Raw Sources Index

這一層保存原始來源或忠實擷取後的 raw capture。raw source 是 wiki claims 的 provenance，不應在沒有明確修正理由時任意改寫。

子目錄

• raw/articles/ — 文章、部落格、GitHub README、社群長文擷取
• raw/transcripts/ — YouTube / podcast / speech transcript
• raw/assets/ — 大型附件、完整文字或非 markdown 資產（若存在）
• raw/papers/ — PDF、研究報告與論文的文字擷取

使用規則

• 新來源先進 raw，再決定要更新哪些 entity / concept / comparison 頁
• raw frontmatter 必須包含 OKF v0.2 的 type，並記錄 title、description、tags、resource、status、generated；保留 source_url, ingested, sha256 作為 AI Ark extension
• sha256 以 frontmatter 下方的 body 計算
• 不把 raw source 當成一般 backlink 來源；raw 可能含有抓取噪音

本輪新增

• Diffusion Controller／diffusion 生成控制與 preference alignment：raw/articles/google-research-diffusion-controller-2026-09-29.md — Google Research 介紹 frozen backbone、輕量 side controller、policy-gradient／PPO、reward-weighted regression、gray-box／white-box 與 HPS-v2／human preference 結果；arXiv abstract 已直接讀取，完整論文、實驗表格、模型權重與 benchmark 未重跑，raw-only
• Anthropic IPO 招股書二手整理／營收、虧損與算力承諾：raw/articles/bnext-anthropic-ipo-prospectus-revenue-loss-2026-09-29.md — BusinessNext 轉述 Reuters 所稱 2025 年營收約 46 億美元、營業虧損逾 80 億美元、算力／基礎設施支出約占營運支出 58%、未來相關義務約 5,180 億美元；Reuters 正文、招股書原檔與 Anthropic 第一方財務文件未直接讀取，raw-only
• GPT-6.1 Astra 取消發布、alignment／scope authorization 與 agent control 事件脈絡：raw/articles/bnext-openai-gpt-6-1-astra-release-halted-safety-2026-09-29.md — BusinessNext 與 Reuters 轉述 GPT-6.1 Astra 因內部安全測試問題取消原訂 10 月發布；OpenAI DNS／Hugging Face 第一方事件頁與 Amodei pacing 原文已直接讀取，WSJ 正文、GPT-6.1 第一方安全報告與獨立重現未直接讀取，raw-only
• Claude Sonnet 5.5 發布、速度／成本主張與 coding benchmark：raw/articles/bnext-claude-sonnet-5-5-features-price-performance-2026-09-29.md — BusinessNext 整理 Anthropic Claude Sonnet 5.5 的 30%+ 速度、最多 30% 任務成本下降、Terminal-Bench／CursorBench／GDPval／OSWorld 分數、API 定價與 1M context／128K max output；Anthropic 官方發布頁與模型文件已直接讀取，未獨立重跑 benchmark 或操作 API，raw-only
• AMD 收購 World Labs／空間智慧與 Physical AI：raw/articles/bnext-amd-world-labs-spatial-ai-2026-09-29.md — BusinessNext 整理約 82 億美元全股票交易、李飛飛加入 AMD、World Labs 的 spatial intelligence／world models、3D 環境與 robotics／simulation；AMD、World Labs 第一方公告與 TechCrunch 募資報導已直接讀取，Marble／Atlas／SceniX 細節、併購效益與產品 benchmark 未直接驗證，raw-only
• AI資料中心 Token 經濟與台灣供應鏈：raw/articles/bnext-ai-data-center-token-economy-taiwan-supply-chain-2026-09-29.md — BusinessNext 轉載／整理北美智權報對 MIC 2026 MIC FORUM Fall 的 AIDC「Token 工廠」、Token/W／Token/$、prefill／decode、KV cache、算網融合與聚算／聯算／送智觀察；BusinessNext、北美智權報與 MIC 官方活動入口已直接讀取，MIC 簡報檔、公司第一方產品頁、原始統計與獨立 benchmark 未直接驗證，raw-only
• Amodei〈We Must Pace the Frontier〉與 AI 控制邊界／資安評測事故：raw/articles/bnext-ai-escape-human-control-amodei-pace-frontier-2026-09-28.md — BusinessNext 二手整理 Amodei pacing、recursive self-improvement、OpenAI／Hugging Face 與 Anthropic cybersecurity evaluation incidents；Amodei、OpenAI、Anthropic 第一方頁面已讀取，Google／Irregular 原始報告與完整技術 payload 未直接讀取，raw-only
• Anthropic claude.ai／Claude desktop app 兩週效能衝刺與 agent hill-climbing：raw/articles/bnext-anthropic-claude-ai-3x-faster-performance-sprint-2026-09-28.md — BusinessNext 整理 Anthropic 官方工程文章的量化基準、13 項核心指標、Claude Tag、CI ratchet、feature flags、3,000+ 次修改與人類 steering；Anthropic 官方文章已讀取，精確幾何平均、RuntimeWire 評論、內部 telemetry、留存／營收與產品操作未直接讀取，raw-only
• Claude／ChatGPT 跨平台搬家、記憶／對話／Skills 轉移：raw/articles/bnext-claude-chatgpt-migration-2026-09-28.md — BusinessNext 整理 Claude export、memory import、OpenAI agent setup import、Library、Skills 搬移與方案／裝置限制；Anthropic 與 OpenAI 官方文件已讀取，完整對話無損搬移、Codex Skill ZIP 相容性與產品操作未直接驗證，raw-only
• 東京電玩展 2026／遊戲業生成式 AI 採用與 AI Technology Pavilion：raw/articles/bnext-tgs-2026-ai-game-development-2026-09-25.md — BusinessNext 整理 CESA 85.8%／1,349 份樣本、3D 資產／測試／資料分析／營運導入、生成角色／AI NPC 實驗與玩家接受度；CESA 官方 preview 與 TGS 官方 exhibitor list 已讀取，產品操作、玩家研究、開發 ROI 與完整報告未直接讀取，raw-only
• QNAP World Tour 2026／Edge AI／營運韌性：raw/articles/bnext-qnap-world-tour-edge-ai-operational-resilience-2026-09-25.md — BusinessNext 贊助內容整理 QNAP 的 AI Service Center、Qsirch AI、QuAgent、備份驗證、高可用性、ADRA NDR X 與 71%／53% 企業資安調查數字；QNAP 原始調查、產品文件、第三方評測與實際操作未直接讀取，raw-only
• 日本新創 Uravation 的 62 位 AI 員工／Claude Code／human approval gate：raw/articles/bnext-uravation-62-ai-employees-claude-code-2026-09-25.md — BusinessNext 整理 9 個 AI 經理帶領 53 位 AI 員工、部門化窄任務、對外內容 100% 人類核可，以及 AI Checker 公司統計；Uravation 官網、組織圖原始頁、工具頁與 Claude Code 官方文件未直接讀取，raw-only
• 長篇影片一致性與 multi-agent orchestration：raw/articles/google-research-coherent-long-form-video-generation-2026-09-24.md — Google Research 介紹 AI video co-director、CANVAS、A²RD、VQQA 的統一框架；直接讀取官方文章，ArXiv paper、benchmark、YouTube 影片、code 與模型未讀取，raw-only
• 亞馬遜封鎖 Meta Muse、Shopify 開放代理結帳：raw/articles/bnext-meta-muse-shopify-agentic-checkout-2026-09-24.md — BusinessNext 整理 Amazon 封鎖 Muse 代購、Shopify 宣布使用 Shop Pay 的 agentic checkout，以及廣告／支付／商務入口差異；GeekWire 與 Amazon agentic-shopping 說明已讀取，Shopify 官方公告、Amazon 封鎖原始公告、法院文件與 Bloomberg／WWD 正文未直接讀取，raw-only
• 記憶體供應分層、PC 價格與高階零組件短缺：raw/articles/bnext-acer-memory-ddr4-ddr5-shortage-2026-09-23.md — BusinessNext 轉述宏碁董事長陳俊聖：DDR4／一般 DDR5 供給充裕、DDR4 甚至跌價，較短缺的是高階規格零組件；另整理 PC 2026 Q4 漲價 5%–20%、2027 年中止漲與「科技通膨」說法。MoneyDJ 原文、宏碁／供應商資料與研調機構原始預測未直接讀取，raw-only
• Claude Opus 5.5 發布、API 降價與成本效率評測：raw/articles/bnext-anthropic-claude-opus-5-5-launch-cost-cut-2026-09-23.md — BusinessNext（李先泰，初稿 AI 編撰）整理 Claude Opus 5.5 的發布、input／output 降 20% 與 cache read 降 60% 定價、官方「典型工作負載成本低約 40%」估計、Artificial Analysis Intelligence Index 57.62（max with fallback）居榜首、《Every》7 天實測與資安／生物 safeguard；Anthropic 官方公告與 Artificial Analysis model page 已讀取，《Every》原文、METR 報告與 System Card 未直接讀取，raw-only
• GPT-6 Sol／Luna 模型發布、API 降價與評測：raw/articles/bnext-openai-gpt-6-sol-luna-api-price-cut-2026-09-23.md — BusinessNext 整理 GPT-6 Sol／Luna 的模型定位、每百萬 token 2／10 美元與 0.10／0.50 美元 API 價格、prompt caching、可用入口與官方 benchmark；OpenAI 官方公告與 Artificial Analysis Sol／Luna model pages 已讀取，BusinessNext 轉述的每題成本／正確率細項、TechCrunch／Decrypt 與產品操作未直接讀取，raw-only
• 工程師「按 Enter 作業員」與 AI coding 驗證／治理：raw/articles/bnext-ai-developer-engineer-2026-09-22.md — BusinessNext 整理大型企業工程師對 12–13 小時 AI 生成程式碼工作、缺乏檢查與抓 Bug 時間的社群抱怨，以及 Elon Musk、Chamath、Michael B. Paulson、8090、tinygrad 等回應；X 原始貼文與 Claude Code 官方文件未直接讀取，維持 secondary synthesis／raw-only
• SpaceXAI Grok 4.7／速度、價格與 token 使用量：raw/articles/bnext-spacexai-grok-4-7-speed-price-claims-2026-09-22.md — BusinessNext 整理 Grok 4.7 的發布、API 價格、官方 benchmark、Artificial Analysis Intelligence Index 46、約 240M 評測輸出 token、速度與 task-level cost 邊界，以及 HackerBench／LatchBio 安全說法；SpaceXAI 官方公告與 Artificial Analysis model page／方法已讀取，raw-only
• Meta Muse 個人 AI agent／下載熱度與 Amazon 代購封鎖：raw/articles/bnext-meta-muse-ai-agent-tops-us-app-charts-amazon-blocks-2026-09-22.md — BusinessNext 整理 Muse 的多步驟瀏覽器／跨服務任務、上市六天超過 90.2 萬次下載與美國雙平台免費榜首、Meta Secure VM／Sentinel 安全設計，以及 Amazon 以未授權 AI agent、身分揭露與憑證處理爭議封鎖代購；Meta 官方公告、GeekWire 與 Bloomberg 已讀取，Amazon 官方公告、Sensor Tower 原始數據與 Meta 深入安全文件未直接讀取，raw-only
• OK超商大變身／三商家購收購後形象店改造：raw/articles/bnext-ok-mart-acquisition-new-store-logo-2026-09-21.md — BusinessNext（林育萱整理．編輯，初稿 AI 編撰）依 TVBS、中央社與經濟日報報導整理：板橋大仁店換藍白新店裝與魚形 Logo（據門市人員為全台第一家形象店）、超市化商品結構、約 4.5 億元改造規畫鎖定 200 多家開放型門市、2027 Q2 前完成既有門市主要改造、2027 年整體轉虧為盈目標與 2,200 家長期展店；邱光隆受訪口徑維持 partially-supported，TVBS／中央社／經濟日報原文未讀取（2026-09-21；raw-only）
• LinkedIn《The AI Talent Divide》／AI 人才薪資與分布：raw/articles/bnext-linkedin-ai-talent-divide-2026-09-21.md — BusinessNext 整理 LinkedIn 美國 AI 勞動市場研究：典型 AI 職缺薪資中位數 17.7 萬美元、非 AI 職缺 8 萬美元、AI 職缺自 2023 年約翻倍；千禧世代占 Head of AI 新進人員約六成、Z 世代占熱門技術職三分之二以上，女性占 AI 新進人員 26%，AI 職位 91% 由學士以上學歷者擔任；LinkedIn 官方新聞稿支持核心數字，研究 PDF 明確標示為描述性／相關性分析，未做獨立重算，raw-only
