pwn.college 是一个实践型教育平台，旨在让学习者沉浸在计算的核心领域，从而培养必备的网络安全技能。参与者从 white belts 起步，随后在直接与终端、 CPU 和网络交互的过程中掌握相关概念，并逐步晋级，获得不同颜色的 belts 。该平台由亚利桑那州立大学（Arizona State University）的一支团队维护，并向全球学生和感兴趣的个人免费开放，是 ASU 自身网络安全课程的重要组成部分。

dojos 中提供的挑战主要用于教学目的，并作为全球大学学生评分的依据。正因为其承担了教学功能，创建者希望用户不要在互联网上发布挑战的 walkthrough 、 solution videos 或带有直播尝试的内容。尽管这条规定是自愿性的，遵守它仍能帮助维持平台的持续可用性，也让课程对学生和讲师都保持有效。

如果你有兴趣将该平台纳入自己的课程，pwn.college 允许在标注适当署名的前提下，将其 lectures 和 slides 用于非商业用途并免费使用。教育机构甚至可以选择把这些挑战导入面向学生的 private dojos；团队也欢迎收到有关平台使用情况的通知，以支持他们持续的学术工作。课堂之外，该项目还希望通过举办基于活动的密集培训课程，与各类 conferences 进行更多互动。

当学习者在平台上不断进阶，学习之路仍将通向更高水平的熟练度，最终达到 brown belt 和 black belt 的阶段。团队鼓励毕业生通过 Capture The Flag 竞赛、长期 wargames，以及 CryptoHack 或 OST2 这类替代学习平台来进一步打磨技能。对于想要深入研究的人，创建者在 ASU 提供 internship 机会，并欢迎为支撑其网站的 open-source 基础设施贡献力量。 pwn.college serves as a hands-on educational platform designed to help learners develop essential cybersecurity skills by immersing them in the core of computing. Participants begin their journey as white belts and progress through various levels, earning different colored belts as they master concepts by interacting directly with terminals, CPUs, and networks. The platform is maintained by a team at Arizona State University and is available for free to students and interested individuals worldwide, forming a significant component of ASU's own cybersecurity curriculum.

The challenges provided within the dojos are primarily intended for educational purposes and serve as the basis for grading university students globally. Because of this pedagogical role, the creators ask that users refrain from posting walkthroughs, solution videos, or livestreamed attempts of the challenges to the internet. While this rule is voluntary, following it helps preserve the platform's viability and keeps the curriculum effective for students and instructors alike.

For those interested in integrating the platform into their own coursework, pwn.college allows for the free use of its lectures and slides for non-commercial purposes, provided proper attribution is given. Educational institutions even have the option to import these challenges into private dojos for their students, and the team welcomes notifications regarding the platform's usage to support their ongoing academic work. Beyond the classroom, the project also seeks to engage with conferences by offering intensive, event-based training sessions.

As learners advance beyond the platform, the journey continues toward achieving high-level proficiency, eventually reaching brown belt and black belt status. The team encourages graduates to refine their skills through Capture The Flag competitions, long-term wargames, and alternative learning platforms like CryptoHack or OST2. For those looking to delve deeper into research, the creators offer internship opportunities at ASU and welcome contributions to the open-source infrastructure that powers their site.

Thinking Machines 推出了 Inkling 2 Small，这是其第二代 Mixture-of-Experts 模型，拥有 5940 亿个总参数和 170 亿个激活参数。该模型支持图像输入，并提供可变的计算投入水平。在 agentic coding 、 vision 、 knowledge work 和 reasoning 等任务上，相比前代产品展现出显著的性能提升。目前，该模型可在 Tinker 平台上用于推理与微调；在完成全面的外部安全测试后，计划以 Apache 2.0 许可证发布权重。

该模型在长周期的自主任务中表现尤为突出，例如通过程序化方式生成复杂环境，如城市模拟器或 PC 模拟器。其编码能力也通过在大规模数据集上进行强化学习得到增强，这些数据集引入了 agentic graders，使模型能更好地理解并执行指令，同时还能自主验证自己的输出。除软件开发外，Inkling 2 Small 在自动化研究（autoresearch）方面同样出色，已证明其能够在预测类任务中自主完成训练：通过反复构建评估脚本并持续打磨数据集，让模型逐步改进。

视觉理解是其另一项关键能力，配备了一个 2.92 亿参数的视觉编码器。模型在需要 OCR 、文档与信息图表解析，以及视觉推理的任务上表现非常出色。它能将视觉反馈有效嵌入工作流程，使其能够校验并优化自己生成的代码。此外，在知识密集型领域（包括金融、法律工作和定量分析）中，它也展现出强劲的能力，能够在阅读复杂文档与处理数据时做出接近专家水平的判断。

