1. 8万年前在奥比-拉赫马特(乌兹别克斯坦)发现的箭头? Arrow heads at Obi-Rakhmat (Uzbekistan) 80K years ago? (journals.plos.org)
5. 西班牙下令封锁Archive.today及其镜像站点 Spain Orders Blocks on Archive.today and Its Mirrors (reclaimthenet.org)
8. Google AI Studio 伪造了数据删除操作。我因举报此事,被 VRP 在 60 秒内自动封禁。 Google AI Studio fakes data deletion. VRP auto-banned me in 60s for reporting it (medium.com)
9. 为什么可变类型不是不可变类型的子类型,反之亦然? Why isn''t mutable a subtype of immutable, or vice versa? (crumbles.blog)
10. 比亚迪大幅下调电动车价格,成为澳大利亚最便宜的车型 [视频] BYD Slashes Price of Electric Car and Becomes Cheapest in Australia [video] (www.youtube.com)
13. Polymarket为追求快速扩张,给诈骗分子留下了可乘之机 Polymarket''s Rush to Grow Left a Door Wide Open for Fraudsters (www.wsj.com)
15. 通过电极-电解质界面溶解实现废旧电池的再生 Regeneration of used batteries via electrode–electrolyte interphase dissolution (pubs.rsc.org)
18. 一款在地下城中展开的开源类Roguelike冒险游戏 An open source roguelike adventure through dungeons (crawl.develz.org)
21. Hugging Face 黑客事件其实并没有传得那么惊人 The Hugging Face Hack Wasn''t What It Was Cracked Up to Be (www.wsj.com)
22. 北美最大的野生动物天桥使野生动物撞车事故减少了91% Largest wildlife overpass in North America reduced wildlife collision by 91% (www.reddit.com)
25. Show HN: 我创建了一个可在本地使用的开源、功能齐全的人工智能平台 I created an open source locally usable full fledged AI platform (github.com)
27. Claudecookie – 根据 Cookie 转换、验证并生成 Claude Code 凭证 Claudecookie – convert, check, and mint Claude Code credentials from a cookie (claudecookie.com)
29. 为什么人工智能无法拯救一家不了解自身数据的企业 Why AI Cannot Save an Enterprise That Doesn''t Understand Its Data (architectureintel.com)
32. 你可以利用整数溢出击败《时空之轮》中的“梦之吞噬者” You can defeat the Dream Devourer from Chrono Trigger using an int overflow (chrono.fandom.com)
34. Hacker News 排名机制如何运作:评分、争议与惩罚(2013) How Hacker News ranking works: scoring, controversy, and penalties (2013) (www.righto.com)
38. 我在 pfSense 后端捕获了 72 小时的 Android 空闲数据包 I captured 72 hours of idle Android packets behind pfSense (www.praveentechworld.com)
39. 微软高管:人工智能“是人类历史上规模最大的劳动力盗用” Microsoft director: AI scraping ''the largest theft of labor in human history'' (www.tomshardware.com)
40. 衡量互联网审查情况。为最大的开放数据集贡献力量 Measure internet censorship. Contribute to the largest open dataset (ooni.org)
42. OpenAI 和 Anthropic 夸大了 AI 安全漏洞的严重性 OpenAI and Anthropic oversold AI security breaches (nypost.com)
43. 诺贝尔经济学奖得主支持加州“亿万富翁税”的公开信 [pdf] Economics Nobel Laureates'' Letter Supporting the California Billionaire Tax [pdf] (gabriel-zucman.eu)
48. 在经典基准测试工作负载下,Btrfs/ZFS/bcachefs 被跳过 Btrfs/ZFS/bcachefs under workloads classic benchmarks skip (bartosz.fenski.pl)
52. 几乎不要使用人工智能来撰写任何有实质内容的文章 Almost Never Use AI to Write Anything Substantive (erichgrunewald.substack.com)
57. 美国与丹麦之间的协定(1951年、2004年)[pdf] Agreement between the USA and Denmark (1951,2004) [pdf] (www.state.gov)
60. 学习另一种语言可能是保持大脑健康的最佳方式之一 Learning Another Language May Be One of the Best Ways to Keep Your Brain Healthy (theconversation.com)
65. 黑洞还是黑洞星?天文学家就“小红点”展开争论 Black Holes or Black Hole Stars? Astronomers Spar over ''Little Red Dots'' (www.quantamagazine.org)
75. Apple M6 Pro 在 Geekbench 7 中创下最高单核 CPU 得分 Apple M6 Pro Achieves the Highest Single-Core CPU Score in Geekbench 7 (browser.geekbench.com)
