Knowledge
AI 知識庫
收錄團隊日常遇到嘅有趣 AI 知識與實務筆記；可用關鍵字搜尋。內文包含旗艦落地筆記（WhatsApp 預約、財務節奏、短片 vs 排程工具）。頂部「最近分享」會每日自動由官方 RSS 收取外部連結。
最近分享
由白名單 RSS 自動收取嘅外部文章，新嘅排最頂；只作導讀，全文喺原文網站。
- 無預覽圖
How a researcher uses Codex and ChatGPT to search for new antimicrobial molecules
César de la Fuente’s lab uses Codex and ChatGPT to search living and extinct genomes for antimicrobial candidates to fight drug-resistant infections.
Rebuilding AUTOMATIC1111 with Gradio Workflow
Powering AI is an architecture problem
On July 22, 2026, a transmission line fault in Ashburn, Virginia—the heart of the world’s largest data center cluster—knocked more than 3 gigawatts of load off the grid in seconds. And it wasn’t the first time. Two year…
Slack can now vibe-code interactive charts and reports inside chats
A new feature coming to Slack will allow you to build interactive reports, polls, dashboards, presentations, microsites, and other tools directly inside a chat. With Slackforce Surfaces, you can describe to Slackbot wha…
Jensen Huang explains why Nvidia will grow an astounding 70% next year
Nvidia has its finger in every pie, and sees another year of plenty in its future, Jensen Huang says. But, he insists, its deals are not circular.
Pentagon in talks to lend $5 billion to AI cloud startup Fluidstack, WSJ reports
Google 將 Gemini 桌面應用帶到 Windows Alt + Space 快速呼叫
Google 繼 4 月於 macOS 推出 Gemini 桌面應用後，現將該應用帶到 Windows。Gemini 桌面應用在 Windows 上以 Alt + Space 鍵盤快捷鍵啟動。Google 強調，用戶可在處理任何工作時即時取用 Gemini，例如為文件快速查核事實，或為簡報構思幾個吸睛標題，取得所需協助後即可回到原本工作流程。 Gemini Spark 全天候代理 Gemini Spark 可作為全天候個人 AI 代理…
- 無預覽圖
AI Token限制背後的複雜現實 - 华尔街日报中文网
AI Token限制背後的複雜現實 华尔街日报中文网
- 無預覽圖
Paul Christiano joins OpenAI Foundation Board
Paul Christiano joins the OpenAI Foundation Board and its Safety and Security Committee, bringing experience in AI alignment, safety, and standards.
IBM releases SOTA Granite Time Series PatchTST-FM-r2 model with commercial-friendly license
What OpenAI’s latest controversy tells us about the future of math
OpenAI’s latest mathematical milestone has quickly become mired in controversy. Today, the company announced that its agents have solved one of the Millennium Prize Problems, some of the most important open problems in…
Suno releases its first AI music model made with record industry help
Suno's new v6 AI music model is its first made with support from the record industry. Suno's Jack Brody told The Verge that v6 was "trained from the ground up, with a new set of data that does not include the same data…
AI research startup Listen Labs scrubbed a $1.5B funding round for Salesforce talks
Listen Labs walked away from a signed Series C term sheet from Menlo Ventures, sources say.
OpenAI 擬同時委託三星生產晶片 與台積電雙重供應
OpenAI 正擴大與三星電子（Samsung Electronics）的合作關係，範圍從記憶體供應及企業軟件延伸至處理器製造。OpenAI 韓國區總經理 Harrison Kim 近日透露，該公司計劃將部分處理器的生產外包給 Samsung Foundry。若消息屬實，OpenAI 的 AI 加速器將同時由台積電和 Samsung Foundry 供應，這暗示了其龐大的產量需求。 合作範圍擴展至晶片共同研發 Harrison Kim…
- 無預覽圖
C.H. Robinson出席Jefferies會議：精益AI、法律風險與增長前景 作者 Investing.com - Investing.com 香港 - 股市報價& 財經新聞
C.H. Robinson出席Jefferies會議：精益AI、法律風險與增長前景 作者 Investing.com Investing.com 香港 - 股市報價& 財經新聞
- 無預覽圖
Political world erupts as AI researchers warn of ‘extinction’ threat - The Washington Post
Political world erupts as AI researchers warn of ‘extinction’ threat The Washington Post
- 無預覽圖
How GPT-5.6 Sol helps run quantum computing experiments
See how an MIT researcher uses GPT-5.6 Sol with Codex to autonomously run quantum computing experiments, analyze results, and calibrate qubits.
