Hacker News Daily — 2026-03-31 Trending
Scraped at 21:43, March 31, 2026 (PDT)
(1) We intercepted the White House app's network traffic
The analysis maps the app's network behavior, showing data flows, endpoints, and potential privacy implications for a high-profile government app. It highlights how mobile apps communicate with backends and third parties in practice.
分析揭示了该白宫应用的网络行为，展示数据流动、接触的端点以及潜在的隐私影响，聚焦政府应用在实践中的后端通信与第三方接入。
(2) Claude Code's source code has been leaked via a map file in their NPM registry
Claude Code’s source code was leaked through a source map file published in their NPM package registry. The incident underscores how publishing artifacts like source maps can unintentionally expose proprietary code and sensitive logic. It’s a reminder for maintainers to scrub sensitive bits from build artifacts and implement stricter registry hygiene and access controls.
Claude Code 的源代码通过其 NPM 包注册表中暴露的一个 source map 文件泄露到公网。事件凸显了在发布构建产物时，source map 可能无意间暴露私有代码和业务逻辑的风险。开发团队需要对发布的产物进行清理，并加强注册表的访问控制与审计。
(3) The Claude Code Source Leak: fake tools, frustration regexes, undercover mode
An analysis of a Claude code-source leak reveals decoy tools and frustrating regex tests, with mentions of an undercover mode. The incident highlights the fragility of model provenance and the risk of leaked tooling being manipulated to mislead researchers. The takeaway is a call for stronger provenance controls and safer, auditable tooling around AI code-generation.
对 Claude 代码源泄露事件的分析显示存在伪工具与令人挫败的正则测试，并提及潜伏模式。此事凸显了模型来源的脆弱性，以及泄露工具被操纵以误导研究者的风险。核心启示是需要加强来源追溯与对 AI 代码生成工具的安全、可审计设计。
(4) OpenAI closes funding round at an $852B valuation
The funding round valued the company at $852 billion, signaling strong investor confidence in AI momentum. The result could influence competitors and the broader pace of AI capital and governance considerations.
此次融资使 OpenAI 的估值达到 8,520 亿美元，显示市场对 AI 领域的强烈投资热情。此轮结果可能影响竞争格局、治理预期及日益资本化的 AI 研发节奏。
Oracle announced cutting about 30,000 jobs, reportedly via an early 6am email, underscoring ongoing belt-tightening in tech as AI shifts reframe headcounts. The move highlights how corporate communications and workforce restructuring are evolving in large software firms.
Oracle宣布裁撤约3万名员工，传以6点的深夜/凌晨邮件通知，折射出在 AI 投资驱动的环境下科技公司持续精简人力。此举凸显大型软件企业在组织沟通和人力结构调整方面的新常态。
A historical uptime review tracks GitHub’s uptime history, highlighting stretches of uninterrupted service and improvements in reliability. For developers, this matters for confidence in CI/CD and collaboration workflows.
回顾 GitHub 的上线时间历史，突出长期无故障段落及可靠性提升。对依赖其工作流的开发者而言，这关乎 CI/CD 与协作的信任度。
(7) Axios compromised on NPM – Malicious versions drop remote access trojan
Malicious Axios releases popped up on npm, embedding a remote access Trojan in certain versions. The attack highlights the ongoing supply-chain risk for widely used libraries, and operators should audit dependencies, pin safe versions, and monitor for unusual network activity after updates.
npm 上出现恶意的 Axios 版本，包含可对受影响环境进行远程访问的木马。此类供应链攻击再次暴露了对广泛使用库的依赖风险，建议锁定版本、进行依赖审计，并在更新后监测异常网络活动。
(8) OkCupid gave 3M dating-app photos to facial recognition firm, FTC says
The FTC says OkCupid/Match shared roughly 3 million dating-app photos with a facial-recognition firm, raising concerns about consent and biometric data usage. The case underscores ongoing regulatory scrutiny of data sharing practices in dating apps.
FTC 指出 OkCupid/Match 将约300万张约会应用照片分享给一个面部识别公司，引发关于同意与生物识别数据使用的担忧。此案凸显对约会应用数据共享行为的持续监管审视。
(9) A dot a day keeps the clutter away
A dot-a-day system is proposed to combat digital clutter by focusing on small, repeatable actions. The approach emphasizes low-friction habits that cumulatively improve organization.
提出每天一个点的管理体系以对抗数字杂乱，强调通过小而持续的行动来提升组织与专注度。低摩擦的习惯可以逐步带来长期效果。
(10) I built a 516-panel financial terminal in 3 weeks using AI
A solo developer built a 516-panel financial terminal in three weeks, leveraging AI to orchestrate data, layout, and alerts across a sprawling dashboard. The project demonstrates how AI-assisted tooling can accelerate heavy UX integrations, while also underscoring hardware, data-ops, and latency challenges in real-time finance use cases.
一个人用 AI 将一个516块屏幕的大型金融终端整合在一起，涉及数据输入、布局和告警的自动化。该项目展示了 AI 辅助工具在实现复杂仪表板方面的潜力，同时也暴露了实时金融工作负载下的硬件、数据处理和延迟挑战。
(11) Microsoft: Copilot is for entertainment purposes only
Microsoft explicitly states in its terms that Copilot is for entertainment purposes only, clarifying its intended use and limitations. This stance raises questions about relying on AI for critical coding or decision-making and highlights the need for clear risk disclosures.
