Ox64 是一款经济实惠的 8 美元 RISC-V 单板计算机，提供了出人意料的强大能力，包括三个不同的核心以及对 Linux 的支持。它看起来像普通微控制器，但实际运行方式更接近标准的 UNIX 机器。不过初次体验并非即插即用：在测试时，以太网、 Wi-Fi 和 SD 卡支持等关键组件尚未能正常工作，而且整个配置过程对用户提出了不少技术挑战。

让这块开发板顺利跑起来的主要障碍在于，它无法像许多常见开发板那样通过 USB 编程。用户只能依赖 UART 引脚。此外，随附的烧录工具 Dev Cube 在运行 Wayland 的 Ubuntu 系统上很容易崩溃，用户需要切换到 Xorg 会话，或调整环境变量来维持稳定性。硬件通信同样不顺，常见的 USB 转串口适配器在握手过程中经常失败。

为了绕开这些连接问题，一个有效的变通办法是改用 RP2040 设备（例如 Adafruit KB2040）来充当计算机与 Ox64 之间的串口桥。通过给 RP2040 刷入 picoprobe 固件，并细致地接好 GPIO 引脚，即可建立稳定连接。即便接上这座桥，原本的标准烧录工具依然存在问题，因此需要改用 bflb-mcu-tool 等替代软件。

最后，要把替代工具用起来，还得手动修改其 Python 代码，移除会冲突的 DTR 和 RTS 信号。完成这些配置调整后，就能成功烧录初始固件，例如 M0 核心对应的底层加载文件。虽然这只是实现完整 Linux 启动的第一步，但 Ox64 与 PC 之间成功建立通信，令人备受鼓舞：说明硬件工作正常、响应灵敏，并已准备好进入更深入的开发阶段。 The Ox64 is a budget-friendly, eight-dollar RISC-V single-board computer that offers a surprising level of capability, including three distinct cores and support for Linux. While it resembles a simple microcontroller, it functions much like a standard UNIX machine. However, the initial experience is not plug-and-play, as essential components like ethernet, Wi-Fi, and SD-card support were not yet functional at the time of testing, and the setup process presents several technical challenges for the user.

A primary hurdle in getting the board running is that it cannot be programmed through USB like many common development boards. Instead, users must rely on UART pins. Furthermore, the provided flashing tool, Dev Cube, is prone to crashing on Ubuntu systems running Wayland, requiring users to switch to an Xorg session or modify environment variables to maintain stability. Hardware communication also proved difficult, as standard USB-to-serial dongles frequently failed during the handshake process.

To bypass these connection issues, a successful workaround involves repurposing an RP2040 device, such as an Adafruit KB2040, to act as a serial bridge between the computer and the Ox64. By flashing the RP2040 with picoprobe firmware and carefully wiring the GPIO pins, a stable connection can be established. Even with this bridge in place, the standard flashing tool remains problematic, necessitating the use of alternative software like the bflb-mcu-tool.

Finally, integrating this alternative tool requires manual modifications to its Python code to strip away conflicting DTR and RTS signals. Once these configuration adjustments are made, it is possible to successfully flash the initial firmware, such as the low-level load files for the M0 core. While this represents only the first step toward achieving a full Linux boot, the successful communication between the Ox64 and the PC serves as a promising confirmation that the hardware is responsive and ready for further development.

Discord 正在进行重大转变，放弃 Electron 框架，转而推出全新、高度优化的 Rust 客户端。公司展示的结果表明，新版本的速度提升非常显著，彻底摆脱了长期以来基于 Electron 的应用程序常见且缓慢的"Starting"界面。在并排对比中，Rust 版本在原客户端完成初始启动之前，就已经成功加载了用户的好友列表，效率差距一目了然。

这次技术重构在内存管理方面同样带来了大幅改进。目前的 Electron 客户端在空闲状态下往往会占用 1GB 到 2GB 的 RAM，而即使同时运行两个用户账户，新的 Rust 重建版本也大约只使用 290MB 内存，内存开销减少约 70% 。 Discord 将这些成果归功于 GPUI——由 Zed 代码编辑器背后的团队开发的开源、 GPU 加速 UI 框架。通过让界面直接在 GPU 上绘制，开发团队绕开了 Electron 依赖的资源密集型浏览器引擎需求。

此次转型也不止于桌面端。 Discord 正在为移动端从 React Native 转向 Rust，用 Rust 统一代码库；同时，Web 端也将通过 WebAssembly 来承载对应的逻辑。这种架构调整确保应用在不同平台上都能保持功能丰富且运行流畅。尽管内部发生了深层变化，Discord 也确认用户界面将与现有体验保持一致，避免老用户需要重新学习如何使用该平台。

Discord 目前正在为新客户端的内部测试做准备。管理层强调，在更大范围发布之前，他们的首要任务是确保稳定性与安全性。此举也反映出科技行业的更大趋势：开发者越来越倾向于抛弃"web app slop"和臃肿的框架，改用更高效、接近原生的方案。随着由 AI 驱动硬件需求带来的成本持续上升，内存使用变得更为关键，Discord 的这一转变也为包括 WhatsApp 和 Teams 在内的其他消息平台树立了高标准——这些平台仍在很大程度上依赖资源密集型的 Web 技术。 Discord is making a significant shift away from the Electron framework in favor of a new, highly optimized Rust-based client. Demonstrations from the company show that the new version is dramatically faster, eliminating the familiar, sluggish "Starting" screen that has long characterized the Electron-based app. In a side-by-side comparison, the Rust version successfully loads a user's friends list before the original client even finishes its initial boot process, highlighting the stark difference in efficiency.

