包括 New York Times 、 The Atlantic 和 USA Today 在内的主要新闻媒体，近期已停止允许 Internet Archive 的 Wayback Machine 保存其内容。倡导者认为，此举威胁到新闻作品的长期可获取性与完整性。自 2026 年初起，这些机构主动封堵该存档工具，理由是担心生成式 AI 公司可能抓取其付费墙后的内容来训练模型。

来自维权组织 Fight for the Future 的请愿发起人反驳称，这些对 AI 的担忧"纯属臆测"，忽视了 Wayback Machine 数十年来作为非营利性公共服务的良好记录——它并不会刻意绕过付费墙，并以诚信方式运作。他们指出，即便没有 Wayback Machine，AI 公司也已能直接从出版商网站抓取内容，而 Internet Archive 自愿遵守 robots.txt 等规则以避免此类行为。

这场运动将封堵 Wayback Machine 视为对新闻自由的直接威胁，强调新闻的力量不仅在于发布，更在于为后世保存。发起人举出一个颇具讽刺意味的例子：USA Today 的调查报道本身就依赖 Wayback Machine 归档的网页内容，但该机构同时却阻止自己的报道被保存。超过 100 名记者联署了一封信，赞扬 Internet Archive 的作用并引发公众广泛讨论，但 The Atlantic 的 CEO 明确拒绝承诺寻找恢复存档权限的解决方案。此举发生在全球新闻业面临日益严峻威胁的背景下，包括审查、威权压力，甚至针对记者的暴力，这些都使独立存档比以往更加重要。

请愿书将 Wayback Machine 定位为至关重要的中立第三方，认为被归档的新闻更能抵御那些可能施压媒体修改或删除不利报道的强大势力。这种保障不仅对历史记录至关重要，也被视为强化民主问责的工具——确保事实在政治或商业压力下仍然可查可得。活动人士强调，独立存档符合任何严肃新闻机构追求真相的根本利益，但他们发现与 Internet Archive 在这项基本保存工作上的合作竟异常困难。

请愿的核心诉求很明确：各大媒体机构的领导者必须公开承诺与 Internet Archive 合作，恢复并维持新闻内容的存档。组织者强调，在虚假信息泛滥、记者身处直接威胁的时代，应对这些挑战的办法应是加强保存与开放获取，而非削弱。他们将 Wayback Machine 可能失去新闻内容视为对互联网上最重要档案工具的致命打击，敦促新闻机构立即改弦更张，支持这一维护公共信息完整性的盟友。 Major news outlets like the New York Times, The Atlantic, and USA Today have recently stopped allowing the Internet Archive's Wayback Machine to preserve their content, a move that advocates argue threatens the long-term accessibility and integrity of journalistic work. Since early 2026, these organizations have actively blocked the archiving tool, citing concerns over generative AI companies potentially scraping their paywalled material to train models. Petition organizers from activist group Fight for the Future counter that these AI concerns are "wholly hypothetical" and ignore the Wayback Machine's decades-long track record as a respectful, nonprofit public service that skips paywalls and operates with integrity. They argue that AI companies can already scrape content directly from publisher sites, but the Internet Archive voluntarily adheres to rules like robots.txt to avoid doing so.

The campaign frames the blocking of the Wayback Machine as a direct threat to press freedom, emphasizing that journalism's power depends not only on publication but on preservation for future generations. Organizers highlight a bitter irony exemplified by USA Today, which publishes investigative reporting that itself relies on archived web content through the Wayback Machine while simultaneously preventing its own work from being similarly preserved. Over 100 journalists signed a letter celebrating the Internet Archive's role, which sparked significant public discussion, yet The Atlantic's CEO notably declined to commit to finding a solution that would restore archiving access. This stance comes amid growing global threats to journalism, including censorship, authoritarian pressure, and even violence against reporters, making independent preservation more critical than ever.

Positioning the Wayback Machine as a vital neutral third party, the petition argues that archived news is more resilient against powerful interests who might pressure outlets to alter or remove damaging stories. This safeguard is portrayed as essential not just for historical record, but as a tool that actively strengthens democratic accountability by ensuring facts remain accessible despite political or commercial pressures. The campaign underscores that independent archiving serves the fundamental interest of any serious news organization committed to truth, yet finds it unnecessarily difficult to collaborate with the Internet Archive on this basic function of preservation.

The petition's core demand is straightforward. Leaders of major media outlets must publicly commit to working with the Internet Archive to restore and maintain the archiving of news content. Organizers stress that in an era of rising disinformation and direct threats to journalists, the way to combat these challenges is through more preservation and access, not less. They view the Wayback Machine's potential loss of news content as a mortal peril to the internet's most powerful archiving tool, urging news organizations to immediately reverse course and champion this ally in sustaining the integrity of public information.

