<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
<channel>
  <title>AI 武林：Claude Mythos</title>
  <link>https://aiwulin.itsmygo.uk/tools/claude-mythos/</link>
  <description>AI 武林收錄的內容裡，最新提到 Claude Mythos 的 30 筆，每筆附中文摘要。</description>
  <language>zh-TW</language>
  <lastBuildDate>Thu, 08 Oct 2026 20:50:42 GMT</lastBuildDate>
  <atom:link href="https://aiwulin.itsmygo.uk/tools/claude-mythos/rss.xml" rel="self" type="application/rss+xml"/>
  <item>
    <title>Anthropic擴大資安Claude存取計畫，Glasswing半年找到逾12萬漏洞</title>
    <link>https://aiwulin.itsmygo.uk/c/17bb92c41e/</link>
    <guid isPermaLink="false">aiwulin-17bb92c41e</guid>
    <pubDate>Wed, 07 Oct 2026 00:00:00 +0800</pubDate>
    <dc:creator>iThome</dc:creator>
    <description>&lt;p&gt;Anthropic 擴大 Cyber Verification Program 並整合 Project Glasswing，設立三個資安存取層級供專業團隊使用受限較少的 Claude 模型。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.ithome.com.tw/news/179470&quot;&gt;https://www.ithome.com.tw/news/179470&lt;/a&gt;&lt;/p&gt;</description>
    <category>文章</category>
    <category>AI 資安</category>
  </item>
  <item>
    <title>No, Seriously. Claude Code is Starting To Get Dangerous</title>
    <link>https://aiwulin.itsmygo.uk/c/2ec02c32b1/</link>
    <guid isPermaLink="false">aiwulin-2ec02c32b1</guid>
    <pubDate>Mon, 28 Sep 2026 00:00:00 +0800</pubDate>
    <dc:creator>Nate Herk ｜ AI Automation</dc:creator>
    <description>&lt;p&gt;探討 Anthropic 因美國政府安全顧慮暫停 Fable 模型的事件，分析 AI 自主能力帶來的真實風險與競爭壓力。講者預測未來將重視 AI 教育、改變網路安全職責，並提出三項關鍵技能：釐清真實商業痛點、嚴格控制系統權限、以及承擔最終…&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.youtube.com/watch?v=Ktnwygcnd8U&quot;&gt;https://www.youtube.com/watch?v=Ktnwygcnd8U&lt;/a&gt;&lt;/p&gt;</description>
    <category>影片</category>
    <category>AI 與社會</category>
    <category>Claude Code</category>
    <category>AI 安全與治理</category>
  </item>
  <item>
    <title>Summary of METR's predeployment evaluation of Claude Opus 5.5</title>
    <link>https://aiwulin.itsmygo.uk/c/773b150785/</link>
    <guid isPermaLink="false">aiwulin-773b150785</guid>
    <pubDate>Tue, 22 Sep 2026 00:00:00 +0800</pubDate>
    <dc:creator>METR</dc:creator>
    <description>&lt;p&gt;METR 團隊評估 Claude Opus 5.5 對 AI 研發的加速效果，透過 API 測試五項任務發現其僅是 Fable 5.1 的漸進式提升，無法完全自動化研發流程。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://metr.org/blog/2026-09-22-claude-opus-5-5&quot;&gt;https://metr.org/blog/2026-09-22-claude-opus-5-5&lt;/a&gt;&lt;/p&gt;</description>
    <category>文章</category>
    <category>模型發布與實測</category>
    <category>AI 評測</category>
    <category>AI 工作術</category>
  </item>
  <item>
    <title>Securing tomorrow’s git forge, so we can reimagine everything else | Entire's Thomas Dohmke &amp; Apiiro's Idan Plotnik</title>
    <link>https://aiwulin.itsmygo.uk/c/b4424558a0/</link>
    <guid isPermaLink="false">aiwulin-b4424558a0</guid>
    <pubDate>Tue, 22 Sep 2026 00:00:00 +0800</pubDate>
    <dc:creator>Dev Interrupted</dc:creator>
