertainty with layer-wise representation evolution and integrates the resulting entropy scores with attention scores during reasoning. Experiments on long-context reasoning benchmarks with Llama-3.1, Llama-3.2, and DeepSeek-R1 demonstrate that InfoKV consistently outperforms existing attention-based KV compression methods in both long prefilling and decoding scenarios.

Video reasoning language models implicitly assume that every input frame is equally reliable. This leads to what we term the Blind Trust Problem: under realistic perturbations such as motion blur, glare, or occlusion, frontier video reasoning models can suffer 15-30%p accuracy drops on real-world embodied benchmarks, while remaining unaware that their visual evidence has been degraded. To address this challenge, we propose Robust-TO, an agentic video understanding framework that explicitly integrates per-frame trustworthiness into every stage of reasoning. Robust-TO organizes heterogeneous visual perception tools under a unified evidence interface. Each tool receives a sub-query derived from the original question and a set of trustworthy frames selected by the reliability-relevance score. It returns evidence in a shared format: a concrete prediction (e.g., a bounding box, motion trajectory, recognized text, or action label), temporal grounding, and a calibrated reliability score. During reasoning, these calibrated scores guide evidence weighting in a three-tier synthesis process (high/medium/low) and define a confidence-cost GRPO reward that jointly optimizes correctness, evidence reliability, and efficiency. On two video reasoning benchmarks spanning eight tasks, Robust-TO achieves 56.4% average accuracy on clean inputs, surpassing the strongest open-source baseline by 10.6%p and outperforming Gemini-2.5-Pro (46.2%). Under five realistic corruption types, Robust-TO maintains 54.3% average accuracy, 5.8%p above the strongest open-source baseline, while exhibiting the smallest clean-to-corrupted accuracy drop among all compared methods.

Marton Eder / Bloomberg : Letter: Austria is pushing the EU to consider hosting Anthropic within its borders, highlighting EU efforts to boost bloc independence from US and Chinese tech —  Austria is pushing the European Union to consider hosting Anthropic PBC within its borders to counter US efforts to block foreigners …

Zvi Mowshowitz / Don't Worry About the Vase : GPT-5.6 system card indicates Sol is well below the level of most worrisome Mythos use cases, suggesting all GPT-5.6 versions could be released without delay —  While we wait for a general release, the system card is the best hint as to what is going on with the new candidate for America's Next Top Model, GPT-5.6.

Qianer Liu / The Information : Sources: Baidu's chip unit Kunlunxin Technology plans a Hong Kong IPO at a $50B target valuation, asking investors to buy chips worth 3-7x their IPO investment —  In China, chip companies may have found a new clientele for their semiconductors: IPO investors.

Jackson Chen / Engadget : Australian government says it plans to double the maximum penalty for any social media company breaking its minimum age law and grant more powers to enforcers —  The fine can now potentially hit 99 million AUD, or $68 million.  —  After becoming the first in the world to implement …

Bloomberg : Australia-based Firmus partners with Nvidia to build its first data center in Batam, Indonesia; the 360 MW Nvidia DSX AI factory campus is developed with DayOne —  Firmus Technologies Pty Ltd. will build its first data center project in Indonesia as part of a partnership with US chip giant Nvidia Corp. …

Nick Lichtenberg / Fortune : An analysis of US payroll data across 730+ occupations: employment among workers ages 22 to 25 in highly AI-exposed jobs is now shrinking by 3.8% per year —  Last August, a team led by Stanford economist Erik Brynjolfsson published a deep look at the impact of AI on jobs, boosted by a …

New York Times : Brazil, which has struggled to standardize soccer talent scouting, is embracing AI-powered scouting apps that assess players by analyzing video clips and more —  Recruitment apps powered by artificial intelligence are gaining ground in Brazil, promising to even the playing field and find more soccer talent.

Sebastian Herrera / Fortune : A profile of Jacob Andreou, the 33-year-old former Snap exec leading Microsoft's consolidated Copilot team efforts to catch up with OpenAI and Anthropic —  Andreou, who once led product at social media firm Snap, is head of Microsoft Copilot.  —  Early this year, Jacob Andreou had a crucial goal.

Tim Higgins / Wall Street Journal : Masayoshi Son questioned Musk's orbital AI data centers, noting electricity is just 7% of costs and the AI race will be won on Earth within a few years —  SoftBank's Masayoshi Son thinks the math doesn't support space-based data centers  —  Some people have expressed skepticism …

Wired : A look at a thriving underground economy for Claude access in China, including “transfer station” sites that buy API tokens abroad and distribute them to users —  As Anthropic tightens restrictions on access to Claude in China, users keep finding new workarounds, from proxy services to fake identities sourced on Telegram.

