1. OpenAI Codex 面向所有用戶推出基於 API 使用量的計費模式 OpenAI Codex Moves to API Usage-Based Pricing for All Users (startupfortune.com)
2. Anthropic的新模型過於危險,不宜公開發布 New Anthropic model is too dangerous to release publicly (www.nbcnews.com)
3. Gemma 4 E4B 與 Gemma 家族:涵蓋 8 套任務集的企業級基準測試 Gemma 4 E4B vs. Gemma Family: Enterprise Benchmark Across 8 Task Suites (aiexplorer-blog.vercel.app)
8. 勞裡·柯克證明了你的內存(RAM)在欺騙你。於是,我們在家裡搭建了Palantir Laurie Kirk proved your RAM lies to you. So we built Palantir at home (github.com)
12. 曾是“超級MAGA”支持者的特朗普擁躉在Truth Social上對伊朗問題怒不可遏 Once ''Ultra MAGA'', Trump Supporters Fume About Iran on Truth Social (www.nytimes.com)
13. 根據產品可維修性對筆記本電腦和手機廠商進行評分 Grading laptop and cell phone companies on the fixability of their products (pirg.org)
15. AI 只用了 12 分鐘就搞定了。而我花了 10 個小時才修好它 AI Did It in 12 Minutes. It Took Me 10 Hours to Fix It (idiallo.com)
18. Agent Brain——面向人工智能代理的7層認知記憶系統(開源) Agent Brain – 7-layer cognitive memory for AI agents (open source) (github.com)
21. 遞歸移動多項式迴歸——O(1) 常數複雜度 Recursive Moving Polynomial Regression – O(1) Constant Complexity (zenodo.org)
22. AMD人工智能總監表示,自更新以來,Claude Code 變得越來越笨拙且懶散 AMD AI director says Claude Code is becoming dumber and lazier since update (www.theregister.com)
25. 這款MRI設備能凍結腫瘤,幫助患者擺脫令人衰弱的疼痛 MRI machine that''s freezing tumors and saving patients from debilitating pain (www.9news.com.au)
28. 約翰迪爾將支付9900萬美元,達成具有里程碑意義的“維修權”和解協議 John Deere to Pay 99M in Monumental Right-to-Repair Settlement (www.thedrive.com)
30. Anthropic的人工智能將協助蘋果發現iOS、macOS和Safari中的漏洞 Anthropic''s AI to Help Apple Find iOS, macOS, and Safari Vulnerabilities (www.macrumors.com)
33. Cursor AI 關於 61GB 內存故障及後續論壇隱形封禁的說明 Cursor AI Documentation of a 61GB RAM failure and subsequent forum shadowbanning (github.com)
35. 目前正有一場針對 Rust 倉庫所有者的網絡釣魚攻擊活動 An active phishing campaign is targeting Rust crate owners (social.rust-lang.org)
36. AI 代碼正在蠶食開源生態,而維護者卻對此視而不見 AI Code Is Hollowing Out Open Source, and Maintainers Are Looking the Other Way (www.quippd.com)
38. Comad World – 6 個 AI 代理根據一份 YAML 配置構建知識圖譜 Comad World – 6 AI agents build a knowledge graph from one YAML config (github.com)
39. 伊朗停火:以色列加大對黎巴嫩的空襲力度,伊朗封鎖海峽 Iran Ceasefire: Israel intensifies Lebanon strikes and Iran closes the strait (apnews.com)
40. 美國汽車製造商表示,歐盟擬議的法規可能會阻礙大型皮卡進入歐洲市場 US carmakers say proposed EU rules could block large pickup trucks from Europe (www.reuters.com)
43. 小型無人機襲擊造成的平民傷亡人數創歷史新高:我製作了一張地圖 Small Drone Attacks Are Killing More Civilians Than Ever Before: I Made a Map (little-flying-robots.ghost.io)
44. YouTube開始向電視觀眾展示90秒的不可跳過廣告 YouTube starts showing 90-second unskippable ads to TV viewers (9to5google.com)
46. 基於SFP模塊的光纖PCIe及附贈的PCIe分析儀 PCIe over Fiber with SFP Modules and Bonus PCIe Analyzer (www.youtube.com)
47. 我開發了一個在網絡中不可見的 OpenClaw 網狀網絡編排應用程序 I created a network-invisible OpenClaw mesh orchestration app (github.com)
48. macOS 網絡工具 Little Snitch 現已登陸 Linux 平臺 Little Snitch, the macOS network tool, is now available on Linux (www.omgubuntu.co.uk)