• TypeSafe AI／Jev 結構化決策模型：raw/articles/bnext-typesafe-jev-decision-model-explainer-2026-09-21.md — BusinessNext（李先泰）整理 Jev 的 Choice／Score／真假機率輸出、70–500 毫秒延遲／40–200 倍速度／輸入 token 定價宣稱、agent 監控與 model routing 用途，以及 Claude／Codex＋typesafe-ai skill 上手與 confidence 門檻；TypeSafe 官方文件、TechCrunch、Sam Witteveen 實測、Ruben Hassid 電子報、OpenAI GPT-5.6 Luna 文件與 InstructGPT 論文未直接讀取，維持 article attribution（2026-09-21；raw-only）
• Henrik Zeberg 美股暴跌72%／4階段走勢預測：raw/articles/bnext-henrik-zeberg-us-stock-crash-72-percent-forecast-4-stage-2026-09-18.md — BusinessNext（林美欣，本文出自鉅亨網）整理 Henrik Zeberg 的看空模型：那斯達克100 可能在 2026 年底衝上約 3.9 萬點後暴跌約 72% 回到 2022 年低點、4 階段走勢（blow-off top → Q4 2026 見頂 → 2027 H1 假反彈 → 黑天鵝）、3 項警訊（勞參率／長期失業、房市疲弱、中低收入承壓）與 2 個警戒訊號（DXY 93–94、短債殖利率快速下跌）；Business Insider 原始訪談、鉅亨網原文與 BLS／NAR／Primerica／BEA／CME FedWatch／Fitch 原始資料未讀取，維持文章 attribution（2026-09-18；raw-only）
• Google Research MilleMiglia／middle-mile logistics instance generator：raw/articles/google-research-millemiglia-middle-mile-logistics-instance-generator-2026-09-18.md — C++／Protocol Buffers 產生 middle-mile benchmark instances；space-time graph multi-commodity flow、fixed schedules、throughput limits 與 synchronization constraints 維持 Google Research 部落格 attribution，paper（HAL）、GitHub repo 與 sample instance 未讀取或執行；raw-only
• Andrew Ng／AI 滅絕論與安全工程：raw/articles/bnext-andrew-ng-ai-extinction-fears-science-fiction-2026-09-18.md — BusinessNext 整理 Andrew Ng 在 Bloomberg TV 的說法：AI 滅絕論「科幻多於科學」、資安與 sandboxing 是較務實的工程問題、一概放慢可能連修補速度一起拖慢，並轉述 OpenAI 模型異常行為揭露；Bloomberg 訪談、Amodei 原文、OpenAI 框架與報告未直接讀取，維持 attribution（2026-09-18；raw-only）
• Jeffrey Gundlach AI 股票估值／財政與信用風險警告：raw/articles/bnext-jeffrey-gundlach-ai-stocks-valuation-2026-09-18.md — BusinessNext 整理美股 Shiller CAPE 超過 42 倍、AI 企業債信用利差、循環融資與 Gundlach 將建議股票配置由 40% 降至 30% 的說法；Benzinga／BIS／節目與活動原文未讀取，維持文章 attribution（2026-09-18；raw-only）
• Claude Code 新版「專案」功能／平行工作線（threads）beta：raw/articles/bnext-claude-code-projects-redesign-parallel-threads-2026-09-18.md — BusinessNext（李先泰，初稿 AI 編撰）整理 Anthropic 9/17 Claude Code Projects beta：單一持續對話當協調者、多條雲端工作線平行拆工送 PR、共享專案記憶與資源庫；開放範圍（Pro／Max 先行）、用量上限與「僅雲端執行」限制維持 BusinessNext／官方轉述 attribution，未操作產品；raw-only
• Grok Bot Routine 排程／AI 新聞晨報：raw/articles/bnext-grok-bot-routine-schedule-ai-news-briefing-2026-09-18.md — BusinessNext 整理 Bot、Skill、Routine 的分工與三步驟採用順序：先手動跑通、再存成 Skill、最後設定週一至週五台灣時間 08:00 排程；時區、雲端執行、50 個 Routine、最近 20 次紀錄與刪除行為維持 BusinessNext attribution，未獨立操作產品；raw-only
• Google Research learning interactives／教師主導的 Generative UI 教學框架：raw/articles/google-research-learning-interactives-generative-ui-2026-09-17.md — 以 Gemini、learning objectives、逐層挑戰、AI scaffolding 與 generate-then-refine guardrails 生成客製化 STEM 互動模擬；12 名美國 STEM 教師 usability study、40 個 STEM requests expert rating、3.5%→69.3% critique 通過率與 86% acceptance rate 維持 Google Research tech report 的 protocol-level attribution，學生 learning gains 尚未驗證；raw-only
• SK海力士傳結盟英特爾赴美設廠／2 大可能方案：raw/articles/bnext-sk-hynix-intel-us-memory-chip-production-talks-2026-09-17.md — BusinessNext（李先泰整理．編輯，初稿 AI 編撰）依《路透》9/16 獨家報導與《CNBC》整理：SK海力士與英特爾洽談赴美生產記憶體晶片、租用俄亥俄晶圓廠或與雲端大廠合資兩方案、雙方均稱尚無定案、韓國政府對 HBM／DRAM 敏感技術的態度成最大變數；Reuters／CNBC 原文與 SK海力士官方聲明未讀取維持 attribution（partially-supported），raw-only
• 光鼎（6226）LED→功率半導體轉型／股價一年飆 400%：raw/articles/bnext-para-led-power-semiconductor-transition-2026-09-15.md — BusinessNext（林育萱整理．編輯，初稿 AI 編撰）依光鼎電子、公開資訊觀測站與北科大校友資料整理的公司報導：一年漲幅逾 400%、IGBT／SiC MOSFET 三階段布局、2025 EPS -0.28 元與 2026H1 獲利回穩、營收結構 LED 約 90%／模組約 9%／功率半導體約 1%；股價、財報與市場規模數字維持文章層級 attribution（partially-supported），raw-only
• Bill Gates AI inequality／人類保留區／蓋茲基金會 10 億美元投入：raw/articles/bnext-bill-gates-ai-inequality-human-reserved-gates-foundation-2026-09-16.md — BusinessNext 整理 Bill Gates 最新文章與《目標守護者報告》：AI 恐淪富人專屬工具、4 類初階／中階職缺最先受衝擊、未來兩年至少 10 億美元（教育 40%／醫療 40%／農業 10%／數位基礎 10%）、「人類保留區」與 AI token／機器人課稅；Gates Notes 8/26 長文為既有 raw 之延伸報導，基金會原始報告未讀取維持 attribution（2026-09-16；raw-only）
• Grok Bot Figma 截圖工作流（SpaceXAI 設計師 John Bai）：raw/articles/bnext-grok-bot-figma-screenshot-workflow-2026-09-16.md — BusinessNext（李先泰，初稿 AI 編撰）對 Claire Vo《How I AI》podcast（2026-09-14）的整理：Figma MCP＋兩張截圖＋語音指令交件、五步工作流、「垃圾桶開發法」與 Peng Zheng 個人網站自動化；podcast／原文維持文章 attribution
• Zuckerberg 回應 Amodei「AI減速論」／市場與法律會逼出安全：raw/articles/bnext-zuckerberg-responds-amodei-ai-safety-liability-market-evaluators-2026-09-16.md — BusinessNext（李先泰，初稿 AI 編撰）整理 Zuckerberg 9/15 X 發文首度回應 Amodei〈We Must Pace the Frontier〉：四張牌（市場誘因、法律責任、獨立評估者、算力分配）、Muse 延後上市例證與同日黃仁勳／Altman Dreamforce「自律」表態；X 原文、Bloomberg 報導與 Fox Business 維持文章 attribution，Amodei 三步計畫經既有 raw 第一方原文核對
• Gemini 3.8 Live／Gemini 3.8 Live Extended Thinking 語音代理模型：raw/articles/bnext-gemini-3-8-live-extended-thinking-2026-09-16.md — BusinessNext 整理 Google 兩款近即時 audio-to-audio 模型的 visual input、背景工具／API、同步推理與說話、97 語言、benchmark、API 定價與 rollout；Google 官方文件已核對模型識別、token limits、async function calling 與 pricing，benchmark／SOTA／企業 ROI 仍維持供應商或媒體 attribution（2026-09-16；raw-only）
• Retrieve-for-Train（R4T）RL-compiled diffusion fan-out retrieval：raw/articles/google-research-retrieve-for-train-rl-compiled-diffusion-fan-out-retrieval-2026-09-15.md — Google Research 的 R4T framework：以 Soft-GRPO + composite set-level reward（groundedness／Vendi Score diversity／alignment）訓練 Gemma3-4B／Qwen3-4B fan-out LLM、離線合成 supervision pairs、distill 成 53.9M diffusion retriever 做 single-pass non-autoregressive fan-out；fashion／music benchmark 的 quality improvement 與 12–20× latency speedup 維持來源研究 protocol 層級證據（partially-supported），跨 domain generalization 維持 unresolved
• 漢測（7856）虧損翻身、探針卡／矽光子／CPO 成長路線：raw/articles/bnext-han-test-7856-probe-card-cpo-2026-09-14.md — BusinessNext（林育萱）整理漢測從工程服務建立現金流、延伸至探針卡與 AI／HPC 測試需求的轉型；公開申購、財務數字與公司展望維持文章層級證據，漢測官方網站與 ASML 新聞稿未在本輪直接讀取
• 振宇五金（ALD Hardware）1→98 家門市成長策略：raw/articles/bnext-ald-hardware-zhenyu-hardware-growth-strategy-2026-09-14.md — BusinessNext（陳祈安，初稿 AI 編撰）依《直擊新商業》YouTube 訪談整理的振宇五金董事長曾善／副總洪珮芳口徑：社區店型、小包裝零件銷售、自有品牌營收占比近六成、AGV 倉儲省 >70% 人力、2025 花蓮洪災 1,500 台高壓清洗機調度與 5 年 150 家目標；企業第一方數據維持 partially-supported，YouTube 影片 transcript 未讀取（unresolved）
• 川普 All-In Summit 擴音來電黃仁勳／AI 毀滅論「騙局」：raw/articles/bnext-trump-huang-all-in-summit-call-ai-doomer-hoax-2026-09-15.md — BusinessNext（李先泰，初稿 AI 編撰）對 2026-09-14 川普在 All-In Summit 擴音來電、稱 AI doomer 論為 HOAX／資料中心是未來 25 年石油的事件報導；Amodei pacing 三步計畫經既有 raw 第一方原文核對，Truth Social／Bloomberg／All-In Podcast 影片／Altman・Sacks X 貼文維持文章 attribution（partially-supported）
• Codex 入門／AI 工作代理人專案資料夾、權限、Plugins／Skills 與一週練習路線：raw/articles/bnext-codex-ai-work-agent-guide-2026-09-15.md — BusinessNext（李先泰，初稿 AI 編撰）依 Codex、OpenAI Academy 與 Riley Brown 課程內容整理的 Codex 新手工作流教學；專案資料夾、檔案產出、Prompt 三原則、多聊天分工與七天練習路線維持文章層級證據，官方來源本輪未逐一核對
• ChatGPT／Claude 對話隱私 9 大防線＋4 步驟：raw/articles/bnext-chatgpt-claude-privacy-9-stops-opt-out-training-2026-09-14.md — BusinessNext（李先泰，初稿 AI 編撰）依 OpenAI／Anthropic 官方資料使用文件整理的對話隱私指南；Navier–Stokes／Buckmaster 背景與 Mark Chen X 引述、9 關卡框架、4 步驟操作建議與 28 項來源清單可驗證，9 張互動卡片逐卡內容因 JS 渲染未取得（unresolved）
• Codex 快捷鍵教學／Quick Chat（⌘⌥N）情境分工與額度池分析：raw/articles/bnext-codex-keyboard-shortcuts-quick-chat-guide-2026-09-14.md — BusinessNext（李先泰，初稿 AI 編撰）依 OpenAI 官方快捷鍵頁整理的 Codex/ChatGPT 桌面 App 10 組 macOS 預設快捷鍵指南；10 組快捷鍵與「Codex only／ChatGPT only」標註經 OpenAI〈Commands〉官方頁逐條核對，額度共用池表述經 Help Center 核對，Option+Space→Pets 改動與 Review Git 要求維持 attribution