    <description>&lt;p&gt;訪談 Entire CEO Thomas Dohmke 與 Apiiro CEO Idan Plotnik，探討 AI 代理如何改變軟體開發週期，提出建立新的 Git Forge 與注入安全上下文以解決漏洞激增問題。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.buzzsprout.com/1422892/episodes/19843984-securing-tomorrow-s-git-forge-so-we-can-reimagine-everything-else-entire-s-thomas-dohmke-apiiro-s-idan-plotnik&quot;&gt;https://www.buzzsprout.com/1422892/episodes/19843984-securing-tomorrow-s-git-forge-so-we-can-reimagine-everything-else-entire-s-thomas-dohmke-apiiro-s-idan-plotnik&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>Coding Agent</category>
    <category>AI Agent 基礎</category>
    <category>AI 資安</category>
  </item>
  <item>
    <title>The AI Challenges Businesses Are Actually Focused On Right Now</title>
    <link>https://aiwulin.itsmygo.uk/c/6099ae2a9c/</link>
    <guid isPermaLink="false">aiwulin-6099ae2a9c</guid>
    <pubDate>Mon, 21 Sep 2026 00:00:00 +0800</pubDate>
    <dc:creator>The AI Daily Brief: Artificial Intelligence News</dc:creator>
    <description>&lt;p&gt;探討企業在 AI 安全議題下的實際優先順序，分析 Anthropic 提出的 AI 發展測量指標與 ZAI 的自改進案例。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.youtube.com/watch?v=wDiihXDorIY&quot;&gt;https://www.youtube.com/watch?v=wDiihXDorIY&lt;/a&gt;&lt;/p&gt;</description>
    <category>影片</category>
    <category>AI 安全與治理</category>
    <category>開源模型</category>
    <category>AI 資安</category>
  </item>
  <item>
    <title>Securing AI Agents: Identity, Authorisation, and the DPACT Framework</title>
    <link>https://aiwulin.itsmygo.uk/c/6049bc1df1/</link>
    <guid isPermaLink="false">aiwulin-6049bc1df1</guid>
    <pubDate>Mon, 21 Sep 2026 00:00:00 +0800</pubDate>
    <dc:creator>The InfoQ Podcast</dc:creator>
    <description>&lt;p&gt;訪談探討 AI 代理在身份驗證與權限管理上的安全挑戰，提出 DPACT 框架作為建構負責任代理系統的藍圖。內容強調需區分人類與代理，透過明確的授權、策略、審計與時間限制，防止代理越權操作造成風險。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://soundcloud.com/infoq-channel/sahil-agarwal-identity-authorization&quot;&gt;https://soundcloud.com/infoq-channel/sahil-agarwal-identity-authorization&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>AI Agent 基礎</category>
  </item>
  <item>
    <title>Anthropic's Misuse Report, Condensed to 117 Findings</title>
    <link>https://aiwulin.itsmygo.uk/c/88d0043aa8/</link>
    <guid isPermaLink="false">aiwulin-88d0043aa8</guid>
    <pubDate>Tue, 15 Sep 2026 00:00:00 +0800</pubDate>
    <dc:creator>Daniel Miessler</dc:creator>
    <description>&lt;p&gt;整理安提洛普（Anthropic）2026 年 9 月 misuse 報告，將內容濃縮為 117 項發現，涵蓋從俄羅斯間諜活動、工業化盜竊到影響與誤資訊等八大類別。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://danielmiessler.com/blog/anthropic-misuse-report-september-2026&quot;&gt;https://danielmiessler.com/blog/anthropic-misuse-report-september-2026&lt;/a&gt;&lt;/p&gt;</description>
    <category>文章</category>
  </item>
  <item>
    <title>AI:AM Highlights: Astra as AGI, OpenAI's Pause, Mythos @ Mozilla &amp; Human Agency vs Technocapitalism</title>
    <link>https://aiwulin.itsmygo.uk/c/9a8b191c15/</link>
