OrangeBot.AI Digest — 2026-06-17

OS runtime unifying GA4, PostHog, + Stripe into a local db

Learn new words from the world around you

Powerful computer built into lightweight see-through glasses

See what AI agents do inside your MCP server

Send, sign & manage documents easily

AI replies that are trained to sound like you

Turn any photo into a beautiful wallpaper

Household chores without the mental load for ADHD

Keep PRs, issues, CI, and docs moving with AI agents

AI email client built for focus. Runs locally on your Mac

With Agents, Branching, Community, and an all-new design

Give agents reliable access to 2,000+ APIs w/ durable state

Voice dictation that adapts to what’s on your screen

Record in both 16:9 and 9:16 at the same time

Models matter. Context matters more. Give your agent a plan.

Looped Transformers scale latent computation by repeatedly applying shared blocks, but sequential looping increases latency and KV-cache memory with the loop count. Parallel loop Transformers (PLT) alleviate this cost through cross-loop position offsets (CLP) and shared-KV gated sliding-window attention, making loop count a practical design choice. We therefore study PLT loop-count selection through a gain--cost view: an extra loop may refine representations, but CLP also introduces a positional mismatch at each loop boundary. We instantiate this study by training LoopCoder-v2, a family of 7B PLT coders with different loop counts, from scratch on 18T tokens, followed by matched instruction tuning and evaluation. Empirically, the two-loop variant delivers broad gains over the non-looped baseline across code generation, code reasoning, agentic software engineering, and tool-use benchmarks, improving SWE-bench Verified from 43.0 to 64.4 points and Multi-SWE from 14.0 to 31.0 points. In contrast, variants with three or more loops regress, revealing a strongly non-monotonic loop-count effect. Our diagnostics show that loop 2 provides the main productive refinement, while later loops yield diminishing, oscillatory updates and reduced representational diversity. Because the CLP-induced mismatch remains roughly fixed as refinement gains shrink, the offset cost increasingly dominates. This gain--cost trade-off explains PLT's saturation at two loops and provides diagnostics for loop-count selection.

Vision-Language-Action (VLA) models benefit from large-scale and diverse embodied data, yet scaling robot trajectory collection is costly and labor-intensive. Recent advances show that large-scale egocentric human videos provide complementary real-world supervision in pretraining. However, joint training on human and robot data remains challenging due to divergences in action spaces, embodiment structures, temporal dynamics, and supervision quality. We introduce ACE-EGO-0, a unified VLA pretraining framework jointly leveraging heterogeneous data sources. To extract large-scale pretraining supervision from egocentric human videos, we build a scalable egocentric video-to-action pipeline that converts raw human videos into robot-format pseudo-action trajectories. To make these labels comparable with robot demonstrations, ACE-EGO-0 uses a unified action representation based on camera-space actions, morphology conditioning, and time-aligned action chunking. To robustly leverage noisy pseudo-action supervision from egocentric human videos, we formulate a reliability-aware training objective with a human auxiliary loss that concentrates supervision on reliable signals. We instantiate ACE-EGO-0 on 4.53K hours of robot and simulation data, together with 1.48K hours of pseudo-action-labeled egocentric human data. Experiments show that incorporating large-scale human supervision under reliability-aware weighting consistently improves both unified joint pretraining and supervised fine-tuning. ACE-EGO-0 achieves state-of-the-art performance on RoboCasa GR1 TableTop and RoboTwin 2.0, while demonstrating strong transfer to real-world bimanual manipulation.

Knowledge distillation transfers a teacher's competence to a small student but is brittle in the small-student regime: forcing the student to imitate logits from a much larger teacher concentrates it on the teacher's sharpest modes, hurting generalization on benchmark families beyond the training corpus. Reinforcement learning (RL) avoids logit imitation by training on the student's own rollouts. However, on questions where every rollout fails-yielding zero advantage and being silently discarded-injecting a stronger teacher's response into the policy gradient breaks the on-policy assumption and induces drift. We introduce Zone of Proximal Policy Optimization (ZPPO), inspired by Vygotsky's zone of proximal development, which keeps the teacher inside the prompt rather than the policy gradient. On hard questions, ZPPO constructs two reformulated prompts: a Binary Candidate-included Question (BCQ) pairs one correct teacher response with one incorrect student response as anonymized candidates the student must discriminate, and a Negative Candidate-included Question (NCQ) aggregates the student's wrong rollouts into a single prompt to surface their shared failure modes. A prompt replay buffer recirculates each hard question until it either graduates-the student's mean rollout accuracy on it reaches half- or is FIFO-evicted under finite capacity, amplifying BCQ and NCQ inside the student's current zone of proximal development. On the Qwen3.5 family at four student scales (0.8B-9B) with a 27B teacher, post-trained as vision-language models and evaluated on a 31-benchmark suite (16 VLM, 10 LLM, 5 Video), ZPPO outperforms off/on-policy distillation and GRPO, with the largest gains at the smallest scale.