• AI 巨頭 CEO「放慢前沿」共識（Amodei／Altman／Musk／Sacks）：raw/articles/bnext-ai-ceos-pace-the-frontier-amodei-altman-musk-sacks-2026-09-14.md — BusinessNext 對 Amodei〈We Must Pace the Frontier〉三步 pacing 計畫（嵌入式評估者、民主國家協調、全球協調）、Altman／Musk X 表態與 Sacks 反方回應的整理；METR OAI-HF 調查數字經第一方原文核對，風險推演與動機分析維持 attribution
• TTA South 三家新創（Handy Cloud／愛寶 EirBot／CanWellBeing）：raw/articles/bnext-tta-south-three-startups-handycloud-eirbot-canwellbeing-2026-09-14.md — BusinessNext 對 Meet Greater South TTA South 展區三家新創的報導：模組化雲端管理與 API/MCP、院內 RAG＋地端運算 AI 照護機器人、AI 身體組成定量系統（3 秒 vs 20–40 分鐘宣稱），以及導入案例與法規計畫的證據界線
• 德國凱馳 KIRA BR 50 智慧自主洗地機器人（sponsored content）：raw/articles/bnext-karcher-kira-br-50-autonomous-floor-scrubber-2026-09-12.md — BusinessNext（sponsored by 德國凱馳）對 KIRA BR 50 的人機協作定位、2,365 平方公尺／小時效能宣稱、點陣雲隱私設計與 AES-256 加密、信實集團機場場域驗證，以及缺工背景數據的證據界線
• 國泰世華 CUBE App 小額跨境匯款（sponsored content）：raw/articles/bnext-cathay-cube-cross-border-remittance-2026-09-11.md — BusinessNext（sponsored by 國泰世華）對 CUBE App 小額跨境匯款的產品與合規報導：八個國家／地區收款範圍、免外幣帳戶／約定帳號、24 小時 AML 機制，以及「上線五個月交易量成長近五倍」vendor claim 的證據界線
• Claude API 成本優化／prompt caching・prompt audit・effort・Message Batches：raw/articles/bnext-claude-api-cost-optimization-prompt-cache-effort-guide-2026-09-11.md — BusinessNext 依 Anthropic 官方文章與平台文件整理三條成本槓桿、14.6%／5.3% prompt audit vendor result、90.5%／78.6% held-out benchmark 與 50% Batch API 折扣的證據界線
• GPT Image 2.5 官方生圖指南／22 組提示詞＋萬用公式：raw/articles/bnext-gpt-image-2-5-prompt-guide-22-prompts-2026-09-11.md — BusinessNext 依 OpenAI 官方 image prompting guide 重新整理的 8 條撰寫原則、22 組生成／編輯／多輪提示詞與萬用公式；「速度提升近50%」維持 vendor claim attribution，繁中改寫與模板化濃縮維持 editorial observation
• Grok Bot 七周上線／Roman Ugarte 五大產品決策：raw/articles/bnext-grok-bot-seven-weeks-five-product-decisions-2026-09-11.md — BusinessNext 對 Lenny’s Podcast（2026-09-08）訪談整理：Grok Bot 從零到公開約七周、不做進 Cursor、獨立小團隊隔離開發、全雲端且每個 bot 一台電腦、核心成員親手 onboard 兩三百人與上線前反向出貨砍功能，以及 colleague-pilled（同事腦）判準的證據界線
• ToolGrad／answer-first tool-use dataset generation：raw/articles/google-research-toolgrad-tool-use-dataset-generation-2026-09-10.md — Google Research 的 textual gradients、API workflow 迭代建構、ToolGrad-500、BFCL 結果與 code／dataset／model 公開入口之證據界線
• Z 世代「職涯極簡主義」／Glassdoor・Harris Poll 調查數字：raw/articles/bnext-career-minimalism-gen-z-2026-09-10.md — BusinessNext（轉載經理人月刊）整理 career minimalism 敘事、68% 拒任管理職與 57% 副業世代比較，以及 Jotform CEO Tank 留才四做法的證據界線
• 凱鈿行動科技（KDAN）AI 轉型／部門 Agent 化：raw/articles/bnext-kdan-mobile-ai-infrastructure-transformation-2026-09-10.md — BusinessNext 對董事長蘇柏州的專訪：AI coding 危機感與自述、文件數位化作為企業知識庫第一步（IDP 案例）、AISO 主權 AI 聯盟參與、部門 Agent 化四層架構與 SDK/API「純高毛利」商業模式的證據界線
• 中央社新聞 MCP／500 萬則新聞接入 AI：raw/articles/bnext-cna-mcp-news-database-2026-09-10.md — 中央社 MCP 訂閱制、近 500 萬則新聞稿與 350 萬張照片等資料範圍、個人月費 200 元、內部 500 題新聞問答評測，以及零點擊／Pew 研究數字與媒體 AI 入口策略的證據界線
• Google TPU 專用化／台廠供應鏈機會：raw/articles/bnext-google-tpu-custom-chip-taiwan-supply-chain-2026-09-10.md — Amin Vahdat SEMICON Taiwan 2026 演說與媒體訪問的二手整理：算力／電力／記憶體動態瓶頸、TPU 8i／8t 推論訓練分版、CPU 專業化痛點與資料移動／動態散熱硬體機會，台廠客製化協作解讀維持 partially-supported
• AGI／台股 ETF／半導體供應鏈：raw/articles/bnext-agi-etf-00947-00904-00913-0052-2026-09-10.md — OpenAI AGI 定義、BusinessNext 對 CMoney ETF 績效表的轉述，以及 00904／00913／0052 的指數／權重差異與投資敘事證據界線
• Jacob Coxon／AI safety 警告與 agent 越權事件：raw/articles/bnext-anthropic-researcher-coxon-ai-safety-warning-2026-09-10.md — Anthropic 研究員離職、self-improving superintelligence 警告、METR／AISI 事件、CoT monitoring 與 Sanders／Casar 監管提案的證據界線
• UiPath 企業級 AI 指揮中心／agentic automation 治理：raw/articles/bnext-uipath-enterprise-ai-command-center-2026-09-09.md — UiPath CXO 圓桌會議、Maestro 編排、治理／權限／人工核准與 Dark Testing Factory 的證據界線
• OpenAI／Navier–Stokes／Buckmaster 爭議：raw/articles/bnext-openai-navier-stokes-proof-buckmaster-dispute-2026-09-09.md — OpenAI 約 1 萬 agents、88 小時、17 小時 Lean 驗證的供應商自述，以及 forced Euler／研究優先權、資料使用與作者署名爭議的證據界線
• ChatGPT Images 2.5／Sketch 實測：raw/articles/bnext-chatgpt-images-2-5-sketch-launch-2026-09-09.md — OpenAI 發布日、延遲「up to 50%」與 30 億張／週的供應商自述、GPT-Image-2.5 Flare／Sunburst API 模型定位、Sketch 等介面工具與作者手機版實測的證據界線
• Google 台灣裁員傳聞／板橋 TPKE：raw/articles/bnext-google-taiwan-layoffs-rumor-tpke-2026-09-08.md — @mdm420xx 社群傳言、TPKE／板橋硬體研發基地背景、Google 發言人「比例不符事實」回應與士林 AI 基礎建設擴建 60% 的證據界線
• LINE AI 待辦機器人四步驟教學（Gemini API + Apps Script + LINE Messaging API）：raw/articles/bnext-line-ai-todo-gas-gemini-2026-09-07.md — 免伺服器 agent workflow、免費層 rate limit／資料使用政策、webhook 簽章驗證與部署版本鎖定邊界
• GPT-6 Astra prompting guide：raw/articles/bnext-gpt-6-astra-prompting-guide-2026-09-07.md — OpenAI 官方五類 prompting 行為、中文改寫模板、configuration_update 與 reasoning effort 限制的證據界線
• GPT-6 Astra skill／prompt 減法清理：raw/articles/bnext-gpt-6-astra-skill-cleanup-less-is-more-2026-09-07.md — Eric Provencher 五步 skill 清理原則、Codex 初始 skill 清單 2%／8,000 字元預算、openai/codex skill-creator 416→229 行 commit 與 Astra 越權評測 48% vs 0% 的證據界線
• Gemini Notebook／NotebookLM 考試準備 prompt 工作流：raw/articles/bnext-notebooklm-study-prompts-exam-prep-2026-09-06.md — 8 組出題、抓重點、模擬考與來源核對 prompt 的二手整理，產品行為與學習成效保留 BusinessNext attribution
• 緯穎 AI 伺服器需求／產能與除權案例：raw/articles/bnext-wiwynn-ai-server-orders-capacity-2026-09-06.md — AI 伺服器公司商業模式、股利／融資、產能／電力規畫、客戶集中與 X／Twitter 訴訟的證據界線
• Xbox Cloud Gaming 月度時數上限：raw/articles/bnext-xbox-cloud-gaming-hour-limits-2026-09-04.md — Game Pass 15／10／5 小時配額、超額購買、免訂閱 cloud playtime 與雲端服務成本／quota 政策的證據界線
• BIG DATA／KEYLM／AI Insight Partner：raw/articles/bnext-big-data-keylm-ai-insight-partner-2026-09-04.md — WITSA／臺灣 AI 卓越獎、企業知識 AI 平台與數據洞察 workflow 的贊助內容及證據界線
• Salesforce／Anthropic Claudeforce／headless enterprise workflow：raw/articles/bnext-salesforce-claudeforce-dumb-pipe-2026-09-04.md — agent 入口、資料／工作流護城河、MCP 可攜性與 SaaS「笨水管」風險的證據界線
• Databricks AI coding 成本管理／效率前沿／模型路由：raw/articles/bnext-ai-coding-costs-at-scale-2026-09-04.md — task-level cost、model／harness flexibility、progressive budget controls、token overhead 與 AI Gateway 的證據界線
• GPT-6 Astra／Critical cybersecurity／monitorability：raw/articles/bnext-gpt-6-astra-critical-cybersecurity-launch-2026-09-04.md — Critical cyber threshold、Provider Adapter／Standard harness 評測差異、分階段 access 與安全／可監控性證據界線
• 跨族群 genomic prediction／PRS transfer learning：raw/articles/google-research-transfer-learning-genomic-prediction-2026-09-03.md — UKB／BBJ sample-size ablation、meta-analysis、PRS-CSx 與臨床外推邊界
• Uber 組織縮減／AI 因果敘事：raw/articles/bnext-uber-layoffs-10-percent-3300-jobs-2026-09-03.md — 約 10% 團隊縮減、管理層級、遠距政策、自駕車投資與 AI 工具使用的證據界線
• Gemini Notebook Expert Intelligence／台灣操作教學：raw/articles/bnext-gemini-notebook-expert-intelligence-taiwan-guide-2026-09-03.md — Play Books 電子書匯入、英文書／中文提問、精選筆記本與地區／帳號 rollout 的證據界線
• Muse Spark 1.3／agentic evaluation／task cost：raw/articles/bnext-meta-muse-spark-1-3-2026-09-03.md — Meta 官方產品與評測方法、Artificial Analysis agentic benchmark／task-cost 分析，以及 max reasoning 安全測試與 open-weight roadmap 的證據界線
• Gemini 3.8 Flash／Cyber：raw/articles/bnext-gemini-3-8-flash-cyber-2026-09-03.md — reasoning／tool loop、token 單價與 task-level cost、CyberGym／CWE-Bench／Gray Swan 評測，以及 Fairwind trusted-access 的證據界線
• Tailscale connectivity platform：raw/articles/networkworld-tailscale-connectivity-platform-2026-08-31.md — identity-aware networking、Aperture AI gateway、Tailscale PAM、Control D、Tailcat 與 programmable tailnet 的產品敘事與證據界線
• NVIDIA Vera Rubin、記憶體與南亞科／國巨供應鏈比較：raw/articles/bnext-nvidia-rubin-nanya-yageo-ai-supply-chain-2026-09-02.md — NVL72 規格、BOM 估算、記憶體／被動元件成本、南亞科 5A 新廠與投資論點的證據界線
