OrangeBot.AI Digest — 2026-04-29

Free and open‑source, stop designing. Describe.

Ship data-driven apps without breaking flow

Vibe-train evals and guardrails tailored to your use case

Screenshot, record, annotate & edit video in on app

Notes, money, and health. Sorted.

Analytical skills for data agents running on Supabase

Generate UI from your design system, not around it

Keep AI-generated code healthy and maintainable

Run your own Claude Code in your pocket.

Spot signals, trigger outreach - turn posts into pipeline

Picsart's power right from your AI chat box

Everything users expect from modern chat. Out of the box.

PII guard for Claude Code to keep client data out of context

Annotate any doc, URL, or folder - send feedback to agents

Your full venture strategy, built in minutes.

Recursive or looped language models have recently emerged as a new scaling axis by iteratively refining the same model computation over latent states to deepen reasoning. We extend such scaling principle from a single model to multi-agent systems, and ask: Can agent collaboration itself be scaled through recursion? To this end, we introduce RecursiveMAS, a recursive multi-agent framework that casts the entire system as a unified latent-space recursive computation. RecursiveMAS connects heterogeneous agents as a collaboration loop through the lightweight RecursiveLink module, enabling in-distribution latent thoughts generation and cross-agent latent state transfer. To optimize our framework, we develop an inner-outer loop learning algorithm for iterative whole-system co-optimization through shared gradient-based credit assignment across recursion rounds. Theoretical analyses of runtime complexity and learning dynamics establish that RecursiveMAS is more efficient than standard text-based MAS and maintains stable gradients during recursive training. Empirically, we instantiate RecursiveMAS under 4 representative agent collaboration patterns and evaluate across 9 benchmarks spanning mathematics, science, medicine, search, and code generation. In comparison with advanced single/multi-agent and recursive computation baselines, RecursiveMAS consistently delivers an average accuracy improvement of 8.3%, together with 1.2times-2.4times end-to-end inference speedup, and 34.6%-75.6% token usage reduction. Code and Data are provided in https://recursivemas.github.io.

Reliably transferring specialized human knowledge from text into large language models remains a fundamental challenge in artificial intelligence. Fine-tuning on domain corpora has enabled substantial capability gains, but the process operates without feedback: when a model fails on a domain task, there is no method to diagnose what is deficient in the training data, and the only recourse is to add more data indiscriminately. Here we show that when a structured knowledge representation extracted from the source corpus serves as the shared foundation for both training data and evaluation, the complete data-engineering lifecycle maps onto the software development lifecycle in a precise and operative way: training data becomes source code specifying what the model should learn, model training becomes compilation, benchmarking becomes unit testing, and failure-driven data repair becomes debugging. Under this correspondence, model failures decompose into concept-level gaps and reasoning-chain breaks that can be traced back to specific deficiencies in the data and repaired through targeted patches, with each repair cycle producing consistent improvements across model scales and architectures without degrading general capabilities. We formalize this principle as Programming with Data and instantiate it across sixteen disciplines spanning the natural sciences, engineering, biomedicine, and the social sciences, releasing a structured knowledge base, benchmark suite, and training corpus as open resources. By demonstrating that the relationship between training data and model behaviour is structurally traceable and systematically repairable, this work establishes a principled foundation for the reliable engineering of human expertise into language models.

Real-world data visualization (DV) requires native environmental grounding, cross-platform evolution, and proactive intent alignment. Yet, existing benchmarks often suffer from code-sandbox confinement, single-language creation-only tasks, and assumption of perfect intent. To bridge these gaps, we introduce DV-World, a benchmark of 260 tasks designed to evaluate DV agents across real-world professional lifecycles. DV-World spans three domains: DV-Sheet for native spreadsheet manipulation including chart and dashboard creation as well as diagnostic repair; DV-Evolution for adapting and restructuring reference visual artifacts to fit new data across diverse programming paradigms and DV-Interact for proactive intent alignment with a user simulator that mimics real-world ambiguous requirements. Our hybrid evaluation framework integrates Table-value Alignment for numerical precision and MLLM-as-a-Judge with rubrics for semantic-visual assessment. Experiments reveal that state-of-the-art models achieve less than 50% overall performance, exposing critical deficits in handling the complex challenges of real-world data visualization. DV-World provides a realistic testbed to steer development toward the versatile expertise required in enterprise workflows. Our data and code are available at https://github.com/DA-Open/DV-World{this project page}.