Game generation is an emerging application of coding agents, requiring models to transform natural-language specifications into playable interactive systems. Unlike traditional coding tasks, game generation takes place within a game engine, where scripts, scenes, assets, rendering, and runtime interactions must jointly produce coherent gameplay. We formalize end-to-end game generation as the problem of producing a complete game artifact that realizes a specification through observable player-game interaction in a target environment. We argue that evaluating this setting requires three desiderata: Engine Grounding, Artifact Completeness, and Interactive Verification. We propose an interaction-grounded evaluation framework that assesses executable gameplay through replayed demonstrations and rubric-guided multimodal judging. We instantiate this framework as GameCraft-Bench, a benchmark comprising 140 Godot tasks across 15 game families. Evaluations of frontier coding agents show that end-to-end game generation remains highly challenging: the strongest agent achieves only 41.46%, and most agents score below 40%. Further analysis reveals that while agents often implement recognizable mechanics, they struggle to deliver complete games with sufficient content, functional visual feedback, and coherent presentation. See https://tongxuluo.github.io/gamecraft-bench-website for demos, code, and data.

Effective personalized AI-assisted learning demands systems that can not only generate accurate learner-specific educational materials, but also dynamically adapt their instruction to diverse learners. However, existing educational agents have primarily focused on lecture content automation and simulations, which often fall short of modelling multimodal and embodied instructional methods tailored for the individual learner. To this end, we propose LectūraAgents - a multi-agent framework that enables personalized learning through end-to-end adaptive embodied teaching. At its core, LectūraAgents mirrors a professor-student relationship, in which a ProfessorAgent leads a collaborative team of specialized subordinate agents through research, planning, review, and embodied delivery of lecture contents that adapt to a learner's needs. The framework offers three main contributions: (1) a hierarchical multi-agent architecture for end-to-end personalized learning; (2) an adaptive embodied teaching mechanism, wherein the ProfessorAgent executes visible and pedagogically motivated teaching actions (e.g., handwrite, highlight, underline, etc.) over contents in a teaching environment; and (3) a Teaching Action-Speech Alignment (TASA) algorithm that employs salience-based heuristics and temporal semantic segmentation to generate coherent teaching action sequences aligned with learner profiles. We evaluate LectūraAgents on diverse courses at high school, undergraduate, and graduate levels using sample-specific rubric-based analysis; with generated lecture materials and teaching actions assessed and validated by expert educators. Experimental results show consistent gains in lecture content quality, embodied teaching quality, assessment, and personalization over existing approaches, positioning LectūraAgents as a pedagogically well-grounded framework for personalized learning at scale.

Clinical early warning systems built on electronic health records, in which clinical observations are recorded as irregularly sampled medical time series (ISMTS), must deliver both calibrated risk scores for patient triage and interpretable rationales that clinicians can verify. Large Language Models (LLMs) have been explored for this task, yet they collapse graded clinical risk into overconfident binary predictions. This risk polarization undermines both calibration and cross-patient comparability. To address this, we propose TRIAGE, a framework that trains an LLM to generate dialectical reasoning over competing clinical outcomes by eliciting outcome-specific rationales. This dialectical formulation mitigates risk polarization, enabling a single LLM to yield continuous risk scores grounded in explicit clinical reasoning. Evaluated on three ISMTS benchmarks, TRIAGE achieves an average AUPRC improvement of 3.3% and reduces calibration error by 81% compared to the competitive baselines. An LLM-as-a-judge assessment further shows that our rationales surpass post-hoc explanations from the baseline by 20% in clinical reasoning quality. The source code is available at https://github.com/HyeongWon-Jang/TRIAGE .

On-policy self-distillation (OPSD) has proven effective for post-training large language models (LLMs), yet its application to diffusion LLMs (dLLMs) remains unexplored. Existing OPSD methods are inherently autoregressive-centric. They inject privileged information via left-to-right prefix conditioning with token-level divergence supervision, a design that fundamentally conflicts with the arbitraryorder generation of dLLMs. We introduce d-OPSD, the first OPSD framework tailored for dLLMs. Our approach makes two core contributions. First, we reframe self-teacher construction by using self-generated answers as suffix conditioning, enabling the student model to learn from "self future-experience" rather than privileged prefixes. Second, we shift supervision from token-level to step-level, aligning training with the iterative denoising process of dLLMs. Experiments across four reasoning benchmarks show that d-OPSD consistently outperforms RLVR and SFT baselines with superior sample efficiency, requiring only around 10% of the optimization steps by RLVR and opening a promising pathway for dLLM posttraining. The code is available at https://github.com/xingzhejun/d-OPSD.