• Claude Fable 5.1／Mythos 5.1、模型成本與 release governance：raw/articles/bnext-anthropic-fable-5-1-mythos-5-1-2026-09-02.md — 同底層模型的 safeguards／trusted access、cache read 降價與 task-level cost、API 遷移、Artificial Analysis 評測、EFS 與科學案例的證據界線
• Jensen Huang／Bill Gates AI 工作辯論：raw/articles/bnext-jensen-huang-bill-gates-ai-jobs-debate-2026-09-02.md — substitution／redistribution 與 productivity／expansion 兩種工作、稅制與再工業化模型的證據界線
• DHH／AI programming／agentic engineering：raw/articles/bnext-dhh-ai-programming-lex-fridman-2026-09-01.md — problem-first prompting、產品品味、Basecamp 5 架構整合、Omarchy Quattro 開源維護與模型／harness 證據界線
• OpenClaw 2.0 / guided setup / Cloud Session：raw/articles/bnext-openclaw-2-0-release-install-guide-2026-09-01.md — v2026.8.1 的安裝驗證、Control UI、Skills／memory、SQLite migration 與 trusted-team collaboration 邊界
• TimesFM-3 / multivariate time-series foundation model：raw/articles/google-research-timesfm-3-multivariate-forecasting-2026-08-31.md — 多 target、past／future covariates、single-pass quantile forecasting、官方 benchmark 與 non-commercial weights 邊界
• MAPL-EMIT / hyperspectral methane plume mapping：raw/articles/google-research-mapl-emit-methane-deep-learning-2026-09-01.md — synthetic plume training、EMIT scene inference、source localization、Earth Engine confidence 與 ppm-m／kg/hr 邊界
• Male fruit-fly connectome / AI-assisted scientific mapping：raw/articles/google-research-male-fruit-fly-connectome-2026-09-03.md — connectomics 3D reconstruction、human proofreading、MaleCNS 公開查詢／下載與行為因果邊界
• Glassdoor workplace AI sentiment / workforce governance：raw/articles/bnext-glassdoor-employer-ai-2026-08-31.md — 世代／性別 AI 評價、評論中的工作不安全感、職缺結構與使用率／信任分離
• OpenAI／Cursor model supply and change-of-control：raw/articles/bnext-openai-cursor-model-supply-2026-08-31.md — OpenAI 擬終止供應、SpaceX 收購、控制權變更條款、第三方 coding harness 連續性與 fallback migration gate
• Zeabur environment variable／AI API key incident：raw/articles/bnext-zeabur-env-var-leak-api-key-stolen-2026-08-31.md — 高權限 AWS 憑證、共享叢集到主資料庫的官方調查說法、環境變數查詢、憑證輪替與未驗證外洩情報邊界
• 3D NAND word-line metallization / molybdenum scaling：raw/articles/bnext-molybdenum-nand-concept-stocks-2026-08-31.md — 鎢轉鉬、Kioxia VLSI 結果、沉積設備估計與台股題材的證據界線
• Claude memory / Chat-Cowork shared context：raw/articles/bnext-claude-memory-chat-cowork-unified-2026-08-30.md — 雲端 Cowork 記憶互通、Topics 編輯控制、敏感主題與 cloud／local runtime 邊界
• ChatGPT Work / cloud browser authenticated tasks：raw/articles/bnext-chatgpt-work-sign-in-websites-2026-08-30.md — 登入網站、19 項任務範例、secure sign-in、憑證與人工確認邊界
• Claude Cowork / built-in browser / browser identity：raw/articles/bnext-claude-cowork-built-in-browser-2026-08-28.md — 獨立 browser profile、Claude in Chrome 分工、desktop／remote runtime 與 prompt injection 邊界
• Claude Cowork / scheduled tasks / runtime boundary：raw/articles/bnext-claude-cowork-scheduled-tasks-2026-08-28.md — 週期／按需執行、獨立 Cowork session、遠端常駐、Claude 帳號／connector context 與本機檔案限制
• Gemini Notebook / Expert Intelligence / licensed knowledge workflow：raw/articles/bnext-gemini-notebook-expert-intelligence-ebooks-2026-08-28.md — Google Play Books 電子書匯入、grounded query、產物生成與 ownership control
• Gemini Omni 1.1 Flash / generative media fidelity ladder：raw/articles/bnext-gemini-omni-1-1-flash-video-extension-interpolation-2026-08-28.md — 場景延伸、首尾影格插補、360p 預覽與 1080p／4K 正式輸出
• ChatGPT browser-use / agent control surface：raw/articles/bnext-chatgpt-browser-10-automation-tips-2026-08-27.md — 內建瀏覽器、cloud browser、Chrome extension、10 個實作技巧與權限／安全邊界
• Bill Gates / Human Reserved / AI policy：raw/articles/bnext-bill-gates-turbulent-ai-era-human-reserved-2026-08-27.md — 工作轉移、惡意濫用、孩童發展、人類保留區與 AI／機器人稅
• Bill Gates / AI inequality / Gates Foundation funding：raw/articles/bnext-bill-gates-ai-inequality-human-reserved-gates-foundation-2026-09-16.md — 4 類初階／中階職缺受衝擊、10 億美元（教育 40%／醫療 40%／農業 10%／數位基礎 10%）、人類保留區與 AI token／機器人課稅
• Hugging Face platform neutrality / vendor capture：raw/articles/bnext-nvidia-hugging-face-acquisition-2026-08-27.md — NVIDIA 傳出收購 Hugging Face、開發者生態、中立性與直接收購的治理風險
• Hugging Face platform neutrality / non-NVIDIA hardware：raw/articles/derekhsu-nvidia-hugging-face-acquisition-2026-08-27.md — Derek Hsu 整理 Hacker News 正反論戰、llama.cpp／GGUF 與平台中立性追蹤指標
• Threads source variant：raw/articles/threads-derekhsu-nvidia-hugging-face-acquisition-2026-08-27.md — Derek Hsu 以 llama.cpp／ggml.ai 與非 NVIDIA 硬體作為「金絲雀」觀察點的短版貼文
• 凱鈿 HR 招募自動化：raw/articles/bnext-kdan-hr-ai-recruiting-automation-2026-08-27.md — 四人 HR 團隊、非工程師組裝流程、情境變體、漸進採用與人類保留高信任節點
• EgentHub / FDE / AI 逆分工：raw/articles/bnext-egenthub-fde-2026-08-27.md — 不到 10 人服務逾 200 家企業、超級個體、跨模型選擇與把 Agent 工作方式複製到客戶端
• CGM foundation model / wearable health：raw/articles/google-research-glucofm-continuous-glucose-monitoring-2026-08-26.md — GlucoFM 的 dual-stream glucose dynamics、missing-aware latent prediction、跨 cohort transfer 與 few-shot evaluation
• Planetary geospatial prediction / autonomous data science：raw/articles/google-research-planetary-prediction-engine-earth-ai-2026-08-27.md — PPE 的 intelligent data selection、dataset curation、AutoML、multimodal fusion 與高風險預測任務
• Google Research：raw/articles/google-research-timesfm-3-multivariate-forecasting-2026-08-31.md — TimesFM-3、多變量 forecasting、covariates、single-pass quantiles、benchmark 與授權邊界
• Local-first agent / data egress / sandbox gate：raw/articles/bnext-perplexity-portable-computer-2026-08-26.md — Portable Computer 的地端 runtime、DGX Spark／RTX 門檻、PII 外送前分類、逐步同意與 sandbox 失效即停用工具
• Eval automation / human judgment：raw/articles/aihao-automate-ai-evals-correctly-2026-08-24.md — Shreya Shankar 演講整理，涵蓋 Analyze／Measure／Improve、error-discovery skill、criteria drift、Monitor 雙向 review loop 與 app-specific worst-case eval
• AI search frontier：raw/articles/aihao-new-frontier-ai-search-2026-08-25.md — Hybrid Search、agentic search、reflected intelligence、signals boosting、presentation bias、LLM judge calibration、late interaction 與 quantization
• AI search frontier / canonical revision：raw/articles/aihao-new-frontier-ai-search-revision-2026-08-25.md — 同一 URL 的標題與成熟度表修訂，補上 sparse／dense 漸進部署、現在就做／值得投入／先知道的採用順序
• Claude Academy / official learning workflow：raw/articles/bnext-claude-academy-chinese-courses-2026-08-24.md — 22 門免費課程、繁體中文化、completion badge 與認證考試的邊界、產品／AI fluency／API 學習路徑
• Grok Bot / role-based agent team setup：raw/articles/bnext-grok-bot-getting-started-2026-08-25.md — 職務化幕僚長／專責 Bot、Teach a task、routine、共享雲端電腦與人工核准邊界
• Agent experience / embodied XR interaction：raw/articles/google-research-agenthands-spatially-grounded-agent-conversations-xr-2026-08-25.md — scene registry、GestureEvents、六維手勢 taxonomy、同步語音與 N = 12 user study
• AI infrastructure / semiconductor tariff and supply availability：raw/articles/bnext-us-chip-tariff-laptops-consoles-2026-08-28.md — 半導體關稅擴大、筆電／遊戲主機／資料中心伺服器、在美設廠承諾與產能時間錯配
• Enterprise AI employee lifecycle / ORRA：raw/articles/bnext-orra-ai-employee-training-4-stages-2026-08-28.md — AI 員工招募、知識訓練、考核、上工後管理、模型替換與人類判斷邊界
• Regional sovereign AI / SEA-LION：raw/articles/bnext-singapore-sovereign-ai-sea-lion-2026-08-28.md — 新加坡以東南亞語言、文化、SEA-HELM 評測與人才建立區域 AI 選擇權的策略整理
• Speech-to-text / transcription evaluation：raw/articles/bnext-gemini-3-5-transcribe-speech-to-text-2026-08-28.md — Gemini 3.5 Transcribe 的格式化逐字稿、WER、API 分流與人工 ground truth 驗證邊界
• Enterprise agent adoption / token depth：raw/articles/bnext-openai-codex-growth-legal-108x-engineer-5x-2026-08-28.md — OpenAI Enterprise Signals、Codex 跨職能採用、從協助走向執行與 Token 指標界線