The technical overhaul brings substantial improvements to memory management as well. While the current Electron client often consumes between 1GB and 2GB of RAM during idle usage, the new Rust rebuild uses roughly 290MB, even when running two user accounts simultaneously. This represents a reduction of approximately 70% in memory overhead. Discord attributes these gains to the use of GPUI, an open-source, GPU-accelerated UI framework developed by the team behind the Zed code editor. By drawing the interface directly on the GPU, the developers have bypassed the resource-heavy browser engine requirements inherent to Electron.

The transition also extends beyond desktop applications. Discord is moving away from React Native for its mobile apps to unify its codebase using Rust, which will also power the logic for its web version via WebAssembly. This architectural change ensures that the app remains feature-rich and performant across different platforms. Despite these deep internal changes, Discord has confirmed that the user interface will remain consistent with the existing experience, ensuring that long-time users will not have to relearn how to navigate the platform.

Discord is currently preparing for private testing of the new client, with leadership emphasizing that their priority is ensuring stability and security before a wider release. The move signals a broader trend in the tech industry where developers are increasingly abandoning "web app slop" and bulky frameworks in favor of more efficient, native-like solutions. As memory usage becomes a more pressing concern due to the rising costs associated with AI-driven hardware demands, this shift by Discord sets a high bar for other messaging platforms like WhatsApp and Teams, which continue to rely on resource-intensive web technologies.

Sam Altman 推广了一种观点：把物理 RAM 留在每台个人电脑里本质上就是低效的做法。对不少消费者来说，这会带来挥之不去的内疚感——他们买了硬件，却没有真正用满。该说法认为，闲置的这些硬件属于资源浪费，本可以被更合理地用在别处。

UploadYourRam.org 这个平台想用一项"革命性"的技术来解决这一困境：让用户把用不上的内存上传起来。与其让多余的容量在本地机器里闲置不动，用户现在可以把它捐给更有产出的用途。这样一来，人们原本觉得"多出来的硬件是负担"的想法，就被转化成了一种数字化的贡献行为。

平台提供多档参与门槛，用户可按自己觉得能腾出的容量来选择。捐出 2 GB 被定位为"有帮助的贡献"，而捐出 4 GB 则被打造成最受欢迎的选项，足以拯救一个 Large Language Model 。若有人特别慷慨，还可以选择 16 GB 的"英雄"档，并承诺能收到 Sam Altman 本人亲笔的感谢信。

该倡议是开发者 TheZoq2 发起的草根项目。整个系统手工搭建，透明公开，完整源代码可在 Codeberg 上供公众查看。它是在硬件所有权格局不断变化、以及向更集中或更高效的计算模式转变的背景下，对现实做出的创造性回应。 Sam Altman has popularized the notion that keeping physical RAM inside individual computers is an inherently inefficient practice. For many consumers, this creates a lingering sense of guilt, as they possess hardware they have already paid for but are not fully utilizing. The concept suggests that this idle hardware is a waste of resources that could be better deployed elsewhere.

The platform UploadYourRam.org aims to solve this dilemma through a revolutionary technology that allows users to upload their unused memory. Instead of letting that extra capacity sit idle within a local machine, individuals can now donate it to a more productive cause. This approach turns the perceived burden of excess hardware into an act of digital contribution.

Users are presented with several tiers of participation depending on how much capacity they feel they can spare. Donating as little as 2 GB is framed as a helpful contribution, while a 4 GB donation is marketed as the most popular option, sufficient to save a single Large Language Model. Those feeling particularly generous can opt for the 16 GB hero tier, which comes with the promise of a personal note of appreciation from Sam Altman himself.

This initiative is a grassroots project created by the developer known as TheZoq2. The entire system is built by hand and remains transparent, with the full source code available for public inspection on Codeberg. It serves as a creative response to the changing landscape of hardware ownership and the push toward more centralized or efficient computing models.