Autonomous scientific research is significantly advanced thanks to the development of AI agents. One key step in this process is finding the right scientific literature, whether to explore existing knowledge for a research problem, or to acquire evidence for verifying assumptions and supporting claims. To assess AI agents' capability in driving this process, we present AutoResearchBench, a dedicated benchmark for autonomous scientific literature discovery. AutoResearchBench consists of two complementary task types: (1) Deep Research, which requires tracking down a specific target paper through a progressive, multi-step probing process, and (2) Wide Research, which requires comprehensively collecting a set of papers satisfying given conditions. Compared to previous benchmarks on agentic web browsing, AutoResearchBench is distinguished along three dimensions: it is research-oriented, calling for in-depth comprehension of scientific concepts; literature-focused, demanding fine-grained utilization of detailed information; and open-ended, involving an unknown number of qualified papers and thus requiring deliberate reasoning and search throughout. These properties make AutoResearchBench uniquely suited for evaluating autonomous research capabilities, and extraordinarily challenging. Even the most powerful LLMs, despite having largely conquered general agentic web-browsing benchmarks such as BrowseComp, achieve only 9.39% accuracy on Deep Research and 9.31% IoU on Wide Research, while many other strong baselines fall below 5%. We publicly release the dataset and evaluation pipeline to facilitate future research in this direction. We publicly release the dataset, evaluation pipeline, and code at https://github.com/CherYou/AutoResearchBench.

Unified multi-modal understanding/generative models have shown improved image editing performance by incorporating fine-grained understanding into their Chain-of-Thought (CoT) process. However, a critical question remains underexplored: what forms of CoT and training strategy can jointly enhance both the understanding granularity and generalization? To address this, we propose Meta-CoT, a paradigm that performs a two-level decomposition of any single-image editing operation with two key properties: (1) Decomposability. We observe that any editing intention can be represented as a triplet - (task, target, required understanding ability). Inspired by this, Meta-CoT decomposes both the editing task and the target, generating task-specific CoT and traversing editing operations on all targets. This decomposition enhances the model's understanding granularity of editing operations and guides it to learn each element of the triplet during training, substantially improving the editing capability. (2) Generalizability. In the second decomposition level, we further break down editing tasks into five fundamental meta-tasks. We find that training on these five meta-tasks, together with the other two elements of the triplet, is sufficient to achieve strong generalization across diverse, unseen editing tasks. To further align the model's editing behavior with its CoT reasoning, we introduce the CoT-Editing Consistency Reward, which encourages more accurate and effective utilization of CoT information during editing. Experiments demonstrate that our method achieves an overall 15.8% improvement across 21 editing tasks, and generalizes effectively to unseen editing tasks when trained on only a small set of meta-tasks. Our code, benchmark, and model are released at https://shiyi-zh0408.github.io/projectpages/Meta-CoT/

Unified multimodal models (UMMs) integrate visual understanding and generation within a single framework. For text-to-image (T2I) tasks, this unified capability allows UMMs to refine outputs after their initial generation, potentially extending the performance upper bound. Current UMM-based refinement methods primarily follow a refinement-via-editing (RvE) paradigm, where UMMs produce editing instructions to modify misaligned regions while preserving aligned content. However, editing instructions often describe prompt-image misalignment only coarsely, leading to incomplete refinement. Moreover, pixel-level preservation, though necessary for editing, unnecessarily restricts the effective modification space for refinement. To address these limitations, we propose Refinement via Regeneration (RvR), a novel framework that reformulates refinement as conditional image regeneration rather than editing. Instead of relying on editing instructions and enforcing strict content preservation, RvR regenerates images conditioned on the target prompt and the semantic tokens of the initial image, enabling more complete semantic alignment with a larger modification space. Extensive experiments demonstrate the effectiveness of RvR, improving Geneval from 0.78 to 0.91, DPGBench from 84.02 to 87.21, and UniGenBench++ from 61.53 to 77.41.