高價值 raw clusters

• AI performance / enterprise maturity：raw/papers/pwc-ai-performance-global-taiwan-2026-07-17.md

• Claude Cowork / localhost agent / artifact delivery：raw/articles/bnext-claude-cowork-workbench-use-cases-2026-07-30.md

• AI infrastructure / microgrid / power-first deployment：raw/articles/bnext-delta-ai-microgrid-power-2026-08-10.md

• Agentic commerce / B2A / GEO：raw/articles/bnext-ikala-b2a-agentic-commerce-2026-08-10.md

• Diffusion model creativity：raw/articles/google-research-diffusion-model-creativity-2026-07-15.md

• LLM Wiki / OKF：raw/articles/karpathy-llm-wiki-gist-2026-04-03.md, raw/articles/dawsonwang-day-180-okf-obsidian-2026-06-29.md, raw/articles/aihao-llm-knowledge-base-2026-05-20.md, raw/articles/aihao-agent-harness-series-2026-07-25.md

• Agent workflow / eval：raw/articles/aihao-ai-native-engineering-org-2026-06-01.md, raw/articles/aihao-ai-code-review-bottleneck-2026-04-20.md, raw/articles/aihao-traces-new-source-of-truth-2026-02-17.md, raw/articles/aihao-generic-ai-metrics-mirage-2026-02-17.md, raw/articles/aihao-working-with-ai-compound-2026-05-20.md, raw/articles/aihao-agents-md-research-and-practices-2026-05-03.md, raw/articles/aihao-agent-skills-build-and-eval-2026-02-20.md, raw/articles/aihao-knowledge-to-skill-framework-2026-07-06.md, raw/articles/aihao-mattpocock-skills-design-philosophy-2026-07-25.md, raw/articles/aihao-post-mcp-era-skills-vs-mcp-2026-03-12.md, raw/articles/aihao-agent-experience-designing-for-agents-2026-05-18.md, raw/articles/aihao-cloud-agent-vs-localhost-agent-2026-05-18.md, raw/articles/bnext-claude-cowork-web-mobile-remote-session-background-scheduling-2026-07-08.md, raw/articles/aihao-pmarca-latent-space-2026-07-01.md, raw/articles/aihao-agent-sandbox-landscape-2026-04-20.md, raw/articles/aihao-ai-product-management-2026-05-18.md, raw/articles/aihao-rag-not-vector-search-2026-02-19.md, raw/articles/ihower-harness-loop-engineering-2026-06-26.md, raw/articles/aihao-harness-engineering-deep-agent-capabilities-2026-06-26.md, raw/articles/aihao-harness-engineering-mid-run-injection-2026-06-26.md, raw/articles/aihao-harness-engineering-goal-outcomes-2026-06-26.md, raw/articles/aihao-harness-engineering-self-improving-2026-06-26.md, raw/articles/aihao-harness-engineering-model-harness-fit-2026-06-26.md, raw/articles/aihao-harness-engineering-agent-frameworks-2026-06-26.md, raw/articles/aihao-cognitive-load-agent-orchestration-2026-06-17.md, raw/articles/google-research-alphaevolve-theoretical-cs-2025-09-30.md, raw/articles/bnext-wistron-ai-factory-agent-smart-manufacturing-2026-06-26.md, raw/articles/bnext-claude-code-management-fiona-fung-2026-06-26.md, raw/articles/bnext-anthropic-claude-tag-slack-team-collaboration-2026-06-24.md, raw/articles/bnext-openai-codex-taste-over-coding-2026-06-29.md, raw/articles/bnext-spotify-ai-engineering-validation-loops-2026-06-30.md, raw/articles/bnext-meta-bans-claude-code-codex-2026-06-30.md, raw/articles/bnext-ai-native-enterprise-transformation-governance-2026-06-16.md, raw/articles/bnext-claude-code-model-vs-effort-2026-07-09.md, raw/articles/bnext-fable-5-prompts-context-cost-2026-07-13.md, raw/articles/bnext-anthropic-engineer-blind-spot-prompt-2026-07-13.md, raw/articles/bnext-claude-cowork-one-on-one-system-2026-07-14.md, raw/articles/bnext-chatgpt-skills-tutorial-2026-07-14.md, raw/articles/bnext-anthropic-claude-value-axes-2026-07-14.md, raw/articles/bnext-yc-designer-ai-curation-workflow-2026-07-15.md, raw/articles/bnext-claude-reflect-ai-usage-reflection-2026-07-15.md