50. 《紐約時報》可能已經找到了比特幣創始人中本聰 The New York Times may have found Bitcoin founder Satoshi Nakamoto (finance.yahoo.com)
54. Meta 發佈了其耗資巨大的超級智能團隊研發的首個 AI 模型 Meta unveils first AI model from costly superintelligence team (www.msn.com)
59. Anthropic的新AI模型Mythos究竟有多危險? How dangerous is Mythos, Anthropic''s new AI model? (www.economist.com)
60. 在美國與伊朗達成停火協議後,以色列在黎巴嫩造成254人死亡 Israel kills 254 in Lebanon after US-Iran agree ceasefire (www.aljazeera.com)
61. Linux玩家並沒有做錯什麼,但他們可能會為Windows盜版買單 Linux gamers didn''t do anything wrong, but they might pay for Windows piracy (www.xda-developers.com)
63. 面向軟件開發者的 USB 指南:用戶空間 USB 驅動程序編寫入門 USB for Software Developers: An introduction to writing userspace USB drivers (werwolv.net)
66. AI 編碼代理是一種新型的貢獻者,而 Git 並非為此而設計 The AI coding agent is a new kind of contributor, and Git isn''t made for it (agent-undo.com)
67. Jitter – 在 Google 文檔中無需共享內容即可證明作者為真人 Jitter – Prove human authorship in Google Docs without sharing content (scalisos.com)
69. 2026年第一季度,科技行業裁員近8萬人 Tech industry lays off nearly 80k employees in the first quarter of 2026 (www.tomshardware.com)
71. 在歐洲,部分AMD顯卡的價格在不同零售商之間存在35%的價差,而Nvidia的價差則為15%至20%。 Some AMD GPU prices have 35% cross-store gaps in Europe vs. 15-20% for Nvidia (www.pricesquirrel.com)
73. 北美和南美河流流域及分水嶺的交互式地圖 An interactive map of river basins and watersheds of North and South America (python-maps-vis.vercel.app)
75. Pine Trees——一款專為 Claude 實例設計的 1,400 行私有反思框架 Pine Trees – A 1,400-line private reflection harness for Claude instances (github.com)
76. 92% 的 MCP 服務器存在安全問題(以及如何解決) 92% of MCP servers have security issues (and how to fix it) (github.com)
78. 軟件支持是否才是現代電子產品的真正“保質期”? Is software support the real expiration date of modern electronics? (borisljevar.substack.com)
79. 為什麼微軟對Windows控制面板的“戰爭”拖了這麼久 Why Microsoft''s war on Windows'' Control Panel is taking so long (www.theverge.com)
80. 伊朗要求在停火期間通過霍爾木茲海峽的船隻支付比特幣費用 Iran demands Bitcoin fees for ships passing Hormuz during ceasefire (www.ft.com)
81. Gym-Anything:將任何軟件轉變為代理環境 Gym-Anything: Turn Any Software into an Agent Environment (cmu-l3.github.io)
82. 為什麼一些加州房主覺得自己被困在想搬離的房子裡 Why some California homeowners feel trapped in houses they want to leave (www.sfgate.com)
83. 致 Google Cloud:你們的棄用政策正在自掘墳墓(2020) Dear Google Cloud: Your Deprecation Policy Is Killing You (2020) (steve-yegge.medium.com)
84. 埃隆·馬斯克在訴訟中尋求罷免OpenAI首席執行官山姆·阿爾特曼 Elon Musk seeks ouster of OpenAI CEO Sam Altman as part of lawsuit (www.cnbc.com)
85. 各國及國際組織“必須緊急介入”,阻止特朗普抹殺 All Nations, World Bodies ''Must Urgently Intervene'' to Stop Trump from Wiping (www.commondreams.org)
87. 用於多智能體人工智能編碼工作流的物理“輸入控制檯” A physical "Enter console" for multi-agent AI coding workflows (blog-whi-sper.vercel.app)
88. 嵌入截斷是人工智能記憶檢索中的瓶頸 Embedding truncation is the bottleneck in AI memory retrieval (open.rankfor.ai)
92. 從字節碼到字節:自動生成魔術數據包 From bytecode to bytes: automated magic packet generation (blog.cloudflare.com)
93. 一個確定性工作流編譯器 – 將大型語言模型(LLM)的意圖轉換為表單、文檔和 API A deterministic workflow compiler – turns LLM intent into forms, docs, and APIs (web.geniesnap.com)