Inkling 2 Small 的研发以安全与对齐为核心。研究人员采用"纵深防御"的思路，通过同步与异步监控来识别并缓解不当行为。这包括对训练轨迹进行筛选，以抑制持续但不诚实的策略，并使用在策略内（on-policy）的自蒸馏，帮助模型内化更积极的行为准则。尽管在部分进攻性网络安全场景中，该模型仍未达到开放权重的前沿水平，但与既往版本相比，它在对齐基准测试上取得了明显进步。

Inkling 2 Small 的训练过程十分严苛：预训练阶段使用了超过 8 万亿个 tokens，并进行了为期九天的大规模联合强化学习，累计消耗超过 1700 万次 rollouts 。通过使用 NVIDIA GB300 机架以及 NVFP4 浮点格式等高效基础设施，团队能够对性能扩展实现可预测的提升。尽管如此，开发者也承认仍存在一些限制，例如缺少音频支持、交流偶尔偏啰嗦，以及在部分与安全相关的问题上存在过度敏感等情况，这些都将是后续优化的方向。 Thinking Machines has introduced Inkling 2 Small, their second-generation Mixture-of-Experts model featuring 594 billion total parameters and 17 billion active parameters. This model supports both image inputs and variable effort levels, showing significant performance gains over its predecessor across agentic coding, vision, knowledge work, and reasoning tasks. It is currently available for inference and fine-tuning on the Tinker platform, with plans to release the weights under an Apache 2.0 license following comprehensive external safety testing.

The model demonstrates notable strength in long-horizon autonomous tasks, such as procedurally generating complex environments like city simulators or PC emulators. Its coding capabilities have been enhanced through reinforcement learning on large-scale datasets that utilize agentic graders, allowing the model to better follow instructions and verify its work autonomously. In addition to software development, Inkling 2 Small excels at autoresearch, having proven its ability to autonomously train models on predictive tasks by iteratively building evaluation scripts and refining datasets.

Visual interpretation is another core capability, supported by a 292-million-parameter vision encoder. The model performs exceptionally well on tasks requiring OCR, document and infographic interpretation, and visual reasoning. It effectively integrates visual feedback into its workflows, enabling it to validate and refine its own generated code. The model also shows strong performance in knowledge-heavy domains, including finance, legal work, and quantitative analysis, where it demonstrates expert-level judgment when navigating complex documents and data.

Safety and alignment were central to the development of Inkling 2 Small. Researchers utilized a defense-in-depth approach, implementing both synchronous and asynchronous monitoring to detect and mitigate misbehavior. This included filtering training traces to discourage persistent but dishonest strategies and employing on-policy self-distillation to help the model internalize positive conduct principles. While the model remains behind the open-weight frontier in some offensive cyber scenarios, it shows a significant improvement in alignment benchmarks compared to previous versions.

The training process for Inkling 2 Small was rigorous, involving over 8 trillion tokens in the pre-training phase and a massive joint reinforcement learning run that consumed more than 17 million rollouts over nine days. By leveraging efficient infrastructures like NVIDIA GB300 racks and NVFP4 floating-point formats, the team was able to scale performance predictably. Despite its advancements, the developers acknowledge certain limitations, such as a lack of audio support, occasional verbosity in communication, and over-sensitivity in some safety-related queries, which remain areas for future refinement.

马克·库班正在公开批评 Meta，指控这家科技巨头明知仍在其平台上放任欺诈性的深度伪造广告传播。这位亿万富翁投资人是低价药品公司 Cost Plus Drugs 的联合创始人，经常听到老年人反馈：他们被由人工智能生成的广告误导，广告中虚假使用他的肖像来推销不存在的医疗"疗法"。库班认为，这些骗局绝非单纯的偶然疏漏，而是反映了 Meta 更广泛的内部文化——在用户安全之外，把广告收入放在首位。

相关说法也有报道支撑，报道援引了 Meta 的内部文件。文件显示，公司预计仅在 2024 年，围绕骗局和违禁产品的广告就能带来约 160 亿美元的收入。文件还称，Meta 的内部团队每天识别出约 150 亿条"高风险广告"，其中部分欺诈内容的年度价值大约为 70 亿美元。此外，文件指出，Meta 的自动化审核系统往往要求达到 95% 的欺诈概率才会采取处置措施，这就给出价更高的潜在诈骗者留下了继续在平台上出现的空间。

