<?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：Opus 4.5</title>
  <link>https://aiwulin.itsmygo.uk/tools/opus-4-5/</link>
  <description>AI 武林收錄的內容裡，最新提到 Opus 4.5 的 21 筆，每筆附中文摘要。</description>
  <language>zh-TW</language>
  <lastBuildDate>Thu, 08 Oct 2026 17:22:41 GMT</lastBuildDate>
  <atom:link href="https://aiwulin.itsmygo.uk/tools/opus-4-5/rss.xml" rel="self" type="application/rss+xml"/>
  <item>
    <title>CoreWeave ARENA: Serverless inference and serverless RL</title>
    <link>https://aiwulin.itsmygo.uk/c/f6a3359d97/</link>
    <guid isPermaLink="false">aiwulin-f6a3359d97</guid>
    <pubDate>Wed, 30 Sep 2026 00:00:00 +0800</pubDate>
    <dc:creator>CoreWeave</dc:creator>
    <description>&lt;p&gt;展示如何利用提示工程讓開源模型在 CoreWeave Serverless Inference 上接近前鋒模型的準確度。透過比較 GLM 5.2 與 Opus 4.5，分析不同提示技術對準確度與成本的影響。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.youtube.com/watch?v=34w2NQsh974&quot;&gt;https://www.youtube.com/watch?v=34w2NQsh974&lt;/a&gt;&lt;/p&gt;</description>
    <category>影片</category>
    <category>推論與部署</category>
  </item>
  <item>
    <title>#501 – DHH: Future of Programming, AI, Agentic Engineering, Vibe Coding &amp; Linux</title>
    <link>https://aiwulin.itsmygo.uk/c/396f18c22d/</link>
    <guid isPermaLink="false">aiwulin-396f18c22d</guid>
    <pubDate>Thu, 27 Aug 2026 00:00:00 +0800</pubDate>
    <dc:creator>Lex Fridman Podcast</dc:creator>
    <description>&lt;p&gt;DHH 分享從手寫程式碼到擁抱 AI 代理工程師的轉變，描述 Opus 4.5 與 Fable 等模型如何讓 AI 從輔助工具進化為主導者。他建議利用多個模型交叉驗證以確保品質，並指出未來程式開發可能不再需要人工逐行編寫。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://lexfridman.com/dhh-2&quot;&gt;https://lexfridman.com/dhh-2&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>Coding Agent</category>
    <category>AI Agent 基礎</category>
    <category>Vibe Coding</category>
  </item>
  <item>
    <title>Stanford MS&amp;E435 Economics of the AI Supercycle | Spring 2026 | Applications, Coding AI</title>
    <link>https://aiwulin.itsmygo.uk/c/ca8fb85d0c/</link>
    <guid isPermaLink="false">aiwulin-ca8fb85d0c</guid>
    <pubDate>Wed, 24 Jun 2026 00:00:00 +0800</pubDate>
    <dc:creator>Stanford Online</dc:creator>
    <description>&lt;p&gt;Guillermo Rauch 分享 Vercel 如何從開發者體驗出發，利用開放原始碼建立基礎設施，並隨著編碼代理（coding agents）的興起，轉型為 AI 應用與代理的部署平台。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.youtube.com/watch?v=HA7lZd7zk3M&quot;&gt;https://www.youtube.com/watch?v=HA7lZd7zk3M&lt;/a&gt;&lt;/p&gt;</description>
    <category>影片</category>
    <category>Coding Agent</category>
    <category>AI Agent 基礎</category>
  </item>
  <item>
    <title>Some ideas for what comes next, May 2026</title>
    <link>https://aiwulin.itsmygo.uk/c/4bb8052776/</link>
    <guid isPermaLink="false">aiwulin-4bb8052776</guid>
    <pubDate>Tue, 26 May 2026 00:00:00 +0800</pubDate>
    <dc:creator>Interconnects</dc:creator>
    <description>&lt;p&gt;Nathan Lambert 探討 2026 年 AI 發展趨勢，指出開放模型尚未達到像 Opus 4.5 般的代理時刻，Gemini 在特定任務上仍無法與 Claude Code 競爭。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.interconnects.ai/p/some-ideas-for-what-comes-next-may&quot;&gt;https://www.interconnects.ai/p/some-ideas-for-what-comes-next-may&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>開源模型</category>