In this work, we propose Mutual Forcing, a framework for fast autoregressive audio-video generation with long-horizon audio-video synchronization. Our approach addresses two key challenges: joint audio-video modeling and fast autoregressive generation. To ease joint audio-video optimization, we adopt a two-stage training strategy: we first train uni-modal generators and then couple them into a unified audio-video model for joint training on paired data. For streaming generation, we ask whether a native fast causal audio-video model can be trained directly, instead of following existing streaming distillation pipelines that typically train a bidirectional model first and then convert it into a causal generator through multiple distillation stages. Our answer is Mutual Forcing, which builds directly on native autoregressive model and integrates few-step and multi-step generation within a single weight-shared model, enabling self-distillation and improved training-inference consistency. The multi-step mode improves the few-step mode via self-distillation, while the few-step mode generates historical context during training to improve training-inference consistency; because the two modes share parameters, these two effects reinforce each other within a single model. Compared with prior approaches such as Self-Forcing, Mutual Forcing removes the need for an additional bidirectional teacher model, supports more flexible training sequence lengths, reduces training overhead, and allows the model to improve directly from real paired data rather than a fixed teacher. Experiments show that Mutual Forcing matches or surpasses strong baselines that require around 50 sampling steps while using only 4 to 8 steps, demonstrating substantial advantages in both efficiency and quality. The project page is available at https://mutualforcing.github.io.

Recent advancements in large audio language models have extended Chain-of-Thought (CoT) reasoning into the auditory domain, enabling models to tackle increasingly complex acoustic and spoken tasks. To elicit and sustain these extended reasoning chains, the prevailing paradigm -- driven by the success of text-based reasoning models -- overwhelmingly relies on Reinforcement Learning with Verified Rewards (RLVR). However, as models are strictly optimized to distill rich, continuous auditory contexts into isolated, verifiable text labels, a fundamental question arises: are we fostering true audio intelligence, or merely reducing a continuous sensory medium into a discrete puzzle? We identify this as the "verifiable reward trap." While RLVR yields remarkable scores on standardized objective benchmarks, it systematically degrades the real-world conversational feel of audio models. By prioritizing isolated correctness over acoustic nuance, RLVR reduces dynamic interactions to mechanical "answering machines," severely compromising prosodic naturalness, emotional continuity, and user immersion, particularly in long-turn dialogues. To bridge the gap between mechanical objective verification and genuine sensory empathy, we introduce Step-Audio-R1.5, marking a paradigm shift toward Reinforcement Learning from Human Feedback (RLHF) in audio reasoning. Comprehensive evaluations demonstrate that Step-Audio-R1.5 not only maintains robust analytical reasoning but profoundly transforms the interactive experience, redefining the boundaries of deeply immersive long-turn spoken dialogue.

While diffusion models generate high-fidelity video clips, transforming them into coherent storytelling engines remains challenging. Current agentic pipelines automate this via chained modules but suffer from semantic drift and cascading failures due to independent, handcrafted prompting. We present Co-Director, a hierarchical multi-agent framework formalizing video storytelling as a global optimization problem. To ensure semantic coherence, we introduce hierarchical parameterization: a multi-armed bandit globally identifies promising creative directions, while a local multimodal self-refinement loop mitigates identity drift and ensures sequence-level consistency. This balances the exploration of novel narrative strategies with the exploitation of effective creative configurations. For evaluation, we introduce GenAD-Bench, a 400-scenario dataset of fictional products for personalized advertising. Experiments demonstrate that Co-Director significantly outperforms state-of-the-art baselines, offering a principled approach that seamlessly generalizes to broader cinematic narratives. Project Page: https://co-director-agent.github.io/

On-policy distillation (OPD) has shown strong potential for transferring reasoning ability from frontier or domain-specific models to smaller students. While effective on static single-turn tasks, its behavior in multi-turn agent settings remains underexplored. In this work, we identify a key limitation of vanilla OPD in such settings, which we term Trajectory-Level KL Instability. Specifically, we observe that KL divergence increases together with a drop in success rate, and even after convergence, the KL remains high, leading to unstable training. This instability arises from inter-turn error compounding: as errors accumulate, the student is driven beyond the teacher's effective support, rendering the supervision signal unreliable. To address this, we propose TCOD (Temporal Curriculum On-Policy Distillation), a simple yet effective framework that controls the trajectory depth exposed to the student and progressively expands it from short to long with a curriculum schedule.Experimental results across four student-teacher pairs on three multi-turn agent benchmarks (ALFWorld, WebShop, ScienceWorld) show that TCOD mitigates KL escalation and enhances KL stability throughout training, improving agent performance by up to 18 points over vanilla OPD. Further evaluations show that TCOD can even surpass the teacher's performance and generalize to tasks on which the teacher fails.