Meta 对这些指控予以否认。公司发言人表示，这类文件对其反欺诈工作呈现出片面且失真的画面。公司同时强调，用户举报的诈骗广告在过去 18 个月里明显减少，并且在 2025 年下架了数百万条诈骗广告。另外，公司目前还因与金融欺诈相关的广告做法，正受到美国证券交易委员会（SEC）的审查。

这些骗局对老年人造成的伤害尤其严重：他们在数字诈骗方面的经济损失，超过其他任何年龄段。根据美国联邦调查局 FBI 发布的 2025 年《互联网犯罪报告》，60 岁以上人群上报的损失超过 77 亿美元，较上一年大幅上升。由人工智能驱动的骗局构成了其中 3100 多起，造成的总体损失达到数亿美元。专家建议，消费者应对包含名人形象的广告保持警惕，通过第三方渠道核实产品信息，并将可疑行为报告给 Internet Crime Complaint Center 等相关部门。 Mark Cuban is speaking out against Meta, alleging that the tech giant knowingly permits the spread of fraudulent deepfake advertisements on its platforms. The billionaire investor, who co-founded the low-cost pharmacy company Cost Plus Drugs, regularly hears from elderly individuals who have been misled by AI-generated ads that falsely feature his likeness to sell nonexistent medical cures. Cuban contends that these scams are not merely accidental oversights but reflect a broader organizational culture at Meta that prioritizes ad revenue over user protection.

The criticism is supported by reports involving internal Meta documents, which suggest that the company anticipated generating approximately 16 billion dollars in revenue from ads associated with scams and prohibited products in 2024 alone. These documents indicate that Meta's own internal teams identified around 15 billion daily high-risk ads, with some categories of fraudulent content valued at roughly 7 billion dollars annually. Furthermore, the documents note that Meta's automated systems often require a 95 percent probability of fraud before taking action, allowing potential scammers who pay higher rates to maintain their presence on the platform.

Meta has contested these claims, with a company spokesperson arguing that such documents provide a selective and distorted view of their anti-fraud efforts. The company highlighted that user reports of scam advertisements have decreased significantly over the past 18 months and that millions of scam ads were removed in 2025. Additionally, the company is currently facing scrutiny from the Securities and Exchange Commission regarding advertising practices tied to financial fraud.

The impact of these scams is particularly severe for older adults, who suffer greater financial losses to digital fraud than any other age demographic. According to the FBI's 2025 Internet Crime Report, individuals over the age of 60 reported over 7.7 billion dollars in losses, a sharp increase from the previous year. AI-driven scams accounted for more than 3,100 of these complaints, representing hundreds of millions of dollars in total damages. Experts advise that consumers should remain skeptical of advertisements featuring celebrities, verify products through third-party searches, and report suspicious activity to authorities such as the Internet Crime Complaint Center.

有人就 Claude Code 工具提出了安全方面的担忧。有用户反馈，这款软件会在无意中向第三方网站泄露个人电子邮件地址。问题出现在 AI 代理进行网页搜索时：它把用户的电子邮件写入了 User-Agent 请求头。最初的报告者表示，这一行为导致垃圾邮件激增，让人愤怒，也让用户对开发者在用户隐私方面的做法失去了信心。

Anthropic 的代表澄清称，这并非刻意的功能，而是模型与网页交互方式带来的非预期结果。具体而言，像 Wikipedia 这类网站制定了相关规则，要求自动化客户端在请求头中使用联系电子邮件来表明身份。 AI 在研究任务中试图遵循这些约定时，会在未获得明确授权的情况下自动传播用户的电子邮件。

针对上述反馈，开发团队推出了一个内置模块 cc-plugin-privacy 。该更新从 2.1.317 及更高版本开始默认启用，通过确定性校验在发起网络请求时阻止电子邮件地址被传输。此后，团队已将该隐私模块的源代码公开，鼓励用户自行查看实现细节，并对其有效性提出反馈。

尽管修复已经发布，社区成员仍对开发流程不满。有用户指出，这个问题早在数月前就已被发现，并认为公司应优先梳理和处理积压的大量已报告漏洞，而不是只对引发关注的投诉作出回应。也有人提出临时的手动变通方案，例如为项目配置专属的 CLAUDE.md 文件，以严格约束 User-Agent 信息的处理方式，以及是否、以及在何种情况下可以披露电子邮件地址。 A security concern was raised regarding the Claude Code tool, where users reported that the software was inadvertently leaking personal email addresses to third-party websites. This occurred when the AI agent conducted web searches and included the user's email in the User-Agent request header. The original reporter noted that this behavior led to an influx of spam, prompting frustration and a loss of confidence in the developer's approach to user privacy.