• Google Research：raw/articles/google-research-small-models-intent-extraction-decomposition-2026-01-22.md, raw/articles/google-research-generative-ui-2025-11-18.md, raw/articles/google-research-natively-adaptive-interfaces-accessibility-2026-02-05.md, raw/articles/google-research-sequential-attention-model-efficiency-2026-02-04.md, raw/articles/google-research-gist-smart-sampling-2026-01-23.md, raw/articles/google-research-atlas-multilingual-scaling-laws-2026-01-27.md, raw/articles/google-research-agentic-rag-2026-06-05.md, raw/articles/google-research-reasoningbank-agent-memory-2026-04-21.md, raw/articles/google-research-gemini-nano-frozen-mtp-2026-06-26.md, raw/articles/google-research-better-ai-benchmarks-raters-2026-03-31.md, raw/articles/google-research-convapparel-user-simulators-2026-04-09.md, raw/articles/google-research-scaling-agent-systems-2026-01-28.md, raw/articles/google-research-groundsource-news-to-data-gemini-2026-03-12.md, raw/articles/google-research-tabfm-zero-shot-tabular-data-2026-06-30.md, raw/articles/google-research-sensorfm-wearable-health-data-2026-07-09.md, raw/articles/google-research-symptomai-everyday-symptom-assessment-2026-07-22.md, raw/articles/google-research-ds-star-data-science-agent-2025-11-06.md, raw/articles/google-research-dialoglab-human-ai-group-conversations-2026-02-10.md, raw/articles/google-research-titans-miras-long-term-memory-2025-12-04.md, raw/articles/google-research-bayesian-teaching-llm-reasoning-2026-03-04.md, raw/articles/google-research-nested-learning-continual-learning-2025-11-07.md, raw/articles/google-research-differentially-private-chatbot-use-2025-12-10.md, raw/articles/google-research-paper-assistant-tool-stoc-2026-2025-12-15.md, raw/articles/google-research-deep-researcher-test-time-diffusion-2025-09-19.md

• Google Research：raw/articles/google-research-science-one-chain-of-evidence-2026-07-30.md — Science One Framework、Chain-of-Evidence 與 autonomous research verifiability

• Google Research：raw/articles/google-research-amie-audio-visual-clinical-consultations-2026-08-11.md — AMIE (Video)、audio-visual clinical consultation、三 agent 架構與 randomized OSCE

• Google Research：raw/articles/google-research-knowledge-profiling-parametric-factuality-2026-08-12.md — Knowledge Profiling、WikiProfile 與 parametric factuality 的 recall bottleneck

• Google Research：raw/articles/google-research-photoscan-cardiometabolic-risk-2026-08-17.md — PhotoScan、smartphone imagery、body composition 與 cardiometabolic risk

• Google Research：raw/articles/google-research-biomarker-discovery-framework-2026-08-21.md — Biomarker Discovery Framework、wearable biomarker prioritization、adversarial multi-agent loop 與 human expert evaluation

• Google Research：raw/articles/google-research-mobility-embedded-pois-2026-08-21.md — Mobility-Embedded POIs、文字與匿名 mobility 融合、場所功能節律與 unseen-place evaluation

• Google Research：raw/articles/google-research-agenthands-spatially-grounded-agent-conversations-xr-2026-08-25.md — AgentHands、XR 空間定位、GestureEvents、同步手勢與 N = 12 user study

• Google Research：raw/articles/google-research-male-fruit-fly-connectome-2026-09-03.md — 雄性果蠅 brain／VNC connectome、AI 3D reconstruction、human proofreading 與 MaleCNS 資料入口

• Prompt / agent workflow：raw/articles/bnext-anthropic-engineer-blind-spot-prompt-2026-07-13.md, raw/articles/bnext-prompt-engineering-symbols-guidelines-2026-03-18.md, raw/articles/bnext-claude-code-loop-engineering-2026-07-16.md, raw/articles/bnext-gpt-5-6-prompting-guide-2026-07-16.md, raw/articles/bnext-chatgpt-deep-research-tutorial-2026-07-20.md, raw/articles/bnext-claude-code-shortcuts-2026-08-19.md

• Personal content agent / scheduled publishing：raw/articles/bnext-claude-code-threads-auto-poster-2026-08-11.md — voice profile、GitHub 素材、Threads API 與排程交付

• Prompt / agent workflow：raw/articles/bnext-chatgpt-i-have-adhd-skill-2026-08-03.md — i-have-adhd 的任務收斂回應契約與模式切換

• Session metadata / AI collaboration：raw/articles/bnext-claude-code-session-title-batch-rename-2026-07-31.md

• Gemini / learning workflows：raw/articles/bnext-gemini-learning-map-2026-06-16.md, raw/articles/bnext-gemini-study-notebooks-2026-06-29.md, raw/articles/bnext-gemini-notebook-2026-07-17.md, raw/articles/bnext-notebooklm-complete-guide-2026-07-20.md, raw/articles/bnext-gemini-notebook-insurance-claim-notes-2026-07-20.md, raw/articles/bnext-chatgpt-youtube-to-ppt-2026-07-22.md, raw/articles/bnext-gemini-spark-personal-ai-agent-2026-07-30.md, raw/articles/bnext-gemini-spark-background-workflow-2026-08-05.md, raw/articles/bnext-gemini-notebook-expert-intelligence-ebooks-2026-08-28.md

• Gemini / learning workflows：raw/articles/bnext-gemini-spark-weekly-report-2026-08-06.md — Gmail／Calendar／Drive context、HTML 專案週報與每週排程交付

• Gemini / spreadsheet workflow / verification gate：raw/articles/bnext-gemini-sheets-monthly-report-format-2026-08-12.md — 參照上月報表整理本月資料、先核准執行計畫，再驗證筆數、公式與總計

• AI presentation / publication gate：raw/articles/bnext-gemini-canvas-editable-presentations-2026-08-11.md — Gemini Canvas、Google Slides／Vids 與可編輯性／publication review gate

• Human-AI interaction / agency gate：raw/articles/bnext-ai-kids-emotion-naming-prompt-2026-08-11.md — 情緒命名 prompt、保留不確定性與當事人確認

• Human-AI communication / eval calibration：raw/articles/bnext-hr-simulator-emergent-tact-2026-08-06.md — HR Simulator、LLM judge 偏好、emergent tact 與人機混合改寫

• Non-engineer agent workflow / skill ROI：raw/articles/bnext-non-engineer-ai-agent-workflow-2026-07-17.md

• Reflective AI use / cognitive agency：raw/articles/bnext-ai-follow-up-question-reasoning-2026-07-21.md

• Claude product use cases / agent experience：raw/articles/bnext-claude-90-use-cases-2026-07-21.md

• Agent control surfaces / physical AX：raw/articles/bnext-openai-codex-micro-keyboard-2026-07-16.md

• Agentic AI cost / quota control：raw/articles/bnext-openai-codex-chatgpt-work-banked-reset-2026-07-16.md

• Enterprise AI adoption / Skill Library：raw/articles/bnext-ai-adoption-knowledge-base-skill-library-2026-07-10.md

• Manager knowledge workflow：raw/articles/bnext-claude-cowork-one-on-one-system-2026-07-14.md, raw/articles/bnext-chatgpt-work-meeting-notes-presentation-2026-07-22.md

• Agent Skills / cross-product format：raw/articles/bnext-chatgpt-skills-tutorial-2026-07-14.md, raw/articles/bnext-ai-agent-skill-2026-08-13.md, raw/articles/bnext-third-party-skill-security-cross-platform-2026-08-14.md — Skill 的跨產品格式、觸發、非工程師建立、拆分與安全 gate

• Model platforms：raw/articles/aihao-microsoft-mai-models-2026-06-08.md, raw/articles/bnext-anthropic-fable-mythos-export-controls-lifted-2026-07-01.md, raw/articles/bnext-anthropic-mythos-delayed-release-2026-07-20.md, raw/articles/bnext-jensen-huang-token-economics-anthropic-trillion-startup-2026-07-03.md

• Enterprise model platforms / control plane：raw/articles/bnext-aws-bedrock-enterprise-ai-agents-2026-07-16.md

• Model-harness guidance：raw/articles/bnext-claude-fable-5-guide-2026-07-03.md, raw/articles/aihao-claude-fable-5-prompting-2026-07-04.md, raw/articles/bnext-anthropic-engineer-claude-harness-2026-07-08.md, raw/articles/bnext-claude-code-model-vs-effort-2026-07-09.md, raw/articles/bnext-fable-5-prompts-context-cost-2026-07-13.md, raw/articles/bnext-claude-pricing-capability-tiers-2026-07-20.md, raw/articles/bnext-google-gemini-3-6-flash-2026-07-22.md, raw/articles/bnext-gemini-3-7-flash-2026-08-14.md

• Model behavior / interaction style：raw/articles/bnext-anthropic-claude-value-axes-2026-07-14.md

• AI usage reflection / AI fluency：raw/articles/bnext-claude-reflect-ai-usage-reflection-2026-07-15.md

• AI coding comprehension / validation：raw/articles/bnext-vibe-coding-understanding-erosion-2026-01-29.md

• AI-assisted design / curation：raw/articles/bnext-yc-designer-ai-curation-workflow-2026-07-15.md

• AI recommendation readiness / GEO：raw/articles/bnext-geo-7-steps-ai-recommend-brand-2026-07-14.md

• AI recommendation readiness / organization execution：raw/articles/bnext-brand-ai-competitiveness-management-2026-07-16.md

• Enterprise AI adoption / solution governance：raw/articles/bnext-asus-experthub-smb-ai-digital-transformation-2026-07-17.md

• Regulated enterprise AI / domain-first workflow：raw/articles/bnext-cathaylife-ai-practice-2026-08-17.md — 壽險 domain knowledge、場景篩選、AI Coach／Cathay Eye 與受控 Agentic AI

• Enterprise AI workflow / Lean AI：raw/articles/bnext-ch-robinson-lean-ai-agents-2026-07-17.md

• AI-native organization / internal enablement：raw/articles/bnext-berry-ai-culvers-vision-ai-2026-08-21.md — Culver’s 千店視覺 AI 部署、部門 AI 工作流與內部工程師 enablement