Memory has become a standard substrate for self-evolving agents, yet retaining experience is not the same as learning how to evolve through it. Existing memory agents can store trajectories, retrieve reflections, or accumulate skills, but often lack the holistic competence to select useful experience, act on it, write reusable knowledge, and maintain a growing repository. We introduce OPD-Evolver, a slow-fast co-evolution framework that cultivates such an agent evolver through on-policy self-distillation. In the fast loop, OPD-Evolver interacts with a four-level memory hierarchy to read, use, write, and maintain experience for rapid test-time evolution. In the slow loop, outcome-calibrated memory attribution and privileged hindsight distill these four abilities into the deployable policy. Across multi-domain benchmarks, OPD-Evolver surpasses memory systems such as ReasoningBank by up to 11.5%, and training-based methods such as Skill0 by ~5.8%. Further analysis shows that OPD-Evolver internalizes high-value experience and memory management, enabling OPD-Evolver-9B to challenge giant counterparts such as Qwen3.5-397B-A17B and Step-3.5-Flash, pointing beyond memory-augmented agents toward genuinely qualified agent evolvers.

Pixel-space diffusion models are trained on full-bandwidth noisy images, yet the useful signal available to the denoiser is strongly frequency dependent. Under rectified-flow diffusion and natural-image power-law spectra, the per-band data-to-noise contour k^{*}(t) = (1-t)^{-2/α} separates a signal-bearing low-frequency region from a noise-dominated high-frequency region at each time t. We show that this implicit coarse-to-fine structure is not merely descriptive: it induces a capacity-allocation problem. A standard pixel-space denoiser must discover the moving bandwidth boundary internally and can spend computation on frequency-time regions where the optimal prediction collapses to deterministic baselines rather than data-distribution modeling. To make this boundary explicit, we introduce Spectral Forcing, a parameter-free, time-conditional 2D-DCT low-pass operator applied to the noisy input before the patch embedder. Its cutoff expands monotonically with the diffusion time and becomes the identity at the data endpoint. Through controlled synthetic experiments, we identify the regime in which the operator is beneficial: coarse patch tokenization and data whose high-frequency content is predominantly noise rather than essential signal. On ImageNet-256 with JiT-700M/32, Spectral Forcing consistently improves both FID and Inception Score across different training epochs, demonstrating robust gains throughout training; at finer tokenization, the spectral forcing is still competitive. We further insert the unchanged operator into SenseNova-U1, a unified text-to-image model, where it improves DPG-Bench and GenEval, showing that the input-side spectral prior transfers beyond class-conditional generation. These results suggest a route to capacity-efficient pixel-space diffusion by showing the signal and hiding the noise.

Vision-language models (VLMs) are typically trained as passive answerers, while their ability to actively ask diverse, non-trivial, visual-centric and grounded questions remains underexplored. Existing visual questioners' performance is bottlenecked by the availability of high-quality training data or the cost of curating them. We show that a VLM can continuously improve itself as a visual questioner without any external supervision. We propose a self-evolving framework that uses a VLM itself as both a proposer and a filter to produce harder, more informative, and visual-centric questions, while maintaining their exploration diversity to avoid training collapse. These questions are then used to train the VLM in both questioner and answerer modes. To evaluate the questioner, we introduce an agentic protocol that assesses questions along perception, reasoning, and diversity dimensions. Experiments across various backbone VLMs show that our method substantially enhances the quality and substantially expands the difficulty boundary of autonomous question generation. Under the same budget, our self-supervision is more effective than training on the static source data. Moreover, the self-evolving questioner remains a competitive or even better answerer.