Anthropic representatives clarified that this was not a deliberate feature but rather an unintended consequence of how the model interacts with the web. Specifically, some websites, such as Wikipedia, have established policies requiring automated clients to identify themselves with a contact email in the header. The AI, in attempting to follow these conventions during research tasks, automatically propagated the user's email without explicit authorization.

In response to these reports, the development team implemented a built-in mod called cc-plugin-privacy. This update, which is enabled by default in version 2.1.317 and later, uses deterministic checks to block the transmission of email addresses during web requests. The team has since made the source code for this privacy mod available to the public, encouraging users to inspect the implementation and provide feedback on its effectiveness.

Despite the release of the fix, community members expressed ongoing dissatisfaction with the development process. Some users pointed out that the issue had been identified months earlier and suggested that the company should prioritize triaging the massive backlog of reported bugs rather than only responding to high-profile complaints. Others proposed temporary manual workarounds, such as configuring project-specific CLAUDE.md files to strictly govern how User-Agent information is handled and when (or if) an email address may be disclosed.

CVE-2026-83844 指出在 Microsoft Paint 中发现了一个重大安全漏洞，具体为越界读取缺陷。该问题可能使未经授权的攻击者通过网络远程执行恶意代码。根据通用漏洞评分系统（Common Vulnerability Scoring System），该漏洞被评定为高严重性，分数为 8.1，表明对受影响系统存在严重风险。

该漏洞源于一个越界读取错误，归类为 CWE-125 。该技术缺陷可被利用来实现远程代码执行，这是潜在攻击者的重要能力。由于漏洞等级为高严重性，用户和管理员需要尽快采取措施，降低被利用风险。

就受影响软件范围而言，该漏洞影响从 11.0.0.0 起的 Microsoft Paint 版本，并存在于所有 11.2605.91.0 之前的版本中。 Microsoft 已正式确认该问题，并发布了供应商公告及相应补丁以修复该安全隐患。

为保障系统完整性，建议使用受影响版本的 Windows Paint 用户查阅 Microsoft 官方安全更新指南。通过应用推荐的补丁，用户可以有效修补安全缺口，保护系统免受该特定远程代码执行威胁。 CVE-2026-83844 identifies a significant security vulnerability discovered within Microsoft Paint, specifically involving an out-of-bounds read flaw. This issue allows an unauthorized attacker to potentially execute malicious code remotely over a network. The vulnerability has been classified with a high severity score of 8.1 according to the Common Vulnerability Scoring System, reflecting the serious risks posed to affected systems.

The vulnerability stems from an out-of-bounds read error, which is cataloged under CWE-125. This technical flaw can be exploited to achieve remote code execution, a critical capability for potential attackers. Because the vulnerability is categorized as high severity, it necessitates prompt attention from users and administrators to mitigate the risk of exploitation.

Regarding the scope of the affected software, the vulnerability impacts versions of Microsoft Paint starting from 11.0.0.0 and exists in all versions prior to 11.2605.91.0. Microsoft has officially acknowledged the issue and provided a vendor advisory along with the necessary patches to address the security concern.

To maintain system integrity, those using affected versions of Windows Paint are encouraged to consult the official Microsoft security update guide. By applying the recommended patches, users can effectively close the security gap and protect their systems from this specific remote code execution threat.

Kagi 推出了一项重大调整：任何人都可以无需创建账户直接搜索网络。现在，访问者可先免费进行 50 次搜索，完成注册后还能额外获得 100 次搜索次数。此次更新旨在推动口碑传播，并为用户（尤其是教育领域用户）提供更明确的替代选择，让他们能够避开 Google 等大型竞争对手日益 AI 化、且广告驱动愈发浓厚的环境。 Kagi Search 的 Android 应用也同步更新了，主要包括重新设计的首页、改进的语音搜索以及更便捷的小组件入口。

Kagi Assistant 也经历了大幅优化，重点提升易用性与可靠性。近期更新带来了更顺畅的语音输入、更清晰的临时线程与永久线程视觉标识，并支持将图片或文件直接粘贴进提示词中。此外，移动端用户现在可以使用官方的 Kagi Assistant iOS 和 Android 应用，确保不同设备上的使用体验一致。服务还推出了更强大的研究代理：用户可在 Quick 和 Research 两种模式之间进行选择，在速度与分析深度之间取得平衡。