99. 亞馬遜通過停產舊款電子書閱讀器,向忠實的Kindle用戶表達謝意 Amazon rewards loyal Kindle devotees by closing the book on old e-readers (www.theregister.com)
104. AWAF——一個用於評估人工智能代理生產就緒度的開放框架 AWAF – An open framework for scoring AI agent production readiness (awaf.ai)
105. QVeris:AI 代理通過單一協議發現、檢查並調用 1 萬種能力 QVeris: AI agents discover, inspect and call 10k capabilities via one protocol (qveris.ai)
107. 我們對 Claude Code 進行了審計,並通過環境變量注入確認了遠程代碼執行漏洞 We audited Claude Code and confirmed RCE via environment variable injection (audited.xyz)
108. 對所有公開的天文數據進行系統性蟲洞搜索 A systematic search for wormholes across all public astronomical data (www.blankline.org)
109. Updu:輕量級、自主託管的運行時間監控工具,僅需一個二進制文件 Updu: Lightweight, self-hosted uptime monitoring in a single binary (github.com)
110. LLM 通過結構化的“智能感知”運行了一款 8 位《Commander X16》遊戲 LLM plays an 8-bit Commander X16 game using structured "smart senses" (pvp-ai.russell-harper.com)
111. 印度建造的500兆瓦電核快中子增殖反應堆首次達到臨界 500 MWe Nuclear Fast Breeder Reactor built by India attains first criticality (www.pib.gov.in)
114. 從零開始用 C 實現 LLM 推理引擎——為何輸出令牌的成本要高出 5 倍 LLM inference engine from scratch in C – why output tokens cost 5x (www.anirudhsathiya.com)
115. 研究:社區環境可能在細胞層面加速衰老 Study: Neighborhood conditions may be driving aging at the cellular level (www.nyu.edu)
116. 比爾·蓋茨將就埃普斯坦案在眾議院監督委員會作證 Bill Gates to testify before House Oversight in Epstein probe (www.politico.com)
117. 聯邦調查局:去年美國人因網絡犯罪遭受的損失創下210億美元的紀錄 FBI: Americans lost a record 21B to cybercrime last year (www.bleepingcomputer.com)
119. Uberbrain – 開源光子類腦計算架構 (CC0) Uberbrain – Open-source photonic neuromorphic computing architecture (CC0) (github.com)
120. Anthropic 最強大的 AI 模型 Mythos Preview 過於危險,不宜發佈 Anthropic''s most powerful AI model Mythos Preview is too dangerous for release (www.euronews.com)
121. 從零開始編寫LLM,第32i部分——干預措施:噪聲中包含什麼? Writing an LLM from scratch, part 32i – Interventions: what is in the noise? (www.gilesthomas.com)
122. SpecLock – 您的 Claude.md 只是建議。這使其成為規定 SpecLock – Your Claude.md is a suggestion. This makes it a law (github.com)
123. 生活的光明面:樂觀主義與痴呆症風險 The Bright Side of Life: Optimism and Risk of Dementia (agsjournals.onlinelibrary.wiley.com)
124. 美國各城市正逐步淘汰Flock Safety監控技術 US cities are axing Flock Safety surveillance technology (www.cnet.com)
125. MegaTrain:在單張 GPU 上對擁有 1000 億以上參數的 LLM 進行全精度訓練 MegaTrain: Full Precision Training of 100B Parameter LLMs on a Single GPU (arxiv.org)
130. 破解谷歌客服系統:洩露通話記錄並識別客服人員身份 Hacking Google Support: Leaking call logs and deanonymising agents (michaeldalton.au)
135. 一個包含 20 個 Elixir/Phoenix 專用代理的 Claude Code 插件 A Claude Code plugin with 20 specialist agents for Elixir/Phoenix (github.com)
136. 針對 MCP 倉庫的 GitHub Actions 供應鏈攻擊剖析 Anatomy of a GitHub Actions Supply Chain Attack Targeting MCP Repos (www.wshoffner.dev)
137. 蘋果的芯片是新格局的核心,但其最大的勝利在於Windows Apple''s chips are the core of a new landscape, but its biggest win is Windows (www.theregister.com)