    <category>Coding Agent</category>
    <category>模型發布與實測</category>
  </item>
  <item>
    <title>Agentic Coding and the Economics of Open Source</title>
    <link>https://aiwulin.itsmygo.uk/c/25f18d16d2/</link>
    <guid isPermaLink="false">aiwulin-25f18d16d2</guid>
    <pubDate>Thu, 02 Apr 2026 00:00:00 +0800</pubDate>
    <dc:creator>Practical AI</dc:creator>
    <description>&lt;p&gt;央歐洲大學經濟學教授 Miklós Koren，探討「Vibe Coding」如何改變軟體開發與開源生態。透過經濟學視角分析激勵機制與注意力稀缺性，並用資料展示 AI 推薦如何增加套件下載卻減少 GitHub 星數。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://share.transistor.fm/s/7d8e0293&quot;&gt;https://share.transistor.fm/s/7d8e0293&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>Coding Agent</category>
    <category>Vibe Coding</category>
  </item>
  <item>
    <title>Ep 82: Behind Legora's $550M Raise, Model Competition, Doubling Revenue Every Quarter, &amp; US Expansion</title>
    <link>https://aiwulin.itsmygo.uk/c/3c7a7a3989/</link>
    <guid isPermaLink="false">aiwulin-3c7a7a3989</guid>
    <pubDate>Wed, 11 Mar 2026 00:00:00 +0800</pubDate>
    <dc:creator>Unsupervised Learning</dc:creator>
    <description>&lt;p&gt;Max Jungestål 與 Jacob Effron 訪談 Legora 團隊，探討其 5.5 億美元融資、Opus 4.5/4.6 模型帶來的產品快速迭代，以及法律產業如何透過競爭壓力與企業客戶需求採用 AI。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://unsupervised-learning.simplecast.com/episodes/ep-82-behind-legoras-550m-raise-model-competition-doubling-revenue-every-quarter-us-expansion-NZdtRFa6&quot;&gt;https://unsupervised-learning.simplecast.com/episodes/ep-82-behind-legoras-550m-raise-model-competition-doubling-revenue-every-quarter-us-expansion-NZdtRFa6&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>AI 商業與創業</category>
  </item>
  <item>
    <title>Cursor's Third Era: Cloud Agents</title>
    <link>https://aiwulin.itsmygo.uk/c/969ea4e52c/</link>
    <guid isPermaLink="false">aiwulin-969ea4e52c</guid>
    <pubDate>Fri, 06 Mar 2026 00:00:00 +0800</pubDate>
    <dc:creator>Latent Space</dc:creator>
    <description>&lt;p&gt;訪談 Cursor 創辦人 Sam Evans 與 Jonas 談論 Cursor 第三代的雲端代理功能，展示模型如何自主測試程式碼、生成操作影片並提供遠端控制檯。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.latent.space/p/cursor-third-era&quot;&gt;https://www.latent.space/p/cursor-third-era&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>AI 編輯器與開發工具</category>
    <category>AI Agent 基礎</category>
  </item>
  <item>
    <title>Opus 4.5 changed everything (Interview)</title>
    <link>https://aiwulin.itsmygo.uk/c/c95b878b4d/</link>
    <guid isPermaLink="false">aiwulin-c95b878b4d</guid>
    <pubDate>Sat, 28 Feb 2026 00:00:00 +0800</pubDate>
    <dc:creator>The Changelog</dc:creator>
    <description>&lt;p&gt;Burke Holland 與 Adam Stacoviak 訪談 Opus 4.5 如何成為 AI 程式設計的突破點，解決舊模型生成錯誤程式碼後卡住的問題。兩人探討未來每個人都可能成為「建構者」，並分享使用 AI 工具重建專案的實際經驗。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://changelog.com/podcast/678&quot;&gt;https://changelog.com/podcast/678&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>Coding Agent</category>
    <category>AI 編輯器與開發工具</category>