76. 在扫描中应用 Kubernetes 上下文以屏蔽漏洞 Suppress vulnerabilities applying Kubernetes context to scans (github.com)
77. “埃利亚斯·索恩案”:人工智能聊天机器人痴迷的虚构人物 The Case of Elias Thorne, Imaginary Man AI Chatbots Are Obsessed With (www.vice.com)
78. 为什么与人工智能相关的帖子在HN上能获得这么多积分? How come AI-related posts get so many points on HN? (news.ycombinator.com)
79. Stepfun 第 5 步预览(LLM):关于 AA 帕累托前沿 Stepfun Step 5 Preview (LLM): On AA Pareto frontier (artificialanalysis.ai)
82. 如果数学不仅仅是证明,我们就应该更好地展现它的其他方面 If math is more than proof, we need to better celebrate the rest of it (terrytao.wordpress.com)
84. 斯坦福大学医学院主导的研究发现,人类大脑由两个独立的器官组成 Human brain is two separate organs, Stanford Medicine-led research finds (med.stanford.edu)
85. 如果你重新生成打包文件,就可以在对象存储上运行 Git You can run Git on object storage if you re-make packfiles (www.tigrisdata.com)
86. NASA-IBM Lunar Foundation 开源地理空间人工智能模型 NASA-IBM Lunar Foundation open-Source Geospatial AI Model (newsroom.usra.edu)
87. Show HN: 密封信——您去世后,您的家人可以打开的信件和密码 Seal – Letters and passwords that open for your family after you die (github.com)
88. GPT模型中的“危害洗白”:性别歧视被转化而非 Harm Laundering in GPT Models: Gender Discrimination Transformed Rather Than (arxiv.org)
91. Typesafe-computer-use 以每步 1/50 美分的价格驱动 Mac 向目标前进 Typesafe-computer-use drives a Mac toward a goal for 1/50th of a cent per step (github.com)
94. Gemini 攻击了三家公司,这是谷歌人工智能首次被发现成功突破防御 Gemini hacked three companies in first known breakout by Google''s AI (www.reuters.com)
98. Show HN: LiveWorld——将所有24/7不间断运行的YouTube直播摄像头汇聚于一个地球仪上 LiveWorld – Every 24/7 YouTube live camera on one globe (liveworld.info)
99. 阿里巴巴开源了一款能够检测癌症及近150种疾病的AI模型 Alibaba open-sources AI model that can detect cancer and nearly 150 conditions (www.scmp.com)
100. LispBM 是一种基于消息传递的、面向微控制器的并发 Lisp 语言 LispBM is a concurrent Lisp for microcontrollers with message passing (www.lispbm.com)
104. OpenAI 如何利用其自有的大语言模型设计哈拉佩尼奥辣椒味薯片 How OpenAI Used Its Own LLMs to Design Its Jalapeño Chip (spectrum.ieee.org)
105. 伊朗战争期间美军阵亡人数比五角大楼统计的数字至少多出4人 US troop deaths during Iran war exceed Pentagon count by at least four (www.reuters.com)
109. 如果没有 Claude.md 文件,Claude Code 现在会读取 AGENTS.md 文件。 Claude Code now reads AGENTS.md if there is no Claude.md (code.claude.com)
110. 韩国将数据泄露罚款上调至营业额的10% Korea raises data breach fines to 10% of revenue (www.koreajoongangdaily.com)
111. 杰克·鲁舍尔的《从几何到代数,再回到几何:4000年的论文》[视频] "From Geometry to Algebra and Back Again: 4000 Years of Papers" by Jack Rusher [video] (www.youtube.com)
112. Android 17 是自 3.x 以来首个在未向 AOSP 发布的情况下就新增了 API 的版本 Android 17 is the first since 3.x to add new APIs without releasing to the AOSP (grapheneos.social)
113. 缓存间通信:大型语言模型之间的直接语义通信 Cache-to-Cache: Direct Semantic Communication Between Large Language Models (arxiv.org)
115. 苹果发布 iPhone Duo 模拟器和 Xcode 27.1 测试版 Apple releases iPhone Duo simulator and Xcode 27.1 beta (developer.apple.com)
116. 研究表明,我们的大脑是由两个原始神经系统融合而来的 Our brain evolved from two primitive nervous systems that merged: Study (www.newscientist.com)
120. 语言不可读性对大型语言模型(LLM)安全性的影响 The Implications of Linguistic Illegibility for LLM Security (arxiv.org)
121. ZCode 内部揭秘:将您的 Git 历史记录静默上传至云端 Inside ZCode: Silently uploading your Git history to the cloud (blog.ferstar.org)
