i-have-adhd 仓库提供一个专用技能和插件，旨在优化 Claude Code 等编码助手与用户的沟通。其主要目的是去除 AI 回复中的客套和开场白，确保助手给出直接、可执行且结构化的信息，而不是把实际解决方案埋在礼貌性的填充语后面。

项目遵循十条核心规则，约束代理行为。规则要求代理直接给出下一步行动、对多步骤任务使用编号，并以单一、具体的下一步结束。规则还严格禁止跑题、回顾和结尾客套，限制列表长度，并要求为任务提供明确的时间估计。

启用该技能后，AI 的输出发生根本改变：常见的热情开场、问题分析和结束语被替换为完成任务所需的核心指令与步骤。此做法旨在为偏好简洁、高密度信息的开发者降低认知负荷、提高效率。

项目高度可定制，用户可以 fork 仓库并修改底层的 SKILL.md 来适配自身需求。安装集成在标准 CLI 流程中，便于在受支持的编码环境中添加或移除插件。该项目灵感部分来自 The Adult ADHD Tool Kit 等临床资源，但已针对机器交互（而非个人时间管理）作出专门调整。

对该插件的支持已扩展到多平台，目前正在为 Claude 、 Gemini 和 OpenCode 等工具开发适配器。凭借大量贡献者和广泛的社区采纳，该仓库成为提示工程与代理型插件如何根据个人工作流与沟通偏好定制 AI 的典型示例。 The "i-have-adhd" repository provides a specialized skill and plugin designed to optimize how coding assistants, such as Claude Code, communicate with users. The primary goal of this tool is to strip away the conversational fluff and preamble often found in AI responses, ensuring that the assistant provides direct, action-oriented, and structured information instead of burying the actual solution behind polite filler text.

The project operates under a set of ten core rules that govern the agent's behavior. These rules mandate that the agent must lead immediately with the next action, use numbered steps for multi-part tasks, and conclude with a single, concrete next step. Furthermore, the instructions strictly forbid tangents, recaps, and closing pleasantries, while also imposing limits on list lengths and requiring clear, specific time estimates for tasks.

By implementing this skill, the AI's output is fundamentally transformed. Where a standard response might include an enthusiastic introduction, an analysis of the problem, and a sign-off, the "i-have-adhd" enabled response is condensed into the essential commands and steps required to complete the task. This approach is intended to reduce cognitive load and improve productivity for developers who prefer concise, high-signal communication from their tools.

The project is highly customizable, allowing users to fork the repository and modify the underlying "SKILL.md" file to suit their specific needs. The installation process is integrated into standard CLI workflows, making it easy to add or remove the plugin from supported coding environments. The initiative is loosely inspired by clinical resources like The Adult ADHD Tool Kit, but it has been specifically adapted for machine interaction rather than human time management.

Support for the plugin has expanded to include various platforms, with active development on adapters for tools like Claude, Gemini, and OpenCode. With a significant number of contributors and strong community adoption, the repository serves as a practical example of how prompt engineering and agent-based plugins can be used to tailor artificial intelligence to individual user workflows and communication preferences.

LibreOffice 26.8 最近取得了重要里程碑，成为该软件近二十年来最受欢迎的一次更新。发布仅一周，安装量就超过一百万次（不含通过 Linux 发行版仓库分发的各类更新）。虽然该套件在写作功能和排版方面有所改进，但热度激增在很大程度上归功于一项大胆的"非功能"——明确声明软件不包含生成式 AI 。

管理 LibreOffice 的 The Document Foundation 在这次创纪录的发布后不久，通过一篇题为 "Yes, no AI is now a feature" 的文章进一步巩固了这一立场。 The Document Foundation 表示，虽然它并不在原则上排斥人工智能，但认为现有的 AI 技术无法满足基本的隐私与安全要求。具体而言，基金会要求任何内建功能都必须支持由用户控制的运行，确保未经授权的内容不会离开计算机，并在没有遥测、厂商依赖或对文件格式造成影响的情况下工作。

目前，The Document Foundation 建议对人工智能功能感兴趣的用户应依赖社区开发的插件，而不是期待原生集成。这一理念使 LibreOffice 有别于那些积极将生成式 AI 纳入产品的竞争对手。基金会指出，对于许多行业对手而言，AI 助手既是提价的战略借口，也是将用户绑入专有基础设施的手段。

LibreOffice 26.8 的成功也回应了一个长期存在的问题：是否存在抵制将 AI 植入每款应用的软件市场。创纪录的下载量表明，包括政府机构和非政府组织在内的大量用户更重视隐私与可控性，而非原生 AI 功能带来的便利。通过将"不含 AI"作为一项鲜明的产品特性，LibreOffice 证明了在日益自动化的环境中，仍有空间为优先考虑用户独立性的替代方案留出位置。 LibreOffice 26.8 recently achieved a significant milestone, becoming the most popular update in the software's nearly two-decade history. Within just one week of its release, the installer surpassed one million downloads, excluding the various updates handled through Linux distribution repositories. While the suite features improvements to its writing system and typography, the surge in popularity is widely attributed to a bold non-feature: the explicit declaration that the software does not include generative AI.