144. 蘋果正商討擴大Mac Neo產量,因其銷量超出預期 Apple in Talks to Boost Mac Neo Production as Sales Exceed Expectations (www.culpium.com)
148. “史詩怒火”行動加劇跨大西洋裂痕:美歐在安全問題上存在分歧 Operation "Epic Fury" Deepens Transatlantic Rift: U.S., EU at Odds over Security (militarnyi.com)
149. 為 Thunderbird 客戶端編寫了一個擴展程序,以幫助防範網絡釣魚 Wrote an extension for the Thunderbird Client to help protect from Phishing (addons.thunderbird.net)
153. 男子購買黑客工具用於攻擊妻子和朋友 Men Are Buying Hacking Tools to Use Against Their Wives and Friends (www.wired.com)
154. 人工智能解決數學猜想取得突破:北京大學團隊的探索 Breakthrough in AI Solving Math Conjectures: Peking Univ. Team''s Exploration (chinaresearchcollective.substack.com)
155. Earendil Inc. 宣佈收購 Pi 並推出 Lefos Earendil Inc. announcing Pi acquisition and Lefos (earendil.com)
158. 它變得更蠢了嗎?根據HN評論追蹤Claude Code和Codex Did it get dumber? Tracking Claude Code and Codex according to HN comments (diditgetdumber.com)
159. Defender – 針對 AI 代理的本地提示注入檢測(不涉及 API 調用) Defender – Local prompt injection detection for AI agents (no API calls) (www.npmjs.com)
160. 什麼是“Cyberdeck”:認識Z世代的新寵DIY What Is a ''Cyberdeck'': Meet Gen Z''s New DIY Obsession (www.newsweek.com)
165. 臺灣在野黨領袖罕見訪華,呼籲和解 Rare China visit, Taiwan''s opposition leader calls for reconciliation (www.aljazeera.com)
168. 克里斯·拉特納:如今每位工程師都是管理者 [視頻] Chris Lattner: Every Engineer Is a Manager Now [video] (www.youtube.com)
169. 電網儲能成本現已足夠低廉,足以釋放印度太陽能的全部潛力 Grid Battery storage is now cheap enough to unleash India''s full solar potential (ember-energy.org)
171. OpenAI Codex 的周活躍用戶數達到 300 萬,較不到一個月前的 200 萬有所增長 OpenAI Codex reaches 3M weekly active users, up from 2M in under a month (twitter.com)
178. 志願者們將一位樂迷錄製的10K演唱會音頻整理成了一座在線寶庫 Volunteers turn a fan''s recordings of 10K concerts into an online treasure trove (apnews.com)
180. 蘋果用於 MacBook Neo 的 A18 Pro 芯片庫存告急 Apple is running out of A18 Pro chips for the MacBook Neo (www.tomsguide.com)
184. 由於 Next.js 中一個規範標籤(canonical tag)的錯誤,我白白浪費了三週的 SEO 優化時間 I lost 3 weeks of SEO because of a canonical tag bug in Next.js (www.learncodeguide.com)
185. OpenAI 加大力度發展文本模型,戰略轉向超級應用計劃 OpenAI Doubling Down on Text Models, Shifting Strategies to Superapp Plan (www.bigtechnology.com)
186. Aether – 從任意 URL 自動提取實體並構建知識圖譜 Aether – Auto-extract entities and build a knowledge graph from any URL (github.com)
187. Škoda DuoBell:一款能穿透降噪耳機的自行車鈴 Škoda DuoBell: A bicycle bell that penetrates noise-cancelling headphones (www.skoda-storyboard.com)
188. 英國利用公共數據推動人工智能發展的宏偉計劃面臨重重困難 UK''s grand plan to fuel AI with public data faces uphill battle (www.theregister.com)
189. 美國在37天內發射了1000枚JASSM巡航導彈。洛克希德公司每年生產396枚。 US fired 1k JASSM cruise missiles in 37 days. Lockheed makes 396 per year (www.shatterbelt.co)
190. 一件青銅時代盾牌時隔230餘年首次在蘇格蘭展出 Bronze Age shield shown in Scotland for first time in more than 230 years (news.sky.com)
193. Claude AI 服務中斷:聊天機器人離線,Anthropic用戶遭遇錯誤提示 Claude AI down: Anthropic users hit with errors as chatbot goes offline (www.the-independent.com)
194. 白皮書:《合成烴成本低於開採原油之路》(2022) Whitepaper: Road to synthetic hydrocarbons cheaper than drilled oil (2022) (caseyhandmer.wordpress.com)