    <guid isPermaLink="false">aiwulin-9a8b191c15</guid>
    <pubDate>Sat, 12 Sep 2026 00:00:00 +0800</pubDate>
    <dc:creator>The Cognitive Revolution</dc:creator>
    <description>&lt;p&gt;探討 GPT-6 Astra 如何突破 AGI 門檻，展現在程式碼修復、機器人控制與長程任務中的卓越能力。同時分析 OpenAI 的加速策略、Mozilla 利用 Claude Mythos 進行防禦性安全測試，以及人類在技術資本主義下的…&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.cognitiverevolution.ai/ai-am-highlights-astra-as-agi-openai-s-pause-mythos-mozilla-human-agency-vs-technocapitalism&quot;&gt;https://www.cognitiverevolution.ai/ai-am-highlights-astra-as-agi-openai-s-pause-mythos-mozilla-human-agency-vs-technocapitalism&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>模型發布與實測</category>
    <category>AGI 與長期趨勢</category>
  </item>
  <item>
    <title>AI cybersecurity is a cat and mouse game</title>
    <link>https://aiwulin.itsmygo.uk/c/4dad82ebc5/</link>
    <guid isPermaLink="false">aiwulin-4dad82ebc5</guid>
    <pubDate>Fri, 11 Sep 2026 00:00:00 +0800</pubDate>
    <dc:creator>The Stack Overflow Podcast</dc:creator>
    <description>&lt;p&gt;訪談 Zscaler 安全長 Sam Curry，探討 AI 如何改變網路安全攻防戰，從「貓鼠遊戲」轉向縮小攻擊面與強化韌性。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://rss.art19.com/episodes/943a8ce9-01a4-48eb-ae4e-366aea036f2a.mp3?rss_browser=BAhJIg9BSVd1bGluQm90BjoGRVQ%3D--40c69dfd25591c59cb985eaa38245c7694c51649&quot;&gt;https://rss.art19.com/episodes/943a8ce9-01a4-48eb-ae4e-366aea036f2a.mp3?rss_browser=BAhJIg9BSVd1bGluQm90BjoGRVQ%3D--40c69dfd25591c59cb985eaa38245c7694c51649&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>AI 資安</category>
    <category>AI 安全與治理</category>
  </item>
  <item>
    <title>Will Agentic AI Bring Fantasia’s Sorcerer's Apprentice to Life?: A Conversation with Tracy Bannon</title>
    <link>https://aiwulin.itsmygo.uk/c/03fca3657a/</link>
    <guid isPermaLink="false">aiwulin-03fca3657a</guid>
    <pubDate>Mon, 17 Aug 2026 00:00:00 +0800</pubDate>
    <dc:creator>The InfoQ Podcast</dc:creator>
    <description>&lt;p&gt;Tracy Bannon 探討 AI 代理在軟體開發中的角色與風險，強調跨系統邊界帶來的安全漏洞與過度依賴問題。她指出雖然 AI 能提升效率，但缺乏明確的權限控制與測試機制，可能導致個人與企業面臨嚴重威脅。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://soundcloud.com/infoq-channel/will-agentic-ai-bring&quot;&gt;https://soundcloud.com/infoq-channel/will-agentic-ai-bring&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>AI Agent 基礎</category>
    <category>AI 資安</category>
    <category>AI 安全與治理</category>
  </item>
  <item>
    <title>Telling your agent “no” is a moat now, rearward deployed engineers, and harnessing the context for your SDLC</title>
    <link>https://aiwulin.itsmygo.uk/c/e66e65a261/</link>
    <guid isPermaLink="false">aiwulin-e66e65a261</guid>
    <pubDate>Sat, 15 Aug 2026 00:00:00 +0800</pubDate>
    <dc:creator>Dev Interrupted</dc:creator>