微软在使用条款中明确表示 Copilot 仅用于娱乐用途，界定了其使用范围与局限。此说法引发了人们对将其用于关键编码或决策场景的可依赖性以及需披露的风险信息的关注。
(12) Show HN: 1-Bit Bonsai, the First Commercially Viable 1-Bit LLMs
.png)
1-Bit Bonsai demonstrates commercially viable 1-bit quantized LLMs, achieving usable performance with dramatically reduced memory and compute. The result challenges the assumption that high-precision weights are mandatory for practical inference, broadening on-device and edge-AI prospects.
1-Bit Bonsai 展示了商业可用的1位量化 LLM，在显著降低内存与计算需求的同时仍保持可用性。这一成就挑战了高精度权重对实用推理的必要性，拓展了设备端和边缘 AI 的前景。
(13) Ministack (Replacement for LocalStack)
Ministack positions itself as a drop-in replacement for LocalStack, enabling local cloud service emulation for development and testing. It aims to simplify setup and improve iteration speed for cloud-native apps.
Ministack 被定位为 LocalStack 的替代方案，方便在本地模拟云服务以便开发与测试。旨在简化设置、提升云原生应用的迭代速度。
(14) Neanderthals survived on a knife's edge for 350k years
New findings suggest Neanderthals endured dramatic climate swings and fluctuating resources for 350k years, surviving by flexible subsistence strategies and regional refugia. The study reframes their extinction as a long, precarious persistence rather than a single collapse, with implications for how we model past human-environment resilience.
新的发现表明尼安德特人曾在长达35万年的气候波动和资源变化中存活，靠灵活的生存策略和地域庇护地维持生存。研究将他们的灭绝重新呈现为长期而脆弱的坚持，而非一次性崩溃，这对我们建模过去的人类-环境韧性具有启示。
(15) Artemis II is not safe to fly
Artemis II safety concerns argue the crewed lunar flyby is riskier than desired given current data, urging more analysis, testing, and risk mitigation before launch.
阿耳忒弥斯II 的安全性引发质疑，认为在当前数据与测试水平下执行载人月球飞行风险较高，需要进一步分析、测试和风险缓解。文章强调推进大胆太空任务时安全评估的重要性。
(16) Open source CAD in the browser (Solvespace)
SolveSpace now runs as a browser-based version, letting you model 2D/3D CAD directly without installing software. The web version demonstrates how a full-featured CAD tool can run client-side, lowering the barrier to entry for quick prototyping and education, though performance and file compatibility may vary across browsers. It’s a reminder of browsers' evolving role as universal engineering workstations.
Solvespace 的浏览器版本无需安装即可进行二维/三维建模。该实现展示了完整 CAD 工具在前端浏览器端运行的可能性，降低了入门门槛，便于快速原型设计与教学，但不同浏览器的性能和文件兼容性可能有所不同。
(17) Tell HN: Chrome says "suspicious download" when trying to download yt-dlp
A user reports Chrome flags the yt-dlp download as suspicious, highlighting false positives in browser security heuristics. The situation underscores the friction between browser safety nets and developers distributing open-source CLI tools.
用户反馈 Chrome 将 yt-dlp 的下载标记为可疑，暴露出浏览器安全启发式判定的误报问题。此情形揭示了浏览器安全机制与开发者分发开源命令行工具之间的摩擦。
(18) Ollama is now powered by MLX on Apple Silicon in preview
Ollama now supports MLX on Apple Silicon in preview, enabling accelerated on-device LLMs on M-series Macs. The preview could bring notable speedups and privacy benefits versus cloud inference, but adopters should be aware of early-stage status and hardware compatibility.
Ollama 现已在 Apple Silicon 支持 MLX，提供本地推理的加速效果。预览版可能带来显著的性能提升和隐私优势，但仍处于早期阶段，需关注兼容性与稳定性。
(19) GitHub backs down, kills Copilot pull-request ads after backlash
GitHub has scrapped Copilot pull-request ads after community backlash, signaling a shift in how sponsor messaging appears inside developer workflows. The move adds to ongoing debates about paid AI assistance in open-source tooling and PR review noise.
GitHub 在社区强烈反对后取消 Copilot 在拉取请求中的广告，体现了开发者工作流中赞助信息呈现的权衡。此举加剧了关于在开源工具中使用付费 AI 助手以及降低 PR 审核噪声的讨论。
(20) Claude Code users hitting usage limits 'way faster than expected'
Users of Claude Code are hitting quota limits far sooner than forecast, indicating surging demand or capacity constraints as AI coding assistants gain traction. The bottleneck could prompt pricing, plan revisions, or scalable infra investments to meet developer workload.
Claude Code 的用户比预期更早达到额度上限，显示需求旺盛或容量受限。此现象可能推动定价、套餐调整以及对可扩展基础设施的投资以应对开发者的工作量。
(21) Cohere Transcribe: Speech Recognition
Cohere launches Transcribe, its new speech recognition service aimed at simplifying transcription workflows for developers and enterprises. It emphasizes accuracy, multilingual support, and easy API integration, positioning itself against established players in the AI speech stack. This matters as AI-powered transcription becomes foundational for analytics, captions, and accessibility tools.
Cohere 推出 Transcribe，面向开发者和企业的语音识别新服务，强调高准确性、多语言支持和便捷的 API 集成，在 AI 语音领域与既有厂商展开竞争。随着自动转写在分析、字幕和无障碍工具中的应用日益普及，这类服务的重要性不断上升。
(22) Ordinary Lab Gloves May Have Skewed Microplastic Data
A study suggests ordinary lab gloves can shed microplastics, skewing measurements in microplastic research. The finding underscores the importance of contamination controls and validation in laboratory methods.