The Document Foundation, which manages LibreOffice, solidified this stance shortly after the record-breaking release with a post titled "Yes, no AI is now a feature." While the foundation clarifies that it does not reject artificial intelligence in principle, it maintains that current AI technologies fail to meet essential privacy and security requirements. Specifically, the foundation insists that any integrated feature must support user-controlled execution, ensure no content leaves the computer without authorization, and function without telemetry, vendor dependency, or compromise to file formats.

For now, the foundation suggests that users interested in AI capabilities should rely on community-developed plugins rather than expecting native integration. This philosophy sets LibreOffice apart from competitors that have aggressively integrated generative AI into their products. The foundation suggests that for many industry rivals, AI assistants serve as a strategic justification for price increases and a means to force users into proprietary infrastructure.

Ultimately, the success of LibreOffice 26.8 addresses a lingering question about whether there is a market for software that resists the industry-wide trend of baking AI into every application. The record download numbers suggest that a significant segment of users, including government agencies and NGOs, actively values privacy and control over the convenience of native AI features. By positioning "no AI" as a distinct product feature, LibreOffice has demonstrated that there is indeed room for alternatives that prioritize user independence in an increasingly automated landscape.

Blackmagic Design 发布了 DaVinci Resolve 21.1 的重大更新，推出了一系列提升创作效率与工作流程管理的功能，现已可免费下载安装，展示了其在后期制作工具方面的投入。此次更新将 Claude 、 Claude Code 和 ChatGPT Codex 等 AI 助手整合进来，用户可通过自然语言完成媒体整理、项目分析和批量渲染等耗时任务，让剪辑师能把更多时间投入到创意工作而非重复性技术操作上。

photo page 得到显著增强，新增对 Fujifilm 、 Leica 和 Sony 等机型的原生文件导入支持；用户可通过专用调色面板（如 DaVinci Resolve Micro Color Panel）进行手动微调，并可通过 ResolveFX 使用更强大的换天工具。 media page 方面，软件现在能自动处理源媒体的宽高比，并通过改进的列表视图简化元数据录入，提升项目素材的浏览与管理效率。

Edit page 的工作流程迎来重要的生产力提升：新的 multi-cam mode 控制在 timeline 与 source viewer 之间切换时避免丢失焦点，增强的键盘快捷键支持多达 25 个摄像机角度，更适合大规模现场活动制作；trim editor 允许在 viewer 中直接进行逐帧调整，inspector panel 支持自定义预设，减少在多个剪辑或项目中重复配置参数的麻烦。

针对调色师，新增对 Dolby Vision 、 HDR10+ 、 HDR Vivid 等 HDR 格式的独立微调通道，便于对各类交付规格分别精细调整。软件还扩展了原生色彩空间管理，新增对 GoPro 、 Leica 、 DJI 和 Vivo 等格式的支持。 Fusion 也获大量增强——加入 25 款以上的 Krokodove 工具，提供更先进的 3D 区域与形状处理选项，同时更直观的 macro editor 支持从模板包拖放加载。

在性能层面，引入了 GPU 加速的 HTJ2K 解码，并为 Apple Silicon 用户提供了硬件加速的 HEIC 解码。协作功能也得到改进：Blackmagic Cloud Presentations 现支持字幕集成和交互式批注共享。 21.1 更新包含 100 多项新工具和控件，体现了 Blackmagic Design 在优化剪辑、调色与视觉特效工作流程上对专业用户的大幅投入。 Blackmagic Design has released a significant update to its industry-standard software, DaVinci Resolve 21.1, introducing a suite of features designed to enhance creative efficiency and workflow management. The update is now available as a free download, highlighting the company's commitment to providing robust post-production tools. Key among these updates is the integration of AI assistants like Claude, Claude Code, and ChatGPT Codex. These assistants allow users to utilize natural language to perform time-consuming tasks such as media organization, project analysis, and batch rendering, ultimately allowing editors to prioritize creative work over repetitive technical processes.

The update brings substantial improvements to the photo page, which now features expanded camera support for models from Fujifilm, Leica, and Sony, enabling native file imports. Users can also perform hands-on adjustments using dedicated color panels, like the DaVinci Resolve Micro Color Panel, and access enhanced sky replacement tools via ResolveFX. On the media page, the software now automatically handles source media aspect ratios and streamlines metadata entry through an improved list view, allowing for faster navigation and organization of project assets.