    <category>AI 工作術</category>
  </item>
  <item>
    <title>Episode 70: 1,400 Production AI Deployments</title>
    <link>https://aiwulin.itsmygo.uk/c/988e9a1d5b/</link>
    <guid isPermaLink="false">aiwulin-988e9a1d5b</guid>
    <pubDate>Thu, 12 Feb 2026 00:00:00 +0800</pubDate>
    <dc:creator>Vanishing Gradients</dc:creator>
    <description>&lt;p&gt;訪談 Zen ML 工程師 Alex Strick van Linterton，探討其維護的 LLM Ops 資料庫中超過 1,400 個生產部署案例。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://hugobowne.substack.com/p/episode-70-1400-production-ai-deployments&quot;&gt;https://hugobowne.substack.com/p/episode-70-1400-production-ai-deployments&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>大型語言模型入門</category>
    <category>AI Agent 基礎</category>
    <category>Coding Agent</category>
  </item>
  <item>
    <title>The AI Vampire</title>
    <link>https://aiwulin.itsmygo.uk/c/dcc13fb6ec/</link>
    <guid isPermaLink="false">aiwulin-dcc13fb6ec</guid>
    <pubDate>Wed, 11 Feb 2026 00:00:00 +0800</pubDate>
    <dc:creator>Steve Yegge（部落格）</dc:creator>
    <description>&lt;p&gt;Steve Yegge 探討 AI 如何讓人過勞，稱其為「AI 吸血鬼」，並指出 Opus 4.5/4.6 與 Claude Code 已大幅提升生產力。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://steve-yegge.medium.com/the-ai-vampire-eda6e4f07163&quot;&gt;https://steve-yegge.medium.com/the-ai-vampire-eda6e4f07163&lt;/a&gt;&lt;/p&gt;</description>
    <category>文章</category>
    <category>AI 工作術</category>
    <category>Claude Code</category>
  </item>
  <item>
    <title>The 80% Problem in Agentic Coding</title>
    <link>https://aiwulin.itsmygo.uk/c/25a13937c9/</link>
    <guid isPermaLink="false">aiwulin-25a13937c9</guid>
    <pubDate>Thu, 29 Jan 2026 00:00:00 +0800</pubDate>
    <dc:creator>Addy Osmani（部落格）</dc:creator>
    <description>&lt;p&gt;探討 AI 代理編碼從 70% 進步到 80% 的現象，指出錯誤從語法錯誤轉向概念性失敗，並引入「理解債務」概念。作者警告過度依賴 AI 會導致編碼能力退化，並強調從命令式轉向宣告式開發的重要性，以維持長期競爭力。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://addyo.substack.com/p/the-80-problem-in-agentic-coding&quot;&gt;https://addyo.substack.com/p/the-80-problem-in-agentic-coding&lt;/a&gt;&lt;/p&gt;</description>
    <category>文章</category>
    <category>Coding Agent</category>
  </item>
  <item>
    <title>Raising An Agent - Episode 9</title>
    <link>https://aiwulin.itsmygo.uk/c/06e3d658b5/</link>
    <guid isPermaLink="false">aiwulin-06e3d658b5</guid>
    <pubDate>Thu, 22 Jan 2026 00:00:00 +0800</pubDate>
    <dc:creator>Amp, Inc.</dc:creator>
    <description>&lt;p&gt;探討 Gemini 3 與 Opus 4.5 帶來的變化，分析 AI 代理如何改變軟體開發與工作流，並分享 Amp 團隊的演進與未來規劃。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.youtube.com/watch?v=2wjnV6F2arc&quot;&gt;https://www.youtube.com/watch?v=2wjnV6F2arc&lt;/a&gt;&lt;/p&gt;</description>
    <category>影片</category>
    <category>AI Agent 基礎</category>
    <category>Coding Agent</category>
  </item>
  <item>
    <title>Ralph OS: a vibe coded operating system</title>
    <link>https://aiwulin.itsmygo.uk/c/0adb083410/</link>
    <guid isPermaLink="false">aiwulin-0adb083410</guid>
    <pubDate>Tue, 20 Jan 2026 00:00:00 +0800</pubDate>
    <dc:creator>Geoffrey Huntley</dc:creator>