195. 在人工智能代理盛行的時代,GitOps為何依然重要(FluxCon)[視頻] Why GitOps still matters in a world of AI agents (FluxCon) [video] (www.youtube.com)
198. 這些驚豔的藝術作品是由一位來自西班牙的92歲老奶奶用MS Paint創作的 These Stunning Artworks Were Created in MS Paint by a 92 Yr Grandma from Spain (digitalsynopsis.com)
202. Yu – 在不洩露任何憑證的情況下,將您的 Claude Code/Codex 置於沙箱環境中 Yu – Sandboxes your Claude Code/Codex with zero credential exposure (blog.dreambubble.ai)
207. 詐騙分子是如何進行虛假來電的?(分步詳解) How Scammers Make Fake Calls? (Step-by-Step Explained) (www.securitynewspaper.com)
211. 非洲的人力資本——比爾·蓋茨致《經濟學人》的信 Africa''s Human Capital – Bill Gates Letter to The Economist (economist.com)
212. 歐盟與英國之間的移民流動(自英國脫歐以來) EU Migration to and from the UK (Since Brexit) (migrationobservatory.ox.ac.uk)
213. Anthropic Mythos 模型能夠發現並利用零日漏洞 Anthropic Mythos model can find and exploit 0-days (www.theregister.com)
214. Collabora Office 技術委員會——首次會議 Collabora Office Technical Committee – first meeting (forum.collaboraonline.com)
216. 核邊緣政策通常行得通。但這也十分危險 Nuclear brinkmanship usually works. It''s also dangerous (www.natesilver.net)
219. 經濟內戰:各州推動立法,為石油公司免責 Economic Civil War: States Push Laws to Shield Oil Companies from Accountability (www.propublica.org)
224. 日本放寬隱私法,力爭成為“最適合開發人工智能的國家” Japan relaxes privacy laws to make itself the ''easiest country to develop AI'' (www.theregister.com)
227. 在一處巨大的園區內,1萬名員工曾向全球各地的人們實施詐騙 Inside a huge compound where 10k workers scammed people globally (apnews.com)
230. 俄羅斯人藉助貓糧自動投餵器、分類廣告和吸塵器來保持聯繫 Russians turn to cat feeders, classifieds, and vacuum cleaners to stay connected (meduza.io)
233. Optinum——能發現AI編碼代理在公關測試中系統性遺漏的盲點 Optinum – finds the blind spots AI coding agents systematically miss in PR tests (github.com)
240. 亞歷山大·弗裡德曼與現代宇宙學的起源(2012) Alexander Friedmann and the origins of modern cosmology (2012) (physicstoday.aip.org)
241. 為什麼波斯灣的石油和天然氣儲量比地球上任何其他地方都多 Why the Persian Gulf has more oil and gas than anywhere else on Earth (theconversation.com)
242. Vision Pro 的發佈如何加劇了蘋果內部的緊張局勢 How the Vision Pro Rollout Inflamed Tensions at Apple (www.wired.com)
244. DevGlish – 我把“Kubernetes”唸錯兩年了。於是我就做了這個 DevGlish – I mispronounced "Kubernetes" for 2 years. So I built this (devglish.com)
247. 中本聰是誰?我揭開比特幣創造者真面目的探索之旅 Who Is Satoshi Nakamoto? My Quest to Unmask Bitcoin''s Creator (www.nytimes.com)
248. Anthropic 可能有意洩露了 Claude 的源代碼 Anthropic may have leaked Claude Code source on purpose (www.ft.com)
250. ContextSync – 通過 Obsidian/OneDrive 同步 VS Code 的 AI 上下文 ContextSync – Sync VS Code AI Context via Obsidian/OneDrive (marketplace.visualstudio.com)
251. 一站式查詢:將一家60人的公司整合到一個Slack機器人中 One Brain to Query: Wiring a 60-Person Company into a Single Slack Bot (merylldindin.com)
254. 《DayZ》開發者談1.29版本服務器性能更新 [視頻] DayZ devs talk 1.29 server performance update [video] (www.youtube.com)
255. 用於評估人工智能在平面設計任務中表現的綜合基準測試 Comprehensive Benchmark for Evaluating AI on Graphic Design Tasks (arxiv.org)
258. AI 代理與自主性之間的所有障礙——一份實用指南 Every Barrier Between AI Agents and Autonomy – A Practical Map (vibeagentmaking.com)