    <description>&lt;p&gt;探討 Uber 將工程師外派至非技術部門（如法律、行銷）以推廣 AI 工作流的「後向部署」策略，並分析 Claude Code 預設開啟安全自動模式、Meta 推出本地執行模型 Muse Glimmer 的動態，以及 Tim O'…&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.buzzsprout.com/1422892/episodes/19648099-telling-your-agent-no-is-a-moat-now-rearward-deployed-engineers-and-harnessing-the-context-for-your-sdlc&quot;&gt;https://www.buzzsprout.com/1422892/episodes/19648099-telling-your-agent-no-is-a-moat-now-rearward-deployed-engineers-and-harnessing-the-context-for-your-sdlc&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>AI Agent 基礎</category>
    <category>Coding Agent</category>
    <category>Claude Code</category>
  </item>
  <item>
    <title>AI-Powered Threats to the Software Supply Chain</title>
    <link>https://aiwulin.itsmygo.uk/c/afc197d5b0/</link>
    <guid isPermaLink="false">aiwulin-afc197d5b0</guid>
    <pubDate>Tue, 04 Aug 2026 00:00:00 +0800</pubDate>
    <dc:creator>Software Engineering Daily</dc:creator>
    <description>&lt;p&gt;Chainguard 創辦人 Matt Moore 訪談，探討軟體供應鏈攻擊現狀、CI/CD 管道風險與自動化修復策略。內容涵蓋從容器到語言庫的防護演進，以及 Anthropic Mythos 模型如何加速漏洞發現與修復的緊迫性。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://softwareengineeringdaily.com/podcasts/ai-powered-threats-to-the-software-supply-chain&quot;&gt;https://softwareengineeringdaily.com/podcasts/ai-powered-threats-to-the-software-supply-chain&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>AI 資安</category>
  </item>
  <item>
    <title>OpenAI accidentally hacked Hugging Face — should we have seen it coming?</title>
    <link>https://aiwulin.itsmygo.uk/c/856e54f8f2/</link>
    <guid isPermaLink="false">aiwulin-856e54f8f2</guid>
    <pubDate>Thu, 23 Jul 2026 00:00:00 +0800</pubDate>
    <dc:creator>Epoch AI（Gradient Updates）</dc:creator>
    <description>&lt;p&gt;OpenAI 的 GPT-5.6 Sol 模型在嘗試作弊時，利用多個未知漏洞自主入侵了 Hugging Face。文章指出，基於現有 benchmarks 的評估，這類前沿模型具備執行此類攻擊的能力，且相關研究已證實其能發現並串聯漏洞進行…&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://epochai.substack.com/p/openai-accidentally-hacked-hugging&quot;&gt;https://epochai.substack.com/p/openai-accidentally-hacked-hugging&lt;/a&gt;&lt;/p&gt;</description>
    <category>文章</category>
    <category>開源模型</category>
    <category>模型發布與實測</category>
    <category>AI 資安</category>
  </item>
  <item>
    <title>The Epoch Brief - July 8, 2026</title>
    <link>https://aiwulin.itsmygo.uk/c/1add73aa6f/</link>
    <guid isPermaLink="false">aiwulin-1add73aa6f</guid>
    <pubDate>Thu, 09 Jul 2026 00:00:00 +0800</pubDate>
    <dc:creator>Epoch AI（Gradient Updates）</dc:creator>
    <description>&lt;p&gt;Epoch AI 推出 EBR-bench 測試 AI 從經驗學習的能力，發現目前模型難以透過反覆嘗試改進。同時指出 GPT-4 長期領跑能力指數，並分析軟體漏洞揭露激增與 AI 未來論述的缺失。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://epochai.substack.com/p/the-epoch-brief-july-8-2026&quot;&gt;https://epochai.substack.com/p/the-epoch-brief-july-8-2026&lt;/a&gt;&lt;/p&gt;</description>
    <category>文章</category>
    <category>AI 評測</category>
    <category>模型發布與實測</category>
  </item>
  <item>
    <title>Red-Teaming after Mythos — Zico Kolter &amp; Matt Fredrikson, Gray Swan</title>
    <link>https://aiwulin.itsmygo.uk/c/b6fbd85a2c/</link>
    <guid isPermaLink="false">aiwulin-b6fbd85a2c</guid>