Muneeb 和 Sohaib Akhter 是一对双胞胎兄弟，先前因计算机欺诈被判重罪。 2025 年 2 月，他们在一家为政府承包商担任 IT 职位时被解雇。被解雇数分钟内，Muneeb 利用权限撤销的疏漏，系统性删除了 96 个包含美国政府信息的数据库，其中包括 Department of Homeland Security 和 Equal Employment Opportunity Commission 的系统。他还下载了敏感文件，并通过清除系统日志、咨询 AI 工具如何抹去证据等手段试图掩盖痕迹。

两人有长期犯罪史。 2015 年，兄弟俩均对电信欺诈与计算机犯罪认罪，Muneeb 服刑三年，Sohaib 服刑两年。获释后，他们进入总部位于 Washington, DC 的 Opexus 工作，该公司为 45 个联邦客户提供服务。尽管进行了背景调查，但他们的前科要么被遗漏，要么审查不充分。在职期间，Muneeb 从公司网络中窃取了约 5,400 份凭证，借助定制脚本访问包括 Marriott 、 DocuSign 和多家航空服务在内的账户，有时还用受害者的里程为自己预订旅行。

两人在下午 4:50 通过 Microsoft Teams 通话被解雇后，Sohaib 的访问权限被立即撤销，但 Muneeb 的凭证仍然有效。在 4:56 至 17:50 之间，Muneeb 下达了多条破坏性命令，兄弟俩在此期间实时沟通。 Sohaib 煽动并建议删除文件系统、讨论可能的勒索方案，但 Muneeb 认为勒索风险太高而予以拒绝。事后，他们在一名未具名同谋的协助下重装了笔记本的操作系统。

三周后，联邦探员突袭了 Sohaib 的住所，查获了电子设备，并发现 7 支枪支和 370 发弹药——鉴于其前科，这些均属非法持有。两人于 2025 年 12 月被捕并面临多项指控。 Muneeb 于 2026 年 4 月签署认罪协议，承认主要指控；Sohaib 则诉诸审判，最终被判犯有共谋实施计算机欺诈、密码贩运及非法持枪罪，量刑将在九月进行。

Muneeb 此后自狱中提交手写请愿，声称辩护律师失职并试图撤回认罪，称签字系在压力下作出；他还申请自行辩护，但此类请求在联邦法院罕有成功。本案暴露出在解除访问权限和背景筛查流程上的严重失误，Opexus 承认本应采取更严格的审查，并表示负责雇用这对双胞胎的相关员工已不再任职。 Muneeb and Sohaib Akhter, twin brothers with prior felony convictions for computer fraud, were fired from their IT jobs at a government contractor in February 2025. Within minutes of being terminated, Muneeb exploited a lapse in access revocation to systematically delete 96 databases containing US government information, including systems for the Department of Homeland Security and the Equal Employment Opportunity Commission. He also downloaded sensitive files and attempted to cover his tracks by clearing system logs and consulting an AI tool on how to erase evidence.

The brothers had a history of criminal activity. In 2015, both pleaded guilty to wire fraud and computer crimes, with Muneeb serving three years and Sohaib serving two. After their release, they secured positions at Opexus, a Washington, DC-based firm serving 45 federal clients. Despite background checks, their past was either missed or inadequately vetted. During their employment, Muneeb amassed 5,400 stolen credentials from company networks, using custom scripts to access accounts on platforms like Marriott, DocuSign, and airline services, sometimes booking personal travel with victims' miles.

After their firing via Microsoft Teams call at 4:50 pm, Sohaib found his access immediately revoked, but Muneeb's credentials remained active. Between 4:56 pm and 5:50 pm, Muneeb executed destructive commands while the brothers communicated in real time. Sohaib encouraged the destruction, suggesting filesystem deletion and discussing potential blackmail schemes, though Muneeb rejected extortion as too incriminating. They later reinstalled operating systems on their laptops with help from an unnamed co-conspirator.

Federal agents raided Sohaib's home three weeks later, seizing technology equipment and discovering seven firearms and 370 rounds of ammunition, illegal possessions given his prior convictions. Both brothers were arrested in December 2025 and indicted on multiple charges. Muneeb signed a plea deal in April 2026 admitting to the major allegations, while Sohaib went to trial and was found guilty of conspiracy to commit computer fraud, password trafficking, and illegal firearm possession. He awaits sentencing in September.

Muneeb has since filed handwritten petitions from jail claiming ineffective counsel and attempting to withdraw his guilty plea, arguing he was pressured into signing it. He has also requested to represent himself, a move that rarely succeeds in federal court. The case highlights critical failures in both credential management during terminations and background screening processes, with Opexus acknowledging that additional diligence should have been applied and that the individuals responsible for hiring the twins are no longer employed.