研究表明普通实验手套可能释放微塑料，从而污染测量数据。这一发现强调在实验室操作中加强污染控制和方法验证的重要性。
(23) Slop is not necessarily the future
Slopware is criticized as not necessarily the future for AI tooling, highlighting concerns about robustness and long-term viability. The piece argues for more disciplined approaches to software design over minimal, rapid prototypes.
文章认为 Slopware 未必是 AI 工具的未来，警示在追求快速迭代时要关注鲁棒性与长期可维护性。提倡更为稳健的工程实践。
(24) I Traced My Traffic Through a Home Tailscale Exit Node
The author traces their traffic as it exits through a home Tailscale exit node, revealing routing paths, latency characteristics, and privacy considerations of personal mesh VPNs. The post spots practical trade-offs of personal exit nodes for private networking.
作者追踪了通过家庭 Tailscale 退出节点传输的网络流量，揭示路由、延迟与隐私方面的权衡。展示了个人出口节点在私有网络中的实际影响。
(25) Universal Claude.md – cut Claude output tokens
Universal Claude.md reports a method to cut Claude's output token usage by about 63%, enabling cheaper and faster responses. The approach focuses on prompt engineering and token-efficient techniques to reduce cost and latency when using Claude at scale.
Universal Claude.md 展示了一种将 Claude 输出令牌消耗降至约63% 的做法，降低成本并提升响应速度。通过更高效的提示设计与令牌管理来提高大模型的可扩展性。
(26) Show HN: Postgres extension for BM25 relevance-ranked full-text search
A Postgres extension implements BM25 relevance-ranked full-text search, offering improved ranking quality for search results within Postgres. It enables teams to tune search relevance without leaving the database.
Postgres 的 BM25 相关性排序扩展实现了更高质量的全文检索排名，帮助在数据库层面优化搜索结果。让开发者不必离开 Postgres 就能提升检索效果。
(27) Fedware: Government apps that spy harder than the apps they ban
A critical look at government apps that prioritize surveillance over user rights, highlighting gaps in procurement and accountability. Calls for stronger auditing and privacy protections.
对政府应用程序过度监控、侵犯隐私的现象进行批评，揭示采购与问责方面的漏洞，并呼吁加强审计与隐私保护。
(28) Why the US Navy won't blast the Iranians and 'open' Strait of Hormuz
The piece argues the US Navy is unlikely to blast or forcibly open the Strait of Hormuz, citing legal, strategic, and humanitarian constraints. It frames open-channel posture as a deterrence and diplomacy-driven approach rather than brinkmanship.
文章认为美军不太可能轰炸或封锁霍尔木兹海峡，原因在于法律、战略与人道等约束，以及以威慑和外交为主的战略取向。
(29) How to turn anything into a router
A practical guide to repurposing everyday devices into network routers, using affordable hardware and open-source tooling. It covers core techniques and trade-offs, from setting up a small single-board computer to routing traffic securely.
将日常设备改造成路由器的实用指南，借助低成本硬件和开源工具实现网络转发。文章介绍从配置小型单板机到路由流量的核心方法，展示 DIY 路由在灵活性和隐私方面的潜力。
(30) GitHub Monaspace Case Study
A case study examining Monaspace on GitHub, focusing on design decisions for scalable monospaced tooling and UI. It discusses architecture trade-offs, performance considerations, and lessons for building developer-focused tooling.
对 GitHub 上的 Monaspace 进行案例研究，聚焦可扩展的等宽工具与 UI 的设计决策，讨论架构权衡、性能考量以及面向开发者工具的经验教训。
(31) Google's 200M-parameter time-series foundation model with 16k context
Google Research unveils TimesFM, a 200M-parameter time-series foundation model with a 16k context window. The model demonstrates that compact architectures can achieve long-range forecasting, unlocking more capable edge and on-device time-series analytics with lower compute budgets.
谷歌研究院发布 TimesFM，一个200M参数的时序基础模型，具备16k上下文窗口。该模型表明紧凑架构也能实现长距离预测，推动边缘和本地时序分析在更低算力预算下实现。
(32) Do your own writing
A call to authors to write themselves rather than rely on AI, with tips for maintaining originality and critical thinking in a crowd-sourced era.
倡导作者坚持亲自写作，而非依赖 AI。文章提供保持原创性与批判性思维的实践性建议，帮助在创作生态中保持独立性。
(33) Combinators
Combinators are higher-order building blocks that compose functions without explicit variables. The TinyAPL documentation likely illustrates how to use them to build complex behavior from simple, reusable pieces, enabling point-free style and functional composition. Mastery helps in understanding language design and meta-programming patterns.
组合子是无需显式变量即可组合函数的高阶构件。文档中的示例展示了如何用它们以点免费风格组合出更复杂的行为，便于理解函数式编程的组合与语言设计思路。
(34) Turning a MacBook into a touchscreen with $1 of hardware (2018)
A 2018 DIY project demonstrates turning a MacBook into a touchscreen using a tiny hardware hack. It showcases how inexpensive components can enable new input modalities on laptops, with caveats about compatibility and usability.
2018 年的一个 DIY 项目展示如何用不到1美元的硬件把 MacBook 变成触控屏，体现用低成本元件扩展笔记本输入方式的可能性，同时需权衡兼容性与可用性。
(35) CodingFont: A game to help you pick a coding font
CodingFont is an interactive game designed to help you pick a coding font by evaluating readability and rhythm. It surfaces metrics and practical tips for choosing fonts that reduce eye strain and improve coding speed.