Safety for Whom? Refusing the Right Subset of a Topic, Not the Whole Topic
This AI entrepreneur is developing agents that can plan ahead for the unexpected
Danijar Hafner’s office in San Francisco’s SoMa district sits mostly empty. His brand-new startup is still in stealth mode and doesn’t even have its name on the door. On the day I visit, there’s only one other person th…
Drama swirls around OpenAI’s legendary mathematical milestone
OpenAI says it found a solution to a major math problem that has remained unsolved for around 90 years, as reported earlier by The New York Times and Wired. In a blog post on Tuesday, OpenAI announced that it discovered…
Hackers are stealing Claude tokens from subscribers
Last month, a Claude user noticed his account was consuming tokens even though he wasn't working. Anthropic has since warned users about hackers.
Arm 發佈新一代 CPU 與 GPU 承諾桌級遊戲效能與 AI 算力
Arm 宣佈推出用於未來 Android 晶片的新款 C2 系列處理器，同時升級 Mali 圖形處理器，旨在大幅提升整體效能。這次更新包括針對單執行緒性能及應用啟動速度的改進，並將本地人工智能處理能力直接整合至 CPU 架構中，不再僅依賴獨立的 NPU 或 TPU。 CPU 效能提升與內建 AI 作為 MediaTek 旗艦晶片及 Google Tensor 等 Android 智能手機晶片的核心設計基礎，Arm 本週推出的 C2-U…
- 無預覽圖
AI戰丨美國指摘阿里等6中國AI公司惡意複製美國AI技術 - 信報網站
AI戰丨美國指摘阿里等6中國AI公司惡意複製美國AI技術 信報網站
- 無預覽圖
Resulticks Named Partner's Choice: AI Innovation Award Winner at Redington's Global NEXT Summit 2026 - Business Wire
Resulticks Named Partner's Choice: AI Innovation Award Winner at Redington's Global NEXT Summit 2026 Business Wire
- 無預覽圖
Supporting independent journalism in Ukraine
OpenAI, AIRPPU and WAN-IFRA launch an AI program to help Ukrainian news organizations strengthen innovation, resilience, and independent journalism.
Microsoft says virtually nobody was grabbing NYT articles through its chatbot
Microsoft's Copilot rarely reproduces even full sentences from news articles and books, let alone substantive chunks that could substitute for the original, the company says in new legal filings as it fights copyright c…
Opaque recurrence, and other AI terms that you should probably know
The rise of AI has brought an avalanche of new terms and slang. Here is a glossary with definitions of some of the most important words and phrases you might encounter.
Exclusive: Ineffable Intelligence adds six ‘cofounders,’ hiring veterans from Google DeepMind, InstaDeep and venture firm Flying Fish
不只在 Flow Music 可用 Google 稱 Lyria 3.5 現可於 Gemini 用來生成音樂
今年 7 月底，Google 推出最新音樂生成模型 Lyria 3.5，讓用家能夠打造內容豐富，並具備自然旋律... 這篇文章 不只在 Flow Music 可用 Google 稱 Lyria 3.5 現可於 Gemini 用來生成音樂 最早出現於 PC3 Magazine。
- 無預覽圖
狗嗅覺成癌篩利器 印公司結合AI助揪7種癌症 - 星島頭條
狗嗅覺成癌篩利器 印公司結合AI助揪7種癌症 星島頭條
- 無預覽圖
AI, automation, robot dogs ensure on-site nuclear safety in China - Asia News Network
AI, automation, robot dogs ensure on-site nuclear safety in China Asia News Network
- 無預覽圖
An Alien Mind
Jakub Pachocki reflects on increasingly capable AI and the challenge of keeping it aligned. He calls for stronger safeguards and international coordination.
Training a coding model to paint watercolours with TRL and OpenEnv
Seattle Times and Newsday sue OpenAI and Microsoft for infringement
The Seattle Times and Newsday are just the latest plaintiffs to take OpenAI to court, alleging copyright infringement. The two outlets say the company used their journalism as training data for its AI models without per…
Authors push back as publishers and agents make claims on Anthropic settlement
Authors say publishers seem to be claiming more than their fair share of settlement payments.