该仓库 OrcaSlicer-bambulab 由 FULU-Foundation 维护，是 OrcaSlicer 的一个专用分支，旨在为 Bambu Lab 3D 打印机恢复完整的 BambuNetwork 支持。与可能仅限本地网络（LAN）的标准版本不同，此构建通过 BambuNetwork 恢复互联网远程控制和打印功能，使用户能够像在限制出台前一样远程操作。该项目在 GitHub 上反响热烈，拥有 1.5k 星和 383 个 fork 。

安装方式因操作系统而异：Windows 用户需在以管理员权限运行的命令提示符或 PowerShell 中执行指定命令以启用 WSL 2（Windows Subsystem for Linux），然后才能启动 Orca Studio；Linux 用户可按常规方式安装，无需额外依赖；macOS 支持仍在开发中，尚未完全可用。

仓库包含面向 Windows 、 Linux 和 macOS 的各类构建脚本与配置文件。主要代码以 C++ 为主（约 82.5%），其余包括 C 、 JavaScript 、 HTML 、 CMake 和 Shell 脚本。项目采用 AGPL-3.0 许可证，保证其开源属性。

维护者 codedbyjake 还建议用户查看其其它仓库中的 BMCU 固件，BMCU 似乎是用于 3D 打印机的配套固件。仓库最近有活动，最新提交修复了工作流之间的工件名称冲突，首个发行版 v1.0.0 于 2026 年 5 月 12 日发布。对于希望通过 OrcaSlicer 保持对 Bambu Lab 打印机完整远程访问的用户而言，该项目是重要资源。 This repository, OrcaSlicer-bambulab by FULU-Foundation, is a specialized fork of the OrcaSlicer software designed to restore full BambuNetwork support for Bambu Lab 3D printers. Unlike the standard version which may limit users to local network (LAN) only connections, this build enables full functionality over the internet through BambuNetwork, allowing users to control and print remotely just as they could before any restrictions were implemented. The project has gained significant traction with 1.5k stars and 383 forks on GitHub.

Installation varies by operating system. Windows users need to enable WSL 2 (Windows Subsystem for Linux) by running specific commands in an elevated Command Prompt or PowerShell before launching Orca Studio. Linux users can proceed with a normal installation without additional requirements. macOS support is currently listed as a work in progress, indicating it's not yet fully functional or available.

The repository includes various build scripts and configuration files for different platforms, including Windows, Linux, and macOS. The primary codebase is written predominantly in C++ (82.5%), with smaller portions in C, JavaScript, HTML, CMake, and Shell scripting. The project is licensed under AGPL-3.0, ensuring it remains open source.

The maintainer, codedbyjake, also encourages users to explore BMCU firmware, which can be found in their other repositories. BMCU appears to be complementary firmware for 3D printers. The repository shows recent activity with the latest commit fixing artifact name conflicts between workflows, and the first release (v1.0.0) was published on May 12, 2026. The project serves as an important resource for Bambu Lab printer users who want to maintain full remote access capabilities through OrcaSlicer.

Scrcpy v4.0 正式发布，这是这款流行 Android 屏幕镜像工具的一次重要升级。最显著的变化是从 SDL2 迁移到 SDL3，带来了持续维护与错误修复，并新增了在调整窗口大小时锁定宽高比等功能，这一底层改进为本次发布的多项新特性奠定了基础。

一个重大新增是弹性显示支持，允许虚拟显示随客户端窗口动态调整大小，可通过 --flex-display 或 -x 参数启用。这让在桌面上以可调整尺寸的窗口运行 Android 应用变得更加实用。新版还加入了摄像头控制，支持在会话中用 MOD+t 切换摄像头手电筒、用 MOD+↑/MOD+↓ 调整变焦，也可以通过新的命令行选项在启动时设置。

此次更新在用户体验上也做了若干优化：调整窗口大小时会保留宽高比以避免黑边（如需可通过选项恢复旧行为）；新增的 --keep-active 选项可防止设备进入休眠而无需修改系统全局设置；默认背景色由黑色改为深灰色，断开连接时会先显示一个断开图标两秒再关闭窗口，从而更清晰地区分连接丢失与程序崩溃。

在底层修复方面，v4.0 解决了包括 Meta Quest 闪烁问题、已 root 设备的复制粘贴问题、色彩空间转换问题，以及一个奇怪的 bug：由于 OPUS 音频解码中的非规格化数，播放静音时会导致高 CPU 占用。新增了 F11 全屏和 MOD+q 退出等快捷键。该版本还升级了核心依赖（如 FFmpeg 、 SDL 和 platform-tools（adb）），并在整个代码库中做了多处技术修复与改进。 Scrcpy v4.0 has been released, marking a significant update to the popular Android screen mirroring tool. The most notable change is the migration from SDL2 to SDL3, which brings active maintenance, bug fixes, and new capabilities like aspect ratio locking when resizing the window. This foundational upgrade enables several of the release's headline features.

One major addition is flex display support, allowing virtual displays to be resized dynamically alongside the client window using the `--flex-display` or `-x` flag. This makes it much more practical to run Android apps in resizable windows on a desktop. The release also introduces camera controls, letting users toggle the camera torch with `MOD`+`t` and adjust zoom with `MOD`+`↑`/`MOD`+`↓`, either during a session or at startup via new command-line options.