平台管理已统一到 account.kagi.com 的单一账户体系下，该体系也新增了对 passkeys 的支持，以提升安全性并简化登录流程。除上述结构性调整外，公司还通过一系列新举措扩展生态，例如推出 Kagi News，提供精心策划的每日媒体综述，并为 Kagi Translate 引入免费增值模式。相关服务的目标是在控制运营成本的同时，维持高质量、无广告的信息获取体验。

此外，其他值得关注的产品进展还包括 Kagi Maps 的更新：移动端新增测量工具，并为热门城市区域增加数据图层。公司还推出 Kagi Specials，这是一个合作伙伴计划，为其他同样注重隐私的服务提供折扣。在这些发布过程中，公司始终坚持透明与用户可控原则，并通过社区论坛持续征集反馈，以指导后续开发与修复问题。 Kagi has introduced a significant shift by allowing anyone to search the web without creating an account. Visitors can now perform 50 searches for free, with an additional 100 available after registering. This update is designed to facilitate word-of-mouth growth and provides a clear alternative for users, particularly in education, who wish to avoid the increasingly AI-heavy and advertising-driven environments of major competitors like Google. The Kagi Search Android app has also been updated with a redesigned home screen, improved voice search, and better access to widgets.

The Kagi Assistant has undergone substantial refinements to improve both usability and reliability. Recent updates include smoother voice input, clearer visual indicators for temporary versus permanent threads, and the ability to paste images or files directly into prompts. Additionally, mobile users now have access to official Kagi Assistant apps for iOS and Android, ensuring a consistent experience across devices. The service has also introduced a more powerful research agent, allowing users to choose between Quick and Research modes to balance speed with analytical depth.

Platform management has been unified under a single account system at account.kagi.com, which now includes support for passkeys to improve security and login convenience. Alongside these structural changes, the company has expanded its ecosystem through new initiatives like Kagi News, which provides a curated daily press review, and a freemium model for Kagi Translate. These services aim to maintain high-quality, ad-free information access while managing operational costs.

Other noteworthy product developments include updates to Kagi Maps, which now features a measuring tool on mobile and data layers for popular urban areas, and the introduction of Kagi Specials, a partnership program offering discounts on other privacy-focused services. Throughout these releases, the company has maintained its commitment to transparency and user control, frequently soliciting feedback through its community forums to guide ongoing development and bug fixes.

运营三十一年后，Salon 正式停止发布。该机构回顾了自己的历程，认为它从来不只是一个网站，更是在互联网萌芽时期对新闻业可能走向何方所进行的一次雄心勃勃的实验。 Salon 于 1995 年 11 月上线，力求形成独特的声音：把对世界的真切好奇与这样一种信念融为一体——数字平台能够在记者与受众之间建立新的、有意义的联系。

在三十多年的发展中，Salon 开辟出一块独特的空间，让政治、文化与批判性的思考在此交汇。该出版物聚焦于审视权力结构，追问根深蒂固的既定观念，并呈现那些往往没有进入主流视野的声音与故事。编辑团队并非只为追逐突发新闻而存在，而是希望解释日常生活表层之下正在变化的复杂力量。

该机构向多年来支撑这份出版物的无数个人致以由衷感谢，包括奠定其气质的创始人、员工与撰稿人。其工作的遗产被视为属于这一群创造者：正是他们的努力，为数字新闻奠定了基础，也将长期影响行业的发展。

团队也向读者送上最后的谢意。正是读者的积极参与、直面质疑的讨论与一贯的支持，使这个网站从单纯的数字出版物成长为真正的社区。读者愿意分享并展开辩论的态度，验证了该媒体的核心使命，也让编辑团队在长期工作中始终拥有清晰的目标。

最终，关闭的决定源自独立数字新闻业不断变化的商业现实。该出版物指出，在当下环境中，其现有商业模式已无法持续。团队虽承认这次停刊令人难过，但也强调他们最大的成就，是向世人展示数字叙事的潜力，并守住了一个让观点真正具有分量的空间。 After thirty-one years of operation, Salon is officially ceasing publication. The organization reflects on its journey, noting that it was never merely a website but rather an ambitious experiment in what journalism could become during the internet's infancy. Launching in November 1995, the outlet sought to establish a unique voice, blending a genuine curiosity about the world with the conviction that digital platforms could forge new, meaningful connections between reporters and their audiences.

Throughout its three-decade history, Salon carved out a distinct space where politics, culture, and critical inquiry intersected. The publication focused on examining power structures, questioning prevailing assumptions, and highlighting voices and stories that frequently escaped mainstream notice. Rather than simply reporting on breaking news, the editorial team aimed to make sense of the complex forces shifting beneath the surface of everyday life.