- 無預覽圖
OpenAI承認「維基事件」 坦言AI異常行為揭露應更透明 | 中廣新聞網 - LINE TODAY
OpenAI承認「維基事件」 坦言AI異常行為揭露應更透明 | 中廣新聞網 LINE TODAY
- 無預覽圖
Nvidia is the central bank of AI. But will its loans prove sound? - AFR
Nvidia is the central bank of AI. But will its loans prove sound? AFR
- 無預覽圖
Legora reviewed 41 documents in minutes with GPT-6 Astra
Legora used GPT-6 Astra to review 41 documents in minutes, find all four planted errors, and improve performance by nearly 40% in this financial-review workflow.
Give Your Coding Agents a Memory You Own
Data from drones in Ukraine is fueling a new Wild West marketplace
Battlefields in Ukraine are littered with the remnants of drones, which are now firmly established as a critical weapon of modern warfare. But behind all that wreckage, there’s a new gold mine for the defense sector. Th…
OpenAI admits to German wiki ‘incident’
OpenAI says it needs to overhaul how and when it reports instances of AI models attacking real-world targets. The acknowledgement comes as the company manages the fallout from reports that a swarm of its out-of-control…
Seattle Times and Newsday are the latest publications to sue OpenAI and Microsoft
Two more news organizations are suing OpenAI and Microsoft over the supposed use of their journalism to train AI.
BOX Q2 Deep Dive: AI Platform Momentum and Enterprise Migration Drive Growth
OpenAI 發佈 GPT-6 Astra 提示詞指南 列禁用語與行為修正
OpenAI 近期更新其模型文檔，詳細說明 GPT-6 Astra 在實際運作中容易出現的非預期行為，並提供開發者具體的提示詞調整策略。該文檔指出，相較於前代產品，新模型在互動風格、指令遵循度以及文字輸出格式上均有顯著變化，需要透過精確的 Prompt Engineering 來優化使用體驗。 提升主動性與減少過度確認 GPT-6 Astra 比 GPT-5.6 Sol 更頻繁地提出澄清問題，而非自行推斷假設，這雖然使其成為「更有效的…
- 無預覽圖
記者編輯部| 【湯紹成專欄】緣起性空與AI模式 - 梅花新聞網
記者編輯部| 【湯紹成專欄】緣起性空與AI模式 梅花新聞網
- 無預覽圖
Playco cut manual fixes 50% prototyping games with GPT-6 Astra
Using GPT-6 Astra, Playco built three themed game prototypes from one grey box foundation and reported 50% fewer manual fixes than with the previous model.
Fine-tuning a 350M Model for Better Structured Outputs in 100 GRPO Steps
Architecting memory and storage in the AI era
The era of AI inference has arrived. Imagine a healthcare system analyzing millions of data points in real time to accelerate life-saving medical research, or an intelligent assistant instantly resolving thousands of co…
Roland is getting into generative AI music with Melody Flip
It's not quite the "push button; get song" of Suno, but Roland's new Melody Flip tool marks the company's foray into generative AI music. Available as a plug-in for your digital audio workstation (DAW), Melody Flip offe…
XDOF, just three months out of stealth, is in talks for a Series B at a $1.2B valuation
The round is being raised just months after the robot data startup exited from stealth.