Alistair Gray / Financial Times : Apple's appeal to a London Court of Appeal ruling begins Monday at the UK Supreme Court; the ruling required it to pay $502M for using Optis' wireless patents —  Battle over licensing of mobile connection protocols could shape global rates for other technologies

Wall Street Journal : Researchers say Z.ai's GLM-5.2 matches latest US models at finding security bugs, as critics question the US' lax approach in restricting Chinese open models —  Clampdown on top U.S. artificial intelligence is fueling concern that Washington is handing Beijing a cyberwarfare advantage

Financial Times : Sources: Google told Meta around March it couldn't offer all the Gemini capacity Meta wanted to buy, disrupting and delaying some of Meta's internal AI projects —  Surging appetite for advanced models is turning computing power into the tech industry's scarcest commodity

Laura Bratton / The Information : Sources: Salesforce staff worry Anthropic's Claude Tag could cannibalize Slackbot and give the AI firm more leverage over the enterprise software industry —  When Anthropic on Tuesday launched a high-profile AI product for businesses that use Slack, some employees at Salesforce, which owns Slack, were confused.

Scharon Harding / Ars Technica : Netflix has been gradually requiring each profile under a Netflix subscription to use a unique email address; the rule doesn't apply to children's profiles —  Recently, my father called me in a panic.  —  There were just a few minutes until Netflix would start streaming a live MMA event, and he couldn't get into my account.

继哥伦比亚大学 Dieter Egli 团队之后，第二个科研团队报告称，利用高精度基因编辑技术改造了人类胚胎 DNA。多名科研人员表示，该实验的成功，让关于胚胎基因编辑的伦理讨论变得更加紧迫。这一次英国剑桥大学的 Kathy Niakan 带领研究人员使用单碱基编辑技术敲除了编码 NANOG 蛋白的基因。实验所用的精子、卵子与胚胎均来自科研捐献，胚胎仅培养约 1 周时间。经过编辑的胚胎无法形成完整外胚层，也就是未来发育成人体各类组织的细胞团；但胚胎仍能分化出胎盘、卵黄囊等胚胎附属支撑结构的前体细胞。

1918 年诺贝尔物理学家得主马克斯普朗克（Max Planck）的两篇论文在 2011 年被撤下。两篇论文于 1940 年代发表在德国期刊《Naturwissenschaften》上，如今隶属于 Springer Nature 集团，《Naturwissenschaften》撤稿的理由是侵犯版权。马克斯普朗克的做法是一稿多发，在以前很常见，但如今被认为“自我剽窃”会引发版权纠纷。魁北克大学物理史学家 Yves Gingras 和科学史学家 Mahdi Khelfaoui 对此做法提出批评。尤其滑稽的是 Springer Nature 以空白 PDF 的形式撤稿，而该空白 PDF 仍然以 39.95 美元的价格出售访问权限。他们怀疑自然集团是使用算法自动撤稿的。

2014 年世界气象组织 (WMO) 协调组织了一场活动，其中包括预测 2050 年的天气。法国主持人 ïvelyne Dhéliat 的预测是到 2050 年该国最高气温将达到 43 摄氏度。本周法国的热浪将这一预测提前 24 年实现，且超过了预测。法国周三 43 个地点有 19 个地点超过了对 2050 年的预测，最高气温高达 44.3 摄氏度。法国 1947 年以来的热浪半数发生在 2010 年之后。法国气象局表示，到 2100 年热浪可能会持续两个月。随着地球暖化，极端高温将变得越来越常见。

韩国计划训练所有军人像使用枪支一样熟练操作无人机。国防部长官安圭伯表示，目标是通过训练士兵像使用“第二件个人武器”一样使用无人机，让无人机成为所有部队的“通用作战工具”。韩国还计划为军事单位配备更多廉价且可消耗的无人机，用于侦察和打击任务，部署更多反无人机激光武器和微波武器。很多国家都在效仿乌克兰训练和装备无人机，韩国并非个例。乌克兰利用无人机和军用机器人弥补其在战场上与俄罗斯庞大军力对抗的劣势。韩国国防部今年将首先向军人提供 1.1 万架训练无人机，目标是到 2029 年部署 6 万架无人机。出于安全考虑，韩国军方希望采购 100% 国产零部件的无人机。