CodingFont 是一款互动游戏，帮助你通过可读性和打字节奏等维度来选择编码字体。它揭示了影响眼睛疲劳和编码速度的字体特性，并给出实用的选型建议。
(36) Android Developer Verification
Google expands Android Developer Verification, requiring identity verification to strengthen ecosystem trust and curb fraudulent apps. The rollout is global, with implications for app review processes and developer onboarding.
Google 将 Android 开发者验证全面向全球开发者铺开，要求完成身份认证以提升生态系统信任并打击假冒应用。此举将影响应用审核流程与开发者注册 onboarding。
(37) RubyGems Fracture Incident Report
RubyGems releases an incident report detailing a fault in the gemstone package ecosystem, its impact on dependencies, remediation steps, and lessons for supply-chain security and deployment pipelines.
RubyGems 发布事件报告，披露宝石包生态系中的故障及其对依赖项的影响、修复步骤与对供应链安全与部署管线的经验教训。
(38) Super Micro Computer Investors Look for Exits
Investors in Super Micro Computer look for exits, signaling a push to realize gains or rebalance portfolios. The move can influence stock sentiment and strategic options in the server hardware space.
投资者开始退出 Super Micro Computer，表明在成长阶段后寻求实现收益或重新配置投资组合。此举可能影响股价走向与服务器硬件领域的并购热度。
(39) Nobody Is Coming to Save Your Career
A sharp, pragmatic prompt urging readers to take ownership of their career, continuously learn, and adapt to a changing tech landscape rather than waiting for external saviors. The piece blends personal accountability with practical advice on skills, networking, and ongoing value creation.
作者直言不讳，劝读者主动掌控职业发展、持续学习、适应变化的技术格局，而非等待外部救援。文中提供技能提升、人脉建设和持续价值创造等实用建议。
(40) Bird brains (2023)
A concise reflection on what bird brains in 2023 reveal about avian cognition and brain evolution.
对2023年鸟类认知研究和大脑进化要点的简要解读。
(41) Scotty: A beautiful SSH task runner
Scotty is a terminal-based SSH task runner with an elegant design, enabling streamlined remote task orchestration. It demonstrates how thoughtful CLI tooling can boost productivity for sysadmins and DevOps engineers.
Scotty 是一个美观的终端 SSH 任务执行器，设计优雅，便于远程任务编排，体现了用心打造的命令行工具如何提升系统管理员与 DevOps 的工作效率。
(42) FTC action against Match and OkCupid for deceiving users, sharing personal data
FTC action accuses Match and OkCupid of deceiving users and sharing personal data with third parties, prompting new privacy commitments and compliance measures. The case underlines ongoing pressure on dating apps to protect user data.
FTC 指控 Match 与 OkCupid 欺骗用户并将个人数据分享给第三方，要求对隐私做出新的承诺与合规措施。这起事件凸显了约会应用保护用户数据的持续压力。
(43) Learn Claude Code by doing, not reading
A practical, do-first guide to learning Claude Code through interactive exercises rather than theory. It emphasizes writing, testing, and iterating prompts and snippets to master Claude's coding capabilities for building AI-assisted tooling.
以互动练习为核心的 Claude Code 实践教学，强调编写、测试和迭代提示与代码片段，以快速掌握 Claude 的编码能力，适合构建 AI 辅助工具的开发者。
(44) Ask HN: Distributed data centers in our basements
Explores DIY or consumer-scale distributed data centers housed in homes or basements, leveraging commodity hardware and open-source tooling. The piece weighs practicality, reliability, power, cooling, and safety considerations as this niche matures.
讨论在家庭或地下室部署的分布式数据中心，利用通用硬件与开源工具。评估可行性与挑战，包括电力、散热、可靠性和安全等要点，以及这一领域的进一步发展。
(45) Vulnerability research is cooked
Argues that vulnerability research is shaped by incentives and disclosure norms, urging more rigorous, transparent practices and better governance of bug disclosures.
认为漏洞研究受激励与披露规范影响，呼吁采用更严格、透明的做法，并改善漏洞披露治理。
(46) Securing Elliptic Curve Cryptocurrencies Against Quantum Vulnerabilities [pdf]
A Google Quantum AI whitepaper on securing elliptic-curve-based cryptocurrencies against quantum vulnerabilities, outlining post-quantum approaches and the migration challenges for wallets and blockchains.
谷歌量子人工智能团队的白皮书探讨对基于椭圆曲线的加密货币进行量子威胁防护，概述后量子时代的方案及钱包和区块链迁移面临的挑战。
(47) OpenGridWorks: The Electricity Infrasctructure, Mapped
OpenGridWorks maps electricity infrastructure, offering an open visualization of grids and critical assets. The project aims to raise transparency, support planning, and highlight risks in energy networks.
OpenGridWorks 正在把电力基础设施绘制成开放数据地图，提高透明度并支持规划与风险分析。此类可视化有助于理解能源网络的脆弱点。
(48) Mathematical methods and human thought in the age of AI
A concise reflection on how AI is reshaping mathematical practice, from automated reasoning to collaborative tools. It argues for maintaining rigor while embracing machine-assisted discovery, and highlights pedagogy implications.
探讨人工智能正在如何重塑数学实践，从自动推理到协作工具的兴起。强调在拥抱机器辅助发现的同时保持严格性，并指出对教学与学习的影响与机遇。
Advocates AT Protocol as a foundational, open, decentralized social protocol, outlining architectural and governance advantages and why builders should consider adopting it.