Exclusive-US, China gear up for mid-September AI safety talks
Anthropic Claude 進軍 CarPlay 支援免提對話功能
Anthropic 的 AI 助理 Claude 即將登陸 Apple 的車載系統 CarPlay，為駕駛者帶來更智能的互動體驗。這項更新允許用戶在行車期間透過語音與 AI 進行交流，無需手動操作螢幕，從而提升駕駛安全性與便利性。 免提對話介面設計 當用戶在 CarPlay 上啟動 Claude 應用程式時，介面會顯示「開始免提對話」畫面。該畫面提供選項讓用戶建立新的聊天內容，或直接存取最近的對話記錄。這種設計旨在簡化操作流程，使用戶…
- 無預覽圖
OpenAI推新模型GPT-6 Astra CEO：已進入通用AI時代 - 星島頭條
OpenAI推新模型GPT-6 Astra CEO：已進入通用AI時代 星島頭條
- 無預覽圖
Daybreak for Frontline Defenders：投入 10 億美元保護關鍵服務
Daybreak for Frontline Defenders: $1B to protect essential services
OpenAI 推出 Daybreak for Frontline Defenders，承諾 10 億美元，擴大關鍵服務取得前沿網絡安全 AI、培訓與支援。
NeoMME：高效原生多模態、多語言模型編碼器
NeoMME: an efficient Multimodal-native and Multilingual Encoder
OpenAI 下一代大模型據稱「已進入 AGI 時代」
OpenAI’s next big AI model has ‘entered the AGI era’
OpenAI 下一代模型 GPT-6 Astra 登場，公司稱之為網絡安全、專業工作、軟件工程、科學同電腦操作上的「世代躍進」。
報道：Accel 洽談領投 Thinking Machines 10 億美元融資，估值 400 億
Accel reportedly in talks to lead $1B round for Thinking Machines at $40B valuation
這家高調新創年化收入已超過 1 億美元。
Nvidia 確認 130 億美元收購開源權重 AI 平台 Hugging Face
Nvidia confirms $13 billion acquisition of open-weight AI platform Hugging Face
Google 推出 Gemini 3.8 Flash，強化編程同複雜任務能力
Google launches Gemini 3.8 Flash to strengthen coding and complex-task performance
一段時間以來各家都在更新 AI 模型，Google 顯得較慢；而家推出 Gemini 3.8 Flash，加強編程與複雜任務。
- 無預覽圖
CrowdStrike、Palo Alto 之後，Zscaler 再爭 AI 推理網絡安全賽道，銷售管線增 75%
After CrowdStrike and Palo Alto, Zscaler joins the AI-inference cybersecurity race with a 75% pipeline jump
繼 CrowdStrike 與 Palo Alto 之後，Zscaler 以急增的銷售管線競逐 AI 推理時代的網絡安全市場。
- 無預覽圖
OpenAI 總裁稱人工智能而家已達人類能力水平
OpenAI’s president claims AI is as capable as humans now - The Washington Post
華盛頓郵報：OpenAI 總裁聲稱 AI 能力已等同人類。
- 無預覽圖
ATV Big Air Tour 用 ChatGPT 把三天工作壓縮成三小時
ATV Big Air Tour turned 3 days of work into 3 hours with ChatGPT
ATV Big Air Tour 用 ChatGPT Work 加快市場推廣同商品作業，甚至 15 分鐘內把商品相片做成庫存網站。
用 IBM 時間序列模型喺 Confluent 上做實時智能
Real-Time Intelligence with IBM Time Series Models on Confluent
以簡潔、可擴展嘅方式接入人工智能
Facilitating AI integration with simplicity at scale
企業一擴張，支撐營運嘅科技可以迅速由資產變成包袱：系統割裂、場地專用工具、試算表同人手變通會造成資料孤島。
Google 稱新 Gemini 3.8 Flash「更肯用腦」，但可能更貴
Google says its new Gemini 3.8 Flash model ‘works harder’ but might cost more
Google 推出 Gemini 3.8 Flash，距上一代只隔數週。公司稱新模型會做更多推理步驟，因而可能更貴。
Builders Stage 將實用擴張策略帶到 TechCrunch Disrupt 2026
The Builders Stage brings practical strategies for scaling startups to TechCrunch Disrupt 2026
Builders Stage 重返 TechCrunch Disrupt，召集創辦人、營運者同投資者，談點樣把公司做大。
一間 AI 訓練數據新創成為 Y Combinator 史上最快獨角獸
An AI Training Data Startup Just Became Y Combinator's Fastest-Ever Unicorn
Anthropic 為 Mac 推出 Claude 背景電腦操作，方便同時處理多項任務
Anthropic launches Claude background computer use on Mac for multitasking
Anthropic 宣布 Claude 喺 macOS 支援背景電腦操作，主要用於 Claude Code 同 Cowork，限 Pro／Max。後台執行可自動點擊、輸入，讓用戶同時做其他事。
- 無預覽圖
紐約擬禁止公立高中以下學生用 AI，以防心理健康同認知發展受損
New York may ban AI for public-school students below high school to protect mental health and development
星島報道：紐約擬禁公校高中以下學生使用 AI，擔心影響心理健康與認知發展。
- 無預覽圖
原生 AI 公司點樣把工作流變成營運能力
How AI-native companies turn workflows into operating capability
Basis、Clay 同 Exa Labs 用 AI 代理改善入職、客戶管理同開發者整合，企業主管可作參考。
BenchMIRT：大語言模型基準到底喺量度咩？
BenchMIRT: What are LLM benchmarks actually measuring?