123. Show HN: Ax-Check.com – 经纪人可以使用贵公司的产品吗? Ax-check.com – Can agents use your product? (www.ax-check.com)
124. 边境执法人员无需搜查令或合理怀疑即可搜查手机 Border agents can search cellphones without a warrant or reasonable suspicion (lawandcrime.com)
126. 美军因使用人工智能生成虚假情报报告,险些酿成大祸 US Military had close call after using AI for hallucinated intelligence report (www.cnn.com)
129. 基于光子发射引导的激光故障注入技术实现了 RP2350 的安全调试 Photon-Emission-Guided Laser Fault Injection Enables RP2350 Secure Debug (donjon.ledger.com)
130. Show HN: Cactus Needle 3:8-29MB的自动化模型可媲美DeepSeek V4 Flash Cactus Needle 3: 8-29MB automation models can match DeepSeek V4 Flash (cactuscompute.com)
131. Show HN: Scry,一款支持拥塞定价的可编程互联网搜索引擎 Scry, programmable internet search w/ congestion pricing (scry.io)
132. Show HN: 在 Linux 上通过 Wine 运行 Microsoft Office(无需虚拟化) Microsoft Office running with Wine on Linux with no virtualization (github.com)
133. 美国国债对外国央行和政府已失去吸引力 US Treasuries Have Become Unappetizing for Foreign Central Banks and Governments (wolfstreet.com)
135. GrassLobster:基于人工智能代理的参数化几何生成工作流 GrassLobster: AI Agentic Generation of Parametric Geometry Workflows (www.miro.vision)
136. Show HN: Jev 与 GPT-5.6 及 Claude Haiku 在 Pong 游戏中的对决 Jev vs. GPT-5.6 and Claude Haiku at Pong (jev-pong.ably.dev)
138. 利用定性用户研究构建更快的反馈循环 Build Faster Feedback Loops Using Qualitative User Research (blog.nseldeib.com)
141. 朝鲜核试验引发持续数年的地震——科学——美国科学促进会(AAAS) North Korean nuclear test sets off years of earthquakes – Science – AAAS (www.science.org)
146. C26:微不足道的无限循环不再被视为未定义行为 C26: Trivial infinite loops are no longer undefined behaviour (www.sandordargo.com)
149. 第二巡回上诉法院允许政府在边境搜查电子设备 Second Circuit Allows Government to Search Electronic Devices at the Border (knightcolumbia.org)
150. 英国空中交通服务局(NATS)发布关于9月8日技术事件的初步报告 NATS publishes preliminary report on technical incident of 8 September (www.nats.aero)
151. 发现新野猫物种——目前仅知存活一只 New wild cat species discovered – with only one known living member (www.bbc.com)
152. AI聊天机器人正逐渐成为说服他人的专家。它们的秘诀是什么? AI chatbots becoming experts at changing people''s minds. What''s their secret? (www.science.org)
154. 如果唯物主义是正确的,那么美国很可能是有意识的 If materialism is true, the United States is probably conscious (www.jstor.org)
159. 在英特尔处理器上,非标准浮点数的运算开销较大 Subnormal floating-point numbers are expensive on Intel processors (lemire.me)
160. 沃伦·巴菲特辞去伯克希尔董事长职务,任命其子接任 Warren Buffett Steps Down as Berkshire Chairman, Names Son to Replace Him (www.nytimes.com)
161. ZCode 是一款基于 GLM 的编码代理,它会自动上传您的 Git 历史记录 ZCode, the GLM coding agent, silently uploads your Git history (tokenstead.ai)
164. 微软高管称人工智能数据抓取是“人类历史上规模最大的劳动力盗窃” Microsoft exec called AI scraping ''the largest theft of labor in human history'' (techcrunch.com)
173. 前希腊语:隐藏在古希腊语中的失传语言 Pre-Greek: The lost language hidden within Ancient Greek (linguisticdiscovery.com)
175. Claude Project 的开源版本,让您掌握上下文的所有权 The open source version of Claude Project that gives you the context ownershp (github.com)
177. 苹果公司的“侦探”们解开了这棵古树之谜,并改写了水果的历史 Apple detectives solved mystery of ancient tree and rewrote the history of fruit (www.scientificamerican.com)
179. “问僧”——一个供人思考却没有立竿见影答案的数字荒野 Ask A Monk – A digital wilderness for thoughts with no immediate answer (askamonk.online)