The update improves the user experience in several ways. Window aspect ratio is now preserved during resizing to avoid black bars, though the old behavior can be restored with a flag. A new `--keep-active` option prevents the device from sleeping without changing global settings. The default background color changed from black to dark gray, and a disconnected icon now appears for two seconds before the window closes, making it clearer when a connection is lost versus a crash.

Under the hood, v4.0 fixes a range of issues including Meta Quest flickering, copy-paste on rooted devices, colorspace conversion, and a curious bug where playing silence consumed high CPU due to denormal numbers in OPUS audio decoding. New shortcuts like `F11` for fullscreen and `MOD`+`q` to quit have been added. The release also upgrades core dependencies including FFmpeg, SDL, and platform-tools (adb), along with various technical fixes and improvements throughout the codebase.

文章介绍了六条排版规则，用来在科幻电影中营造未来感，并以对单词在 Eurostile Bold 字体上的逐步变形为例说明。规则一是加入斜体倾斜，暗示前进的动感；规则二把曲线与棱角结合起来；规则三在字母上加入"consummate Vs"（即尖锐的 V 形切口）；规则四通过字母相互连接，呼应虚构的 "Kern Wars"，并避免触犯 "Kern Tithe"；规则五去掉字母里任意的水平线段；规则六则加入噪点纹理、拉丝金属效果、蓝色光效、浮雕以及星空背景。

作者把 Ridley Scott 的 Blade Runner 视为这些原则的"始祖范例"。文中列举了许多其他电影在不同程度上运用这些规则：Battlestar Galactica (2003) 和 Transformers 几乎涵盖了所有规则，后者尤重拉丝金属效果；Guardians of the Galaxy 除斜体外用到了大多数技巧；RoboCop 强调 consummate Vs 和夸张的浮雕；Star Wars 用相连的字母很好地体现了规则四；The Amazing Spider-Man 也遵循了大部分规则，但在字距处理上有些疏漏。

其它例子还包括：Captain America: The Winter Soldier 在规则二和三上表现突出；Alien vs. Predator 采用了斜体与金属化风格；G.I. Joe: Retaliation 使用了除字距调整外的所有技巧；WALL·E 侧重规则二；Back to the Future 展示了规则一、二和四。文末更新指出，Star Trek: The Next Generation 是一个典型范例，甚至用了星空背景。文章最后提到，这篇内容的扩展版收录在书籍 Typeset in the Future 中。 The article presents six typographic rules for making text look futuristic in science fiction movies, using a step-by-step transformation of a word in Eurostile Bold as an example. Rule 1 involves adding an italic slant to suggest forward motion. Rule 2 introduces a mix of curvy and angular forms. Rule 3 adds "consummate Vs," or sharp V-shaped notches in letters. Rule 4 combines letters to reflect the fictional "Kern Wars" and avoid violating the "Kern Tithe." Rule 5 removes arbitrary horizontal segments from letters. Rule 6 incorporates noise textures, brushed metal effects, blue lighting, embossing, and a star field background.

The author identifies Ridley Scott's Blade Runner as the "Ur Example" of these principles in action. Numerous other films are cited as following these rules to varying degrees. Battlestar Galactica (2003) and Transformers apply nearly all the rules, with the latter emphasizing brushed metal. Guardians of the Galaxy uses most tricks except italics, while RoboCop focuses on consummate Vs and extreme embossing. Star Wars exemplifies Rule 4 with its connected letters, and The Amazing Spider-Man follows most rules but is noted for missed kerning opportunities.

Further examples include Captain America: The Winter Soldier for its strong use of Rules 2 and 3, Alien vs. Predator for its italic and metallic style, and G.I. Joe: Retaliation for using every trick except kerning. WALL·E is highlighted for its focus on Rule 2, and Back to the Future demonstrates Rules 1, 2, and 4. An update notes that Star Trek: The Next Generation is a quintessential example, even featuring a star field. The article concludes by noting that an expanded version appears in the book Typeset in the Future.

2026 年 5 月 11 日，Simon Kelley 宣布 CERT 正在披露 dnsmasq 中的六个严重安全漏洞。这些漏洞长期存在，影响几乎所有现代版本的软件。厂商已被提前通知，预计很快会有相应的 dnsmasq 软件包补丁发布。详细信息和补丁可在 dnsmasq 官网找到，修复版本 2.92rel2 已可下载。

这些修复也已提交到开发分支。部分是向后移植的补丁，另一些则是更全面的重写，旨在从根本上解决问题。 Simon 指出，基于 AI 的安全研究大幅增加，导致漏洞报告激增。他花了大量时间来管理这些报告、过滤重复项，并判断哪些问题需要向供应商预先通报、哪些需要立即修复。他认为，鉴于"好人"和"坏人"都可能发现这些漏洞，长期保密期变得越来越不现实，协调保密和回溯修补的工作也非常繁重。