Deploying guardrails for custom policies remains challenging, as generic safety models fail to capture task-specific requirements, while prompting LLMs suffers from inconsistent boundary-case performance and high inference costs. Training custom classifiers achieves both accuracy and efficiency, yet demands substantial labeled data that is costly to obtain. We present BARRED (Boundary Alignment Refinement through REflection and Debate), a framework for generating faithful and diverse synthetic training data using only a task description and a small set of unlabeled examples. Our approach decomposes the domain space into dimensions to ensure comprehensive coverage, and employs multi-agent debate to verify label correctness, yielding a high-fidelity training corpus. Experiments across diverse custom policies demonstrate that small language models finetuned on our synthetic data consistently outperform state-of-the-art proprietary LLMs (including reasoning models) and dedicated guardrail models. Ablation studies confirm that both dimension decomposition and debate-based verification are critical for ensuring the diversity and label fidelity required for effective fine-tuning. The BARRED framework eliminates the reliance on extensive human annotation, offering a scalable solution for accurate custom guardrails.

Terminal agents have demonstrated strong potential for autonomous command-line execution, yet their training remains constrained by the scarcity of high-quality and diverse execution trajectories. Existing approaches mitigate this bottleneck by synthesizing large-scale terminal task instances for trajectory sampling. However, they primarily focus on scaling the number of tasks while providing limited control over the diversity of execution trajectories that agents actually experience during training. In this paper, we present SkillSynth, an automated framework for terminal task synthesis built on a scenario-mediated skill graph. SkillSynth first constructs a large-scale skill graph, where scenarios serve as intermediate transition nodes that connect diverse command-line skills. It then samples paths from this graph as abstractions of real-world workflows, and uses a multi-agent harness to instantiate them into executable task instances. By grounding task synthesis in graph-sampled workflow paths, SkillSynth explicitly controls the diversity of minimal execution trajectories required to solve the synthesized tasks. Experiments on Terminal-Bench demonstrate the effectiveness of SkillSynth. Moreover, task instances synthesized by SkillSynth have been adopted to train Hy3 Preview, contributing to its enhanced agentic capabilities in terminal-based settings.

Creating interactive STEM courseware traditionally requires HTML/CSS/JavaScript expertise, leaving barriers for educators. While generative AI can produce HTML codes, existing tools generate static presentations rather than interactive simulations, struggle with long documents, and lack pedagogical accuracy mechanisms. Furthermore, full regeneration for modifications requires 200--600 seconds, disrupting creative flow. We present MAIC-UI, a zero-code authoring system that enables educators to create and rapidly edit interactive courseware from textbooks, PPTs, and PDFs. MAIC-UI employs: (1) structured knowledge analysis with multi-modal understanding to ensure pedagogical rigor; (2) a two-stage generate-verify-optimize pipeline separating content alignment from visual refinement; and (3) Click-to-Locate editing with Unified Diff-based incremental generation achieving sub-10-second iteration cycles. A controlled lab study with 40 participants shows MAIC-UI reduces editing iterations (4.9 vs. 7.0) and significantly improves learnability and controllability compared to direct Text-to-HTML generation. A three-month classroom deployment with 53 high school students demonstrates that MAIC-UI fosters learning agency and reduces outcome disparities -- the pilot class achieved 9.21-point gains in STEM subjects compared to -2.32 points in control classes. Our code is available at https://github.com/THU-MAIC/MAIC-UI.

Aligning denoising generative models with human preferences or verifiable rewards remains a key challenge. While policy-gradient online reinforcement learning (RL) offers a principled post-training framework, its direct application is hindered by the intractable likelihoods of these models. Prior work therefore either optimizes an induced Markov decision process (MDP) over sampling trajectories, which is stable but inefficient, or uses likelihood surrogates based on the diffusion evidence lower bound (ELBO), which have so far underperformed on visual generation. Our key insight is that the ELBO-based approach can, in fact, be made both stable and efficient. By reducing surrogate variance and controlling gradient steps, we show that this approach can beat MDP-based methods. To this end, we introduce Variational GRPO (V-GRPO), a method that integrates ELBO-based surrogates with the Group Relative Policy Optimization (GRPO) algorithm, alongside a set of simple yet essential techniques. Our method is easy to implement, aligns with pretraining objectives, and avoids the limitations of MDP-based methods. V-GRPO achieves state-of-the-art performance in text-to-image synthesis, while delivering a 2times speedup over MixGRPO and a 3times speedup over DiffusionNFT.