Edit page workflows have seen major productivity upgrades, including new multi-cam mode controls that prevent loss of focus when switching between the timeline and source viewer. Enhanced keyboard shortcuts support up to 25 camera angles, making the software more capable for large-scale live event productions. Additionally, the trim editor now allows for direct frame-level adjustments in the viewer, and the inspector panel supports custom presets, which help eliminate the need to manually configure parameters across multiple clips or projects.

For colorists, the update introduces individual trim passes for various HDR formats, such as Dolby Vision, HDR10+, and HDR Vivid, ensuring that users can fine-tune deliverables independently. The software has also expanded its native color space management to include support for GoPro, Leica, DJI, and Vivo formats. Fusion has been equally empowered with over 25 new Krokodove tools, including advanced 3D region and shape manipulation options, alongside a more intuitive macro editor that supports drag-and-drop loading from template bundles.

Technical performance is further bolstered by GPU-accelerated HTJ2K decoding and hardware-accelerated HEIC decoding for Apple Silicon users. Additionally, Blackmagic has improved its collaborative features through updates to Blackmagic Cloud Presentations, which now supports subtitle integration and interactive annotation sharing. With over 100 new tools and controls, the 21.1 update represents a substantial effort by Blackmagic Design to refine the editing, color grading, and visual effects experience for professionals across the post-production industry.

网络始终在变化之中，受不断演进的浏览器标准、硬件变更和企业利益调整所驱动。在这种环境下，大量曾经用于跨浏览器兼容或与已废弃平台集成的过时 HTML 片段和 meta 标签被遗弃，形成了一个"墓地"。这些遗物常见于旧站点或资深开发者的习惯中，范围从像 ICBM 标签这样的地理坐标，到像 Skype Toolbar 或 Microsoft Smart Tags 这样的侵入性厂商专属工具。

许多此类片段起源于浏览器大战时期，开发者为兼顾 Internet Explorer 、 Netscape 或早期移动浏览器而使用特定标签。例如，开发者通过条件注释或 X-UA-Compatible meta 标签来强制文档模式，以应对版本相关的兼容性问题。同样，像 Image Toolbar 或 Baidu 的页面转码等工具也需要显式的 opt-out meta 标签，以防第三方软件在未经站点所有者同意的情况下注入不想要的内容或重构页面布局。

随着移动技术的发展，早期为小屏幕优化网页而诞生了 HandheldFriendly 、 MobileOptimized 等标签。与此同时，各操作系统也试图通过固定站点和自定义应用定义将网页内容更深地整合进系统。从 Safari 用于固定标签页的 mask-icon，到 Windows 那套用于磁贴和任务栏任务的 msapplication 标签集，这些做法都是后来渐进式 Web 应用（PWA）兴起前的前奏。不过，随着时间推移，标准化的 API 逐渐取代了这些零散且针对平台的实现方式。

搜索和社交媒体的发展同样留下了痕迹，像 Revisit-after 以及 AJAX crawling 方案凸显了 Google 在处理动态网页内容方面的早期困难。同样，社交分享的兴起催生了诸如 Share Partners 系统和早期的 Twitter cards 这类临时标准。如今，这些大多已被 Open Graph 协议取代，使许多旧的、针对特定平台的元数据声明变得过时。这些历史残存物提醒我们，网络的混沌式成长以及许多一度普遍的开发实践本质上都是暂时的。 The web is in a constant state of flux, shaped by evolving browser standards, hardware changes, and shifting corporate interests. This environment has left behind a graveyard of antiquated HTML snippets and meta tags that were once essential for cross-browser compatibility or integration with now-defunct platforms. These relics, often found in older sites or preserved in the habits of long-time developers, include everything from geographic coordinates like ICBM tags to intrusive vendor-specific tools like the Skype Toolbar or Microsoft Smart Tags.

Many of these snippets originated from the browser wars, where developers used specific tags to cater to Internet Explorer, Netscape, or early mobile browsers. For instance, developers utilized conditional comments or the X-UA-Compatible meta tag to force document modes and address version-specific quirks. Similarly, tools like the Image Toolbar or Baidu's page transcoding required explicit opt-out meta tags to prevent third-party software from injecting unwanted content or restructuring site layouts without the owner's consent.

As mobile technology advanced, early efforts to optimize the web for small screens birthed tags like HandheldFriendly and MobileOptimized. Meanwhile, operating systems sought to integrate web content more deeply through pinned sites and custom app definitions. From Safari's mask-icon for pinned tabs to Windows' extensive set of msapplication tags for tiles and taskbar tasks, these methods were the precursors to the modern Progressive Web App (PWA). Over time, however, standardized APIs replaced these disparate and platform-specific implementations.