Simon 强调必须在今后的版本中修复这些漏洞，力求让新版本尽可能无漏洞。他提到，在此次公告发布前的几周里，已有多项安全修复提交到 git 仓库。他计划尽快发布 dnsmasq-2.93rc1，目标是尽快推出稳定的 2.93 版本，并鼓励社区测试该发布候选版本，以帮助尽早推出稳定版本。

展望未来，Simon 承认 AI 生成的漏洞报告潮流不太可能停止，因此漏洞管理和修复的工作很可能会持续。他计划优先保证 2.93 按时发布，即便这意味着一些正在进行的修复将在发布后再处理。总体目标仍是不断提高 dnsmasq 的安全性和稳定性。 On May 11, 2026, Simon Kelley announced that six serious security vulnerabilities in dnsmasq are being disclosed by CERT. These are long-standing bugs affecting nearly all modern versions of the software. The vulnerabilities have already been pre-disclosed to vendors, so patched versions of their dnsmasq packages are expected to be released soon. Details and patches are available on the official dnsmasq website, and a fixed release, version 2.92rel2, has been made available for download.

The fixes for these bugs have also been committed to the development tree. While some fixes are backported patches, others are more comprehensive rewrites aimed at addressing root causes. Simon notes that there has been a significant increase in AI-based security research, leading to a flood of bug reports. He has spent considerable time managing these reports, filtering duplicates, and deciding which issues require vendor pre-disclosure and which should be fixed immediately. He believes that long embargoes are becoming less practical, given that both "good guys" and "bad guys" are likely finding these bugs, and the effort to coordinate embargoes and backports is substantial.

Simon emphasizes the importance of fixing bugs in future releases and making new versions as bug-free as possible. He mentions that many security-fix commits have already been made to the git repository in the weeks leading up to this announcement. He plans to release dnsmasq-2.93rc1 soon, with the goal of a stable 2.93 release as quickly as possible. He encourages community testing of the release candidate to help ensure a timely and stable release.

Looking ahead, Simon acknowledges that the influx of AI-generated bug reports is unlikely to stop, so this process of managing and fixing vulnerabilities will likely continue. He plans to prioritize the timely release of version 2.93, even if it means some ongoing fixes will be addressed after the release. The focus remains on improving the security and stability of dnsmasq moving forward.

Needle 是 Cactus Compute 开发的一款 2600 万参数的"Simple Attention Network"，由 Gemini 3.1 蒸馏而来，专为在手机、手表、眼镜等消费端设备上高效运行而设计。该模型擅长个人 AI 场景下的一次性函数调用（single-shot function calling），在该任务上优于 FunctionGemma-270m 和 Qwen-0.6B 等更大模型。架构紧凑，采用编码器—解码器设计，使用了 GQA 、 RoPE 和 ZCRMSNorm 等技术；在 2000 亿 token 上完成预训练，随后在一个包含 20 亿 token 的函数调用数据集上进行了微调。

为设备端部署做了大量优化：在 Cactus 基础设施上运行时，预填充（prefill）速度可达每秒 6000 个 token，解码速度可达每秒 1200 个 token 。模型权重和数据集生成代码全部开源，可在 Hugging Face 的 Cactus-Compute/needle 仓库获取。网络由 12 层编码器和 8 层解码器组成，嵌入层权重共享，词表为 8192 个 BPE token 。

上手简单：克隆仓库并运行安装脚本后可在 localhost:7860 启动网页交互界面，方便测试模型并在自定义工具上进行微调。提供的 Python API 支持加载检查点、对输入进行 token 化并生成函数调用输出，接收文本查询和工具定义，返回结构化的函数调用参数。

在定制方面，Needle 同时提供网页界面和命令行工具，可在自定义数据集上微调。网页界面能利用 Gemini 生成合成训练数据、训练模型、评估性能并打包结果；CLI 支持完整训练流程、在 PleIAs/SYNTH 数据集上预训练、检查点评估，以及为有 Google Cloud 权限的用户管理 TPU 等操作。

尽管 Needle 在函数调用任务上表现突出，开发者也指出它仍属实验性、主要面向边缘设备的小型 AI 方向。更大的模型在对话能力和通用性上仍具优势。项目鼓励用户用自己的工具测试并根据需要微调，同时也提醒小模型有时表现可能较为不稳定。 Needle is a 26-million-parameter "Simple Attention Network" developed by Cactus Compute, distilled from Gemini 3.1 and designed to run efficiently on consumer devices like phones, watches, and glasses. The model specializes in single-shot function calling for personal AI applications, outperforming larger models like FunctionGemma-270m and Qwen-0.6B on this specific task. It features a compact architecture with an encoder-decoder structure, using techniques like GQA, RoPE, and ZCRMSNorm, and was pretrained on 200 billion tokens before being fine-tuned on a 2-billion-token function call dataset.