Existing image editing methods can be generally categorized into textual instruction-based and visual prompt-based ones. Textual instructions are semantically expressive, but are limited by the coarse granularity of spatial control of the editing results. In contrast, visual prompts such as drag and point can provide precise spatial guidance, but are limited by the inherent ambiguity in semantic intent. To unify the strength of textual and visual prompts, we present Text-Vision Co-Instructed Image Editing, which jointly models textual instructions as semantic intent and sparse visual instructions as spatial guidance, aiming to achieve precise and intent-faithful image manipulation. To this end, we first construct a textual-visual instruction paired dataset with more than 23K samples derived from dynamic videos, enabling aligned supervision for cross-modal instruction. We then propose TV-Edit, a Textual-Visual instruction unified Editing framework to contextualize drag or point-based visual instructions with image-text semantics and lift them into semantic-aware control representations for pretrained editing backbones. By integrating semantic intent and spatial constraints, TV-Edit leads to more precise spatial control, less instruction ambiguity, and stronger structural consistency than text-only or drag-based alternatives. Finally, we establish TV-Edit-Bench, a deliberately designed benchmark to evaluate semantic faithfulness, spatial alignment, and visual consistency with ground-truth references and controlled textual-visual variations for reliable assessment. Our experiments across multiple editing backbones demonstrate that TV-Edit consistently yields more precise and intent-faithful edits, significantly outperforming state-of-the-art instruction-based and drag-based baselines.

Modern language models increasingly adopt hybrid architectures that combine full attention with efficient attention modules, such as sliding-window attention (SWA) and recurrent sequence mixers. However, how these efficient modules shape model capabilities remains poorly understood. To address this gap, we conduct a systematic analysis across hybrid architectures from three perspectives: scaling behavior, mechanism analysis, and architecture design. First, from a scaling perspective, we find that efficient-attention design primarily affects how fast long-context capability emerges, while different hybrids eventually converge to comparable long-context performance under sufficient training. Second, mechanistically, we show that long-range retrieval is mainly carried by full attention, whereas efficient attention shapes its optimization trajectory. This explains a counter-intuitive phenomenon we call Large-Window Laziness: larger SWA windows can delay the formation of retrieval heads in full-attention layers. Third, guided by this mechanism, we show that applying NoPE to only the full-attention layers of a small-window SWA hybrid substantially improves long-context performance with negligible impact on short-context performance.

The shift from video generation to interactive world modeling places new demands on data: beyond captioned videos, world models require temporally aligned video-action-language trajectories grounded in the actions, camera motion, states, and events that drive future scene changes. However, such data is difficult to obtain at scale. Web video datasets offer broad visual coverage but lack executable actions and reliable states; robotic datasets provide action and state supervision but are costly and limited in scene diversity; and existing simulators often lack large-scale human-driven interaction trajectories. In this paper, we introduce EgoCS-400K, a large-scale replay-grounded egocentric Counter-Strike dataset for world models, built from public professional CS and CS2 match demos that preserve human gameplay trajectories and enable parsing, replaying, rendering, and temporal alignment. We extract player states, view directions, movements, keyboard/button inputs, view-angle changes, weapon usage, game events, and round-level context, and render clean first-person videos from the same trajectories. EgoCS-400K contains over 400,000 first-person videos and 10,000 hours of gameplay from more than 1,000 matches and 40,000 rounds, covering 13 maps and 10 player viewpoints per round. It supports a range of interactive visual modeling tasks, including action-conditioned future prediction, state- and event-aware scene rollout, replay-grounded captioning, and agent egocentric action understanding. By connecting visual observations with human actions, camera motion, game states, and events at scale, EgoCS-400K serves as a practical bridge between passive web videos, controllable game simulation, and costly real-world embodied data.

Agentic search over large corpora relies on retriever-mediated interfaces (e.g., BM25 or ColBERT) for scalable candidate discovery. While effective at ranking relevant documents, these interfaces expose evidence only as ranked results or bounded document views, limiting agents' ability to reorganize material and verify constraints across documents. Direct Corpus Interaction (DCI) addresses this limitation by exposing shell-executable corpus operations for flexible search, filtering, comparison, and verification. However, full-corpus terminal commands become slow and unstable as the corpus grows, degrading performance and efficiency. We introduce DR-DCI, a retriever-steered DCI framework that treats retrieval as an agent-callable action for expanding a local workspace. Rather than operating directly over the full corpus, the agent dynamically pulls relevant documents into an evolving workspace and conducts DCI operations within it. This design combines retriever-level recall with DCI-style precision: retrieval keeps exploration scalable, while DCI preserves the local operations needed for effective evidence resolution. Experiments show that DR-DCI is both effective and efficient across scales. On Browsecomp-Plus, DR-DCI reaches 71.2\% accuracy, improving over raw DCI and ablated variants by up to 8.3 points while reducing tool usage, wall time, and estimated cost. With workspace-preserving context reset, accuracy further improves to 73.3\%. In corpus-scaling experiments, DR-DCI remains effective from 100K to 10M documents, whereas raw DCI becomes unstable and BM25 performs substantially worse. DR-DCI also scales to a 20M-scale file-per-document Wiki-18 QA setting, achieving an average score of 63.0 across six benchmarks and outperforming retrieval-based and trained search-agent baselines. Ablation analysis further shows that ranked previews and inter-document DCI are key to performance.