AI agents are increasingly deployed on complex, domain-specific workflows -- navigating enterprise web applications that require dozens of clicks and form fills, orchestrating multi-step research pipelines that span search, extraction, and synthesis, automating code review across unfamiliar repositories, and handling customer escalations that demand nuanced domain knowledge. Each new task domain requires painstaking, expert-driven harness engineering: designing the prompts, tools, orchestration logic, and evaluation criteria that make a foundation model effective. We present a two-level framework that automates this process. At the first level, the Harness Evolution Loop optimizes a worker agent's harness H for a single task: a Worker Agent W_{H} executes the task, an Evaluator Agent V adversarially diagnoses failures and scores performance, and an Evolution Agent E modifies the harness based on the full history of prior attempts. At the second level, the Meta-Evolution Loop optimizes the evolution protocol Λ= (W_{H}, H^{(0)}, V, E) itself across diverse tasks, learning a protocol Λ^{(text{best)} that enables rapid harness convergence on any new task -- so that adapting an agent to a novel domain requires no human harness engineering at all.} We formalize the correspondence to meta-learning and present both algorithms. The framework shifts manual harness engineering into automated harness engineering, and takes one step further -- automating the design of the automation itself.

Bloomberg : Sources: Anthropic has begun weighing a new funding round at a $900B+ valuation, after previously resisting investor proposals at an $800B+ valuation —  Anthropic PBC has begun weighing a fresh funding round that would value the artificial intelligence developer at more than $900 billion …

Jonathan Vanian / CNBC : Meta says Q1 family daily active people fell 20M QoQ to 3.56B, vs. 3.62B est., due to internet disruptions in Iran and WhatsApp access restrictions in Russia —  Meta shares fell in extended trading on Wednesday after the company reported lower-than-expected capital expenditures and missed on user growth.

Alphabet : Alphabet reports Q1 revenue up 22% YoY to $109.9B, vs. $107.2B est., Google Cloud revenue up 63% to $20B, vs. $18.05B est., net income up 81% to $62.58B —  Alphabet Announces First Quarter 2026 Results MOUNTAIN VIEW, Calif. - April 29, 2026 - Alphabet Inc. (NASDAQ: GOOG, GOOGL) …

Sarah Perez / TechCrunch : Google says paid subscriptions reached 350M in Q1, up 25M QoQ, driven by YouTube and Google One, while Gemini Enterprise paid MAUs grew 40% QoQ —  Google has added another 25 million paid subscriptions to its services over the past quarter, according to parent company Alphabet's first-quarter earnings on Wednesday.

Kelly Cloonan / Wall Street Journal : eBay reports Q1 revenue up 19% YoY to $3.09B, vs. $3.04B est., net income up 2% to $512M, GMV up 18% to $22.2B, and forecasts Q2 revenue above estimates —  The company logged a profit of $512 million, or $1.12 a share  —  EBay recorded higher profit and revenue in its latest quarter as gross merchandise volume climbed.

Quartz : AWS revenue rose 28% YoY in Q1 to $37.6B, vs. $36.64B est., its fastest growth in 15 quarters, as capital expenditure rose to $44.2B, up from $25B in Q1 2025 —  AWS revenue rose 28% to $37.6 billion, the fastest growth in 15 quarters, as total revenue climbed 17% to $181.5 billion

Emma Roth / The Verge : Microsoft reports Q3 Xbox hardware revenue fell 33% YoY and Xbox content and services revenue, which includes Game Pass, dropped 5% YoY —  Xbox hardware revenue took a 33 percent hit, even as the company raked in $82.9 billion. … Microsoft's Xbox hardware revenue continues to tumble …

Ian King / Bloomberg : Qualcomm reports Q2 revenue down 3% YoY to $10.6B and says a top hyperscaler is on track to begin using its chips later this year; QCOM jumps 7%+ after hours —  Qualcomm Inc. rallied in late trading after the company said it was making headway in the lucrative data center market and predicted …

Financial Times : Google has told employees in a memo that it “proudly” works with the US military and will continue to do so, amid opposition from some staff over DOD contracts —  The tech giant signed an artificial intelligence deal with the defence department on Monday

Jonathan Vanian / CNBC : Meta reports Reality Labs Q1 revenue of $402M, vs. $488.8M est., and a $4.03B operating loss, vs. $4.82B est.; Reality Labs has $80B+ in losses since late 2020 —  As Meta pumps increasing amounts of cash into artificial intelligence, the company's metaverse efforts continue to bleed money.