作为一个开放、去中心化的社交协议基础，阐述 AT Protocol 的架构与治理优势，并解释为何开发者应考虑采用这一协议。
(50) 7,655 Ransomware Claims in One Year: Group, Sector, and Country Breakdown
A security industry dataset tallies 7,655 ransomware incidents over a year, breaking down by threat actor, sector, and geography. The stats highlight persistent perimeters and sector-specific risk, informing defenders where to focus controls and response planning.
安全行业的统计显示，在一年内发生了7,655起勒索软件事件，按攻击团体、行业和地区进行细分。数据揭示各行业与地区的脆弱点，帮助防守方有针对性地部署防护与应急计划。
(51) Accidentally created my first fork bomb with Claude Code
The author reports accidentally creating a fork bomb with Claude Code, illustrating how AI-assisted coding can produce dangerous constructs. The incident underscores the need for sandboxing, guardrails, and threat modeling when experimenting with code-generation models.
作者自述在使用 Claude Code 时意外创建了 Fork Bomb，展示了 AI 辅助编码可能带来的危险构造。此事强调在进行代码生成实验时需要沙箱、安全护栏与威胁建模。
(52) Iran says it will target US tech companies in Middle East
Iran signals potential actions targeting US tech firms operating in the Middle East, raising geopolitical and cybersecurity concerns for the tech industry and international supply chains.
伊朗表示可能对在中东运营的美国科技公司采取行动，提升了对科技行业和国际供应链的地缘政治及网络安全担忧。
(53) Ask HN: Academic study on AI's impact on software development – want to join?
An academic study is seeking contributors to join a project examining AI's impact on software development, inviting researchers and practitioners to help collect and analyze data.
一项学术研究正在招募参与者，旨在研究 AI 对软件开发的影响，向研究者与从业者开放参与数据收集与分析的机会。
(54) Good CTE, Bad CTE
Good CTE, Bad CTE distinguishes when common table expressions help readability and maintainability versus when they become performance traps. Practical guidance emphasizes avoiding over-nesting and unnecessary wrapping of expensive queries.
好 CTE，坏 CTE 区分了 CTE 提升可读性与维护性的情形与成为性能陷阱的情形。给出实际建议，避免嵌套过深和对昂贵查询的无谓包装。
(55) Mr. Chatterbox is a Victorian-era ethically trained model
Mr. Chatterbox is a Victorian-era ethically trained model, using guardrails inspired by 19th-century norms to shape its outputs. The experiment probes how historical ethical frames influence modern AI behavior and what that means for real-world safety design.
Mr. Chatterbox 是一个以维多利亚时代道德准则为引导的伦理训练模型，用19世纪的规范来约束输出。这一实验探讨了历史道德框架如何影响现代 AI 行为，以及对现实世界安全设计意味着什么。
(56) Objections to systemd age-attestation changes go overboard
Debate over systemd's age-attestation changes has drawn strong objections, highlighting concerns about compatibility, policy, and the impact on Linux distributions.
关于 systemd 的年龄认证变更引发强烈反对，凸显了兼容性、政策及对 Linux 发行版影响的担忧。
(57) Car Seats as Contraception
Car Seats as Contraception analyzes the provocative idea of using car seating contexts as a form of contraception, exploring its social and policy implications. The piece frames broader questions about access, behavior, and reproductive health.
汽车座椅作为避孕工具探讨了一种大胆的设想，讨论其社会与政策层面的影响，以及与获取、行为和生殖健康相关的更广泛问题。
(58) Incident March 30th, 2026 – Accidental CDN Caching
The incident on March 30, 2026, examined accidental CDN caching that led to issues such as stale content or mis-delivery. The postmortem stresses the need for robust cache invalidation, monitoring, and rollback plans in CDN configurations.
2026年3月30日的事故回顾：CDN 缓存的意外配置导致内容未及时刷新等问题。事后强调必须加强缓存失效、监控和回滚机制在 CDN 配置中的应用。
A practical guide to recovering access to Apple Keychain, covering common issues, recovery steps, and security considerations. It highlights the risks of losing keychain access and how to regain credentials.
提供恢复 Apple Keychain 的实用指南，涵盖常见问题、恢复步骤与安全注意事项，强调丢失密钥串的风险及找回凭证的做法。
(60) William Blake, Remote by the Sea
A roundtable on William Blake's seascape imagery and its relation to Romantic imagination, exploring how the sea motifs reflect spirituality and critique of industrial modernity.
围绕威廉·布莱克的海洋意象及其与浪漫主义想象之间的关系展开讨论，探究海洋题材如何映射他的精神世界与对工业现代性的批判。
(61) Show HN: I turned a sketch into a 3D-print pegboard for my kid with an AI agent
An indie Show HN project demonstrates turning a hand-drawn sketch into a 3D-printed pegboard for a kid, using an AI agent to bridge sketch to CAD to printer. It showcases end-to-end automation and the practicality of agent-assisted maker workflows.
Show HN：把手绘草图经由 AI 助手转化成给孩子用的3D打印穿孔板，展示从草图到 CAD 再到 3D 打印的端到端自动化。演示了 AI 驱动的创客工作流的可行性。
(62) Safeguarding cryptocurrency by disclosing quantum vulnerabilities responsibly
Google Research advocates responsible disclosure of quantum vulnerabilities to protect cryptocurrency, proposing collaboration with the crypto ecosystem to shore up quantum resilience before attacks become feasible. The post emphasizes proactive, coordinated action.