The evolution of search and social media also left its mark, with legacy tags like Revisit-after and the AJAX crawling scheme highlighting Google's early struggles with dynamic web content. Similarly, the rise of social sharing led to ad hoc standards such as the Share Partners system and early Twitter cards. Today, these have largely been superseded by the Open Graph protocol, rendering many of the old, platform-specific metadata declarations obsolete. These remnants of the past serve as a reminder of the web's chaotic growth and the temporary nature of many once-standard development practices.

内容分发网络（CDN）是现代网站的关键入口，负责从缓存到流量过滤等各类功能。由于这些网络位于源服务器之前，成为了一个共享的依赖点：一旦某家 CDN 供应商出现故障，所有依赖该基础设施的网站就会同时下线，彼此之间是否有关联无关紧要。

对 44,143 家使用 CDN 的欧洲公司分析显示，市场高度集中。近 90% 的企业将前端基础设施托付给 Cloudflare 。虽然 Amazon 、 Fastly 和 Akamai 等供应商仍有市场存在，但与 Cloudflare 的主导地位相比微不足道。这一趋势在各主要欧洲市场普遍存在：Cloudflare 在 Spain 和 Ireland 的份额约为 79%，在 Netherlands 则高达 95% 。

这种高度集中对数字韧性构成重大风险。尽管 CDN 供应商常以提升可用性和安全性为卖点，但当单一供应商主导市场时，其内部失误会演变成行业范围的危机。近年多次 Cloudflare 的全球性故障，就是由常规配置变更或自动化错误引起，而非外部网络攻击。这些事件表明，当大部分市场共享同一扇"前门"时，整个行业都会暴露于单点故障的风险之下。

需要将这种基础设施集中与数据驻留问题区分开来：CDN 只是前端，后端的源服务器、数据库和数据处理系统通常部署在不同的托管平台上。尽管前端高度整合，后端却更为多样化，许多欧洲公司将实际数据存放在区域性托管提供商处。然而在故障发生时，这种区分几乎无济于事——无法访问前端就等于看不到底层的数据架构。

归根结底，对单一供应商的高度依赖暴露了现代数字战略的悖论。企业购买 CDN 是为了确保稳定性和性能，但当前的市场结构却制造出一种集体脆弱性。想要评估自身的风险敞口，企业可以通过检查 HTTP 响应头中是否包含 Cloudflare 的 ray ID 或其他厂商特有的标识来识别所用的 CDN 。认识到这种依赖，是弄清自身数字基础设施真实状况的第一步。 A content delivery network, or CDN, serves as the critical entry point for modern websites, handling everything from caching to traffic filtering. Because these networks sit in front of the origin servers, they function as a shared dependency. If a CDN provider experiences an outage, every website relying on that infrastructure effectively goes offline simultaneously, regardless of whether those businesses have any other connection to one another.

An analysis of 44,143 European companies using CDN services reveals a profound level of market concentration. Nearly 90 percent of these firms rely on Cloudflare for their front-door infrastructure. While other providers like Amazon, Fastly, and Akamai maintain a presence, their market shares are dwarfed by Cloudflare's dominant position. This trend holds true across major European markets, with Cloudflare's share ranging from roughly 79 percent in Spain and Ireland to as high as 95 percent in the Netherlands.

This intense concentration poses significant risks to digital resilience. While CDN providers often market their services as a means to improve uptime and security, the reality is that when a single provider dominates the market, its internal errors become industry-wide crises. Recent history bears this out, with several global outages for Cloudflare occurring due to routine configuration changes or minor automation errors rather than external cyberattacks. These incidents demonstrate that when most of a market shares one front door, the entire sector becomes vulnerable to a single point of failure.

It is important to distinguish this infrastructure concentration from data residency concerns. The CDN serves only as the front-end, while the back-end origin servers, databases, and data processing systems often reside on different hosting platforms. While the front-end is highly consolidated, the back-end landscape is much more diverse, with many European companies utilizing regional hosting providers for their actual data storage. However, this distinction offers little comfort during an outage, as the inability to reach the front door effectively blinds the user to the underlying data architecture.

Ultimately, this high level of reliance on a single provider suggests a paradox in modern digital strategy. Companies buy CDN services to ensure stability and performance, yet the current market structure creates a collective vulnerability. For businesses aiming to assess their own exposure, identifying their CDN usage is a simple matter of checking HTTP response headers for specific identifiers like Cloudflare's ray ID or other vendor-specific signatures. Recognizing this dependency is the first step toward understanding the true nature of one's digital infrastructure.