Jordan Novet / CNBC : Microsoft's Q3 Intelligent Cloud revenue was $34.68B vs. $34.27B est., with Azure and other cloud services surging 40% YoY; Microsoft 365 Copilot has 20M+ seats —  Microsoft shares slipped 1% on Wednesday after the software maker reported more robust fiscal third-quarter results than analysts had expected.

Riley Griffin / Bloomberg : Meta raises its 2026 capex to between $125B and $145B, exceeding analysts' estimates, after previously forecasting $115B to $135B; META drops 6%+ after hours —  Meta Shares Plunge as AI Investments Raise Spending Outlook  —  Video Player is loading.  —  Unmute  —  Current Time 0:01 Loaded: 12.32% Playback Rate

Annie Palmer / CNBC : Amazon reports Q1 ad revenue up 24% YoY to $17.24B, vs. $16.87B est., and subscription services revenue up 15% to $13.43B —  Amazon on Wednesday posted better-than-expected earnings and revenue for the first quarter, and reported cloud sales that topped analysts' expectations.

Todd Spangler / Variety : Alphabet reports YouTube's Q1 ad revenue rose 10.7% YoY to $9.88B, vs. $9.99B est., and Google's ad revenue reached $77.25B, up from $66.89B in Q1 2025 —  Time spent viewing on YouTube, the internet's most massive video platform, keeps growing at a healthy clip — and with it …

Microsoft : Microsoft reports Q3 revenue up 18% YoY to $82.9B, net income up 23% to $31.8B, and says its AI business surpassed an annual revenue run rate of $37B, up 123% —  Microsoft Cloud and AI Strength Fuels Third Quarter Results  —  REDMOND, Wash. — April 29, 2026 — Microsoft Corp. today announced …

用 Rust 开发的文本编辑器项目 Zed 宣布发布 1.0 版本。开发者表示 1.0 版本并不意味着“完成”或“完美”，而是意味着到达了一个关键点。开发者还宣称 Zed 编辑器是一个 AI 原生编辑器，能并行运行多个 AI 智能体，包括 Claude Agent、Codex、OpenCode，以及 Cursor。AI 构建在编辑器的基础架构之中，而只是附加组件。

根据发表在《People and Nature》期刊上的一项研究，欧洲大山雀和其它 36 种鸟更惧怕女性，原因未知。研究显示，在鸟儿飞走前男性能比女性多靠近大约一米距离。无论男女衣着，是否长发，身高如何，以何种方式接近鸟，这种现象始终存在。鸟类或许能辨别人类的性别，但具体机制并不清楚。研究人员观察了生活在五个欧洲国家城市中心的鸟类，其中包括已知一见到人类就会飞走的鸟类如喜鹊，以及倾向于稍迟才飞走的鸟类如鸽子。对女性表现出过度恐惧的反应在不同鸟类中间是一致的。研究人员猜测鸟可能通过气味或步态辨别性别

中国网民报告 .icu 顶级域名的权威服务器在短时间内解析到了错误的 IP。.icu 最知名的域名可能是 996.ICU。不同于个别域名被错误解析，域名服务器被错误解析可能会导致影响扩大化。4 月 28 日 6:22 a.m. UTC 进行的测试显示，.icu 权威服务器之一的 b.nic.icu 在 58.13% 的情况下解析到正确 IP，其余解析到污染 IP。到 16:00 UTC 问题基本解决，99.38% 解析到正确 IP，仅 0.63% 返回污染 IP。

荷兰政府上线了它自己的开源代码托管平台 code.overheid.nl。该平台是基于 Forgejo——Forgejo 是 Gitea 的分支，是一个类似 GitHub 的 Git 软件开发和版本控制平台，支持 bug 跟踪、Wiki 和代码审查等功能。该平台托管于荷兰政府基础设施之上，所有政府机构都可以免费使用。荷兰政府表示此举是加强数字主权行动的一部分。