谷歌研究院倡导对量子漏洞进行负责任披露，以保护加密货币体系的量子鲁棒性，主张与加密生态系统协作在攻击可行前强化防护措施。
(63) In Expanding de Sitter Space, Quantum Mechanics Gets More Elusive
New insights into quantum mechanics in curved, expanding spacetimes reveal deeper conceptual hurdles and paradoxes. The work discusses how expansion and horizons in de Sitter space complicate standard quantum treatments, with implications for cosmology and quantum gravity research.
对在弯曲、膨胀时空中量子力学的最新见解揭示了更深层的概念难题和悖论。狄士特时空的膨胀与视界使传统量子理论的处理变得复杂，这对宇宙学和量子引力研究具有重要意义。
(64) Show HN: Coasts – Containerized Hosts for Agents
Coasts provides containerized hosts for agents, enabling isolated, reproducible environments for autonomous agents. It simplifies running multiple agents with clean state and easy scaling.
Coasts 为代理提供容器化的宿主环境，实现代理的隔离、可重复的运行环境。它便于以干净状态运行多个代理并支持快速扩展。
(65) A sea of sparks: Seeing radioactivity
A modern spinthariscope project visualizes radiation as sparks, offering a tangible glimpse into atomic processes. It blends vintage science outreach with hands-on electronics to make radiation more approachable.
这个现代的 spinthariscope 项目用火花直观展示放射性，帮助公众理解原子过程。它把复古科普与现代电子技术结合起来，使放射性变得更易理解。
(66) Build123d: A Python CAD programming library
Build123d is a Python-based CAD programming library that enables parametric, script-driven 3D modeling. By treating CAD as code, it aims to streamline automation, testing, and design exploration.
Build123d 是一个用 Python 编写 CAD 的编程库，支持参数化、脚本驱动的三维建模。将 CAD 视为代码有助于实现自动化、测试和设计探索。
(67) Sony halts memory card shipments due to NAND shortage
Sony is halting memory card shipments amid a NAND shortage, illustrating how semiconductor supply constraints ripple into consumer accessories. The pause signals broader risks for camera gear and portable devices relying on flash storage.
由于 NAND 闪存短缺，索尼暂停记忆卡出货，显示半导体供应紧张如何波及到消费外设。此举也反映了相机与便携设备对闪存的依赖以及潜在风险。
(68) Cherri – programming language that compiles to an Apple Shortuct
Cherri is a programming language that compiles to Apple Shortcuts, letting developers write automation logic in a higher-level, typed syntax and then produce the iOS automation graph. It lowers the barrier to complex automations beyond the native Shortcuts UI, while still targeting the same automation platform.
Cherri 是一门可编译为 Apple Shortcuts 的编程语言，允许开发者用更高层、带类型的语法编写自动化逻辑，再生成 iOS 的 Shortcuts 工作流。它让实现复杂自动化的门槛低于原生 Shortcuts 界面，同时仍然面向同一自动化平台。
(69) We're pausing Asimov Press
Asimov Press announces a pause in operations, signaling a temporary halt in publishing or distribution of its science-fiction content. The move raises questions about indie presses' sustainability and how small publishers navigate market shifts and author relations.
Asimov Press 宣布暂停运营，标志着其科幻出版/发行工作的临时停摆。此举反映独立出版社在市场变化面前的生存挑战，以及作者关系与发行模式的调整。
(70) Audio tapes reveal mass rule-breaking in Milgram's obedience experiments
New archival audio tapes reveal widespread rule-breaking by participants in Milgram's obedience experiments, suggesting obedience was not monolithic and that many subjects questioned or resisted authority. The findings add nuance to historical debates about obedience, ethics, and experimental design in social psychology.
来自档案的音频记录显示，参与者在米尔格拉姆的服从实验中普遍打破规定，对权威的顺从并非一刀切。研究为历史中的服从与实验伦理带来新的细微差别，凸显了实验设计及参与者行为的复杂性。
(71) R3 Bio pitched “brainless clones” to serve the role of backup human bodies
R3 Bio pitched “brainless clones” as backup human bodies, a controversial idea that raises questions about consent, identity, and the ethics of radical life-extension biotech.
R3 Bio 推出“无脑克隆人”作为备份人体的设想，这一有争议的想法涉及同意、身份与极端长寿生物技术的伦理。文章分析了资金叙事与风险。
(72) Clojure: The Documentary, official trailer [video]
The documentary trailer highlights the language's history, philosophy, and influence on modern software culture. The film frames Clojure as a case study in Lisp heritage and practical concurrency.
《Clojure: The Documentary》正式预告片聚焦语言的历史、哲学及对现代软件文化的影响。影片把 Clojure 视作 Lisp 遗产与并发编程实践的案例。
(73) What we learned building 100 API integrations with OpenCode
Lessons learned from building 100 API integrations with OpenCode—patterns for integration design, developer experience, and troubleshooting at scale. Practical guidance for teams facing many external services.
总结在 OpenCode 平台构建百余个 API 集成的经验，涵盖集成设计、开发者体验和大规模故障排查的模式与要点，为连接多方服务的团队提供实用建议。
(74) Claude Code bug can silently 10-20x API costs
A bug in Claude Code quietly inflates API costs by 10-20x due to caching mishaps, underscoring how subtle engineering flaws can erode cost efficiency in AI toolchains. The fix is in progress as the vendor addresses the bug.
Claude Code 的缓存漏洞悄然将 API 成本提升 10-20 倍，暴露了微妙工程缺陷对 AI 工具成本效益的侵蚀。厂商正在修复并改进监控与计费。
(75) 4D Doom
A project titled 4D Doom explores a four-dimensional version of the classic Doom, pushing ideas about rendering and navigation in higher dimensions. It serves as a playful showcase of higher-dimensional computation and visualization.