    <description>&lt;p&gt;展示由 Mark Probst 開發的 Ralph OS 作業系統，該系統以 Rust 編寫並執行於 x86-64 架構，具備協同多執行緒功能。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.youtube.com/watch?v=IeTe3kbwmAU&quot;&gt;https://www.youtube.com/watch?v=IeTe3kbwmAU&lt;/a&gt;&lt;/p&gt;</description>
    <category>影片</category>
    <category>Coding Agent</category>
  </item>
  <item>
    <title>EP123 - 馬斯克談未來、Anthropic 上市能買嗎、蘋果認輸、ChatGPT 廣告</title>
    <link>https://aiwulin.itsmygo.uk/c/8c86d3425a/</link>
    <guid isPermaLink="false">aiwulin-8c86d3425a</guid>
    <pubDate>Mon, 19 Jan 2026 00:00:00 +0800</pubDate>
    <dc:creator>科技浪</dc:creator>
    <description>&lt;p&gt;快速回顧 Anthropic 推出最強編碼模型 Opus 4.5 及 Cowork 功能，並討論蘋果轉向使用 Gemini 的 Siri 策略，以及 OpenAI 在 ChatGPT 推出廣告。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.youtube.com/watch?v=PlqxDubmaGg&quot;&gt;https://www.youtube.com/watch?v=PlqxDubmaGg&lt;/a&gt;&lt;/p&gt;</description>
    <category>影片</category>
    <category>模型發布與實測</category>
    <category>AI 工作術</category>
    <category>開源模型</category>
  </item>
  <item>
    <title>Claude Code and What Comes Next</title>
    <link>https://aiwulin.itsmygo.uk/c/4e544687a2/</link>
    <guid isPermaLink="false">aiwulin-4e544687a2</guid>
    <pubDate>Thu, 08 Jan 2026 00:00:00 +0800</pubDate>
    <dc:creator>Ethan Mollick（部落格）</dc:creator>
    <description>&lt;p&gt;示範 Claude Code 如何自主編寫程式、部署網站並進行使用者測試，展現 AI 編碼工具的自主能力躍升。讀者可了解其運作原理、克服語意窗限制的技巧（如壓縮對話、使用技能與子代理），以及實際操作建議。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.oneusefulthing.org/p/claude-code-and-what-comes-next&quot;&gt;https://www.oneusefulthing.org/p/claude-code-and-what-comes-next&lt;/a&gt;&lt;/p&gt;</description>
    <category>文章</category>
    <category>Claude Code</category>
  </item>
  <item>
    <title>What AI Engineering Looks Like at Meta, Coinbase, ServiceTitan and ThoughtWorks</title>
    <link>https://aiwulin.itsmygo.uk/c/57f1cdf405/</link>
    <guid isPermaLink="false">aiwulin-57f1cdf405</guid>
    <pubDate>Tue, 06 Jan 2026 00:00:00 +0800</pubDate>
    <dc:creator>The AI Native Dev</dc:creator>
    <description>&lt;p&gt;邀請 Meta、Coinbase 等公司工程師，分享如何將 AI 融入工程團隊並避免 hype。內容涵蓋使用 Cursor 與 Claude Code 提升生產力、利用多代理模式測試新模型，以及 ServiceTitan 如何透過標準化任…&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.buzzsprout.com/2375985/episodes/18461019-what-ai-engineering-looks-like-at-meta-coinbase-servicetitan-and-thoughtworks&quot;&gt;https://www.buzzsprout.com/2375985/episodes/18461019-what-ai-engineering-looks-like-at-meta-coinbase-servicetitan-and-thoughtworks&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>AI 輔助軟體工程</category>
  </item>
  <item>
    <title>Claude Code is Amazing... Until It DELETES Production</title>
    <link>https://aiwulin.itsmygo.uk/c/a02c225523/</link>
    <guid isPermaLink="false">aiwulin-a02c225523</guid>
    <pubDate>Mon, 05 Jan 2026 00:00:00 +0800</pubDate>
    <dc:creator>IndyDevDan</dc:creator>
    <description>&lt;p&gt;示範如何透過 Claude Code Hooks 建立防護機制，防止 AI 代理因幻覺執行刪除資料庫或檔案等毀滅性指令。觀眾可學習設定本地、全域及提示式 Hook，並利用 patterns.yaml 檔規範權限，確保生產環境安全。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.youtube.com/watch?v=VqDs46A8pqE&quot;&gt;https://www.youtube.com/watch?v=VqDs46A8pqE&lt;/a&gt;&lt;/p&gt;</description>