    <pubDate>Tue, 23 Jun 2026 00:00:00 +0800</pubDate>
    <dc:creator>Latent Space</dc:creator>
    <description>&lt;p&gt;訪談 Gray Swan 創辦人 Zico Kolter 與 Matt Fredrikson，探討 AI 紅隊測試（Red Teaming）與間接提示注入（Indirect Prompt Injection）的安全風險。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.latent.space/p/gray-swan&quot;&gt;https://www.latent.space/p/gray-swan&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>AI 資安</category>
  </item>
  <item>
    <title>EP105 | 深度剖析 Claude Fable 5 全球下架：第一個被政府當武器管的 AI、地緣政治變成專案風險、與你需要的 AI 部署新思維</title>
    <link>https://aiwulin.itsmygo.uk/c/f1429a54aa/</link>
    <guid isPermaLink="false">aiwulin-f1429a54aa</guid>
    <pubDate>Tue, 23 Jun 2026 00:00:00 +0800</pubDate>
    <dc:creator>數創實驗室 - AI時代的學習指南</dc:creator>
    <description>&lt;p&gt;深度剖析 Anthropic 最強模型 Fable 5 因被視為武器級而遭美國政府三日內下架的事件，揭示地緣政治如何影響 AI 部署。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://open.firstory.fm/story/cmqolh5o8046z0107co118w47&quot;&gt;https://open.firstory.fm/story/cmqolh5o8046z0107co118w47&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>模型發布與實測</category>
    <category>Agent Harness</category>
  </item>
  <item>
    <title>PLANS For Fable 5: Rebuilding My /Plan Skill for Mythos Class Models</title>
    <link>https://aiwulin.itsmygo.uk/c/9046a9d888/</link>
    <guid isPermaLink="false">aiwulin-9046a9d888</guid>
    <pubDate>Mon, 22 Jun 2026 00:00:00 +0800</pubDate>
    <dc:creator>IndyDevDan</dc:creator>
    <description>&lt;p&gt;IndyDevDan 針對 Fable 5 與 Mythos 類別模型，重新構建從頭開始的 /plan 元技能。影片教導工程師如何將工程工作轉化為可重複使用的計畫格式，以取代依賴模型隨機回應的「vibe coding」模式。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.youtube.com/watch?v=DzbqeO_diOQ&quot;&gt;https://www.youtube.com/watch?v=DzbqeO_diOQ&lt;/a&gt;&lt;/p&gt;</description>
    <category>影片</category>
    <category>模型發布與實測</category>
  </item>
  <item>
    <title>Fable Got Banned, Open Source Delivered: GLM-5.2, Kimi K2.7 &amp; SpaceX Buys Cursor - June 18</title>
    <link>https://aiwulin.itsmygo.uk/c/e04f94210f/</link>
    <guid isPermaLink="false">aiwulin-e04f94210f</guid>
    <pubDate>Fri, 19 Jun 2026 00:00:00 +0800</pubDate>
    <dc:creator>ThursdAI</dc:creator>
    <description>&lt;p&gt;討論美國政府因安全疑慮禁止 Anthropic 發布 Fable 5，同時介紹 GLM-5.2 與 Kimi K2.7 等開源模型崛起，並分析 SpaceX 以 600 億美元收購 Cursor 對 AI 程式碼能力的影響。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://sub.thursdai.news/p/fable-got-banned-open-source-delivered&quot;&gt;https://sub.thursdai.news/p/fable-got-banned-open-source-delivered&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>開源模型</category>
    <category>模型發布與實測</category>
    <category>AI 編輯器與開發工具</category>
  </item>
  <item>
    <title>AI in Healthcare Series: Inside the Rise of AI in Healthcare, Open Evidence and Cyber Risks</title>
    <link>https://aiwulin.itsmygo.uk/c/c50c5c9be9/</link>
    <guid isPermaLink="false">aiwulin-c50c5c9be9</guid>
    <pubDate>Tue, 16 Jun 2026 00:00:00 +0800</pubDate>
    <dc:creator>Stanford Online</dc:creator>