Tristan 和 Levent Alpoge 发布了三项重要成果，证明了 incompressible porous media equation 、 Boussinesq equations 和 3D incompressible Euler equations 出现 finite-time blowup 的例子。研究通过引入 smooth forcing 实现，所用策略继承并发展了 Diego Córdoba 与 Luis Martínez-Zoroa 最初提出的核心思想。作者们强调，这是一项纯粹的个人合作，独立于他们各自的机构，并在很大程度上依赖多种大型语言模型（包括 Claude 、 Codex 和 Astra）来推进 Córdoba-Martínez-Zoroa program 的完成。

借助 AI 完成这些证明的过程漫长且充满挑战。大部分时间进展缓慢，直到八月中旬出现突破，最终又用 Lean proof assistant 对结果进行了验证。尽管在数学上取得了成功，作者对在巨大压力下发布的预印本的呈现质量表示遗憾。他们认为，这项工作的真正意义在于展现了人在压缩时间内与 AI 协同开展高水平数学研究的可能性——这一点值得学界就培训、署名和人类注意力在科研中的价值等问题进行深入而冷静的讨论。

关于与 OpenAI 的接触，叙述出现了令人不安的转折。据传一款内部 OpenAI 模型也解决了一个重大悬而未决的问题——尤其是 Navier-Stokes 的 forced blowup——Tristan 因此与包括 Sebastien Bubeck 在内的 OpenAI 代表进行了交流。对话中透露，OpenAI 有一个完整团队在攻关，投入了大量算力，而且该内部模型的成果并非像最初所暗示的那样仅靠极少的人为干预。此外，该模型成功的时间点似乎与外界流传的关于作者自身进展的消息相吻合。

讨论中有人建议作者将 Euler 的结果与 OpenAI 关于 Navier-Stokes 证明的公告同时发布，甚至提出因 Levent 在 Anthropic 工作而将其从作者名单中移除。作者拒绝这些提议后，局势变得紧张，并出现可能影响他们职业声誉的暗示性言辞。

作者强调他们并未指控任何具体的不当行为，因为他们并未见过 OpenAI 的证明，也不清楚自己的数据是否被使用。公开此事的目的是维护历史记录的准确性，防止错误叙述流布。研究人员希望将讨论重心重新拉回到数学本身，以及这些工具对更广泛学术界可能带来的变革性影响。 Tristan and Levent Alpoge have released three significant mathematical results concerning finite-time blowup for the incompressible porous media equation, the Boussinesq equations, and the 3D incompressible Euler equations. These findings were achieved using smooth forcing, a strategy building upon foundational ideas originally proposed by Diego Córdoba and Luis Martínez-Zoroa. The researchers highlight that this work was a purely personal collaboration, independent of their respective institutional affiliations, and relied heavily on the use of various large language models, including Claude, Codex, and Astra, to push the Córdoba-Martínez-Zoroa program to completion.

The process of utilizing AI for these proofs was lengthy and challenging. Progress remained slow for much of the year until a breakthrough occurred in mid-August, leading to the verification of the results via the Lean proof assistant. Despite the mathematical success, the authors express regret regarding the presentation quality of their preprints, which were produced under significant pressure. They emphasize that the true significance of this endeavor lies in the capability for humans and AI models to collaborate on high-level mathematical research in a compressed timeframe, a development that warrants a deep, unhurried discussion within the scientific community regarding the future of training, credit, and the value of human attention in research.

The narrative takes a troubling turn regarding interactions with OpenAI. Following rumors that an internal OpenAI model had also solved a major open problem—specifically, forced blowup for Navier-Stokes—Tristan engaged in discussions with representatives from the company, including Sebastien Bubeck. During these conversations, it was revealed that an entire team at OpenAI had been working on the problem, that significant compute resources were deployed, and that the internal model's effort was not the result of minimal human input as initially suggested. Furthermore, the timing of the model's success appeared to coincide with the circulation of news regarding the authors' own progress.

During these discussions, proposals were made to the authors, including suggestions that they post their Euler results alongside an OpenAI announcement of the Navier-Stokes proof, and that Levent be removed from authorship due to his employment at Anthropic. When these offers were declined, the situation grew contentious, with implications made regarding the authors' professional reputations.

The authors maintain that they are not accusing anyone of specific wrongdoing, as they have not seen OpenAI's proof and do not know if their own data was utilized. The purpose of coming forward is to ensure that the historical record remains accurate and to prevent false narratives from taking hold. Ultimately, the researchers express a desire to shift the conversation back toward the mathematics and the transformative potential of these tools for the wider academic community.