自 macOS 13 起，Apple 已悄然内置了一个复杂的后台系统，该系统由两个主要守护进程（daemons）组成，其共同目的是构建用户的详细个人知识图谱。该系统并非孤立运行，而是会汇总 Mac 上来自多种来源的数据，包括通讯录（Contacts）、日历（Calendar）、照片（Photos）、消息（Messages）、邮件（Mail）、位置历史记录以及屏幕使用时长（Screen Time）。通过交叉比对这些输入，系统生成一个可查询且持久化的数据库，用于追踪你的身份信息、家庭与社交关系、行为模式，甚至你的身体活动与日常作息。

这一机制的核心位于目录 ~/Library/IntelligencePlatform，其中包含 51 个 SQLite 数据库。最关键的是 graph.db，这是一个三元组存储（triple store），用 181 个实体类和 278 个谓词来组织信息。该图谱记录实体之间的关系，并为每条连接附上置信度分数，同时注明该推断的来源。尽管系统也会采用较直接的方法，比如从联系人名片或日历条目复制数据，但它还使用了更先进的技术，包括基于 CoreML 的 XGBoost 分类器、嵌入模型以及本地 LLM，用于对关系进行分类、并识别个人身份。

该系统特别敏感的一点在于，它能够在没有明确用户输入的情况下推断亲属关系与社交亲密程度。例如，Apple 预装了 XGBoost 分类器，通过分析 144 个不同的行为特征（如消息发送频率、照片共现情况、沟通发生的时间等），来判断某人是否属于家人或亲密朋友。由于系统的设计目标是把各种数据点串联起来，它甚至能仅凭元数据与时间戳推断出隐私信息，比如家庭中孩子的年龄、他们所在的学校或托儿所身份，以及设备使用者的规律性睡眠时间。

尽管这种监控的深度令人震惊，但几乎没有官方文档解释这些守护进程如何工作，或用户如何管理系统收集到的数据。删除机制也进一步加剧了隐私风险：当用户删除某个联系人后，实时图谱中的对应记录会被替换，但相关数据往往仍会留在 expired_stable_graph 表中，这意味着推断出的关系以及敏感推断可能会在用户以为已经移除之后很久仍然存在。

为帮助用户理解系统从自己身上推断出了什么，研究人员开发了一款名为 iKnow 的开源工具。该工具提供本地仪表板，允许用户直接在自己的设备上浏览个人知识图谱，让人得以看清操作系统在个人、群组与日常活动之间建立的连接。通过这种可视化，工具揭示了一个现实：Apple 的个人知识图谱本质上是把个人生活集中存放在一个高风险的仓库里，一旦访问控制出现问题就极易被利用，但对被画像的人而言，它在很大程度上仍然是不透明的。 Since macOS 13, Apple has quietly shipped a sophisticated background system consisting of two primary daemons whose collective purpose is to construct a detailed personal knowledge graph of the user. This system does not function in isolation, but instead aggregates data from a wide range of sources on the Mac, including Contacts, Calendar, Photos, Messages, Mail, location history, and Screen Time. By cross-referencing these inputs, the system creates a queryable, persistent database that tracks who you are, your family and social connections, your behavioral patterns, and even your physical movements and routines.

The heart of this mechanism is located within the directory ~/Library/IntelligencePlatform, which houses fifty-one SQLite databases. The most significant of these is graph.db, a triple store that organizes information using a vocabulary of 181 entity classes and 278 predicates. This graph records relationships between entities, assigning each connection a confidence score and documenting the source of the inference. While the system uses straightforward methods like copying data from contact cards or calendar entries, it also employs more advanced techniques, including CoreML-powered XGBoost classifiers, embedding models, and local LLMs to categorize relationships and identify individuals.

A particularly sensitive aspect of this system is its ability to infer kinship and social closeness without explicit user input. For instance, Apple includes pre-shipped XGBoost classifiers that analyze 144 distinct behavioral features, such as message frequency, photo co-occurrence, and communication timing, to determine whether someone is a family member or a close friend. Because the system is designed to link various data points, it can deduce private information, such as the age of a child in a household, the identity of their school or daycare, and the habitual sleep schedules of the device owner, simply by analyzing metadata and timestamps.

Despite the depth of this surveillance, there is almost no official documentation explaining how these daemons function or how users can manage the data they collect. Privacy is complicated further by the way the system handles deletions; when a user deletes a contact, the corresponding record in the live graph is replaced, but the data often lingers in an expired_stable_graph table, meaning that inferred relationships and sensitive inferences can persist long after the user believes they have been removed.

To help users understand what their Mac has inferred about them, researchers have developed an open-source tool called iKnow. This local dashboard allows users to browse their own knowledge graph directly on their machine, providing transparency into the connections the OS has drawn between individuals, groups, and daily activities. By providing this visibility, the tool highlights the reality that Apple's personal knowledge graph acts as a high-stakes, centralized store of personal life that is ripe for exploitation if access controls were ever compromised, yet it remains largely opaque to the person it is profiling.