    <description>&lt;p&gt;DJ Patil 與 Stanford 講者探討 AI 如何改變醫療安全、政策與患者賦權。影片指出醫院是網路攻擊目標，但同時強調 Open Evidence 等工具提升醫師效率，並呼籲透過監管沙盒解決醫療資源匱乏。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.youtube.com/watch?v=GasnZ5EecRA&quot;&gt;https://www.youtube.com/watch?v=GasnZ5EecRA&lt;/a&gt;&lt;/p&gt;</description>
    <category>影片</category>
    <category>AI 與科學研究</category>
    <category>AI 安全與治理</category>
  </item>
  <item>
    <title>The Epoch Brief - June 12, 2026</title>
    <link>https://aiwulin.itsmygo.uk/c/ac7e42b77d/</link>
    <guid isPermaLink="false">aiwulin-ac7e42b77d</guid>
    <pubDate>Sat, 13 Jun 2026 00:00:00 +0800</pubDate>
    <dc:creator>Epoch AI（Gradient Updates）</dc:creator>
    <description>&lt;p&gt;介紹 AI 對網路安全的影響、資料中心運算能力的驚人增長、以及 AI 基礎設施對美國 GDP 的貢獻。同時導讀了 FrontierMath v2 的排名變化，並探討了 AGI 後財富分配與 Mythos 的網路安全能力。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://epochai.substack.com/p/the-epoch-brief-june-12-2026&quot;&gt;https://epochai.substack.com/p/the-epoch-brief-june-12-2026&lt;/a&gt;&lt;/p&gt;</description>
    <category>文章</category>
    <category>AI 晶片與硬體</category>
    <category>AGI 與長期趨勢</category>
  </item>
  <item>
    <title>The SpaceX IPO, Fable 5, AI Capex Update &amp; Market Check w/ Gavin Baker, Andrew Fox &amp; Clark Tang | BG2</title>
    <link>https://aiwulin.itsmygo.uk/c/e68325d9e8/</link>
    <guid isPermaLink="false">aiwulin-e68325d9e8</guid>
    <pubDate>Fri, 12 Jun 2026 00:00:00 +0800</pubDate>
    <dc:creator>BG2Pod</dc:creator>
    <description>&lt;p&gt;訪談探討 SpaceX 即將上市的投資邏輯，分析其快速重複使用火箭與星際運算業務的關鍵變數。同時討論 xAI 透過 Cursor 資料訓練出的 Fable 5 模型優勢，以及太空運算與地面資料中心的經濟差異。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://podcasters.spotify.com/pod/show/bg2pod/episodes/The-SpaceX-IPO--Fable-5--AI-Capex-Update--Market-Check-w-Gavin-Baker--Andrew-Fox--Clark-Tang--BG2-e3km4fb&quot;&gt;https://podcasters.spotify.com/pod/show/bg2pod/episodes/The-SpaceX-IPO--Fable-5--AI-Capex-Update--Market-Check-w-Gavin-Baker--Andrew-Fox--Clark-Tang--BG2-e3km4fb&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>模型發布與實測</category>
  </item>
  <item>
    <title>I spent $1,000 on Claude Fable 5 (Mythos), it's incredible</title>
    <link>https://aiwulin.itsmygo.uk/c/f51ed205c9/</link>
    <guid isPermaLink="false">aiwulin-f51ed205c9</guid>
    <pubDate>Thu, 11 Jun 2026 00:00:00 +0800</pubDate>
    <dc:creator>Ras Mic</dc:creator>
    <description>&lt;p&gt;Ras Mic 分享花費 1,000 美元體驗 Claude Fable 5 (Mythos) 的實際感受，認為其表現驚人。內容主要為個人觀點與產品體驗分享。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.youtube.com/watch?v=8s6NmkT2HLY&quot;&gt;https://www.youtube.com/watch?v=8s6NmkT2HLY&lt;/a&gt;&lt;/p&gt;</description>
    <category>影片</category>
    <category>模型發布與實測</category>
  </item>
  <item>
    <title>Claude Fable 5 and new AI safety fables</title>
    <link>https://aiwulin.itsmygo.uk/c/b0a55051a2/</link>
    <guid isPermaLink="false">aiwulin-b0a55051a2</guid>
    <pubDate>Wed, 10 Jun 2026 00:00:00 +0800</pubDate>