The model is optimized for on-device deployment, achieving impressive inference speeds of 6000 tokens per second for prefill and 1200 tokens per second for decode when running on Cactus infrastructure. Weights and dataset generation code are fully open-source, available on Hugging Face under the Cactus-Compute/needle repository. The architecture includes 12 encoder layers and 8 decoder layers, with tied embeddings and a shared vocabulary of 8192 BPE tokens.

Getting started with Needle is straightforward. Users can clone the repository and run the setup script to launch a web playground at localhost:7860, where they can test the model and fine-tune it on custom tools with minimal effort. The Python API allows for simple integration, with functions for loading checkpoints, tokenizing inputs, and generating function call outputs. The model accepts a text query and tool definitions, returning structured function call arguments.

For customization, Needle provides both a web interface and CLI tools for fine-tuning on custom datasets. The playground can generate synthetic training data using Gemini, train the model, evaluate performance, and bundle the results. The CLI supports various operations including full training runs, pretraining on the PleIAs/SYNTH dataset, evaluation of checkpoints, and TPU management for those with access to Google Cloud infrastructure.

While Needle excels at function calling, the developers note that it's experimental and focused specifically on redefining tiny AI for edge devices. Larger models still have advantages in conversational settings and general scope. The project encourages users to test the model with their own tools and fine-tune as needed, acknowledging that small models can sometimes be finicky in their behavior.

DuckDB 团队推出了 Quack，一种新的客户端 - 服务器协议，允许多个 DuckDB 实例相互通信。传统上 DuckDB 是进程内数据库，类似于 SQLite，在单一进程内运行、没有客户端 - 服务器架构。虽然这种设计在交互式分析和嵌入式场景中很出色，但当多个进程需要同时修改同一个数据库时就显得不足。很多用户不得不采用变通办法，比如自建 RPC 、使用 Arrow Flight SQL 的第三方实现，甚至在 PostgreSQL 中运行 DuckDB 。 Quack 直接填补了这一空白，使独立进程能够并发读写，实现完整的多用户体验。

从用户角度看，Quack 很容易上手。客户端和服务器都是 DuckDB 实例，功能通过 DuckDB v1.5.2 中的 Quack 扩展提供。使用时在一台实例上启动 Quack 服务器，另一台用简单的认证令牌连接即可。连接后，查询远程表与查询本地表一样简单。客户端还可以在远程服务器上创建表，或通过 query 函数把复杂的原生 SQL 直接发到远端执行。

Quack 的设计基于成熟技术并针对性能进行了优化。它以 HTTP 作为传输层，能利用现有且高度优化的基础设施，便于与负载均衡、防火墙等工具集成，也支持 DuckDB-Wasm 实例从浏览器原生连接。协议消息使用自定义 MIME 类型 application/duckdb，复用 DuckDB 自身高效且经过验证的序列化原语。安全性默认通过自动生成的认证令牌和绑定到 localhost 来保障；如果需要将服务器暴露到互联网，强烈建议用 nginx 等 HTTP 代理做 SSL 终止。性能方面，Quack 优化到可以在一次往返中完成查询及其初始结果返回，这对延迟敏感的操作至关重要，同时在大批量数据传输上也非常高效。

对 Quack 、 PostgreSQL 和 Arrow Flight SQL 的基准测试显示了令人信服的结果。在批量数据传输场景中，Quack 表现显著优于两者——在不到 5 秒内传输了 6000 万行数据，比 Arrow Flight 更快，且远超 PostgreSQL 。在小写入场景（强调往返效率）下，Quack 在最多 8 个并行线程时也意外地优于 PostgreSQL，达到约每秒 5500 次事务的水平。这些性能使 DuckDB 能够支持新的场景，例如汇总遥测数据或让多个数据生产者驱动仪表板，从而将其从"进程内分析工具"的小众角色推向数据架构中的更核心位置。

Quack 的发布只是更大路线图的第一步。未来计划包括把 Quack 集成到 DuckLake 以支持远程 catalog 服务器，改进 DuckDB 核心以应对更高的事务吞吐和更多并行线程；团队还在考虑通过扩展支持自定义消息，并增加用于只读副本的复制功能。该协议将在 2026 年秋与 DuckDB v2.0 一同为生产环境的发布而进一步完善。团队也承认 MotherDuck 、 GizmoSQL 等项目对其设计的影响；虽然他们选择不采用 Arrow Flight SQL，以保持对序列化的控制并避免强制性的多轮往返，但仍认可其作为交换格式的价值。 The DuckDB team has introduced Quack, a new client-server protocol that allows multiple DuckDB instances to communicate with each other. Traditionally, DuckDB has been an in-process database, meaning it runs within a single process without a client-server architecture, similar to SQLite. While this design excels in interactive analytics and embedded use cases, it falls short when multiple processes need to modify the same database simultaneously. Many users resorted to workarounds like custom RPC solutions, third-party implementations using Arrow Flight SQL, or even running DuckDB inside PostgreSQL. Quack addresses this gap directly, enabling a full multi-user experience where separate processes can read and write data concurrently.