几十年来计算加速了数学的进步。50 年前数学家利用计算机证明了四色定理，以一种人类几乎不可能实际验证的方式证明任何地图可以用不超过四种颜色着色。但在整个计算时代，人类数学家的作用仍然至关重要。人类凭借直觉提出猜想，凭借创造力和经验设计证明策略，最终验证证明是否正确。今天 AI 正在挑战这一模式。短短几年内，大模型就从只知道复述的“随机鹦鹉”演变成高级数学推理机器。UCLA 教授陶哲轩认为 AI 能作为催化剂推动向他所谓的“大数学”的转变，设想未来人类与机器将进行大规模、去中心化的协作，复杂的数学任务被分解，人类负责创造性部分，而 AI 则承担大部分技术性工作。陶哲轩已在实践这一理念。AI 正迫使数学家思考数学对他们的意义。一位数学家称，数学塑造了其思维方式，使其能以非常逻辑和理性的方式思考，对生活各个方面都有帮助。随着 AI 改变数学，研究人员想知道未来的数学家是否也能这样说。

在要求 OpenAI 分阶段发布其新模型之后，美国政府显然也对 Anthropic 提出了相同的要求。根据商务部长 Howard Lutnick 给 Anthropic 的信函，美国政府允许 Anthropic 重新发布其新模型 Mythos 5，但现阶段只提供给少数客户。Mythos 5 的访问权限将提供给约 100 家机构，其中包括政府机构和私营公司。OpenAI 的新模型 GPT-5.6 也采用类似的发布模式。

BBC 宣布于 6 月 27 日停播 Radio 4 长波广播，6 月 30 日关闭长波平台。BBC 将此归咎于维护过时技术的成本过高。Radio 4 长波设备使用的电子管早就停产，而 BBC 早在 1997 年就将相关信号塔出售给了私营公司，这些设备由私营公司运营。随着越来越多的人转向流媒体，维持全国性的广播和电视覆盖在经济上意义不大。在关闭长波之后，英国接着准备淘汰地面电视，相比使用人数很少的长波，地面电视的关闭将会影响很多人。英国政府计划最早 2034 年最晚 2044 年关闭地面电视。

欧盟航空安全局（EASA）下令对阿联酋航空和澳洲航空运营的 16 架空客 A380 飞机进行紧急检查，此前部分 A380 飞机的机翼部件发现了裂纹。EASA 表示，裂纹是在此前对机翼翼梁结构的检查中发现的，它认定这些裂纹可能会降低机翼的结构完整性。为解决潜在安全隐患，空客必须进行额外的专项细查。16 架 A380 客机中有 15 架由阿联酋航空运营，1 架由澳洲航空运营。A380 是载客量最高的民航客机，共制造了 254 架，目前已经停产。

英国伦敦大学玛丽皇后学院的研究人员分析了 1960-2023 年间发行的逾 38 万首歌曲的歌词后发现，流行音乐中使用的情感语言和道德语言发生了显著变化。表达关怀和体面等道德美德的词语随时间推移变得越来越少见，而与伤害、欺骗、颠覆和堕落相关的语言逐渐增多。研究人员指出，“音乐不仅仅是娱乐。它是社会讲述自身故事的方式之一。通过分析几十年来的歌词，我们可以开始看到情感表达和道德叙事随时间如何演变。”研究还发现，女性艺术家更多与关爱和忠诚等美德联系在一起，而男性艺术家和男女混合组合则更多与反映伤害、颠覆和堕落等负面主题联系在一起。

根据发表在《Communications Biology》期刊上的一项研究，大猿的笑声节奏可能与现代人类相似，而这一现象已持续了至少 1500 万年。研究结果还表明，在大猿的演化过程中，笑声变得更快、变化更多，且越来越受到所处情境的影响。所有大猿（人科动物）都会笑，包括与人类亲缘关系较近的物种，如倭黑猩猩，以及亲缘关系较远的物种，如婆罗洲猩猩。然而笑声的节奏随时间如何演变，及其可能与人类语言的演化有何关联，此前尚不清楚。 在研究中，英国华威大学的研究人员分析了 4 只婆罗洲猩猩（Pongo pygmaeus）、两只大猩猩（Gorilla gorilla）、3只倭黑猩猩（Pan paniscus）、4只黑猩猩（Pan troglodytes）以及4个人的笑声录音，这些个体的年龄在6个月至7岁之间。 科学家研究了140段笑声序列，并测量了每次发声之间的时间间隔。研究发现，所有物种的笑声都遵循一种规律的节律模式，连续发声之间的间隔均匀。由于这种模式在所有研究物种中均存在，研究人员推测，这种有节奏的笑声可能早在 1500 万年前就已存在于它们的共同祖先身上。 他们还推断，随着时间的推移，笑声变得更快、更多样化，比如人类会根据情境改变笑声的速度，如被挠痒时发出的笑声比玩耍时更快，而其他猿类则不会。此外，与人类亲缘关系越近的猿类，其笑声节奏的变化性就越大。 这些发现表明，在大猿和人类的演化过程中，发声的灵活性和控制力可能逐渐增强，作者推测这可能促成了语言的出现，未来需要通过更大样本量的研究证实这些发现。