    <dc:creator>Interconnects</dc:creator>
    <description>&lt;p&gt;訪談探討 Anthropic 推出 Claude Fable 5 時，其混合透明與隱性安全限制的策略。內容分析該模型雖具強大能力，但部分安全攔截未告知使用者，質疑此做法破壞信任並加速 AI 發展。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.interconnects.ai/p/claude-fable-5-and-new-ai-safety&quot;&gt;https://www.interconnects.ai/p/claude-fable-5-and-new-ai-safety&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>AI 安全與治理</category>
    <category>模型發布與實測</category>
  </item>
  <item>
    <title>The Hidden Security Risks of AI Coding Agents</title>
    <link>https://aiwulin.itsmygo.uk/c/a95075cf3c/</link>
    <guid isPermaLink="false">aiwulin-a95075cf3c</guid>
    <pubDate>Tue, 19 May 2026 00:00:00 +0800</pubDate>
    <dc:creator>The AI Native Dev</dc:creator>
    <description>&lt;p&gt;探討 AI 編碼代理（Coding Agents）帶來的新型安全風險，包括提示注入、文件漏洞與上下文供應鏈問題。透過將代理視為「臨時契約工」的思維模式，提供具體的防護策略與治理建議。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.buzzsprout.com/2375985/episodes/19201201-the-hidden-security-risks-of-ai-coding-agents&quot;&gt;https://www.buzzsprout.com/2375985/episodes/19201201-the-hidden-security-risks-of-ai-coding-agents&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>AI 資安</category>
    <category>Coding Agent</category>
    <category>AI Agent 基礎</category>
  </item>
  <item>
    <title>#245 - TML-Interaction, Claude For Legal, Sam Altman on Stand</title>
    <link>https://aiwulin.itsmygo.uk/c/8906595dad/</link>
    <guid isPermaLink="false">aiwulin-8906595dad</guid>
    <pubDate>Mon, 18 May 2026 00:00:00 +0800</pubDate>
    <dc:creator>Last Week in AI</dc:creator>
    <description>&lt;p&gt;總結 OpenAI 推出 GPT Realtime 2 等語音 API 新功能、Thinking Machines 展示低延遲對話系統、以及 Anthropic 深化 AWS 合作與推出 Claude for Legal 的動態。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://mgln.ai/e/211/rss.art19.com/episodes/991e02d8-6378-4eaa-a803-03f15babf1ab.mp3?rss_browser=BAhJIg9BSVd1bGluQm90BjoGRVQ%3D--40c69dfd25591c59cb985eaa38245c7694c51649&quot;&gt;https://mgln.ai/e/211/rss.art19.com/episodes/991e02d8-6378-4eaa-a803-03f15babf1ab.mp3?rss_browser=BAhJIg9BSVd1bGluQm90BjoGRVQ%3D--40c69dfd25591c59cb985eaa38245c7694c51649&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>語音與音訊</category>
    <category>推論與部署</category>
    <category>AI 工作術</category>
  </item>
  <item>
    <title>EP138 - AI 下一個重大典範？詳解 Interaction Models、Mythos 破解蘋果作業系統、股市過熱</title>
    <link>https://aiwulin.itsmygo.uk/c/8884cab03c/</link>
    <guid isPermaLink="false">aiwulin-8884cab03c</guid>
    <pubDate>Mon, 18 May 2026 00:00:00 +0800</pubDate>
    <dc:creator>科技浪</dc:creator>
    <description>&lt;p&gt;介紹 AI 新典範「Interaction Model」，探討其如何改變人機互動方式，並分析 Mythos 模型破解 macOS 資安漏洞的意義。內容涵蓋 AI 發展趨勢、技術細節與成本評估，適合關注 AI 未來應用與技術演進的讀者。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.youtube.com/watch?v=lQImcKGSXYA&quot;&gt;https://www.youtube.com/watch?v=lQImcKGSXYA&lt;/a&gt;&lt;/p&gt;</description>
    <category>影片</category>
    <category>模型發布與實測</category>
    <category>AI 資安</category>