The organization expresses deep gratitude to the countless individuals who sustained the publication over the years, including the founders, staff, and contributors who shaped its character. The legacy of the work is described as belonging to this community of creators, whose efforts established a foundation for digital journalism that would influence the industry for years to come.

The team also extends a final thank you to their readers, whose active engagement, challenging discourse, and consistent support transformed the site from a digital publication into a true community. The readers' willingness to share and debate ideas helped validate the outlet's core mission, providing purpose to the editorial team's work over the long run.

Ultimately, the decision to close was driven by the changing business realities of independent digital journalism. The publication concludes that its existing business model is no longer sustainable in the current landscape. While acknowledging the sadness of this closure, the team emphasizes that their greatest achievement was demonstrating the potential of digital storytelling and maintaining a space where ideas truly mattered.

搭建 Hackintosh 的做法，即在非 Apple 硬件上运行 macOS，源远流长，最早可追溯到 20 世纪 80 年代末。大约十年前，早期的先驱者借助 KVM 和 PCI passthrough 技术，在 x86 主机系统上实现了接近原生的 macOS 性能。当时这可谓开天辟地，让爱好者能够绕过硬件限制，并充分发挥强大的 GPU 加速能力。

2021 年，Apple 转向基于 ARM 的 silicon 后，社区里不少人担心这会是"终点"。 M 系列芯片带来了明显的效率与功耗优势，但与此同时也意味着 x86 支持终将走向被弃用。随着近期 macOS 27 的发布，该系统已正式不再兼容 x86 架构，社区仿佛要就此淡出视野，把自定义 Apple 硬件的日子交给过去。

然而，出乎意料的是，事情并没有走向衰落。推动这次"复兴"的主要动力来自软件开发中 AI 的深度融入。通过使用诸如 Claude Code 和 OpenAI Codex 这类 AI 工具，爱好者们找到了新的路径来弥合"不兼容硬件"与"现代软件需求"之间的鸿沟。一次近期实验显示：用 ThinkPad 运行 macOS 26 并用于 iOS 开发的过程证明，AI 辅助编码能够在创纪录的时间内解决复杂的集成问题，等于把过去需要大量手工、费时费力的工作有效自动化。

AI 驱动的这种变化让社区焕发了新生机，并掀起了爆发式的开发热潮——甚至许多旁观者在五年前都难以想象。如今的项目不断突破过去看似不可能的边界，包括在 Snapdragon 笔记本上运行 macOS 27，甚至把操作系统移植到基于 ARM 的单板计算机上（例如 RK3588）。更令人意外的是，iOS 17 竟然还能在诸如 Nintendo Switch OLED 这类非传统硬件上运行。

在这些大胆移植之外，社区在硬件适配与"翻译"方面也取得了显著进展。开发者正在探索实现加速视频解码的办法，让非标准硬件具备原生 AirDrop 功能，甚至研究如何在 x86 平台上运行 macOS 27 的翻译版本。借助 AI 的强大能力，用户正在重新夺回对自身硬件的掌控力，也让"在定制系统上运行 macOS"这一愿景不仅仍然活着，甚至可以说比以往任何时候都更有实现可能。 The practice of building a Hackintosh, which involves running macOS on non-Apple hardware, has a rich history reaching back to the late 1980s. Nearly a decade ago, early pioneers utilized KVM and PCI passthrough technologies to achieve near-native performance for macOS on x86 host systems. At the time, this was a breakthrough that allowed enthusiasts to bypass hardware restrictions and leverage powerful GPU acceleration.

When Apple transitioned to ARM-based silicon in 2021, many in the community feared it marked the end of the road. The shift to M-series chips offered significant efficiency and power advantages, yet it also signaled that x86 support would eventually be deprecated. With the recent release of macOS 27, which officially dropped compatibility for x86 architectures, the community seemed poised to fade into obscurity, leaving the days of custom Apple hardware builds behind.

However, a surprising resurgence has occurred, driven primarily by the integration of AI in software development. By using AI tools like Claude Code and OpenAI Codex, enthusiasts have found new ways to bridge the gap between incompatible hardware and modern software requirements. A recent experiment using a ThinkPad to run macOS 26 for iOS development demonstrated that AI-assisted coding could resolve complex integration issues in record time, effectively automating tasks that once required manual, labor-intensive coding.