久坐是一种健康风险，但对久坐行为的干预需要考虑可行性和有效性。根据发表在《British Journal of Sports Medicine》期刊上的一项研究，研究人员评估了每隔 30 分钟、60 分钟或 120 分钟就站起来步行 5 分钟的干预措施。有 19342 名成年人参与了研究，其中 11484 人分成三组执行上述三种不同的干预方法。结果显示，所有干预组参与者报告疲劳和负面情绪显著降低，正面情绪显著提升。在考虑了可行性和有效性等因素之后，研究人员指出每小时站起来走 5 分钟在可行性和有效性之间取得了最佳平衡。

美光 CEO Sanjay Mehrotra 在最新的财报电话会议上披露，该公司与 16 家大客户签署了“战略客户协议”，大部分协议涵盖的时间从 2026 年一直持续到 2030 年，客户承诺购买一定数量的产品，支付价格处于设有最低和最高价格的定价区间内。这意味着如果内存价格进一步上涨，客户基本不会受到的影响。美光 CEO 称，客户意识到，内存和存储设备的供应短缺需要相当长的时间才能缓解。美光预计供应将在 2028 年逐步改善，但目前无法预测内存供应何时才能赶上持续增长的需求。他说客户同意预付款项，该公司将利用这笔资金扩建晶圆厂。

上海交通大学医学院附属第一人民医院的一支研究团队利用了英国生物样本库（UK Biobank）的数据，最终纳入了 82826 名基线时无眼部疾病的参与者。这些参与者均连续 7 天佩戴了配有高分辨率光传感器的腕带式加速度计，以客观记录其个人光照暴露情况。研究结果显示，在晚间时段（晚上20:00至23:30），当参与者所处环境的平均光照强度超过1000勒克斯时，与其后眼部退行性疾病的发病风险显著升高相关。其中，年龄相关性黄斑变性的患病风险增加了31%，白内障风险增加了18%，而原发性开角型青光眼的风险则大幅增加了47%。研究人员还观察到了显著的时间-剂量反应关系。在极高强度（如超过2250勒克斯）的光照下暴露时间越长，发生整体年龄相关性眼病和青光眼的风险就越高。

Bohemia Interactive 在 GPL v3.0 许可证下公开了《Arma: Cold War Assault》重制版源代码，项目托管在 GitHub 上。《Arma: Cold War Assault》于 2001 年以《Operation Flashpoint: Cold War Crisis》的名字发布，游戏提供了 12.5 km × 12.5 km 开放世界地图，它对于现代化立体化野战的真实模拟为它赢得了一大批军事游戏爱好者拥趸。游戏的开放性以及强大的脚本编程能力，也给它带来了大量 MOD。重制版代码已现代化至 C++20，使用 CMake 和 Clang 构建，并支持 Windows x64 和 Linux x64 等平台。Bohemia Interactive 称，游戏代码是自由软件，但名字和商标并不能自由使用，而且模型、纹理、音效、任务和语音等游戏数据也都没有公开，需要另外购买。

Windows 10 于 2025 年 10 月 14 日结束支持，微软原本此后不再提供免费的安全更新，但 Windows 10 仍然有大量用户使用，软件巨人去年宣布将提供免费安全更新一年。如今还有几个月时间才到期，微软又将免费安全更新延长一年，Windows 10 用户不需要做任何事就能再享受一年免费安全更新。最新的扩展安全更新将于 2027 年 10 月 12 日到期。根据 StatCounter 的统计，有 26% 的 PC 仍然运行 Windows 10，由于微软提高了 Windows 11 的硬件需求，大部分 Windows 10 PC 无法升级到 Windows 11。