4D Doom 项目将 Doom 推向四维空间，探索在高维空间中的渲染与导航。是对高维计算和可视化的一种有趣概念实验。
(76) What major works of literature were written after age of 85? 75? 65?
A data-driven look at late-life literary production, examining whether major works emerge after 65, 75, or 85 and how prolific authors remain with age. The analysis shows rare but notable late-life masterpieces, while overall output tends to wane as authors age.
本文用数据探讨文学巨匠是否在65、75、85岁后仍有重大作品问世，以及作者随年龄增长的产出变化。结论是虽有罕见的晚年杰作，但总体产出往往随年龄上升而减少。
(77) Seeing like a spreadsheet
A historical piece argues that spreadsheets reshaped American institutions by standardizing data, workflows, and governance. The essay connects simple tabular tools to broad changes in policy and everyday decision-making, with implications for designing data-driven systems.
本文主张电子表格通过标准化数据、工作流和治理，重塑了美国产业与机构。作者将简单的表格工具与政策与日常决策的广泛变革联系起来，对设计数据驱动系统具有启示作用。
(78) From 300KB to 69KB per Token: How LLM Architectures Solve the KV Cache Problem
The piece explains how LLM KV caches have been dramatically shrunk from around 300KB per token to about 69KB, enabled by architectural innovations. It covers memory reuse, compression, and smarter caching strategies that cut inference costs and allow larger contexts.
文章阐述了大模型 KV 缓存每个 token 的内存占用从约 300KB 降到约 69KB 的关键架构创新，涉及内存重用、压缩和更聪明的缓存策略，显著降低推理成本并扩展上下文容量。
(79) Agents of Chaos
An in-depth look at chaotic agents in technology and security, exploring their methods and the risks they pose to systems and governance.
对技术与安全领域中的混乱代理人进行深度探讨，分析其作法及对系统与治理造成的风险。
(80) Researchers find 3,500-year-old loom that reveals textile revolution
Bronze Age loom discovery sheds light on textile production advances, showing early mechanization and social shifts tied to cloth trade. The find helps historians understand technological diffusion and labor organization in ancient economies.
Bronze Age 织机发现揭示早期纺织生产的进步，体现了与布料贸易相关的早期机械化与社会变革。研究有助于理解古代经济中的技术扩散与劳动组织。
(81) Roulette Computers: Hidden Devices That Predict Spins
An investigative look at hidden hardware and methods used to predict roulette spins, illustrating how specialized devices can tilt odds.
揭秘用于预测轮盘旋转的隐藏硬件与方法，展示了专用设备如何改变赌博概率。文章还讨论赌场防护的局限性以及对赌博科技与安全研究的影响。
(82) TruffleRuby
TruffleRuby demonstrates how Ruby can run atop the Graal/Truffle ecosystem, delivering performance improvements and better interoperability in polyglot runtimes. It offers practical options for Ruby workloads that need to coexist with other languages in the same VM.
TruffleRuby 展示 Ruby 如何在 Graal/Truffle 生态中运行，带来性能提升并提升多语言虚拟机中的互操作性。对于需要与其他语言在同一虚拟机中共存的 Ruby 工作负载具有现实意义。
Oscar Reutersvärd, recognized as a pioneer of impossible figures, prefigured Escher with early drawings that bend perspective. The piece situates his work in the lineage of mathematical art and visual paradox.
奥斯卡·雷特斯瓦德被视为“不可能图形”之先驱，其早期作品在透视与几何上扭曲，为埃舍尔等后来的视觉悖论艺术奠定了基础。
(84) Show HN: Forkrun – NUMA-aware shell parallelizer (50×–400× faster than parallel)
Forkrun introduces NUMA-aware parallelization for shell commands, delivering dramatic speedups on multi-socket servers. The project demonstrates how memory locality and CPU affinity can unlock large performance gains for everyday CLI tasks. It could inspire a new wave of performance-focused shell tooling.
Forkrun 提供对 shell 命令的 NUMA 感知并行化，在多插槽服务器上实现显著的速度提升。该项目展示了内存局部性与 CPU 亲和性如何为日常命令行任务带来巨大的性能提升，或将推动新一轮面向性能的 shell 工具设计。
(85) In math, rigor is vital, but are digitized proofs taking it too far?
As mathematicians lean on digitized, formally verified proofs, the article questions whether formalization can erode intuition and accessibility. It weighs the gains in reliability against the potential for steeper learning curves and closed formats.
随着数学家越来越依赖形式化、可验证的证明，本文质疑过度形式化是否会削弱直觉与可读性。权衡了可靠性提升与学习成本、可访问性下降之间的取舍。
(86) TinyLoRA – Learning to Reason in 13 Parameters
TinyLoRA demonstrates that reasoning can be learned with a tiny, parameter-efficient adaptation. It shows that small, targeted changes can unlock reasoning capabilities without full fine-tuning, prompting renewed consideration of where prompts end and training begins.
TinyLoRA 展示通过极少量参数的高效适配就能为模型带来推理能力。小而定向的改动即可实现有用的推理能力，这促使人们重新思考提示工程与微调之间的界限。
(87) Teenage Engineering's PO-32 acoustic modem and synth implementation
Teenage Engineering's PO-32 acoustic modem and synth implementation exposes open-source work that expands the device's use beyond standard musicians' toolkit. It enables data transmission and synthesis experiments without wiring.