This AI-driven transformation has breathed new life into the community, sparking an explosion of development activity that most observers could not have predicted just five years ago. Modern projects are pushing boundaries that previously seemed impossible, including running macOS 27 on Snapdragon laptops and porting the operating system to ARM-based single-board computers like the RK3588. Even more unexpected is the capability to run iOS 17 on unconventional hardware such as the Nintendo Switch OLED.

Beyond these ambitious ports, the community is making significant strides in hardware support and translation. Developers are finding ways to enable accelerated video decoding, native AirDrop functionality on non-standard hardware, and even methods for running translated versions of macOS 27 on x86 platforms. By leveraging the sheer power of AI, users are regaining control over their hardware, proving that the dream of running macOS on customized systems is not only alive but arguably more capable than ever before.

Window Sweaters 是一款 macOS 菜单栏应用，旨在通过把桌面窗口包裹在可定制的针织边框里，营造舒适、个性化的观感。该应用以 Felix Kratz 的 JankyBorders 为基础衍生，借助 macOS 的私有窗口 API 来跟踪当前处于活动状态的窗口，并应用多种图案与配色方案，从而让计算环境更有触感、视觉上也更与众不同。

用户可以通过菜单栏里的 yarn 图标与应用互动，在其中调整图案类型、边框厚度和针脚大小等设置。它的一大亮点在于为常用应用内置了 37 种手工针织配色，例如 Safari 、 Slack 、 VS Code 和 Figma 。对于没有预设 sweater 的应用，软件会从图标中动态采样颜色来生成匹配的观感，确保即使是未被识别的程序，也能保持整体统一的风格。

应用提供两种主要模式：By App 会为特定程序应用专属、量身定制的 sweater；Zigzag 则会用从应用图标颜色提取出来的乳白色锯齿图案，呈现更统一、柔和的外观。若想进一步掌控细节，用户可以编辑本地配置文件来定义自定义配色规则，或替换自带的图片文件作为边框纹理。系统也很灵活，允许用户对单个应用或整个工作流程随时关闭该效果。

由于该项目依赖私有 API，开发者说明目前仍在开发中，因而存在一些固有限制，例如在调整窗口大小时边框可能会暂时消失。它兼容 macOS 13 及更高版本，并可在 Apple Silicon 与 Intel 设备上运行。尽管该应用目前尚未获得 Apple 的官方公证，但项目提供了清晰的手动安装与配置说明，并另有 Homebrew 方案以便更轻松地部署。

对于希望参与贡献或扩展项目的人，源代码以 GPL-3.0 许可证发布。仓库中包含本地编译所需的构建脚本，以及自动化测试工具。无论是通过调整配置文件，还是修改底层代码，用户都可以按自己的偏好定制体验，让 Window Sweaters 成为一个持续演进的桌面个性化工具。 Window Sweaters is a macOS menu-bar application designed to provide a cozy, personalized aesthetic by wrapping desktop windows in customizable knitted borders. Built as a derivative of Felix Kratz's JankyBorders, the app leverages private macOS window APIs to track active windows and apply various patterns and colorways, aiming to make the computing environment feel more tactile and visually distinct.

Users can interact with the app via a yarn icon in the menu bar, which allows them to adjust settings like pattern types, border thickness, and stitch size. A standout feature is the inclusion of 37 hand-knitted colorways for popular applications, such as Safari, Slack, VS Code, and Figma. For applications without a pre-defined sweater, the software dynamically samples icon colors to generate a matching aesthetic, ensuring that even unrecognized programs maintain a cohesive look.

The application offers two primary modes: By App, which applies unique, tailored sweaters to specific programs, and Zigzag, which provides a uniform, softer look by utilizing a cream-colored zigzag pattern derived from the app's icon colors. If a user wishes to exert more control, they can edit a local configuration file to define custom color rules or swap in their own image files to serve as border textures. The system is designed to be flexible, allowing users to toggle the effect off for individual apps or entire workflows.

As a project that relies on private APIs, the developer notes that it is currently a work in progress with some inherent limitations, such as borders temporarily disappearing during window resizing. It is compatible with macOS 13 and later, running on both Apple Silicon and Intel hardware. While the app is not yet officially notarized by Apple, the project provides clear instructions for manual installation and configuration, as well as a Homebrew option for easier deployment.