经营一家专注于 Internet and Technology law 、且完全使用 Free and Open Source Software（FOSS）的精品律师事务所。过去五年多来，这种模式已被证明既可持续又务实。目前的配置仍以 Debian 为主要操作系统，安装在 Lenovo ThinkPad X1 Yoga 上。虽然硬件出现过一些端口问题，但软件生态依然稳健：既有 LibreOffice 、 Thunderbird 这类图形化应用，也有 vim 、 mutt 以及多种命令行界面等终端工具。之所以更偏好终端工作流，是因为它们更高效、干扰更少。

通过自动化，文档制作与管理流程变得越来越精简。 espanso 等文本扩展工具能节省大量时间；自定义脚本，以及 pandoc 、 typst 、 reveal.js 等工具，则可以直接从 Markdown 文件生成达到专业水准的文档和演示。文件管理由 Nautilus 负责，并辅以 vimwiki 等本地维基工具来进行知识管理。安全始终是核心关注点：使用 LUKS 进行磁盘加密；借助 Cryptomator 通过 Syncthing 同步敏感文件；并通过 YubiKey 硬件令牌增强的 Bitwarden 来管理密码。

服务器端的运维已转向在 Intel NUCs 上进行自托管，体现出对更简洁、更可靠方案的选择。 2026 年的技术栈中，最显著的变化之一是用 Syncthing 与 radicale 的组合取代了 Nextcloud，大幅提升了文件同步性能。通信基础设施也进一步优化：Jitsi 已被停用，改用 Signal 来进行安全会议，因为客户的需求常常要求与 Teams 或 Zoom 等主流平台保持兼容。对业务至关重要的功能，如开票和案件管理，则通过 InvoicePlane 、 EspoCRM 等专用 FOSS 应用来持续支撑。

发布与行政维护所依赖的是高度自动化的流程。网站采用 mkdocs 生成静态站点，而博客改用 BSSG——一种基于 bash 的生成器，相比先前工具具有更高的稳定性。合规与行政事务由 paperless-ngx 等用于扫描的工具以及 LiberaForms 等用于安全处理客户问卷的工具提供支持。 Uptime Kuma 等监控服务，以及通过 Wazuh 和 OpenVAS 进行的安全扫描，确保内部基础设施始终稳定且具备韧性。

移动端连接由运行 GrapheneOS 的 Pixel 设备承担。尽管 Android 生态本身存在固有局限，它仍被视为替代主流平台的最可行选择。整个运营过程中，使用 Free software 在与仅使用专有 Microsoft 或 Apple 产品的客户协作时几乎不会带来任何摩擦。事务所坚持开源理念：完全不使用基于 AI 的工具，并将重点放在以人为本、可验证且透明的技术实践上。 Running a boutique law firm focused on internet and technology law using exclusively Free and Open Source Software (FOSS) has proven to be a sustainable, practical model for over five years. The current setup continues to rely on Debian as the primary operating system, installed on a Lenovo ThinkPad X1 Yoga. While the hardware has encountered some port issues, the software ecosystem remains robust, utilizing a mix of graphical applications like LibreOffice and Thunderbird alongside terminal-based tools like vim, mutt, and various command-line interfaces. This preference for terminal-based workflows stems from their efficiency and reduced potential for distraction.

The document production and management process has become increasingly streamlined through automation. Text expansion tools like espanso save significant time, while custom scripts and tools like pandoc, typst, and reveal.js allow for professional-grade document and presentation creation directly from Markdown files. File management is handled through Nautilus, complemented by local wiki tools like vimwiki for knowledge management. Security remains a foundational concern, addressed through LUKS for disk encryption, Cryptomator for syncing sensitive files via Syncthing, and Bitwarden for password management, bolstered by YubiKey hardware tokens.

Server-side operations have shifted toward self-hosting on Intel NUCs, reflecting a move toward simplicity and reliability. Notable changes in the 2026 stack include replacing Nextcloud with a combination of Syncthing and radicale, which has significantly improved file synchronization performance. Communication infrastructure has also been refined, with Jitsi being retired in favor of Signal for secure conferencing, as client demands frequently necessitate compatibility with mainstream platforms like Teams or Zoom. Business-critical functions, such as invoicing and matter management, are maintained through dedicated FOSS applications like InvoicePlane and EspoCRM.

The infrastructure for publishing and administrative upkeep relies on highly automated processes. The website is a static site generated via mkdocs, while the blog now utilizes BSSG, a bash-based generator that offers superior stability compared to previous tools. Compliance and administrative tasks are supported by tools like paperless-ngx for scanning and LiberaForms for handling client questionnaires securely. Monitoring services like Uptime Kuma and security scanning via Wazuh and OpenVAS ensure that the internal infrastructure remains stable and resilient.

Mobile connectivity is managed through a Pixel device running GrapheneOS, which is chosen as the most viable alternative to mainstream platforms despite the inherent limitations of the Android ecosystem. Throughout these operations, the reliance on Free software has presented zero friction when collaborating with clients who exclusively use proprietary Microsoft or Apple products. The firm remains committed to its open-source philosophy, opting out of AI-based tools entirely and maintaining a focus on human-centric, verifiable, and transparent technological practices.