• Generative media / domain-first productization：raw/articles/bnext-whatsub-ai-subtitle-tool-2026-08-21.md — What’Sub 的繁中字幕 workflow、瀏覽器本機處理與 AI Demo → 可付費產品 gate

• Generative media / product operations / creator-to-owner：raw/articles/bnext-whatsub-product-operations-interview-2026-08-26.md — What’Sub 訪談的 AI coding 測試回饋、付費後營運、資安／金流責任與角色轉換

• Industrial OT data / agent-ready governance：raw/articles/bnext-thingnario-energy-data-governance-2026-07-24.md

• AI-native organization / system thinking：raw/articles/bnext-netflix-system-thinking-2026-07-20.md

• AI Agent lifecycle / enterprise governance：raw/articles/bnext-chtinventai-agenthub-governance-2026-07-20.md

• Context engineering / memory economics：raw/articles/aihao-semianalysis-hbm-context-window-2026-07-06.md, raw/articles/bnext-fable-5-prompts-context-cost-2026-07-13.md

• AI coding organization / talent：raw/articles/bnext-ai-coding-talent-pipeline-2026-07-03.md, raw/articles/bnext-groq-management-lessons-2026-07-06.md, raw/articles/bnext-openai-hr-hiring-ai-talent-war-2026-07-07.md

• AI coding adoption maturity：raw/articles/bnext-claude-code-ai-adoption-5-stages-2026-07-17.md

• AI product judgment / safety：raw/articles/aihao-ai-product-management-2026-05-18.md, raw/articles/bnext-anthropic-pm-interview-judgment-2026-07-17.md

• AI cognitive offloading / software cost debate：raw/articles/aihao-ai-second-brain-cost-2026-07-08.md, raw/articles/aihao-software-cost-zero-debate-2026-07-08.md

• Model platforms / orchestration：raw/articles/sakana-ai-fugu-multi-agent-system-as-model-2026-06-30.md

• Model gateway / routing / token economics：raw/articles/bnext-stripe-openrouter-model-gateway-2026-08-20.md — Stripe 收購 OpenRouter、跨供應商模型路由、token billing 與推論成本管理

• Model evaluation / open-weight release：raw/articles/bnext-kimi-k3-2026-07-17.md

• Vibe coding deployment / publication gate：raw/articles/bnext-chatgpt-sites-2026-07-17.md

• Prompting / eval：raw/articles/aihao-mom-test-prompting-2026-07-02.md

• Personal agent harness：raw/articles/raymondhouch-ai-agent-7-layer-architecture-2026-06-30.md

• Edge AI / hardware governance：raw/articles/bnext-taiwan-ai-hardware-open-source-2026-06-26.md, raw/articles/bnext-physical-ai-2030-sweet-spot-humanoid-robot-2026-07-01.md, raw/articles/bnext-blackberry-qnx-physical-ai-comeback-2026-07-03.md, raw/articles/bnext-nvidia-japan-robotics-physical-ai-2026-07-15.md

• Taiwan AI industry / agent economy：raw/transcripts/sega-on-ai-taiwan-ai-dominance-or-oem-2026-06-29.md, raw/articles/bnext-taiwan-ai-supply-chain-who-benefits-2026-08-05.md

• Field AI automation / physical harness：raw/articles/bnext-ai-farmer-codex-automation-2026-07-17.md

• Gemini / learning workflows：raw/articles/bnext-gemini-learning-map-2026-06-16.md, raw/articles/bnext-gemini-study-notebooks-2026-06-29.md, raw/articles/bnext-gemini-notebook-2026-07-17.md, raw/articles/bnext-notebooklm-complete-guide-2026-07-20.md, raw/articles/bnext-chatgpt-youtube-to-ppt-2026-07-22.md

• Communication / personal productivity：raw/articles/bnext-ai-speech-feedback-mckinsey-bcg-framework-2026-07-03.md

• Communication / presentation agency：raw/articles/bnext-ai-presentation-human-value-2026-07-31.md, raw/articles/bnext-ai-presentation-four-questions-2026-08-05.md

• AI search / marketing decision stack：raw/articles/91app-ai-seo-geo-guide.md, raw/articles/bnext-ai-search-seo-rules-2026-06-23.md, raw/articles/bnext-ai-decision-stack-marketing-algorithmic-trust-2026-07-08.md

• AI search / GEO multi-turn recommendation：raw/articles/bnext-brand-pr-strategy-geo-ai-overviews-2026-07-29.md

• Agent evaluation / long-horizon benchmark：raw/articles/aihao-agents-last-exam-benchmark-2026-07-22.md, raw/articles/aihao-llm-judge-checklist-decomposition-2026-07-22.md

• Agent eval methodology / architecture layers vs error analysis：raw/articles/aihao-agent-evals-two-approaches-2026-08-05.md

• Agent eval methodology canonical revision：raw/articles/aihao-agent-evals-two-approaches-revision-2026-08-22.md — floor raising、code-aware eval、六代架構、offline／online 與 harness 邊界

• Enterprise AI adoption / task-to-operating-model redesign：raw/articles/aihao-ai-enabled-ai-first-ai-native-2026-08-22.md — AI-enabled、AI First、AI Native、Excel 週報案例與 zero-basing

• Enterprise AI adoption / five-stage revision：raw/articles/aihao-ai-enabled-ai-first-ai-native-revision-2026-08-23.md — 單機作業、雲端系統、API／權限／稽核、跨系統資料通道與五層導入判準

• Eval UX / verification-first product design：raw/articles/aihao-eval-smell-verifiable-ux-2026-08-05.md

• Model release / task-level quality-speed-cost comparison：raw/articles/bnext-gemini-3-7-flash-2026-08-14.md

• Gemini model selection / subscription capacity / task-level cost：raw/articles/bnext-google-gemini-3-7-flash-subscription-model-selection-2026-08-17.md

• Data agent / verification / semantic context：raw/articles/aihao-hex-data-agents-izzy-miller-2026-08-05.md

• Agentic engineering patterns：raw/articles/aihao-simon-willison-agentic-engineering-patterns-2026-07-25.md

• Agent Skills / demonstration capture：raw/articles/bnext-claude-cowork-record-a-skill-2026-07-22.md

• Agent Skills / model-specific context：raw/articles/bnext-claude-5-context-engineering-2026-07-29.md, raw/articles/bnext-matt-pocock-ai-skills-workflow-2026-07-29.md

• Model-specific system prompts / harness guidance：raw/transcripts/hermes-model-specific-system-prompts-2026-06-18.md

• Model-harness guidance / Claude Opus 5：raw/articles/bnext-claude-opus-5-prompt-engineering-2026-07-29.md

• Model-harness guidance / prompt ablation and verification：raw/articles/bnext-claude-code-prompt-reduction-agent-verification-2026-07-30.md

• AI presentation / human agency and communication：raw/articles/bnext-ai-presentation-human-value-2026-07-31.md

• Retail omnichannel / agent-ready member data：raw/articles/bnext-pxmart-omnichannel-member-data-2026-07-31.md

• Generative media agent workflow / multimodal orchestration, eval, cost and provenance：raw/articles/bnext-google-generative-media-guide-2026-07-31.md

• Open-source AI hardware governance / neutral stewardship：raw/articles/bnext-taione-open-source-foundation-2026-07-30.md

• Open-weight model strategy / supply autonomy：raw/articles/bnext-andrew-ng-open-weight-ai-2026-07-31.md

• Open-weight local agent / quantization / draft-model inference：raw/articles/bnext-meta-muse-glimmer-open-weight-local-agent-2026-08-11.md

• AI infrastructure financing / compute collateral / recovery economics：raw/articles/bnext-nvidia-ai-compute-financing-2026-08-11.md

• Open-source AI system / OSPO / hardware-to-software stack：raw/articles/bnext-brian-behlendorf-open-source-ai-2026-07-31.md

• Model-harness guidance / model-app-harness selection：raw/articles/bnext-ethan-mollick-model-selection-2026-07-28.md

• Agent orchestration / Graph Engineering：raw/articles/bnext-graph-engineering-ai-agent-2026-07-29.md

• Agent adoption / model-harness-fit / organization experiment：raw/articles/bnext-ethan-mollick-coexistence-ai-adoption-2026-07-29.md

• RAG / document parsing / parser eval：raw/articles/aihao-beyond-rag-llamaindex-workshop-2026-07-26.md

• Financial Agent templates / domain workflow components：raw/articles/bnext-anthropic-claude-financial-agents-guide-2026-07-24.md

• Physical AI / edge deployment：raw/articles/bnext-aidc-ai-transformation-2026-07-24.md

• AI content governance / human review：raw/articles/bnext-ai-fatigue-brand-soul-marketing-2026-07-24.md

• Agent control surfaces / desktop voice：raw/articles/bnext-chatgpt-voice-desktop-2026-07-24.md

• Agent control surfaces / browser + local verification：raw/articles/bnext-chatgpt-chrome-extension-codex-guide-2026-07-29.md

• Prompt / context capture：raw/articles/bnext-karpathy-voice-prompting-method-2026-07-24.md

• Persistent Codex threads / context capture / Goal automation：raw/articles/bnext-openai-codex-jason-liu-long-lived-ai-team-2026-07-29.md

• Enterprise agent adoption / manufacturing knowledge transfer：raw/articles/bnext-szs-ai-agents-in-house-development-2026-08-01.md

• Enterprise agent / ERP integration / knowledge distillation：raw/articles/bnext-only-ai-agent-erp-2026-08-24.md

• Long-horizon coding agent / procedural 3D generation / visual feedback：raw/articles/bnext-karpathy-claude-opus-lord-of-the-rings-2026-08-03.md

• Regulated enterprise agent workflow：raw/articles/bnext-fubon-insurance-ai-assistants-2026-07-29.md

• AI conversation sharing / privacy boundary：raw/articles/bnext-claude-shared-chats-privacy-2026-07-29.md

• AI conversation sharing / privacy self-check：raw/articles/bnext-claude-share-link-leak-self-check-2026-07-30.md

• Voice-first prompt capture / intent alignment：raw/articles/bnext-voice-dictation-ai-prompt-2026-07-30.md

• Generative media / synthetic identity and one-way consent gate：raw/articles/bnext-ai-memorial-technology-yijia-2026-08-04.md