Current world models face a fundamental tension: faithful long-horizon simulation demands deep computation, but deeper models are expensive to deploy and prone to compounding errors. We resolve this by introducing Looped World Models (LoopWM), which are the first looped architectures for world modelling. Our method iteratively refines latent environment states through a parameter-shared transformer block. This yield up to 100x parameter efficiency over conventional approaches with adaptive computation that automatically scales depth to match the complexity of each prediction step. Orthogonal to scaling model size and training data, LoopWM establishes iterative latent depth as a new scaling axis for world simulation, which might significantly push the community forward.

Arseniy Shestakov : Apple's planned move of Hide My Email aliases to private.icloud.com will let services easily distinguish them from normal iCloud email addresses and block them —  a small and unimportant announcement appeared in Apple developer news: New domain for Sign in with Apple and iCloud+ Hide My Email.

Sarah Perez / TechCrunch : Mastodon adds email newsletters, letting writers send their posts directly to subscribers' inboxes, even to readers without a Mastodon account —  Mastodon, the open, decentralized alternative to Big Tech apps like X and Threads, is betting that email could help solve the open social web's biggest problem: audience growth.

Rolfe Winkler / Wall Street Journal : In an interview, Tim Cook says Apple price hikes are “unavoidable” to offset surging memory and storage chip costs, and “the situation has become unsustainable” —  The CEO tells the Journal in an exclusive interview that soaring costs make price increases ‘unavoidable’

Lucas Ropek / TechCrunch : Snap's stock closed down 8.14% on Wednesday, after the company launched the $2,195 Specs AR glasses on Tuesday; SNAP is down ~41% YTD —  Snap's long-awaited AR glasses, Specs, didn't have the best debut.  —  The company's stock hasn't been on the healthiest trajectory lately.

Joey Cappelletti / Associated Press : Bernie Sanders proposes legislation to create a sovereign wealth fund financed via a one-time 50% stock tax on AI companies that reach $200M in annual AI sales —  As artificial intelligence companies reshape the economy and race toward trillion-dollar valuations, Sen. Bernie Sanders …

Mark Gurman / Bloomberg : Sources: Apple is testing a second-generation iPhone Air, planned for spring 2027, with a second rear camera for ultrawide-angle photos and better battery life —  Apple Inc. is preparing a second-generation iPhone Air for spring 2027, aiming to boost the appeal of the slimmed-down device …

Michael Nuñez / VentureBeat : Anthropic updates Claude Design with design system imports, bidirectional integration with Claude Code, lower token consumption, and more export destinations —  When Anthropic quietly released Claude Design in April as a “research preview,” it generated the kind of instant traction …

Andy Robinson / Video Games Chronicle : Epic says Unreal Engine 6, planned for early access in late 2027, will unify UE5 and Unreal Editor for Fortnite, add integrations with Claude and Gemini, more —  EPIC CLAIMS UE6 IS “GOING TO CHANGE A LOT ABOUT HOW GAMES ARE MADE”  —  Epic has detailed Unreal Engine 6 …

Financial Times : Sources: Amodei, Altman, and Hassabis called for US-led collaboration on AI rules at the G7 summit; Macron and Modi raised concerns over the US block on Mythos —  Dario Amodei is supported by rival Sam Altman in call for international co-operation  —  Anthropic chief executive Dario Amodei …

Kai Nicol-Schwarz / CNBC : CEOs of Anthropic and Google DeepMind call for U.S.-led AI coalition in meeting at G7 … Anthropic CEO Dario Amodei and Google DeepMind's Demis Hassabis called for a U.S.-led coalition to shape rules and standards around artificial intelligence at a meeting with tech leaders and heads of state …

Lawrence Abrams / BleepingComputer : A newly discovered data leak has exposed what appears to be a collection of Fortinet and FortiGate VPN credentials for 73,932 firewall URLs across 194 countries —  A newly discovered data leak dubbed “FortiBleed” has exposed what appears to be a collection of Fortinet and FortiGate VPN credentials …

Ryan Merket / RuntimeWire : Joshua Baer, the founder and CEO of Texas accelerator Capital Factory, died on Tuesday night in a business jet crash in Laredo, Texas —  Joshua Baer (@joshuabaer), the founder and CEO of Capital Factory, died Tuesday night in a business jet crash in Laredo, Texas, KXAN reported …