Underground 2 Revived 是一个由粉丝主导的项目，旨在将经典的 PC 赛车游戏 Need for Speed: Underground 2 带入现代网页浏览器。该项目由一位爱好者和他的兄弟作为业余项目开发，目标是通过为 Chrome 和 Firefox 等平台重新编译原始游戏代码，让这款作品再次变得更易访问。项目完全独立，不隶属于 Electronic Arts，也未获得其背书；它更关注保存与社区可访问性，而非商业收益。

该项目提供了一系列技术增强功能，让这款已有二十年历史的游戏达到现代标准。玩家可在宽屏模式下以最高 120 帧每秒的速度畅玩，并全面支持控制器、赛车方向盘以及触控操作。开发团队不仅完成了还原式移植，还加入了多人游戏能力，允许用户与朋友一起进行自由漫游，并参与线上比赛。为保持公平环境，项目运行过程中不提供微交易或付费特权，而是通过自愿捐款来覆盖托管与开发成本。

在兼容性与性能方面，该浏览器端平台借助 WebGPU 实现图形渲染。核心玩法仍保持对原作的忠实还原，同时开发团队还加入了多种来自社区的模组，以提升日常体验，例如宽屏修复、恢复并保留未删减的音轨，以及覆盖国际玩家的多语言本地化。此外，平台还提供云存档功能，可让用户在不同设备之间同步其生涯进度与个人资料。

该项目在原版单人体验之外，还加入了更强的社交元素，包括车队系统、好友列表以及快捷聊天菜单。通过这些功能，玩家能够组织车聚、参与突发的比赛，并在游戏世界中展开交流。用户还能通过细粒度的设置菜单自定义体验，菜单涵盖从 HUD 位置与帧率上限到具体游戏规则等各类选项，例如显示隐藏比赛，以及调整交通与 AI 车手的行为。

对技术细节感兴趣的玩家可以在 GitHub 上获取源代码。项目基于 Talha Müderrisoğlu 的 NFSU2Recomp（一种静态重编译器）构建，而该方案又从其他知名的游戏重编译项目中汲取了思路。通过托管方式保持简洁务实，并借助专门的 Discord 社区提供完善支持，Underground 2 Revived 以协作形式让这段怀旧的游戏历史得以持续运行，也让新老玩家都能再次体验这款作品。 Underground 2 Revived is a fan-led project that brings the classic PC racing game Need for Speed: Underground 2 into modern web browsers. Developed as a passion project by an enthusiast and his brother, the initiative aims to make the title accessible again by recompiling the original game code for platforms like Chrome and Firefox. The project is entirely independent and not affiliated with or endorsed by Electronic Arts, focusing on preservation and community accessibility rather than commercial gain.

The project features a suite of technical enhancements that bring the two-decade-old game up to modern standards. Players can experience the game in widescreen at up to 120 frames per second, with full support for controllers, racing wheels, and touch controls. Beyond simple emulation, the developers have integrated multiplayer functionality, allowing users to engage in free-roam sessions and online races with friends. To maintain a fair environment, the project operates without microtransactions or paid perks, relying instead on voluntary donations to cover hosting and development costs.

Compatibility and performance are central to the user experience, with the browser-based platform leveraging WebGPU for graphics. While the core gameplay remains faithful to the original, the developers have implemented various community-sourced mods to improve quality of life. These include widescreen fixes, restored and uncensored soundtracks, and an array of language localizations that support international players. Additionally, the platform includes cloud saving features, allowing users to synchronize their career progress and profiles across different devices.

The project goes beyond the original single-player experience by introducing social features such as crew systems, friend lists, and a quick-chat menu. These additions enable players to organize car meets, participate in impromptu races, and communicate within the game world. Users can customize their experience through a granular settings menu that governs everything from HUD positioning and frame-rate limits to specific game rules, such as showing hidden races or adjusting the behavior of traffic and AI drivers.

For those interested in technical transparency, the source code is available on GitHub. It builds upon Talha Müderrisoğlu's NFSU2Recomp, a static recompiler, which itself draws inspiration from other notable game recompilation projects. By maintaining a clean, no-frills approach to hosting and providing comprehensive support through a dedicated Discord community, Underground 2 Revived serves as a collaborative effort to keep a nostalgic piece of gaming history functional for new and returning players alike.

Linux 7.3 的发布标志着 Asahi Linux 项目取得了显著进展，其特点是在上游化工作方面实现了实质性进步。包括 macaudio 和 AVD 组件在内的主要驱动程序已提交接受审核。尤其是 AVD 驱动程序，已在创纪录的时间内从下游树转移到审核列表。此外，面向 M4 Pro 、 M4 Max 和 M5 的设备树开发工作也已启动。在策略上，团队正转向对新硬件支持（如 Thunderbolt 和 TI USB-PD 驱动）优先走上游路线，这种做法被证明非常高效。