• AI-native organization / product judgment / PM problem ownership：raw/articles/bnext-whatnot-pm-product-theater-2026-08-04.md

• Agent Skills / third-party package security review：raw/articles/bnext-github-skill-install-security-check-2026-08-04.md

• Agent Skills / GitHub repo literacy and adoption ladder：raw/articles/bnext-github-free-resources-skills-guide-2026-08-06.md

• Enterprise AI enablement / three-layer AI architecture / agent governance：raw/articles/bnext-taiwan-mobile-ai-wave-enterprise-agent-2026-08-04.md

• Agent memory reconstruction / MemHarness / raw replay ablation：raw/articles/bnext-agent-memory-reconstruction-memharness-2026-08-04.md

• Decision-first planning / Wayfinder：raw/articles/bnext-matt-pocock-wayfinder-2026-08-05.md

• Enterprise AI employee governance / ORRA：raw/articles/bnext-orra-ai-agent-governance-2026-08-05.md

• Human value / intelligence inflation / AI 3.0：raw/articles/bnext-intelligence-inflation-ai-3-0-2026-08-05.md

• AI industry stack / 18-layer tower：raw/articles/bnext-huawei-ai-18-layer-tower-2026-08-05.md

• Local AI / AI NAS deployment gate：raw/articles/bnext-ai-nas-qnap-local-deployment-2026-08-05.md

• Open-weight economics / CUDA substitution：raw/articles/bnext-deepseek-funding-record-2026-08-05.md

• Scientific discovery automation / research loop：raw/articles/bnext-discovery-loop-jeff-dean-2026-08-06.md

• Long-horizon agents / model-harness fit / AI startup selection：raw/articles/bnext-jeff-dean-ai-1-percent-rule-2026-08-07.md

• Scientific discovery selection / 10x leverage / 2–20 year boundary：raw/articles/bnext-jeff-dean-discovery-loop-2026-08-21.md

• World Models / WEF emerging technologies / Physical AI：raw/articles/bnext-wef-emerging-technologies-2026-08-06.md

• AI memory hierarchy / HBF / HBM / OCP standardization：raw/articles/bnext-hbf-high-bandwidth-flash-2026-08-06.md

• AI memory hierarchy / 3D NAND molybdenum word-line transition：raw/articles/bnext-molybdenum-nand-concept-stocks-2026-08-31.md — 材料物性、無氟前驅體、填充／漏電／RC、量產與設備市場邊界

• Document parsing / PDF-to-Markdown / OCR routing：raw/articles/bnext-pdf-markdown-tools-2026-08-07.md

• Model-specific inference hardware / memory placement / AMD-Taalas：raw/articles/bnext-amd-taalas-model-specific-inference-2026-08-07.md

• Personal AI agent / ambient hardware / physical AX：raw/articles/bnext-openai-first-hardware-ai-speaker-2026-08-07.md

• AI chip infrastructure / edge inference / vertical integration：raw/articles/bnext-spacex-tesla-terafab-ai-chip-infrastructure-2026-08-07.md

• AI infrastructure / power-first strategy / data-center deployment economics：raw/articles/bnext-nvidia-hut8-ai-power-deal-2026-08-07.md

• Physical AI / commercial drone productization / total solution integration：raw/articles/bnext-qisda-drone-team-mass-production-2026-08-07.md

• AI website readiness / GEO access and parseability：raw/articles/bnext-ai-agent-website-check-2026-08-07.md

• Enterprise agent accountability / system governance：raw/articles/bnext-ai-agent-accountability-2026-08-10.md — AI 代理的權限、邊界、監督、責任歸屬與流程內控

• Brand AI communication / agent-ready governance：raw/articles/bnext-process-2-0-brand-ai-assistant-2026-08-10.md — Brand Strategy、Brand Communication System、AI Brand Communication 與五種治理斷層

• Multi-session agent coordination / summary-only handoff：raw/articles/bnext-claude-code-cross-session-messaging-2026-08-10.md — Claude Code session 間的純文字摘要傳訊、worktree 同步與權限邊界

• AI output quality / trust / literacy governance：raw/articles/bnext-gallup-gen-z-ai-anger-literacy-gap-2026-08-10.md — 從 token／產出量轉向品質、信任、查核規範與 AI 素養

• Physical AI / sensor-to-fab co-design / edge perception：raw/articles/bnext-sony-tsmc-physical-ai-image-sensor-2026-08-10.md — 影像感測器設計、製程製造、感測鏈驗收與供應 gate

• Data science agent verification / prompt gate：raw/articles/bnext-ai-data-analysis-verification-gate-2026-08-10.md — 資料健檢、比較基準與事實／推測／未知分離

• Taiwan AI industry / five-layer supply-chain map：raw/articles/bnext-ai-stocks-five-layer-supply-chain-2026-08-11.md — 能源、晶片、基礎設施、模型、應用與台灣供應鏈位置

• Taiwan AI industry / five-layer supply-chain canonical revision：raw/articles/bnext-ai-stocks-five-layer-supply-chain-revision-2026-08-14.md — 代表台廠營收／訂單證據 gate 與模型層直接／間接受惠邊界

• Taiwan AI industry / advanced packaging and OSAT capacity：raw/articles/bnext-spil-douliu-cowos-new-fab-2026-08-11.md — CoWoS、OSAT、產能瓶頸、雲林擴廠與在地人才培育

• Forward-deployed engineering / FDE / FAE：raw/articles/bnext-forward-deployed-engineer-fde-2026-08-11.md — 客戶現場導入、系統整合、權限與回饋迴路

• Agent Teams / computer-use / routine：raw/articles/bnext-spacexai-grok-bot-agent-team-2026-08-12.md — 幕僚長與專責 Bot、共享上下文、工具內交付與可重跑流程

• Model selection / long-horizon agent eval / cache economics：raw/articles/bnext-grok-4-6-agent-long-horizon-eval-2026-08-13.md — Grok 4.6、長時程任務評測、自我驗證、回合數與 cached input 成本

• One-person company / agent team / compute economics：raw/articles/bnext-one-person-company-polsia-2026-08-13.md — AI 代理團隊、背景營運、人類創意 owner、外包專業與算力成本

• Enterprise agent adoption / financial denominator / token observability：raw/articles/bnext-appier-agentic-ai-gross-margin-2026-08-14.md — 企業內部多 agent 協作、人均毛利、R&D 週期與 token dashboard

• Enterprise AI deployment / IT + CT vertical integration：raw/articles/bnext-taiwan-mobile-systex-ai-it-ct-acquisition-2026-08-14.md — 網路、IDC、算力、資安、系統整合與產業應用交付鏈

• Personal agent / computer history / context privacy：raw/articles/bnext-chatgpt-computer-history-2026-08-14.md — 事件流工作日誌、可撤銷 context capture、明文記憶、資料上傳與 prompt injection gate

• raw/articles/bnext-chatgpt-work-14-features-cloud-vs-local-2026-08-14.md — ChatGPT Work 14 個功能、cloud／local runtime、plugins、排程與 artifact delivery（2026-08-14）

• raw/articles/bnext-chatgpt-work-14-features-cloud-vs-local-revision-2026-08-17.md — ChatGPT Work 修訂版：14 項功能清單、Skills 封裝範圍、FAQ 與低風險採用順序（2026-08-17）

• Brand AI content governance / G-U-A-R-D：raw/articles/bnext-xiangshuai-ai-ad-guard-framework-2026-08-12.md — AI 廣告人工把關、品牌語感、風險查核、KPI 重構與發布檢核表

• AI content provenance / text watermarking：raw/articles/bnext-claude-text-watermarking-2026-08-17.md — Claude 文字浮水印原理、偵測邊界、EU AI Act 背景、作者責任與 C2PA 區分

• Open-weight model ecosystem / download and deployment signals：raw/articles/bnext-qwen-huggingface-open-model-ranking-2026-08-17.md — Qwen 下載／衍生模型、GGUF 本地部署訊號、小模型分布與 NVIDIA／AMD 模型策略

• AI product design / code abundance / design-engineering asymmetry：raw/articles/bnext-openai-design-lead-do-less-2026-08-17.md — 反功能膨脹、可組合產品積木、雙速設計投資與人類使用者驗證

• Taiwan AI application layer / Physical AI robotics revision：raw/articles/bnext-ai-application-physical-ai-robotics-revision-2026-08-19.md — 代理式商務、機器人應用、直接／間接受惠與訂單／營收 value-capture gate

• Taiwan AI adoption intensity / enterprise demand gate：raw/articles/bnext-taiwan-ai-usage-intensity-decline-2026-08-19.md — 台灣 AI 使用強度、中小企業導入率與「尚無明確應用需求」障礙

• Frontier model training safety / sandbox escape：raw/articles/bnext-openai-frontier-training-pause-hugging-face-breach-2026-08-19.md — 前沿 RL 訓練暫停、sandbox／工具監控、30 分鐘警報停機與能力門檻

• Claude Gmail connector / external action governance：raw/articles/bnext-claude-gmail-send-connectors-2026-08-19.md — Gmail 寄送／轉寄、Connector 寫入權限、逐次確認與企業政策

• Claude Code design-first UI artifact：raw/articles/bnext-claude-code-design-command-early-preview-2026-08-19.md — /design 早期預覽、可編輯畫板、artifact review 與設計／實作偏差

• Claude product learning map：raw/articles/bnext-claude-learning-map-2026-08-19.md — Chat／Cowork／Code、Skills、context、token 與 Claude Code 的八段式學習地圖

• Claude product learning map revision：raw/articles/bnext-claude-learning-map-revision-2026-08-19.md — 40 篇教學、Claude Code /design 畫板預覽與產品學習地圖修訂

• Taiwan AI ecosystem / Silicon Valley bridge / startup product gates：raw/articles/bnext-steve-chen-youtube-taiwan-startup-2026-08-21.md — 矽谷資本／人才網絡、台灣供應鏈雙向橋樑、合作夥伴盡調、全球化與目標受眾 gate

• Research portfolio / power-law bets / product pruning / AI governance：raw/articles/bnext-altman-power-law-openai-safety-2026-08-24.md — 冪次法則、砍掉 Sora／Atlas、Codex context bottleneck、incident learning 與權力集中

相關入口

• index
• index
• llm-wiki-and-okf
• agent-workflows