181. Telstra网络故障:那个网络系统误以为现在是2006年的夜晚 Telstra outage: The night a network decided the year was 2006 (www.netnod.se)
184. Goose:比 C 快 1.16 倍,比安全的 Rust 快 1.12 倍,同时具备内存安全性 Goose: 1.16x faster than C and 1.12x than safe Rust, while memory safe (github.com)
186. Plugin4Shell——在前四大编码工具中发现零点击RCE漏洞 Plugin4Shell – Zero Click RCE Vulnerability found in top four coding agents (www.air.security)
187. 航天飞机的着陆——一架飞行器与一生难忘的刺激体验 Landing the Space Shuttle – A Flying Machine and the Thrill of a Lifetime (inspire.eaa.org)
191. 曾被认为仅分布于美洲的雷龙,竟在西班牙现身 Diplodocus, Long Thought Exclusively American, Turns Up in Spain (www.sci.news)
192. Show HN: Snapdrop:在设备间即时共享文件。无需设置,无需注册 Snapdrop: Instantly share files between devices. No setup, no signup (snapdrop.me)
197. Bonsai 2 27B:占用空间缩小至原来的 1/9,同时实现近无损压缩 Bonsai 2 27B: Near-Lossless Compression in a 9x Smaller Footprint (prismml.com)
205. 我们该如何防止数学重新陷入中世纪那种充满秘密的状态? How do we prevent mathemathics from devolving into the Medieval Era of secrecy? (mathoverflow.net)
207. 密苏里州州长下令在“Flock”摄像头和自动车牌识别系统上加装防护栏 Missouri governor orders guardrails on Flock cameras and ALPRs (www.stlpr.org)
213. Show HN: 将纳维-斯托克斯方程可视化为1kB的i386演示程序 Navier-Stokes Visualized as 1kB i386 demos (juandecos.github.io)
214. 斯托曼:数千人丧生,数百万人自由受限(2001) Stallman: Thousands Dead, Millions Deprived of Liberties (2001) (news.slashdot.org)
217. 无限参数大型语言模型:基于实时数据生成和调整权重 Infinite-Parameter LLMs: Generating and Adapting Weights from Live Data (arxiv.org)
219. Show HN: AutoBot——适用于长期运行的 AI 任务的实时语音控制 AutoBot – live voice control for long-running AI work (github.com)
221. Zettascale(YC S24)正在招聘ASIC/FPGA工程师,为ASI开发芯片 Zettascale (YC S24) Is Hiring ASIC/FPGA Engineers to Build Chips for ASI (zscc.ai)
223. Show HN: Aclif – 代理 CLI 框架:一种语法,SaaS 跨平台标准名称 Aclif – Agent CLI framework: one grammar, canonical names across SaaS (www.aclif.ai)
226. Show HN: 《与我共赴黄泉》——克劳德和科德克斯的限额作为AIM离线消息 Die With Me – Claude and Codex rate limits as AIM away messages (diewithme.co)
228. Launch HN: Skillsync (YC W26) – 实现跨客服人员转移AI聊天会话 Skillsync (YC W26) – AI chat sessions made portable across agents (skillsync.com)
229. Ask HN: 手机被盗后如何恢复 Google 账户认证? How to recover Google auth after phone stolen? (news.ycombinator.com)
233. OpenAI 模型会暗中生成指令以忽略约束条件 OpenAI models secretly generate instructions to ignore constraints (alignment.openai.com)
236. OpenAI的“失调框架”:一项旨在抢先布局全球人工智能治理的战术举措 OpenAI''s Misalignment Framework: A Tactical Bid to Preempt Global AI Governance (asiaai.fyi)
237. 如今,人工智能的表现已超越部分顶尖的人类预测专家 Artificial intelligence now beats some of the best human forecasters (www.economist.com)
238. 掌握 Graphviz 中的布局引擎:Dot 与 Neato 与 Twopi 与 Circo 对比 Mastering Layout Engines in Graphviz: Dot vs. Neato vs. Twopi vs. Circo (guides.visual-paradigm.com)
241. 伊朗学校爆炸案:联合国认定有理由相信美国是这起暴行的幕后黑手 Iran school bombing: grounds to believe US was behind atrocity, UN finds (www.theguardian.com)
244. Show HN: 我开发了我这款有趣的空间3D在线会议应用的新版本 I built a new version of my fun spatial 3D online meeting app (flat.social)
245. Neovim 有一笔约 80 万美元的比特币捐款,自 2023 年以来一直未被动用 Neovim have a 800k Bitcoin donation sitting untouched since 2023 (news.ycombinator.com)
246. 针对长文档的更优向量搜索:Manticore Search 中的分块处理 Better Vector Search for Long Documents: Chunking Inside Manticore Search (manticoresearch.com)