M3 系列机器的支持已达到普遍可用阶段，这意味着用户不再需要依赖实验性的安装程序标志。针对 M3 实现的一系列改进（例如启用 DCP）带来了更好的续航和性能。这些更新还使功能完善的睡眠模式成为可能，并能充分利用 MacBook 凹槽周围的区域。与此同时，团队正着手解决 M5 系列芯片带来的复杂虚拟化难题：这些芯片使用 Exclaves 和 Tightbeam 协议。通过在 m1n1 下的 EL1 中成功启动多个虚拟机，团队现在能够对 M4 和 M5 系列硬件进行逆向工程。

GPU 开发也在稳步推进，贡献者已成功初始化 M3 GPU 并渲染出三角形。尽管 Apple 独特的设计选择带来了技术门槛，团队仍在努力将这些成果整合进上游的 Mesa 项目。此外，Thunderbolt 支持在中断两年后终于回归。目前能力虽然仅限于 USB3 、 XDomain 和网络，但这次集成仍然是重要的里程碑。团队在处理 Apple 专有控制器架构时遇到了不小的困难——该架构偏离了 Intel 所确立的行业标准，不过他们现已成功合入所需补丁。

通过二进制翻译在 Linux 上实现更顺畅的游戏体验，相关工作取得了切实的改进。团队利用最初为 Android 设计的 Per Process Page Size（P3S）补丁集，成功让各种 x86 应用能够在受限的 muvm 虚拟机环境之外运行，包括更老的游戏作品，如 Fallout: New Vegas 和 Grand Theft Auto IV 。虽然性能已明显提升，但这一领域依然复杂。团队目前正在等待明确 Google 对 P3S 补丁集的长期规划，因为依赖非官方内核修改来实现内存管理并不是适合长期上游维护的策略。

在上述核心功能之外，项目也进一步完善了现有基础设施。用于麦克风阵列的 Triforce beamformer 引入了经过 SIMD 优化的代码，因此获得了显著的性能提升。此外，团队还在上游发布的同时成功推出了 Fedora Asahi Remix 45，这为其发行工作树立了新的标杆。这些逐步累积的改进，让不断增长的用户群体获得更精致、更强大的桌面体验。

最后，项目正在对法律框架做出重大调整。为保护他们的工作成果不被恶意行为者和试图在未做出适当署名的情况下将成果商业化的机构侵害，Asahi Linux 正在将其仓库从宽松的 MIT 许可证迁移到 GNU General Public License version 2 或更高版本。此举体现了项目对 Free and Open Source Software 原则的坚持。通过采用强 copyleft 路线，团队希望避免代码被切分和被剥削，确保他们的努力继续服务于更广泛的生态系统，而不是被专有分支据为己有、封闭起来。 The release of Linux 7.3 marks a period of significant progress for the Asahi Linux project, characterized by substantial advancements in upstreaming efforts. Major drivers, including macaudio and the AVD component, have been submitted for review. The AVD driver, in particular, moved from the downstream tree to the review list in record time. Furthermore, development on M4 Pro, M4 Max, and M5 device trees has begun. In a strategic shift, the team is now favoring an upstream-first approach for new hardware support, such as Thunderbolt and TI USB-PD drivers, which is proving to be a highly effective methodology.

Support for M3 series machines has reached general availability, meaning users no longer need to rely on the experimental installer flag. Improvements to the M3 implementation, such as the enablement of DCP, have resulted in better battery life and performance. These updates also allow for functional sleep modes and the utilization of the area surrounding the MacBook notch. Meanwhile, the team is tackling the complex virtualization challenges posed by M5 series chips, which utilize Exclaves and the Tightbeam protocol. By successfully spawning multiple virtual machines in EL1 under m1n1, the team can now perform reverse engineering on both M4 and M5 series hardware.

GPU development continues to move forward, with contributors successfully initializing the M3 GPU and rendering a triangle. Despite the technical hurdles introduced by Apple's unique design choices, the team is working on integrating these developments into the upstream Mesa project. Additionally, after a two-year hiatus, Thunderbolt support has arrived. While current capabilities are limited to USB3, XDomain, and networking, the integration marks a major milestone. The team encountered significant difficulties navigating Apple's proprietary controller architecture, which deviates from industry standards set by Intel, but they have now successfully incorporated the necessary patches.

The pursuit of seamless gaming on Linux via binary translation has seen meaningful improvements. By leveraging a Per Process Page Size (P3S) patchset originally designed for Android, the team has managed to run various x86 applications, including older titles like Fallout: New Vegas and Grand Theft Auto IV, outside of the restrictive muvm virtual machine environment. While performance is significantly improved, this remains a complex area. The team is currently awaiting clarification on Google's long-term plans for the P3S patchset, as relying on unofficial kernel modifications for memory management is not a sustainable long-term strategy for upstream maintenance.

Beyond these core features, the project has refined its existing infrastructure. The Triforce beamformer, used for microphone arrays, received significant performance optimizations thanks to the introduction of SIMD-optimized code. Additionally, the team successfully launched the Fedora Asahi Remix 45 simultaneously with the upstream release, representing a new benchmark for their distribution efforts. These incremental gains contribute to a more polished and capable desktop experience for the growing user base.