这项实现公开了 PO-32 的声学调制解调器与合成功能，扩展了设备的用途，支持无线数据传输与声音实验。
(88) One of the largest salt mines in the world exists under Lake Erie
A massive salt mine under Lake Erie ranks among the world’s largest, formed by ancient evaporites. It underpins regional winter road salt supply and industrial needs, with geotechnical and environmental considerations shaping its ongoing operations.
伊利湖底隐藏着全球最大的盐矿之一，源自主风化海床的蒸发作用。该矿支撑区域冬季道路用盐和工业需求，同时也要考虑地质与环境因素。
(89) Show HN: Loreline, narrative language transpiled via Haxe: C++/C#/JS/Java/Py/Lua
Loreline is a project that uses Haxe to transpile a narrative language into multiple targets, enabling cross-language deployment for storytelling or scripting contexts.
Loreline 是一个使用 Haxe 将叙事语言转译为多种目标语言的项目，便于在讲述性或脚本情境中跨语言部署。
(90) Use string views instead of passing std:wstring by const&
Advocates using string_view to avoid copying and to enable more flexible string handling, with caveats around lifetimes, encoding, and interoperability with wide strings. The post shares practical guidance and pitfalls when refactoring code paths to use string_view.
提倡使用 string_view 以避免拷贝并提升字符串处理的灵活性，同时需注意生命周期、编码和与 std::wstring 的互操作性等坑点。文中给出将代码路径改用 string_view 的实用建议与常见陷阱。
(91) Inside the 'self-driving' lab revolution
Nature's report surveys how autonomous driving labs are modularizing development with standardized datasets, safety protocols, and rapid prototyping with simulations. It highlights the shift toward modular, test-driven workflows and the challenges of translating lab results to real-world deployment.
Nature 的报道探讨了自动驾驶实验室通过标准化数据集、安全协议和仿真快速原型化来实现模块化开发的趋势，强调向以测试驱动的工作流转变以及将实验结果落地到现实世界的挑战。
(92) A Love Letter to 'Girl Games'
A love letter celebrating the 'Girl Games' genre, exploring creativity, community, and the value of inclusive indie games for players and developers alike.
致敬“Girl Games”这一体裁，探讨其创意、社区与对玩家与开发者共同体的包容性独立游戏价值。
(93) A Primer on Long-Duration Life Support
This primer outlines core concepts behind long-duration life support systems, including air and water recycling, waste management, and power budgeting for extreme environments. It highlights the engineering challenges of sustaining life-support ecosystems over months or years and the cross-disciplinary work required.
本入门文章介绍长时生命维持系统的核心概念，如空气与水的循环利用、废物管理与能源预算，聚焦在极端环境中维持生命支持生态系统的工程挑战，以及所需的跨学科协作。
(94) Back to FreeBSD – Part 2 – Jails
Explores FreeBSD's jail feature as a lightweight OS-level virtualization technique, detailing setup, security considerations, and how jails compare to modern containers. The piece provides practical guidance for deploying stable, isolated services on FreeBSD.
深入探讨 FreeBSD 的 Jail 作为轻量级操作系统级虚拟化技术，涵盖搭建、安全性考量以及与现代容器的差异对比，并给出在 FreeBSD 上部署稳定、隔离服务的实用建议。
(95) Multiple Sclerosis
A concise exploration of multiple sclerosis, its symptoms, and the current state of research and treatment options.
对多发性硬化的简要概述，涵盖其病理、症状以及当前的治疗进展。
(96) Unit: A self-replicating Forth mesh agent running in a browser tab
Unit demonstrates a self-replicating Forth mesh agent that runs inside a browser tab, offering a provocative look at distributed autonomous code in the wild. The project raises questions about browser sandboxing, security boundaries, and what constitutes self-replication today.
Unit 展示了一个在浏览器标签页中运行的自复制 Forth 网格代理，揭示了边缘环境中分布式自治代码的可能性。该项目引发对浏览器沙箱、安全边界以及何谓自我复制的讨论。
(97) Nematophagous Fungus
Nematophagous fungi use specialized structures to trap and digest nematodes, an ecological strategy with potential biocontrol applications in agriculture. The post surveys examples and the biology behind these natural predators.
线虫捕食真菌利用专门的结构捕捉并消化线虫，具有在农业中作为自然生防手段的潜力。介绍了相关生物学与生态学要点。
(98) Learn Something Old Every Day, Part XVIII: How Does FPU Detection Work?
Discusses how CPUs detect floating-point unit availability and features, historical quirks, and the implications for software querying CPU capabilities. Highlights how detection techniques evolved and their relevance to low-level systems programming.
讲解 CPU 如何检测浮点单元（FPU）的可用性与特性、历史遗留问题，以及软件查询 CPU 能力时的含义。强调检测技术的演变及其在底层系统编程中的相关性。
(99) 30 Years Ago, Robots Learned to Walk Without Falling
A retrospective on a landmark milestone in humanoid robotics, detailing how early research enabled stable bipedal walking and the progress since. The milestone helped unlock real-world mobility for robots and underscored the long journey toward robust, energy-efficient locomotion.
回顾人形机器人领域的里程碑，介绍早期研究如何实现稳定的双足行走及其后的发展。该突破推动机器人真正具备地面机动能力，并凸显实现鲁棒、节能步态的长期挑战。
(100) IronGlass Brings Legendary Soviet Cinema Lenses to Mirrorless Cameras
IronGlass brings legendary Soviet cinema lenses to mirrorless cameras via adapters, offering photographers a distinct vintage look with modern bodies. The tech enables creative experimentation without buying new glass.
IronGlass 通过适配器把传奇的苏联电影镜头带入无反相机，给予摄影师独特的复古质感与现代工艺结合的可能。