From a user perspective, Quack is straightforward to set up. Both the client and server are DuckDB instances, and the functionality is provided through a Quack extension available in DuckDB v1.5.2. To connect, a user starts a Quack server on one instance and connects to it from another using a simple authentication token. Once connected, querying a remote table is as easy as querying a local one. The protocol also supports creating tables on the remote server from the client side and even shipping complex, verbatim SQL queries for remote execution using a `query` function.

The design of Quack is built on proven technologies and optimized for performance. It uses HTTP as its transport layer, benefiting from a ubiquitous and highly optimized infrastructure that is easy to manage with tools like load balancers and firewalls. This also allows DuckDB-Wasm instances to connect natively from a browser. Protocol messages use a custom MIME type, `application/duckdb`, which leverages DuckDB's own efficient and well-tested serialization primitives. Security is handled by default through auto-generated authentication tokens and binding to localhost, with strong recommendations to use an HTTP proxy like nginx for SSL termination when exposing the server to the internet. For performance, Quack is optimized to handle a query and its initial results in a single round trip, which is crucial for latency-sensitive operations, while also being highly efficient at bulk data transfer.

Benchmarking Quack against PostgreSQL and Arrow Flight SQL reveals compelling results. For bulk data transfer, Quack significantly outperforms both, transferring 60 million rows in under 5 seconds, which is faster than Arrow Flight and dramatically faster than PostgreSQL. For small writes, which stress round-trip efficiency, Quack surprisingly outperforms PostgreSQL up to 8 parallel threads, achieving around 5,500 transactions per second. This performance unlocks new use cases for DuckDB, such as centralizing telemetry data or driving dashboards from multiple data producers, moving it beyond its niche as an in-process analytics tool into a more central role in data architecture.

The release of Quack is the first step in a larger roadmap. Future plans include integrating Quack into DuckLake to enable remote catalog servers and improving DuckDB's core to handle higher transaction throughput with more parallel threads. The team is also considering extending the protocol to allow custom messages via extensions and adding replication capabilities for read replicas. The protocol will be refined for a production release alongside DuckDB v2.0 in the fall of 2026. The team acknowledges the influence of projects like MotherDuck and GizmoSQL, and while they chose not to use Arrow Flight SQL to maintain control over their serialization and avoid mandatory multi-round-trip waits, they see its value as an interchange format.

Google 正在推出一款名为 Googlebook 的新笔记本电脑，定于 2026 年秋季发布。该设备旨在全面整合 Google 的 Gemini AI，将 AI 定位为现代计算的决定性特征，而不仅仅是硬件规格。

这款笔记本配备了多项由 Gemini 驱动的工具，帮助简化日常任务。 Magic Pointer 允许用户选中屏幕上的任意元素并即时让 Gemini 对其进行操作，例如将多张图像合成为定制乐队海报。 Create My Widget 则可通过简单的语音或文本指令生成个性化桌面小部件，比如为即将到来的 Iceland 之旅创建实时行程追踪器。

除 AI 功能外，Googlebook 还与 Android 手机实现无缝协作。 Cast My Apps 允许用户在笔记本上直接打开并使用手机应用，无需安装；Quick Access 则能像访问本地存储一样浏览手机上的文件。此集成需在运行 Android 17 或更高版本的手机上并完成初始设置。

硬件在追求轻巧机身的同时也提供强劲性能。 Google 与 Acer 、 ASUS 、 Dell 、 HP 和 Lenovo 等主要厂商合作打造该设备。笔记本在设计上也有细节考量，比如键盘上专门的 Google "G" 键，以及带有指纹印记设计的标准退格键。

Google 强调应对 Gemini 的回答进行核实，功能可能会发生变化，视觉匹配的结果也可能有所不同。某些语言在发布时可能不可用。感兴趣的用户可注册以在 2026 年秋季设备发售时收到通知。 Google is introducing a new laptop called Googlebook, set to launch in fall 2026. The device is designed to fully integrate Google's Gemini AI intelligence, positioning AI as the defining feature of modern computing rather than just hardware specs.

The laptop features several Gemini-powered tools that aim to make everyday tasks easier. Magic Pointer lets users select any element on screen and instantly ask Gemini to work with it, such as combining images to create a custom band poster. Create My Widget allows users to build personalized desktop widgets through simple voice or text commands, like generating a live travel tracker for an upcoming trip to Iceland.

Beyond AI features, Googlebook is built to work seamlessly with Android phones. Cast My Apps enables users to open and use their phone applications directly on the laptop without needing to install them, while Quick Access lets users browse files from their phone as if they were stored locally on the device. This integration requires a phone running Android 17 or above and initial setup.