Finally, the project is making a significant change to its legal framework. To protect the integrity of their work against bad-faith actors and entities seeking to monetize their efforts without proper attribution, Asahi Linux is transitioning its repositories from the permissive MIT license to the GNU General Public License version 2 or later. This decision reflects the project's dedication to Free and Open Source Software principles. By adopting a strong copyleft approach, the team aims to prevent the fragmentation and exploitation of their code, ensuring that their efforts continue to serve the broader ecosystem rather than being enclosed by proprietary forks.

提供的输入页面不包含文章正文。它看起来是网站的安全验证界面，阻止我访问实际内容。由于没有可供分析或归纳的材料，我目前无法完成你的请求。请把你希望我总结的文章文本发来，我会很乐意帮你整理。 The provided input page does not contain the text of an article. It appears to be a security verification screen from a website, which prevents access to the actual content. Because there is no source material to analyze or summarize, I am unable to fulfill the request at this time. Please provide the text of an article you would like me to summarize, and I will be happy to assist you.

一家联邦法院已批准 RealPage Inc. Rental Software Antitrust Litigation 系列和解协议，总额近 3.6 亿美元。相关司法进展标志着一场从 2022 年末和 2023 年初启动的集体诉讼进入了关键阶段。该案指控软件提供商以及约 50 家主要公寓业主和运营商使用算法定价工具，非法协调租金上调。

和解的群体包括在 2018 年 10 月至 2025 年 11 月期间，向使用 RealPage 的 Revenue Management Solutions 、 YieldStar 或相关 AI 定价工具的业主或运营商支付租金的美国多户住宅租客。尽管被告同意支付这些数额巨大的款项，但他们并未承认任何不当行为或承担任何责任。根据协议内容，各公司必须停止向 RealPage 提供用于价格建议的非公开数据，并停止以依赖竞争对手非公开数据来影响定价策略的方式使用该软件。

法律专家表示，这些和解协议虽为未来的定价模式提供了明确指引，但未必能完全阻止企业再想出新的方式来实现价格协调。 University of Wisconsin-Madison 的反垄断专家、法学名誉教授 Peter Carstensen 指出，目前仍不清楚这些裁定在长期内究竟能在多大程度上限制相关企业之间的协调与沟通。

推动批准的过程并非一帆风顺。包括 District of Columbia 、 Kentucky 、 Maryland 、 New Jersey 和 Washington 在内的多个司法辖区的总检察长对和解协议提出挑战，认为此类协议可能会非法限制他们自身的独立诉讼。但最终，法官 Waverly Crenshaw 驳回了他们拒绝批准和解的请求，裁定这些释放条款并不会损害各州独立的主权利益。

该案凸显了联邦层面和解与各州主导的反垄断行动之间日益加剧的张力。据 Peter Carstensen 所述，各州对联邦政府处理这类案件的方式表现出更强的质疑，从而出现联邦协议一旦达成后，各州仍继续提起诉讼的情况。

尽管这起重大的联邦集体诉讼告一段落，但关于算法租金设定的法律局势仍未完全尘埃落定。法院确认，驳回仅适用于达成和解的被告，这意味着针对未达成和解的其他当事方的潜在索赔，或未来仍可能提起的诉讼仍然存在。因此，行业仍将持续面临来自其他州及地方诉讼的审查。 A federal court has approved a series of settlements totaling nearly $360 million in the RealPage Inc. Rental Software Antitrust Litigation. These legal developments conclude a significant phase of class-action lawsuits filed in late 2022 and early 2023, which accused the software provider and approximately 50 major apartment owners and operators of using algorithmic pricing tools to illegally coordinate rent hikes.

The settlement class encompasses U.S. multifamily renters who paid rent to owners or operators utilizing RealPage's Revenue Management Solutions, YieldStar, or related AI pricing tools between October 2018 and November 2025. While the defendants have agreed to these substantial payouts, they have not admitted to any wrongdoing or liability. As part of the agreement, the companies must refrain from providing nonpublic data to RealPage for use in price recommendations and stop using the software in ways that rely on competitors' nonpublic data to influence pricing strategies.

Legal experts suggest that while these settlements provide clear guidance on future pricing models, they may not entirely prevent firms from finding new ways to coordinate pricing. Peter Carstensen, an antitrust expert and professor of law emeritus at the University of Wisconsin-Madison, noted that it remains unclear how effectively these decrees will limit coordination and communication among these firms in the long run.

The path to approval was not without conflict. Attorneys general from several jurisdictions, including the District of Columbia, Kentucky, Maryland, New Jersey, and Washington, challenged the settlement agreements, arguing that they might unlawfully restrict their own independent lawsuits. However, Judge Waverly Crenshaw ultimately denied their request to reject the settlements, ruling that the releases do not harm the states' independent sovereign interests.

This case highlights a growing tension between federal settlements and state-led antitrust efforts. According to Carstensen, states have shown increasing skepticism toward the federal government's approach to these types of cases, leading to situations where states continue to pursue litigation even after federal agreements have been finalized.