Ingrid Lunden / Resilience Media : Paris-based Comand AI, which is developing AI-based command-and-control software for military operations, raised a €32M Series A led by Blossom Capital —  French startup emerges at a time when tech sovereignty has become a major talking point in defence

Michael Peel / Financial Times : Studies: Mira, an AI medical tool developed by researchers in Germany, and Google's Amie matched or surpassed doctors on diagnostic and treatment decisions —  Two health models displayed clinical value across a range of diagnostic and treatment decisions, studies show

Tim Bradshaw / Financial Times : Q&A with Luciana Lixandru, who co-leads Sequoia's global early-stage investment business, on why it is time for “act two” for Europe's tech sector, AI, and more —  The global co-lead of the US venture capital firm's early-stage investment business talks about why it is time for ‘act two’ for Europe's tech sector

Epic Games 宣布了新版本控制系统 Lore，源代码采用 MIT 许可证托管在 GitHub 上。Git 是最流行的版本控制系统，但它最初的是为 Linux 这一大型去中心化项目设计的，并没有为游戏或封闭环境下的大型私有软件开发优化。Git 不太适合游戏公司的纹理、3D 模型、音频等文件的协同开发，因此游戏领域流行的版本控制系统是私有的 Perforce，开源的 Lore 瞄准的就是该私有软件。Epic Games 称，“Lore是一个集中式、内容寻址的版本控制系统，使用默克尔树和不可变的版本链来表示仓库状态，并针对二进制优先存储、重复数据删除以及大规模的稀疏/按需数据水合进行了优化。”

根据 WordPress VIP 的报告《Future of the Web Report》，六成美国消费者对品牌信息中的 AI 表示反感。74% 的消费者认为今天的互联网没有 10 年前有人味；普通人冲浪 40 分钟就会产生在线互动缺乏真实感的感受——这被称为 Bot fatigue；16% 的消费者认为没有品牌真正有效利用了 AI，六成消费者认为品牌信息中的 AI 会让人倒胃口。

大量研究表明 GLP-1 药物不仅仅能减肥，它几乎无所不能。根据发表在《Criminology》期刊上的一项新研究，GLP-1 减肥药有助于抑制暴力冲动。研究人员强调这是一项观察性研究，并没有证明两者之间存在因果。GLP-1 药物在减轻体重过程中除了降低食欲外还会对行为产生影响，比如遏制对酒精的渴望。这一结果可能源于药物对冲动控制和奖赏处理感知的影响。而冲动和酒精饮用都是公认的暴力行为风险因素。研究人员分析了 7521 名美国成年人的调查数据，其中 821 人曾服用过 GLP-1 减肥药，597 人正在服用该药，受访者被询问了饮酒和冲动行为。结果显示正在服用 GLP-1 药物的人中冲动行为和暴力行为之间的关联减弱了 62%，饮酒行为与暴力行为之间的关联性减弱了 52%。

俄罗斯安全公司卡巴斯基对中俄 Steam 用户发出警告，恶意墙纸正在 Steam 创意工坊快速扩散，其目的是劫持他们的账号。攻击者利用了热门墙纸应用 Wallpaper Engine 创意工坊分享功能的漏洞，恶意程序隐藏在分享的壁纸包中。运行被感染的壁纸会导致 Steam 账号被盗，或者系统被植入后门或加密货币挖矿程序。安全研究人员在创意工坊发现了数十款恶意壁纸，每一款都被下载了数千次，甚至数万次。黑客主要针对中国 Steam 用户，墙纸的艺术风格和标题都专门针对中国玩家量身定制，中国玩家的下载量最多，占到了总下载量的  89.4%，其次是俄罗斯的 5.5%，新加坡 (1.4%)、香港 (0.9%)、德国 (0.9%)、越南 (0.9%)、印度 (0.5%) 和加拿大 (0.5%)。Steam 目前已经移除了包含恶意程序的墙纸。

Firefox 浏览器从 v151 开始，Gzip 压缩/解压缩就依赖于 zlib-rs 库，用 Rust 语言开发的版本替代了 C 语言版本改进了性能，提供了更好的内存安全性，以及带来了英特尔第 13 代/第 14 代酷睿 CPU 不稳定导致的崩溃问题。致力于用 Rust 语言重写关键库的非盈利组织 Trifecta Tech Foundation 在 2024 年夏天就与 Mozilla 讨论在浏览器中集成 zlib-rs，但从测试到落地花了两年时间，一个重要原因就是 zlib-rs 触发了臭名昭著的英特尔 CPU bug。测试中 zlib-rs 中的一些代码导致英特尔 Raptor Lake CPU 频繁崩溃，开发者最终发现问题与 Huffman 编码写入内存的一个特定指令相关，识别问题之后解决起来就容易了，开发者通过加入一段“不安全代码”修复了该问题。