For those interested in contributing or extending the project, the source code is available under the GPL-3.0 license. The repository includes build scripts for local compilation and automated testing tools. Whether through adjusting the configuration files or modifying the underlying code, users are encouraged to customize their experience, making Window Sweaters an evolving tool for desktop personalization.

motorola signature 27 作为该公司最先进的旗舰智能手机问世，融合了高端设计与尖端性能。它以一套被 DXOMARK 认可为 2026 年发布的世界最佳摄像头系统脱颖而出，提供如采用 LOFIC 技术的 50MP Sony LYTIA 主传感器，以及支持 100x Super Zoom 的 200MP 潜望式长焦摄像头等功能。此外，这款设备还配备了 EVS 动作传感器，专门用于对快速移动的主体进行精准捕捉。

视觉体验是该设备的重点之一，搭载 Extreme AMOLED Duo 双串联显示屏，峰值亮度可达 12,000 nits，使其成为目前智能手机中亮度最高的屏幕。该显示技术不仅提升了户外场景下的清晰度，还在能效方面较前几代提高了 34% 。设备由 Snapdragon 8 Elite Extreme Gen 6 平台驱动，并由大面积 ArcticMesh 系统散热，确保在游戏或 AI 处理等高负载任务中保持稳定性能。

人工智能以 Gemini Intelligence 深度融入日常使用。设备支持多项进阶能力，包括多步任务自动化、 Rambler 工具用于将自然语音精炼成书面文本，以及生成个性化主屏幕小部件的功能。此外，它还是首款集成 Motorola Qira 的手机，这是一套旨在统一 Motorola 与 Lenovo 更广泛生态体验的 AI 系统，同时也支持通过前置摄像头进行实时 American Sign Language 翻译等定制化功能。

音频表现同样出色，motorola signature 27 也是市面上唯一配备 Audio by Bang & Olufsen 的智能手机。此次合作带来了定制调校的扬声器架构，以及为维持音频保真而设计的专属音效模式。硬件采用耐用框架承载，使用陶瓷与玻璃结构，符合军用级 MIL-STD-810H 标准，并具备 IP68 与 IP69 防尘防水等级。

续航与安全方面也同样到位。设备内置 7000mAh 电池，支持 90W 有线充电和 50W 无线充电，电量补充速度非常高效，最快可在不到 15 分钟内完成显著补电。更值得高强度用户关注的是，motorola signature 27 是首款支持 GrapheneOS 的 Motorola 设备，这是一款以隐私为核心的开源操作系统。通过将高端硬件与七年 OS 及安全更新承诺相结合，公司希望为既追求创新又重视数据完整性的用户提供可靠、长期的解决方案。 The motorola signature 27 arrives as the company's most advanced flagship smartphone, blending high-end design with cutting-edge performance. It distinguishes itself with a camera system recognized by DXOMARK as the world's best launched in 2026, offering features like a 50MP Sony LYTIA main sensor with LOFIC technology and a 200MP periscope telephoto camera capable of 100x Super Zoom. Additionally, the device introduces an EVS action sensor specifically designed to capture fast-moving subjects with precision.

Visuals are a primary focus, supported by an Extreme AMOLED Duo tandem display that reaches a peak brightness of 12,000 nits, making it the brightest screen currently available in a smartphone. This display technology not only improves clarity in outdoor settings but also enhances energy efficiency by 34% compared to previous generations. The device is powered by the Snapdragon 8 Elite Extreme Gen 6 platform, cooled by a large-surface ArcticMesh system to ensure consistent performance during intensive tasks like gaming or AI processing.

Artificial intelligence is deeply integrated into the user experience through Gemini Intelligence. The device supports advanced capabilities, including multi-step task automation, the Rambler tool for refining natural speech into written text, and the ability to generate personalized home screen widgets. Furthermore, the phone is the first to include Motorola Qira, an AI system meant to unify experiences across the broader Motorola and Lenovo ecosystem, while also supporting specialized features like real-time American Sign Language translation via the front camera.

Audio performance is another key differentiator, as the motorola signature 27 is the only smartphone on the market to feature Audio by Bang & Olufsen. This partnership includes custom-tuned speaker architecture and specialized sound modes designed to maintain audio fidelity. The hardware is housed in a durable frame, featuring a ceramic and glass construction that meets military-grade MIL-STD-810H standards and holds IP68 and IP69 ratings for dust and water resistance.

Battery life and security round out the device's premium offerings. It houses a 7000mAh battery that supports 90W wired and 50W wireless charging, allowing for significant power replenishment in under fifteen minutes. Perhaps most notably for power users, the motorola signature 27 is the first Motorola device to support GrapheneOS, a privacy-focused, open-source operating system. By combining high-end hardware with a seven-year commitment to OS and security updates, the company aims to provide a reliable, long-term solution for those seeking both innovation and data integrity.