    <category>影片</category>
    <category>Claude Code</category>
  </item>
  <item>
    <title>TOP 2% Engineering: /PLAN 2026</title>
    <link>https://aiwulin.itsmygo.uk/c/f6c0475e4a/</link>
    <guid isPermaLink="false">aiwulin-f6c0475e4a</guid>
    <pubDate>Mon, 22 Dec 2025 00:00:00 +0800</pubDate>
    <dc:creator>IndyDevDan</dc:creator>
    <description>&lt;p&gt;探討在 Gemini 3 Flash、Opus 4.5 與 GPT-5.2 等強大模型下，工程師如何透過建立信任來成為頂尖 2%。內容聚焦於具體的代理程式設計策略與工具呼叫技巧，幫助讀者掌握未來 AI 開發的關鍵方向。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.youtube.com/watch?v=u-SQ0Jsv4mI&quot;&gt;https://www.youtube.com/watch?v=u-SQ0Jsv4mI&lt;/a&gt;&lt;/p&gt;</description>
    <category>影片</category>
    <category>模型發布與實測</category>
    <category>Coding Agent</category>
    <category>AI 輔助軟體工程</category>
  </item>
  <item>
    <title>Auto-grading decade-old Hacker News discussions with hindsight</title>
    <link>https://aiwulin.itsmygo.uk/c/0bf64185fe/</link>
    <guid isPermaLink="false">aiwulin-0bf64185fe</guid>
    <pubDate>Wed, 10 Dec 2025 00:00:00 +0800</pubDate>
    <dc:creator>Andrej Karpathy（部落格）</dc:creator>
    <description>&lt;p&gt;Andrej Karpathy 示範如何用大型語言模型回溯分析十年前的 Hacker News 討論，自動評分預測準確度。讀者可學習如何構建「未來預測者」系統，並了解大模型處理歷史資料的實際應用與成本。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://karpathy.bearblog.dev/auto-grade-hn&quot;&gt;https://karpathy.bearblog.dev/auto-grade-hn&lt;/a&gt;&lt;/p&gt;</description>
    <category>文章</category>
    <category>大型語言模型入門</category>
  </item>
  <item>
    <title>Build Crew Live: Episode 8: Prompt Shootout and the importance of feedback loops</title>
    <link>https://aiwulin.itsmygo.uk/c/ad7d26d96c/</link>
    <guid isPermaLink="false">aiwulin-ad7d26d96c</guid>
    <pubDate>Sat, 06 Dec 2025 00:00:00 +0800</pubDate>
    <dc:creator>Amp, Inc.</dc:creator>
    <description>&lt;p&gt;Thorsten Ball 在 Build Crew Live 第 8 集中，透過實際編碼示範 AMP 工具除錯終端機問題，並進行提示詞比對實驗。觀眾可學習如何建立回饋迴路以最佳化效能，並掌握提示詞對代理驅動開發的影響。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.youtube.com/watch?v=rCLvd_Csp-8&quot;&gt;https://www.youtube.com/watch?v=rCLvd_Csp-8&lt;/a&gt;&lt;/p&gt;</description>
    <category>影片</category>
    <category>提示工程</category>
    <category>Coding Agent</category>
  </item>
  <item>
    <title>Effective harnesses for long-running agents</title>
    <link>https://aiwulin.itsmygo.uk/c/c2414e3a01/</link>
    <guid isPermaLink="false">aiwulin-c2414e3a01</guid>
    <pubDate>Wed, 26 Nov 2025 00:00:00 +0800</pubDate>
    <dc:creator>Anthropic Engineering Blog</dc:creator>
    <description>&lt;p&gt;探討長程 AI 代理如何在有限上下文視窗下持續工作，提出由初始化代理與編碼代理組成的雙重解決方案。透過建立功能清單、Git 歷史與進度檔案，讓代理能 incremental 地開發並留下乾淨狀態，避免重複勞動與錯誤累積。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.anthropic.com/engineering/effective-harnesses-for-long-running-agents&quot;&gt;https://www.anthropic.com/engineering/effective-harnesses-for-long-running-agents&lt;/a&gt;&lt;/p&gt;</description>
    <category>文章</category>
    <category>AI Agent 基礎</category>
    <category>Coding Agent</category>
  </item>
</channel>
</rss>