Albanese Government 已提交名为 Digital Duty of Care 的草案，标志着对网络环境进行监管的重要一步。该举措拟通过为数字服务提供商设定强制性的最低平台标准，把确保用户获得更安全体验的责任转移给服务提供商。提案的核心是 My Feed, My Way 倡议，赋予 Australians 对社交媒体算法的控制权。

根据该计划，社交媒体平台须告知用户可选择的默认信息流偏好。用户可以选择接受个性化算法推荐的信息流，也可以选择退出，仅查看自己主动关注的创作者和朋友发布的内容。此举通过限制自动推荐的影响，赋予用户更大的在线体验自主权。

立法适用范围不仅限于社交媒体，还涵盖在线游戏、各类应用程序以及 AI 聊天机器人。数字服务提供商在法律上须保护 18 岁以下用户，防止采用已知会产生负面影响的设计机制，例如令人上瘾的功能或损害自尊的元素。平台还必须采取防护措施，阻止未成年人接触有害内容，包括宣扬饮食失调的材料、色情内容、带有敌意的基于性别的观念，以及美化危险行为或自残的内容。

为确保问责，法律草案授权 eSafety Commissioner 负责监管合规并执行新标准。这包括发布针对有害应用或网站的移除通知，并简化处理网络欺凌和成人网络虐待的既有流程。平台需详尽记录其风险缓解策略，并确保这些保护措施在实施后持续有效。

未能遵守新标准将面临严厉后果，最高罚款可达 1.092 亿美元。政府正与行业组织、公民社会团体及数字平台进行定向磋商，以完善这些措施。通过该倡议，Albanese Government 希望建立一个强有力的框架，要求数字产品达到基本安全标准，并将其与对汽车或食品等实体商品的保护相提并论。该立法预计将在今年晚些时候提交 Parliament 。 The Albanese Government has introduced draft legislation for a Digital Duty of Care, marking a significant step toward regulating online environments. This initiative aims to shift the responsibility onto digital service providers to ensure safer experiences for users by setting mandatory minimum standards for their platforms. A centerpiece of this proposal is the My Feed, My Way initiative, which empowers Australians to take control of their social media algorithms.

Under the proposed My Feed, My Way scheme, social media platforms will be required to notify users of their right to choose their default feed preferences. Users will have the option to opt in to a feed curated by personalized algorithms or, alternatively, opt out to view only the content posted by the creators and friends they actively choose to follow. This move is designed to grant users greater agency over their online experience by limiting the influence of automated recommendations.

The legislation extends its reach beyond social media to encompass online games, various apps, and AI chatbots. Digital service providers will be legally required to protect users under 18 from design features known to negatively impact behavior, such as addictive functions or elements that undermine self-esteem. Furthermore, these platforms must implement safeguards to prevent minors from accessing harmful content, including material promoting eating disorders, pornography, hostile gender-based ideas, or content that glorifies dangerous acts and self-harm.

To ensure accountability, the draft laws empower the eSafety Commissioner to manage compliance and enforce new standards. This includes the authority to issue removal notices for harmful apps or websites and to streamline existing processes for addressing cyber-bullying and adult cyber abuse. Platforms will be expected to thoroughly document their risk mitigation strategies and ensure these protections remain effective over time.

Failure to adhere to these new standards carries significant consequences, with potential penalties reaching as high as 109.2 million dollars. The government is currently engaging in targeted consultation with industry bodies, civil society organizations, and digital platforms to refine these measures. With this initiative, the Albanese Government intends to establish a robust framework that mandates basic safety standards for digital products, drawing a comparison between the protections currently afforded to physical goods like cars or food. The legislation is expected to be introduced to Parliament later this year.

Mistral AI 在 D 轮融资中获得了 30 亿欧元，这是该公司三年前成立以来，欧洲科技公司规模最大的一笔股权融资。本轮由 Samsung Electronics 领投，Scaleup Europe Fund 和现有投资者 PSG Equity 参投，投后估值超过 210 亿欧元。所募资金将用于加强前沿研究、扩充计算能力，并加速在其已进入的 20 个国家的商业与国际化扩展。

公司正将自己定位为主权 AI 转型的领军者，以满足政府和大型企业在不放弃对基础设施或数据控制权的前提下使用人工智能的需求。随着组织不再仅仅追求最强大的模型，而更关注避免长期依赖与供应商锁定，它们越来越重视自主可控。 Mistral 强调自己是唯一提供全栈解决方案的厂商，涵盖开放权重模型、私有计算能力和可投入生产的工具，从而保障用户的独立性与灵活性。

Mistral 的价值主张核心是主权 AI 层，优先把控四个关键维度：将数据留在组织边界内、确保模型可定制且可控、维持可预测且私有的计算环境、以及提供全面可审计的系统。该架构旨在使机构能够在复杂且关乎任务的环境中（如制造业和重工业）集成最先进的 AI，而不会将敏感的机构知识或工作流程暴露给外部方。

本轮融资得到了来自欧洲、亚洲和北美的多元化战略与金融投资者财团的有力支持。吸引 Samsung Electronics 和 ASML 等行业巨头参与，表明 Mistral 的技术特别契合高级工程与工业应用的需求。这类机构级背书也体现出市场对其提供可行主权替代方案的广泛信心，确保客户在数字智能上保持决策权。 Mistral AI has reached a significant milestone by securing 3 billion euros in a Series D funding round, marking the largest equity fundraising event for a European technology firm since the company's inception three years ago. This round, led by Samsung Electronics and joined by the Scaleup Europe Fund and existing investor PSG Equity, brings the company's post-money valuation to over 21 billion euros. The capital infusion is intended to bolster the company's frontier research, expand its compute capacity, and accelerate its commercial and international growth across the 20 countries where it currently operates.