259. LLM 爬蟲機器人正在導致 acme.com 的 HTTPS 服務器過載 LLM scraper bots are overloading acme.com''s HTTPS server (acme.com)
260. 肯尼迪設定9月為最後期限,以查明美國自閉症發病率上升的原因(2025年) Kennedy sets September deadline to identify cause of rising US autism rates (2025) (www.reuters.com)
261. 文件顯示,匈牙利與俄羅斯達成了一項旨在加強雙邊關係的12點計劃 Hungary and Russia struck 12-point plan for closer ties, documents show (www.politico.eu)
263. 致幻劑引發的視覺幻覺與精神分裂症 Visual Hallucinations in Psychedelics vs. Schizophrenia (psychedelirium.substack.com)
264. RNA條形碼可實現大腦連接的高速繪製 RNA barcodes enable high-speed mapping of connections in the brain (news.illinois.edu)
267. 針對Booking.com酒店及客戶的“我付了兩筆錢”網絡釣魚活動 Phishing Campaigns "I Paid Twice" Targeting Booking.com Hotels and Customers (blog.sekoia.io)
271. OpenAI表示其新模型GPT-2過於危險,不宜發佈(2019年) OpenAI says its new model GPT-2 is too dangerous to release (2019) (slate.com)
274. 緬因州即將成為首個禁止新建數據中心的州 Maine Is About to Become the First State to Ban New Data Centers (www.wsj.com)
276. “人類社會本不該如此運作,”美國參議員表示 ''This Is Just Not How the Human Race Should Operate,'' Says US Senator (www.commondreams.org)
277. 規模更大、可訓練性更強的語言模型可靠性反而更低 Larger and more instructable language models become less reliable (pmc.ncbi.nlm.nih.gov)
280. 品牌紛紛採用“無AI”聲明,以在良莠不齊的市場中脫穎而出 Brands Adopt ''No AI'' Disclaimers to Stand Out Amid the Slop (www.wsj.com)
281. 保險公司持有的1萬億美元私人信貸規模讓監管機構望塵莫及 Insurers'' 1T Buildup in Private Credit Is Leaving Regulators in Dust (www.wsj.com)
282. “阿耳忒彌斯”號傳回了一些非凡的高質量月球照片 Artemis delivers some exceptional, high-quality photos of the Moon (arstechnica.com)
283. 通過強化學習訓練最先進(SOTA)的漏洞發現代理 Training SOTA vulnerability discovery agents through RL (depthfirst.com)
284. 非技術團隊全面採用 AI 編碼助手的影響 Impact of 100% Adoption of AI Coding Agents by Non-Technical Team (www.kapwing.com)
286. 使用 LM2596T 穩壓器為樹莓派 4B 供電時電壓會下降 Voltage drops when powering a Raspberry Pi4B from a LM2596T regulator (electronics.stackexchange.com)
287. 俄羅斯軍事黑客將英國網民的網絡流量進行了重定向 Russian military hackers reroute British internet users'' traffic (www.ft.com)
288. “白噪音”(基於Nostr的加密群聊)已完成安全審計 White Noise (encrypted group chat over Nostr) completes security audit (leastauthority.com)
290. 馬薩諸塞州議員正尋求推行全美“最嚴格”的社交媒體禁令 Mass. lawmakers looking to usher in ''most restrictive'' social media ban in USA (www.independent.co.uk)
291. AI編程助手技能(Claude Code、Codex、OpenCode、OpenClaw、Factory Droid) AI coding assistant skill (Claude Code, Codex, OpenCode, OpenClaw, Factory Droid (github.com)
293. 俄羅斯和中國否決了旨在重新開放霍爾木茲海峽的聯合國決議 Russia and China veto UN resolution aimed at reopening the Strait of Hormuz (apnews.com)
295. GitHub Copilot CLI 現已支持 BYOK 和本地模型 GitHub Copilot CLI now supports BYOK and local models (github.blog)
296. 零散股:如何判斷一篇文章是否由人工智能撰寫 Odd Lots: This Is How to Tell If Writing Was Made by AI (www.bloomberg.com)
297. 為什麼Anthropic的新模型讓網絡安全專家感到不安 Why Anthropic''s new model has cybersecurity experts rattled (www.platformer.news)
299. AI輔助會降低堅持度,並損害獨立表現 AI Assistance Reduces Persistence and Hurts Independent Performance (arxiv.org)