Despite the conclusion of this major federal class action, the legal landscape surrounding algorithmic rent-setting is far from settled. The court confirmed that the dismissal applies only to the settling defendants, meaning other potential claims against non-settling parties or future actions remain possible. As such, the industry continues to face ongoing scrutiny through various other state and local lawsuits.

libghostty 1.0 的发布标志着终端仿真领域的一项重要里程碑：它建立了一个可移植、高性能的 C 库，为软件开发者提供了稳定的基础。该库从 Ghostty 终端应用程序的核心中提取而来，目前提供稳定的 C ABI，确保项目能够放心进行集成。自推出以来，库已从内部项目发展为生产级工具，服务于数百万用户，为 JetBrains 等主要集成开发环境中的终端提供支持，也覆盖从 Neovim 到 OrbStack 等商业工具的各类平台。

本次 1.0 版本的首要关注点是长期稳定性与兼容性。通过采用不透明句柄设计并进行细致的版本管理，库使开发者能够构建在底层库持续演进过程中仍能保持可用的软件。为确保这一点，项目通过 libabigail 等工具对 CI 进行严格约束，验证每次更新都符合既定的 ABI 稳定承诺。项目的这一做法旨在让各类 Linux 发行版和操作系统上的软件包维护者充满信心，能够将该库作为标准依赖进行发布。

无障碍使用与便于分发是本次发布的核心。项目现已提供完善的构建工具、源代码压缩包以及依赖清单，便于集成到官方操作系统仓库中。此外，库在许可授权方面坚持简洁且合规，尽量让依赖树不包含 copyleft 代码。借助 simdutf 和 Wuffs 等技术，库在保持轻量、模块化以及无需依赖重型标准库即可运行的同时，实现了高性能与图像解码能力。

可移植性同样是其亮点之一：库支持涵盖 Linux 、 macOS 、 Windows 、 Android 和 WebAssembly 在内的大量平台与架构。各类构建会在 CI 环境中通过自动化测试进行验证，以确保在不同硬件上（从强大的服务器到资源受限的嵌入式系统）表现一致。其效率也很突出：库被设计为在尽可能低的内存与计算开销下管理大量终端会话，因此是现代、以终端为中心的应用的灵活选择。

面向未来，项目希望在稳定性与终端协议的持续实现之间取得平衡，让用户既能获得丰富的功能集，又能在必要时切换特定能力。作为一个统一的通用库，它减少了行业中重复开发的需求，并推动形成共享的高质量生态。借助为使用 Go 、 Rust 、 JavaScript 、 Swift 以及更多语言的开发者提供的广泛语言绑定，libghostty 为构建现代、功能丰富的终端界面提供了一条务实、低成本的路径。 The release of libghostty 1.0 marks a significant milestone for terminal emulation, establishing a portable, high-performance C library that provides a stable foundation for software developers. Extracted from the core of the Ghostty terminal application, the library now offers a stable C ABI, ensuring that projects can integrate it with confidence. Since its introduction, the library has moved from an internal project to a production-grade tool used by millions, powering terminals in major integrated development environments like those from JetBrains, as well as various other platforms ranging from Neovim to commercial tools like OrbStack.

A primary focus of this 1.0 release is long-term stability and compatibility. By implementing an opaque handle design and careful versioning, the library allows developers to build software that remains functional even as the underlying library evolves. To maintain this, the project utilizes strict CI enforcement through tools like libabigail to ensure that every update adheres to the promised ABI stability. This rigorous approach is intended to provide confidence to package maintainers across various Linux distributions and operating systems who wish to ship the library as a standard dependency.

Accessibility and ease of distribution are central to the current release. The project now provides comprehensive build tooling, source tarballs, and dependency manifests to facilitate integration into official operating system repositories. Furthermore, the library is designed with a focus on clean, permissible licensing, keeping its dependency tree free of copyleft code. By utilizing technologies like simdutf and Wuffs, the library achieves high performance and image decoding capabilities while remaining lightweight, modular, and capable of operating without reliance on heavy standard libraries.

Portability remains a standout feature, with the library supporting an extensive array of platforms and architectures, including Linux, macOS, Windows, Android, and WebAssembly. These builds are verified through automated testing in CI environments to ensure consistent behavior across different hardware, from powerful servers to resource-constrained embedded systems. Its efficiency is notable, as it is designed to manage high volumes of terminal sessions with minimal memory and compute overhead, making it a versatile choice for modern, terminal-based applications.

Looking ahead, the project aims to balance stability with the ongoing implementation of terminal protocols, ensuring that users have access to a rich feature set while maintaining the ability to toggle specific capabilities. By serving as a unified, generalized library, it reduces the need for redundant development efforts across the industry and promotes a shared, high-quality ecosystem. With extensive language bindings available for developers working in Go, Rust, JavaScript, Swift, and many others, libghostty provides a practical, low-effort path to building modern, feature-rich terminal interfaces.