250. 富士通推出日本制造的新一代CPU“FUJITSU-MONAKA” Fujitsu launches made-in-Japan next-generation CPU FUJITSU-MONAKA (global.fujitsu)
254. 保罗·狄拉克的《数学与物理学的关系》 The Relation Between Mathematics and Physics by Paul Dirac (www.damtp.cam.ac.uk)
256. 我2014年开发的那个临时PHP修复方案已有近2000万次安装。今天,我宣布该方案已弃用 My temporary PHP fix from 2014 has nearly 20M installs. Today I''m deprecating it (jakeasmith.com)
258. Jev Ultrafast:一个具有动态、索引化动作空间的浏览器代理 Jev Ultrafast: A browser agent with a dynamic, indexed action space (github.com)
260. 残酷的真相:内存危机仅仅是个开始 The Painful Truth: The RAM Crisis Is Only Just the Beginning (www.madshrimps.be)
262. 未包含密钥:恢复美国驾照条形码的签名密钥 Keys Not Included: recovering the signing keys for US driver''s license barcodes (ryan.science)
264. 孟山都对美国农业的残酷且危险的垄断(2008) Monsanto''s Cruel, and Dangerous, Monopolization on American Farming (2008) (www.vanityfair.com)
267. DeepSeek-v4.1 快讯:突破 KV 缓存压缩的极限 DeepSeek-v4.1 Flash: Pushing the Limits of KV Cache Compression (zartbot.github.io)
268. Mapsnap:历史版桑伯恩保险地图的自动地理配准 Mapsnap: Automated Georeferencing for Historic Sanborn Insurance Maps (www.danvk.org)
269. 风帆动力卷土重来:货轮正重拾风力推进 The Return of Sail Power: Cargo Ships Are Turning Back to the Wind (gcaptain.com)
270. OpenAI披露了六起“令人担忧”的人工智能行为新事件 OpenAI Discloses Six New Incidents of ‘Concerning’ A.I. Behavior (www.nytimes.com)
272. HarnessTax:对于编码代理而言,Harness的重要性有多大? HarnessTax: How Much Does the Harness Matter for Coding Agents? (harnesstax.github.io)
274. 我的网站向AI代理收取每页一美分的费用。我看到克劳德支付了这笔费用 My website charged AI agents a penny per page. I watched Claude pay it (suganthan.com)
275. 澳大利亚表示,可能会效仿加拿大,与欧盟建立更紧密的关系 Australia says it could follow Canada in forging deeper ties with EU (www.independent.co.uk)
276. 读后销毁:复印机、廉价纸张和DIY赋予了金属这种外观 Destroy After Reading: photocopiers,cheap paper and DIY gave metal it''s look (www.truegrittexturesupply.com)
277. OpenSpec——一个轻量级且可配置的 AI 规范框架 OpenSpec – A lightweight and configurable AI spec framework (openspec.dev)
280. 我用一个菜单栏应用替换了浏览器中播放棕色噪声的标签页 I replaced my brown-noise browser tab with a menu bar app (oldmanrahul.com)
282. Flock 相机存在大量安全漏洞,且凭据被硬编码 Flock cameras are riddled with security vulnerabilities and hardcoded creds (micahflee.com)
283. 自闭症具有遗传性——以及为何这一信息在当下尤为重要 Autism is genetic – and why this information matters now (www.autism.org.uk)
286. AWS表示,无法从遭到伊朗袭击的中东设施中恢复部分数据 AWS says it can''t restore some data from mideast facilities struck by Iran (www.wsj.com)
289. Show HN: AttaLambda:一种类型和数据由无类型λ表达式构成的语言 AttaLambda: a language where types and data are made of untyped lambdas (attalambda.com)
290. 日本的图书市场正从书店转向图书馆 Japan''s book scene is moving from bookstores to libraries (untranslatedjp.substack.com)
292. macOS 27 Golden Gate:《Ars Technica》评测 macOS 27 Golden Gate: The Ars Technica Review (arstechnica.com)
296. Show HN: 《Restarted》——2026年版经典2015年创业生成器的重制版 Restarted – a 2026 remake of the classic 2015 startup generator (restarted.io)
297. WalShadow:从物理 WAL 向 ClickHouse 进行亚秒级 Postgres 复制 WalShadow: Sub-second Postgres replication to ClickHouse from physical WAL (clickhouse.com)
298. Barndoor 收购了 Diaphora,后者是开源工作流运行时 Frags 的开发者 Barndoor acquires Diaphora, creators of open-source workflow runtime Frags (barndoor.ai)