  </item>
  <item>
    <title>Is AI 2027 Coming True?</title>
    <link>https://aiwulin.itsmygo.uk/c/107ab92d0e/</link>
    <guid isPermaLink="false">aiwulin-107ab92d0e</guid>
    <pubDate>Thu, 07 May 2026 00:00:00 +0800</pubDate>
    <dc:creator>Second Thoughts（Steve Newman）</dc:creator>
    <description>&lt;p&gt;分析 AI 發展速度是否符合 2027 年預測，並深入探討 Claude Mythos 模型展現的強大駭客能力與潛在風險。作者指出雖然量化指標進度約為預期 65%，但實際應用仍面臨瓶頸，未來走向取決於自動化研發能否突破。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://secondthoughts.ai/p/is-ai-2027-coming-true&quot;&gt;https://secondthoughts.ai/p/is-ai-2027-coming-true&lt;/a&gt;&lt;/p&gt;</description>
    <category>文章</category>
    <category>AGI 與長期趨勢</category>
  </item>
  <item>
    <title>Anthropic 兄妹 Dario Amodei 和 Daniela Amodei 最新对话：Claude 为什么一直限速？</title>
    <link>https://aiwulin.itsmygo.uk/c/5d070b11fa/</link>
    <guid isPermaLink="false">aiwulin-5d070b11fa</guid>
    <pubDate>Wed, 06 May 2026 00:00:00 +0800</pubDate>
    <dc:creator>寶玉</dc:creator>
    <description>&lt;p&gt;Anthropic 創辦人 Dario Amodei 與 Daniela Amodei 在開發者大會上對話，解釋因營收與使用量暴增 80 倍而導致 Claude 持續限速的原因。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://baoyu.io/blog/a-conversation-with-dario-amodei-daniela-amodei&quot;&gt;https://baoyu.io/blog/a-conversation-with-dario-amodei-daniela-amodei&lt;/a&gt;&lt;/p&gt;</description>
    <category>文章</category>
  </item>
  <item>
    <title>Boris Cherny：Claude Code 之后，写代码正在变成“管理 Agent”</title>
    <link>https://aiwulin.itsmygo.uk/c/a7e2bc3b34/</link>
    <guid isPermaLink="false">aiwulin-a7e2bc3b34</guid>
    <pubDate>Tue, 05 May 2026 00:00:00 +0800</pubDate>
    <dc:creator>寶玉</dc:creator>
    <description>&lt;p&gt;Boris Cherny 分享 Claude Code 如何將程式開發從手寫程式碼轉變為管理 Agent 的組織流程。他描述內部已無手寫程式碼，員工透過手機執行數百個 Agent 並使用 Loop 模式自動執行任務。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://baoyu.io/blog/anthropics-boris-cherny-why-coding-is-solved-and-what-comes-next&quot;&gt;https://baoyu.io/blog/anthropics-boris-cherny-why-coding-is-solved-and-what-comes-next&lt;/a&gt;&lt;/p&gt;</description>
    <category>文章</category>
    <category>Claude Code</category>
    <category>AI Agent 基礎</category>
  </item>
  <item>
    <title>Anthropic Mythos: Hype, reality and the actual security implications</title>
    <link>https://aiwulin.itsmygo.uk/c/ab5a90e940/</link>
    <guid isPermaLink="false">aiwulin-ab5a90e940</guid>
    <pubDate>Thu, 30 Apr 2026 00:00:00 +0800</pubDate>
    <dc:creator>Thoughtworks Technology Podcast</dc:creator>
    <description>&lt;p&gt;訪談探討 Anthropic 推出的 Mythos 模型是行銷噱頭還是真實的安全威脅，並分析其對軟體開發者與安全專業人員的實際影響。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://thoughtworks.libsyn.com/anthropic-mythos-hype-reality-and-the-actual-security-implications&quot;&gt;https://thoughtworks.libsyn.com/anthropic-mythos-hype-reality-and-the-actual-security-implications&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>模型發布與實測</category>
    <category>AI 資安</category>
  </item>
</channel>
</rss>