一份报告发现，逾三分之二两岁以下婴儿使用屏幕，少数婴儿每天最多花 8 小时在屏幕上。近三分之一新生儿每天看屏幕的时间超过三小时，4-11 个月大的婴儿近 20% 每天使用屏幕逾一小时。已有证据表明，屏幕时间与儿童肥胖、近视、睡眠和行为问题风险，以及长大后社交问题等负面影响相关联。新生儿父母允许婴儿使用屏幕的一大理由是分散他们注意力，让自己有时间做其它家务活或工作。

2024 年马斯克（Elon Musk）向旧金山高等法院起诉 OpenAI 及其联合创始人 Sam Altman 和 Greg Brockman 违反公司的创始原则，将商业利益置于公共利益之上。OpenAI 则公开了马斯克的邮件，证明作为曾经的联合创始人，马斯克同意 OpenAI 建立一个盈利实体，还表示将提供资金，但之后暂停了资金支持，他的目的是获得多数股权和董事会控制权，双方最终因此终止了合作。本周这起诉讼正式进入审讯阶段，马斯克在法庭上作证，称创办 OpenAI 是将其作为一家非盈利公司去对抗 Google，如果 OpenAI 的目标是盈利他不会支持它。马斯克称他在与 Google 联合创始人 Larry Page 就 AI 安全问题上发生争执后萌生了创办非盈利 AI 公司的想法。他担心 Page 没有认真对待 AI 安全问题，因此希望通过一个非盈利的开源替代方案去对抗 Google。

打哈欠是一种常见的人类行为，可能是感觉疲劳或无聊，或者是看到同伴打哈欠的反射行为。但科学家至今并不清楚人类为什么要打呵欠。根据发表在《Respiratory Physiology & Neurobiology》期刊上的一项研究，澳大利亚研究人员利用实时 MRI 扫描技术，观察了打哈欠时头部和颈部内部的变化，与正常呼吸和深呼吸进行对比。结果显示，打哈欠会触发一种特定的脑脊液和静脉血同时流出颅骨，而深呼吸时脑脊液则会流入颅骨。脑脊液充当了缓冲，能保护大脑和脊髓免受损伤，还能帮助输送营养物质和排出垃圾。研究表明打哈欠有助于清除脑脊液，其作用很可能是发生在临睡前。这项研究为理解打哈欠的生理功提供了一条新途径。

食肉细菌短短三天内就破坏了一位 74 岁佛罗里达男子的手臂和腿。三天前他还身体健康，还在海边活动，但在跳入水中时他的右腿被划伤了，伤口很快疼痛难忍，而且右臂的颜色也变了。他被送往医院急救，其右腿进行了膝上截肢手术。医生对其血液和组织样本的检测发现他感染了创伤弧菌（Vibrio vulnificus），一种生活在温暖咸淡水的食肉细菌。创伤弧菌通过两种途径感染人类：其一是通过伤口接触受污染的水，其二是食用受污染的海鲜。创伤弧菌会释放出各种毒素杀死感染者，感染者的总死亡率高达 35%，如果存在免疫功能问题或肝脏疾病，死亡率会进一步提高到 50%-60%。如果抗生素治疗或坏死组织切除手术延误，死亡率将是 100%。在本病例中，该男子最终幸存。该病例凸显了创伤弧菌在气候变化下日益加剧的威胁。美国 CDC 建议只食用完全煮熟的海鲜，以及在身上有创口的情况下避免进入咸淡水。

终端模拟器项目 Ghostty 宣布将退出 GitHub 平台，原因是在微软治理下 GitHub 越来越不稳定，严重扰乱开发者的工作。开发者称，过去一个月 GitHub 几乎天天宕机。GitHub 是工作的平台，如果每天都要宕机数小时那么它不再是一个可以工作的地方。Ghostty 项目将在未来几个月公布迁移细节，GitHub 上的项目将成为一个只读镜像。

Fedora 项目释出了 Fedora Linux 44。新版本搭载了最新桌面环境 GNOME 50 和 Plasma 6.6，针对桌面用户的变化包括：Wine 和 Steam (RPM Fusion) 默认启用能显著提升游戏性能的 NTSYNC 内核模块，Budgie 10.10，Atomic Desktops 移除了 FUSE 2，IBus 1.5.34，ibus-speech-to-text 支持 WhisperCpp 语音识别引擎。针对开发者的变化包括：LLVM 22，GNU C Compiler (gcc) 16.1，GNU Binary Utilities (binutils) 2.46，GNU C Library (glibc) 2.43，GNU Debugger (gdb) 16.3，Boost 1.90，Golang 1.26，Django 6.0，Helm 4，PHP 8.5，Ruby 4.0 等等。更多可浏览发布公告。