泄漏财务数据显示 2025 年 OpenAI 净亏损约 80 亿美元。数据显示，OpenAI 的营收从 2024 年的 37 亿美元增至 2025 年的 130.7 亿美元。研发支出从 2024 年的 78.1 亿美元飙升至 2025 年的 191.8 亿美元，其中仅支付给微软的研发费用就高达 105.9 亿美元。产品生产和分销支出从 2024 年的 26.5 亿美元增至 2025 年的 75 亿美元。销售和市场营销支出从 2024 年的 11.1 亿美元增至 2025 年的 57.3 亿美元。OpenAI 的运营亏损从 2024 年的 87.8 亿美元增至 2025 年的 209.2 亿美元，净亏损从 2024 年略高于 50 亿美元飙升至 2025 年的近 390 亿美元。但其中包含了一笔大约 300 亿美元的从非盈利结构转为盈利性结构的估值相关会计支出，如果不计入这笔费用，OpenAI 在 2025 年净亏损约为 80 亿美元。OpenAI 披露 ChatGPT 周活跃用户逾 9 亿，但付费用户只有 5000 万。

根据内分泌学会年会上发表的报告，多项研究显示 GLP-1 减肥药有助于提高男性睾酮水平和精子质量。一项研究对 1600多 名开具减肥药处方的男性患者的电子健康记录进行了分析，发现在接受 GLP-1 药物或双重激素受体激动剂治疗后，参与者的睾酮水平增加了约 30%。另一项回顾性研究同样分析了 215 名接受减肥药物治疗男性的记录，发现治疗后他们的平均睾酮水平比治疗前高出约 20%。睾酮是精子产生和维持生育能力不可或缺的激素，而肥胖会降低睾酮水平已是医学界的共识。脂肪细胞中含有高水平的酶，能将睾酮转化为主要的女性性激素雌二醇。此外肥胖引起的代谢变化和体内炎症水平升高也会直接影响睾酮的产生。当 GLP-1 药物帮助患者有效减重时，这些负面因素也随之减弱，从而促使生殖激素网络恢复正常。

根据发表在《科学》期刊上的一项研究，地下真菌网络长度达到 11 万万亿公里(或 110 京公里，1 京等于 1 千万亿)，是地日距离的 7.5 亿倍。丛枝菌根真菌（Arbuscular mycorrhizal fungi）是由被称为菌丝的管状细胞构成的网络。它们通过与逾七成的植物建立共生关系维系着地球上的生命。这种网络已存在约 4.75 亿年，它们通过向植物提供养分和水分换取植物产生的碳，它们还通过将碳吸收到土壤中帮助调节气候。Society for the Protection of Underground Networks(Spun)组织的研究团队利用机器学习模型，结合世界各地逾 16000 个土壤样本的数据，绘制出第一张丛枝菌根真菌网络的全球地图。研究人员称，仅仅一茶匙土壤就可能存在长达 10 米的菌根网络。研究还发现，农耕会破坏真菌网络，农田菌根网络密度平均比野生生态系统低 47.3%。草原地区拥有最密集的菌丝系统，但这些地区缺乏保护，正日益退化。

Mozilla 在宣布 Firefox 152 的同时，公布了将在未来推出的一系列新功能，其中包括：更新 UI 的 Project Nova；自定义快捷键；改进 PDF 编辑功能——支持在浏览器上直接拆分、合并和重组 PDF 文档；Multi-Account Containers 从扩展变成原生功能；移动版本将内置免费 VPN（可能只限于少数国家）；通过语音向浏览器提问获得 AI 生成答案的 Quick Answers；隐私 AI 浏览 Smart Window；省电模式（Power Saving Mode）识别手机上消耗资源最多的标签页，自动降低其资源占用，从而延长电池续航时间，等等。