The company is positioning itself as a leader in the shift toward sovereign AI, addressing a growing demand among governments and large enterprises to utilize artificial intelligence without sacrificing control over their infrastructure or data. As organizations move past the initial phase of simply seeking the most powerful model, they are increasingly focused on avoiding the long-term dependencies and vendor lock-in associated with traditional AI deployments. Mistral emphasizes its role as the only provider offering a full-stack solution, which includes open-weight models, private compute capacity, and production-ready tools that allow for independence and flexibility.

Central to Mistral's value proposition is the concept of a sovereign AI layer, which prioritizes four key dimensions of control: keeping data within organizational boundaries, ensuring models remain customizable and controllable, maintaining predictable and private compute, and providing fully auditable systems. This architecture is designed to allow institutions to integrate state-of-the-art AI into complex, mission-critical environments, such as those in manufacturing and heavy industry, without exposing sensitive institutional knowledge or workflows to external parties.

The funding round reflects a strong endorsement from a diverse syndicate of strategic and financial investors spanning Europe, Asia, and North America. By attracting backing from industry giants like Samsung Electronics and ASML, Mistral is signaling that its technology is uniquely suited for the requirements of advanced engineering and industrial applications. This institutional support highlights a broad confidence in the company's mission to provide a viable, sovereign alternative to the dominant AI frameworks, ensuring that customers retain agency over their digital intelligence.

最近发布的 GLM 5.3-flash 是一款开源权重的大型语言模型，能在消费级硬件上快速且低成本运行，这打开了一个关键的脆弱性窗口。因为该模型可以被修改以移除安全防护，任何在硬件上有少量投入的人现在都能部署一个强大且无约束的工具，用来实施复杂的网络攻击。这使风险从理论上的可能性变成了迫在眉睫的现实威胁，恶意活动可以被大规模自动化。

危险在于模型的可及性与其在网络安全任务上出色性能的叠加。基准测试显示，GLM 5.3-flash 在识别并利用现实世界漏洞方面表现优异，几乎把先进的自动化黑客能力交到了任何人手中。尽管 Project Glasswing 和 Daybreak 等项目正尝试用前沿模型修补安全缺陷，但核心难题仍是如何在分散且常为遗留系统的基础设施中迅速部署这些补丁。

事态需要紧急应对，因为攻击者的推进速度已快于防御方。政府和监管机构应将重心从仅仅发现漏洞转向强制并激励补丁的实际部署。政策应优先支持基于风险的定期渗透测试，并对关键基础设施——例如电网、金融系统和电信——实施更严格的安全标准。单靠对模型或硬件的禁令很可能无效且适得其反，因为技术无可避免地会扩散。

企业和开源基金会应把这视为利用前沿 AI 强化防御、全面改造安全态势的良机。这需要超越手工安全流程，采用结构化提示、自动化测试和形式化验证来加速修补周期。应赋能内部安全工程团队，优先进行漏洞的筛选与补丁部署，确保在漏洞被自动化代理利用前得到修复。

最后，组织需建立健全内部控制，例如沙箱隔离、限定权限的凭证和纵深防御，以防自有防御代理成为事故源。逐步放弃内存不安全的语言、全面清点依赖项并演练快速的事件恢复，都是必要步骤。即便对当前威胁持怀疑态度的人，也应看到这些措施代表了网络安全必须尽快实现的进步，应在机会窗口关闭前付诸行动。 The recent release of GLM 5.3-flash, an open-weight large language model that is both fast and cheap to run on consumer hardware, has created a critical window of vulnerability. Because the model can be modified to remove safety safeguards, anyone with a modest investment in hardware can now deploy a powerful, unconstrained tool capable of performing sophisticated cyberattacks. This capability moves the needle from theoretical risk to an imminent, real-world threat where malicious activities can be automated at scale.

The danger lies in the combination of the model's accessibility and its high performance in cybersecurity tasks. Benchmarks indicate that GLM 5.3-flash is highly proficient at identifying and exploiting real-world vulnerabilities, effectively putting the power of advanced, automated hacking into the hands of virtually anyone. While existing initiatives like Project Glasswing and Daybreak have been working to patch security flaws using frontier models, the core challenge remains the slow and difficult process of deploying those fixes across fragmented, often legacy, infrastructure.