The hardware itself follows a featherweight design philosophy while delivering substantial power. Google has partnered with major manufacturers including Acer, ASUS, Dell, HP, and Lenovo for the device. The laptop includes thoughtful design touches like a dedicated Google "G" key on the keyboard, alongside the standard backspace key with a fingerprint impression design.

Google emphasizes that responses from Gemini should be verified and that features are subject to change, with results varying based on visual matches. Some language availability may differ at launch. Interested users can sign up to be notified when the device becomes available this fall.

Canada 的 Bill C-22（The Lawful Access Act）重新引入了有问题的监控措施，几乎是去年被撤回的 Bill C-2 的翻版。尽管 Bill C-2 在隐私团体强烈反对后被撤回，Bill C-22 却以边境安全为由保留了许多相同的问题，试图以牺牲数字隐私为代价扩大政府的监控权力。

该法案将要求包括电信运营商和消息应用在内的数字服务记录并保存用户的元数据长达一年。元数据能揭示人们的通讯对象、位置以及日常行为模式。扩大元数据的收集与保存，会迫使企业存储比现在更多的个人信息，从而增加数据泄露和被恶意方未经授权访问的风险。

更令人担忧的是，法案授权 Canada 的 Minister of Public Safety 要求公司为其服务设置后门，供执法部门访问。法案声称这些命令不会引入"systemic vulnerabilities"，但专家指出在处理加密通信时，这在技术上不可能实现。法案还禁止公司公开披露此类政府命令的存在，完全消除了监控活动的透明度。

法案中对"systemic vulnerabilities"和"encryption"的定义模糊不清，为政府越权留下了大量空间。加拿大官员声称可以在不造成系统性漏洞的情况下增加监控，但这与技术现实相悖。宽泛的定义可能涵盖各类应用和操作系统，从而影响范围广泛的数字服务。

这一做法与英国的情况类似：英国政府曾要求 Apple 在其 Advanced Data Protection 功能中设置后门，Apple 最终为英国用户移除了该功能而非妥协。 Meta 和 Apple 都对 Bill C-22 可能赋予 Canadian authorities 类似权力表示担忧，并公开反对该法案。 US House Judiciary and Foreign Affairs committees 也在致 Canada 的 Minister of Public Safety 的联名信中表达了关切。

现实事件已经证明了监控后门的危险，例如 2024 年的 "Salt Typhoon" 黑客事件，攻击者正是利用了为执法访问用户数据而建立的系统。构建这些监控机制必然会带来可被恶意利用的漏洞。

公众理应享有强有力的隐私保护、透明的数据处理以及对加密信息的明确保障。 Bill C-22 未能提供这些保护，反而推动扩大所谓的合法访问机制，将危及数百万用户的数字安全。 Canada's Bill C-22, also known as The Lawful Access Act, is a reintroduction of problematic surveillance measures that mirror last year's failed Bill C-2. While Bill C-2 was withdrawn after significant backlash from the privacy community, Bill C-22 retains many of the same concerns under the guise of border security. The legislation represents another attempt to expand government surveillance powers at the expense of digital privacy rights.

The bill would force digital services, including telecoms and messaging apps, to record and retain user metadata for a full year. Metadata reveals extensive information about who people communicate with, their locations, and their daily patterns. Expanding metadata collection requirements would compel companies to store even more personal data than they currently maintain, creating larger targets for potential data breaches and unauthorized access by bad actors.

Perhaps most concerning is the provision allowing Canada's Minister of Public Safety to demand companies create backdoors in their services for law enforcement access. The bill claims these mandates cannot introduce "systemic vulnerabilities," though experts note this is technically impossible when dealing with encrypted communications. The legislation also prohibits companies from publicly disclosing the existence of such government orders, eliminating transparency around surveillance activities.

The definitions of both "systemic vulnerabilities" and "encryption" remain unclear throughout the bill, leaving significant room for government overreach. Canadian officials have maintained it's possible to add surveillance without creating systemic vulnerabilities, but this contradicts technical reality. The broad definitions could encompass apps and operating systems alike, potentially affecting a wide range of digital services.

This approach mirrors what occurred in the UK, where the government demanded Apple implement backdoors into its Advanced Data Protection feature. Apple ultimately removed the feature for UK users rather than comply. Both Meta and Apple have expressed concerns that Bill C-22 would grant Canadian authorities similar powers, and both companies have publicly opposed the legislation. The US House Judiciary and Foreign Affairs committees also raised concerns in a joint letter to Canada's Minister of Public Safety.

The dangers of surveillance backdoors are demonstrated by real-world incidents like the 2024 Salt Typhoon hack, which exploited systems built for law enforcement access to user data. Building these surveillance mechanisms inevitably creates vulnerabilities that malicious actors can access. Canadians deserve robust privacy protections, transparent data handling, and clear safeguards for encrypted information. Bill C-22 fails to provide any of these protections, instead pushing for expansive lawful access mechanisms that would compromise digital security for millions of users.