根据 Sensor Tower 的《State of AI Report for 2026》报告，在 ChatGPT 发布三年半之后，其市场份额首次跌破 50%，而用户正在 Google Gemini、Anthropic Claude 等不同 AI 助手之间切换。ChatGPT 是最快达到 10 亿月活用户的应用，它的月活用户目前超过 11 亿，之后是 Gemini 的 6.62 亿 和 Claude 的 2.45 亿。ChatGPT 在今年 1 月市场份额还超过 50%，但到了 5 月底降至 46.4%，Gemini 占 27.7% 和 Claude 占 10.3%，Grok、Perplexity、DeepSeek 和 Meta AI 都低于 5%。 Sensor Tower 估计，2026 年上半年，AI 应用下载量预计将接近 23 亿次，用户支出将超过 42 亿美元。相比之下 2025 年上半年的 AI 支出为 18.3 亿美元——这表明 AI 行业正将重心从增长转向盈利。但下载量和支出增长率均已放缓，表明即使绝对数量在继续攀升，市场可能正走向成熟。中国和印度的 AI 应用下载量出现了下滑，2026 年第一季度亚洲下载量下降了 3.3%。

最大化词元使用（tokenmaxxing）对微软的 AI 工具 Copilot 产生了不利影响，软件巨人正在考虑使用 DeepSeek 的开源模型以降低成本。微软考虑使用的是 DeepSeek-V4 自托管版本的修改版，它将作为一种低成本选项用于驱动微软的 Copilot Cowork。Copilot Cowork 目前运行在 Anthropic 和 OpenAI 的模型上，两家公司不断涨价，Copilot 也从无限量使用切换到了基于使用量的定价模式，此举招致了用户的强烈不满。更便宜的型号有助于降低成本让用户满意，但可能会让特朗普政府不满意。

黑客行动主义者曝光了 Peter Thiel 于 2006 年创办的秘密组织 Dialog 的成员信息和内部记录。Dialog 每年都会组织非公开的活动，邀请美国官员、外国政府官员和硅谷高管参加。2026 年活动定于 8 月 12 日至 16 日在爱尔兰都柏林的 Powerscourt Hotel 举行，注册的与会者共 222 人，并标明了他们是活跃会员还是特邀嘉宾（guest），会议讨论的主题包括“金钱能买到幸福吗？”“恢复核能”“应对第三次世界大战”“性生活”“建立邪教（Build-a-Cult）”“建立一个党（Build-a-Party）”等等。2025 年上任的北约欧洲盟军指挥官 Alexus Grynkewich 将军自 2021 年起就参与了 Dialog 活动。名单中的名人还有 Palantir 联合创始人 Joe Lonsdale、美财长 Scott Bessent、参议员 Ted Cruz、陆军部长 Dan Driscoll、参与监管 Pantir 的众议员 Jim Hime，以及 Google 和 DeepMind 高管、经济学诺奖得主 Roger Myerson 等。Dialog 还扮演了某种媒人的角色，询问与会者是否在寻找爱情，还提供了约会应用，该应用的口号是“为杰出人士建立有意义的联系”。Dialog 在 2014 年的活动邀请了 Jeffrey Epstein，但并不知道他有没有出席。

气候变化和城市化加剧了热岛效应，城市地区的温度显著高于农村地区，而更高的温度又推动了制冷需求和加剧了电网压力，形成某种恶性循环。日本大阪府大学 Jihui Yuan 副教授领导的团队调查了垂直绿化等城市降温策略。他们的研究显示，朝南绿墙可将室内热条件改善最多 1.7°C；低反照率外表面能改善室外热舒适度最多 1.5°C；高反照率外表面则有助于降低室内温度。

GLP-1 减肥药如 Ozempic、Wegovy、Rybelsus 能快速降低体重，此前有担忧认为快速的体重下降可能导致骨质疏松，增加骨折风险。然而最新研究发现，相比其它起效较慢的减肥药，GLP-1 减肥药能将骨折风险降低 15%。研究人员承认需要更多研究去证实相关性。研究人员分析了逾 59,000 名患者，其中 26,324 名服用了 GLP-1 减肥药，对照组的 33,555 人服用的是非 GLP-1 减肥药。结果显示，实验组发生 794 例骨折，对照组则发生 1045 例。

根据亚马逊公布的数据，它的数据中心在 2025 年使用了 25 亿加仑的水。电商巨人声称它的用水量远低于主要竞争对手。亚马逊称，其数据中心用水量为每千瓦时 0.12 升（L/kWh），称微软在 2025 年的用水量为每千瓦时 0.27 升，Meta 在 2024 年的用水量为每千瓦时 0.19 升，Google 最糟糕达到每千瓦时 1.15 升。亚马逊表示，其设施约 90% 的时间都采用“自然空气冷却”，即引入室外空气使其流经服务器吸收热量，无需用水——但在最炎热的天气里会使用水蒸发降温。