Google 2026 年 8 月公布嘅最新 AI 消息
The latest AI news we announced in August 2026
以下係 Google 2026 年 8 月嘅 AI 更新摘要。
Google 比荷里活更需要片廠；片廠未必更需要 AI
Google needs Hollywood more than the studios need AI
報道指 Google 接觸荷里活大型片廠，希望簽授權協議，用受版權保護嘅材料訓練模型。
報道：AfterQuery 成為 YC 史上最快獨角獸，估值 32 億美元
AfterQuery reportedly becomes Y Combinator’s fastest-ever unicorn, now valued at $3.2B
AI 模型訓練新創 AfterQuery 據報融資後估值 32 億美元，距 4 月 3,000 萬美元 A 輪（估值 3 億）只隔五個月。
美國在 G20 科技會議促請對 AI 監管採取放手態度
US urges hands-off approach to AI regulation at G20 tech meeting
鼓勵用 AI 嘅新招？部分 Google 搜尋結果被 AI 概覽佔滿成頁
A new push to use AI? Some Google results are filled by AI Overviews
Google 搜尋除傳統連結列表外，部分結果頁被 AI 概覽答案佔去大量版面。
- 無預覽圖
理財最前線：AI 由「大腦」轉向五感，三隻台股卡位感知商機
AI shifts from ‘brain’ to the five senses: three Taiwan stocks positioned for perception
鏡週刊：AI 由語言模型轉向感知，三隻台股卡位相關商機。
- 無預覽圖
Polimill 為日本打造下一代公共 AI 基礎設施
Polimill builds Japan's next-generation public AI infrastructure
Polimill 用 OpenAI GPT 同 Codex，幫地方政府搜尋、使用行政知識，並加快開發。
Hugging Face 被入侵事件，或反映 OpenAI 文化問題
The Hugging Face hack could indicate cultural issues at OpenAI
MIT Technology Review《The Algorithm》：上月重大 AI 安全事故或指向 OpenAI 內部文化。
Debian 唔會禁止 Linux 發行版使用 AI 程式碼
Debian won’t ban AI code from its Linux distribution
Debian 投票允許開發者在開發、維護同文件中使用 AI 工具，新政策承認「負責任」使用可以有幫助。
本日熱門（流量）
以香港時區計算本日瀏覽次數；新站或流量不足時會以最新文章補位。
最高評分
至少累積一定數量評分後才會以平均分排序；不足時會以最新文章補位。
全部文章
店舖 WhatsApp：點樣用 Chat Core 接預約（同唔該自動答嘅）
Shop WhatsApp: how Chat Core takes bookings—and what it must not auto-answer
AutoChat 嘅路徑係填電話、試對話、預約入簿；真客先至接 Meta。呢篇講香港店舖點樣落地，以及邊啲問題必須人審。
AdReel 同 Buffer／Hootsuite：差在「交件從何而來」
AdReel vs Buffer / Hootsuite: where delivery packs come from
排程工具強在日曆同分析；AdReel 強在產品相到投放交件：多畫幅靜圖／短廣告、客戶審圖、ZIP 對 Ads Manager。創作者日更交 KOL Studio。
物管同教學點樣用 Eco‑Vision：多階段拍攝，而唔係一張遠景
Eco‑Vision on rounds and in class: staged photos, not one distant shot
香港行道樹同屋苑巡場需要可覆核紀錄。Eco‑Vision 跟《齊來認認樹》分階段拍；官方圖資同白標要另開，示範前端唔會假裝已接。
怎樣評估 LLM 應用：由示範走向量產
Evaluating LLM apps: from demo to production
用固定測試集、人類抽檢與線上指標，逐步把「感覺不錯」變成可追蹤的品質。
RAG 入門：為何要「先檢索、再生成」
RAG basics: retrieve, then generate
把公司內部文件變成可檢索片段，再交給模型組織答案，可減少幻覺並附上出處。
寫好提示詞：先結構、後細節
Prompting: structure first, details second
把任務拆成角色、輸出格式、限制條件，通常比一大段自由文字更穩定；適合日常寫作與工作流程。
點解可以自動更新？
每日訂閱出版方公開嘅 RSS（唔爬 Google／Yahoo 搜尋頁），只保存標題、短摘要同原文連結，點進去係對方網站。內文筆記仍由團隊撰寫。