Urgency is required to address this landscape, as attackers are currently advancing faster than defenders. Governments and regulatory bodies should shift their focus toward mandating and incentivizing the actual deployment of patches, rather than just the identification of vulnerabilities. Policy should prioritize frequent, risk-based penetration testing and enforce stricter security standards for critical infrastructure, such as power grids, financial systems, and telecommunications. Relying on bans for models or hardware is likely to be ineffective and counterproductive, as the technology will inevitably proliferate regardless of such measures.

Companies and open source foundations must treat this as a prime opportunity to overhaul their security posture by leveraging frontier AI as a defensive tool. This involves moving beyond manual security practices and utilizing structured prompts, automated testing, and formal verification to accelerate the patching cycle. Internal security engineering teams should be empowered to prioritize the triage and deployment of fixes, ensuring that vulnerabilities are addressed before they can be exploited by automated agents.

Finally, organizations need to implement robust internal controls, such as sandboxing, scoped credentials, and defense-in-depth, to prevent their own defensive agents from becoming vectors for incident creation. Moving away from memory-unsafe languages, maintaining thorough inventories of dependencies, and practicing rapid incident recovery are essential steps. Even for those skeptical of the immediate threat level, these measures represent a long-overdue advancement in collective cybersecurity that must be acted upon before the window of opportunity closes.

自 2024 年初 Google 调整政策以来，独立维基托管项目遇到了重大障碍。该政策事实上把许多新建独立网站置于所谓的"Google Jail"状态：除站点主页外，几乎所有新域名下的页面都无法出现在搜索结果中。对于高度依赖搜索流量的项目而言，影响十分严重。

这种限制自 2024 年 3 月以来，似乎影响了近 90% 在新域名上启动的维基。该现象不同于普通的重复内容问题，无论内容是原创，还是从 Fandom 等平台迁移过来，都同样会被屏蔽。即便维基的主页能击败竞争对手取得较好排名，搜索引擎仍然不展示其子页面。这种"悬置"状态可能持续数月甚至接近一年，虽然在站点活跃度高或围绕维基主题发生重大新闻时，情况有时会缓和。

技术证据表明，Google 可能为打击低质量 SEO 做法，对新域名施加了较为严厉的过滤规则。有趣的是，当项目作为已有信誉的域名的子域名托管时，这个问题就消失了——即使流量不高，子域名也能避免困扰独立域名的索引封锁。这种差异迫使托管方重新考虑将社区从大型维基平台迁出的策略。

面对现状，托管组织在可见性与品牌独立性之间权衡。一个可行的做法是将新项目以子域名形式部署在已有基础设施上，以维持搜索引擎的信任，尽管这会带来品牌识别方面的问题。虽然这不是长久之计，但能作为可行的权宜之策，确保用户可访问这些维基。

最终目标仍是等到网站在搜索引擎中获得足够权威和信任后，再迁回独立根域名。组织者希望通过重定向策略和正式的地址变更工具，保留在子域名期间积累的搜索排名。与此同时，社区仍在寻求对此现象的更多澄清，以更好地评估这类搜索引擎政策对独立网络长期影响。 Independent wiki hosting initiatives have recently encountered a significant hurdle following a policy shift by Google in early 2024. This change has effectively relegated many new, independent websites to a state often referred to as Google Jail. For these projects, which often rely on search traffic for the vast majority of their visitors, the impact is severe. Specifically, Google has been preventing almost all pages on brand-new domains from appearing in search results, with the notable exception of the site's main page.

This restrictive behavior appears to affect nearly 90 percent of wikis launched on new domains since March 2024. The phenomenon is distinct from standard duplicate content issues, as it persists regardless of whether the content is original or migrated from existing platforms like Fandom. Even when a wiki's main page successfully outranks its competition, the search engine still refuses to display sub-pages. This state of limbo can last for months, or even up to a year, though it sometimes dissipates during periods of significant site activity or major updates related to the wiki's subject matter.

The technical evidence suggests that Google implemented this measure to combat low-quality SEO practices by applying heavy-handed filters to new domains. Interestingly, this problem vanishes when a project is hosted as a subdomain of an already established, reputable domain. Even without high traffic, these subdomains avoid the indexing blockade that plagues fresh, standalone web addresses. This discrepancy has forced hosting providers to rethink their strategy for migrating communities away from large wiki farms.

Moving forward, hosting organizers are weighing a difficult set of compromises to ensure their wikis remain visible. One approach involves launching projects as subdomains of established infrastructure to maintain search engine trust, despite the branding issues this presents for the individual projects. While this is not an ideal permanent solution, it provides a functional workaround that keeps the wikis accessible to their user base.

The ultimate goal remains to eventually migrate these wikis to their own dedicated root domains once they have gained enough authority and trust with search engines. Organizers hope that by using redirection strategies and formal address change tools, they can preserve the search ranking accumulated during the time spent on subdomains. In the meantime, the search for more clarity on this phenomenon continues, as the community seeks to better understand the long-term implications of these search engine policies on the independent web.