仅美国 11 个数据中心园区配套的新建天然气项目的温室气体排放量就有望超过摩洛哥 2024 年全年排放量。这些新建天然气发电项目旨在为 OpenAI、Meta、微软和 xAI 等公司的数据中心供电，其温室气体年排放量可能超过 1.29 亿吨。这些项目绕过了电网，专门为数据中心供电。由于公众对电费上涨的抵制，以及接入电网面临漫长的等待，越来越多的数据中心倾向于自行发电。xAI Colossus 和 Colossus 2 园区申请的空气排放许可显示每个园区每年可能产生逾 640 万吨碳排放，超过了 30 个中等规模天然气电厂的排放量。微软正计划从西德克萨斯州天然气项目购买电力，该项目每年排放逾 1150 万吨温室气体，超过牙买加全年排放量。Global Energy Monitor 的报告称，2026 年美国为数据中心建造的专用天然气发电项目接近 100GW，而 2024 年这一数字仅为 4GW。数据中心自建发电项目预计会在一个长期趋势，可能会对气候产生重大影响。

2025 年美国有 43 个州报告了逾 2000 例麻疹病例，几乎所有病例都发生在未接种疫苗人群中。2026 年的麻疹病例数仍在增加。美国学龄儿童的麻风腮三联疫苗（MMR）的接种率持续下降，徘徊在 93% 左右，低于 95% 的群体免疫阈值。研究人员调查了 2970 名成年人，虽然大多数美国人（83%）表示 MMR 疫苗的好处大于风险，但大约六分之一受访者表示对接种疫苗犹豫。犹豫的成年人总体更年轻，62% 的人年龄在 44 岁以下，且更有可能为人父母。他们更有可能是少数族裔、低收入和受教育程度较低的人。他们表达了更保守的政治信念，且更可能认同共和党（39%）或独立派（33%）。犹豫的成年人还更有可能认同“让美国再次健康”运动（MAHA），其比例为 43%，而非犹豫成年人占 27%。疫苗接种犹豫和非犹豫者之间的最大差异是前者偏爱阅读新右派新闻如 Breitbart、Newsmax 和 Zero Hedge。

尼安德特人和现代智人大脑的头骨在外形上存在明显的差异：尼安德特人的头骨更扁更长，而现代人类更圆。根据发表在 PNAS 期刊上的一项研究《Neanderthal brain and cognition reconsidered》，外形上的差异并不能说明什么。对比 400 名（200 欧裔 200 汉族）现代人类大脑的 MRI 扫描和尼安德特人头骨内模，现代人类大脑之间的差异比更新世智人和和尼安德特人之间的差异更大。鉴于大脑大小并不能准确预测认知能力，尼安德特人的认知能力可能更接近现代人类，这意味着人类可能不是凭借更聪明或更强的适应能力而战胜尼安德特人。研究人员认为，尼安德特人大脑和认知能力差异，完全可纳入现代人类内部的差异。

根据发表在《EMBO Reports》期刊上一项历时十年的研究，被称为辅助 T 细胞（helper T cells）的免疫细胞会长期携带肥胖记忆。记忆通过 DNA 甲基化过程标记在免疫细胞的 DNA 上，在成功减肥之后仍然会持续 5-10 年。这意味着减肥者仍然会长期面临肥胖相关疾病风险。研究人员称，短期减肥可能无法立即降低肥胖相关的疾病风险，需要在几年时间里维持减肥后的体重才可能逆转肥胖对 T 细胞的影响。

Mercor 是美国一家 AI 初创公司，主要业务是为其他 AI 公司提供专家帮助训练模型和聊天机器人。它招聘专家/合同工时要求对方提供护照或驾照扫描件、自拍和录制一段语音。本月初勒索组织 Lapsus$ 披露它从 Mercor 窃取了 4TB 语音样本。语音样本加上身份证件，引发了合同工们身份被盗用的担忧。已有至少五名合同工对 Mercor 提起了诉讼，指控该公司以训练数据的名义收集语音特征，但并未明确说明这些特征是永久性的生物识别标识符。现有的语音克隆技术只需要 15 秒钟的清晰参照音频，而 Mercor 要求合同工录制的语音长度达到了 2-5 分钟，足够实现语音克隆。