<?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：Gemini 3</title>
  <link>https://aiwulin.itsmygo.uk/tools/gemini-3/</link>
  <description>AI 武林收錄的內容裡，最新提到 Gemini 3 的 17 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/gemini-3/rss.xml" rel="self" type="application/rss+xml"/>
  <item>
    <title>If we want them to do Knowledge Work, design them as Knowledge Agents — Benjamin Clavié, Mixedbread</title>
    <link>https://aiwulin.itsmygo.uk/c/fb9df5429c/</link>
    <guid isPermaLink="false">aiwulin-fb9df5429c</guid>
    <pubDate>Wed, 16 Sep 2026 00:00:00 +0800</pubDate>
    <dc:creator>AI Engineer</dc:creator>
    <description>&lt;p&gt;Benjamin Clavié 指出當前 AI 代理多專注於程式設計任務，但知識工作（如法律、醫療）更複雜且依賴隱含語境。他提出應設計「知識代理」而非僅限「程式設計代理」，並透過工具迴圈與組織迴圈最佳化搜尋效率。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.youtube.com/watch?v=O84lhGc1OOI&quot;&gt;https://www.youtube.com/watch?v=O84lhGc1OOI&lt;/a&gt;&lt;/p&gt;</description>
    <category>影片</category>
    <category>AI Agent 基礎</category>
    <category>Coding Agent</category>
  </item>
  <item>
    <title>1020: How to Choose Model Size and Effort Level: The Two Critical Dials</title>
    <link>https://aiwulin.itsmygo.uk/c/2136ce2758/</link>
    <guid isPermaLink="false">aiwulin-2136ce2758</guid>
    <pubDate>Fri, 21 Aug 2026 00:00:00 +0800</pubDate>
    <dc:creator>Super Data Science</dc:creator>
    <description>&lt;p&gt;解析大型語言模型中「模型大小」與「努力程度」兩個關鍵設定，釐清前者決定能力上限，後者控制思考深度而非單純延遲。透過 Anthropic、OpenAI 與 Google 的實例，說明如何根據任務難易度與成本效益，精準搭配這兩個開關以獲得最佳…&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.podtrac.com/pts/redirect.mp3/chrt.fm/track/E581B9/arttrk.com/p/VI4CS/pscrb.fm/rss/p/traffic.megaphone.fm/SUPERDATASCIENCEPTYLTD6826796686.mp3&quot;&gt;https://www.podtrac.com/pts/redirect.mp3/chrt.fm/track/E581B9/arttrk.com/p/VI4CS/pscrb.fm/rss/p/traffic.megaphone.fm/SUPERDATASCIENCEPTYLTD6826796686.mp3&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>大型語言模型入門</category>
    <category>從零打造語言模型</category>
    <category>推論與部署</category>
  </item>
  <item>
    <title>Gemma 4 in Action: Bringing Frontier AI to the Edge</title>
    <link>https://aiwulin.itsmygo.uk/c/e4df4645da/</link>
    <guid isPermaLink="false">aiwulin-e4df4645da</guid>
    <pubDate>Tue, 30 Jun 2026 00:00:00 +0800</pubDate>
    <dc:creator>Google for Developers</dc:creator>
    <description>&lt;p&gt;Gemma 4 讓先進 AI 能在無網路環境的裝置上執行，解決連線障礙。影片展示其在烏幹達醫療、克丘亞語文化儲存等領域的實際應用，並強調開源模型對技術民主化的重要性。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.youtube.com/watch?v=HcwMTu1xQDw&quot;&gt;https://www.youtube.com/watch?v=HcwMTu1xQDw&lt;/a&gt;&lt;/p&gt;</description>
    <category>影片</category>
    <category>開源模型</category>
  </item>
  <item>
    <title>Nano Banana 2 is Here! Gemini-3 Shutdown &amp; The AI Layoff Myth | EP99.36</title>
    <link>https://aiwulin.itsmygo.uk/c/658b3f7347/</link>
    <guid isPermaLink="false">aiwulin-658b3f7347</guid>
    <pubDate>Fri, 27 Feb 2026 00:00:00 +0800</pubDate>
    <dc:creator>This Day in AI</dc:creator>
    <description>&lt;p&gt;探討 Google 新推出的 Nano Banana 2 影像模型，指出其價格降低約 50% 且速度提升，並示範了基於註記的圖片編輯與幻燈片生成功能。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://share.transistor.fm/s/f5de1415&quot;&gt;https://share.transistor.fm/s/f5de1415&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>圖像生成</category>
    <category>模型發布與實測</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>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>愛好 AI Engineer 電子報 🚀 2025 AI 年度回顧 #34</title>
    <link>https://aiwulin.itsmygo.uk/c/3186e6f5d2/</link>
    <guid isPermaLink="false">aiwulin-3186e6f5d2</guid>
    <pubDate>Mon, 19 Jan 2026 00:00:00 +0800</pubDate>
    <dc:creator>ihower（張文鈿）</dc:creator>
    <description>&lt;p&gt;這篇文章是 2025 年 AI 領域的年度回顧，涵蓋 Andrej Karpathy 對 AGI 時間表的看法、OpenAI 研發長 Mark Chen 的訪談，以及 Agent Labs 與 Model Labs 的產業分類。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://ihower.tw/blog/13612-aie-2025-year-in-review&quot;&gt;https://ihower.tw/blog/13612-aie-2025-year-in-review&lt;/a&gt;&lt;/p&gt;</description>
    <category>文章</category>
    <category>AI 輔助軟體工程</category>
    <category>AGI 與長期趨勢</category>
    <category>Coding Agent</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>Secrets to unlock Gemini 3's hidden power...</title>
    <link>https://aiwulin.itsmygo.uk/c/1fe95d7173/</link>
    <guid isPermaLink="false">aiwulin-1fe95d7173</guid>
    <pubDate>Fri, 02 Jan 2026 00:00:00 +0800</pubDate>
    <dc:creator>AI Jason</dc:creator>
    <description>&lt;p&gt;介紹如何解鎖 Gemini 3 的隱藏功能與強大能力。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.youtube.com/watch?v=oL6bLQOXwAY&quot;&gt;https://www.youtube.com/watch?v=oL6bLQOXwAY&lt;/a&gt;&lt;/p&gt;</description>
    <category>影片</category>
    <category>模型發布與實測</category>
  </item>
  <item>
    <title>The State of A.I. 2025 (ft Richardson Dackam)</title>
    <link>https://aiwulin.itsmygo.uk/c/ed2c532156/</link>
    <guid isPermaLink="false">aiwulin-ed2c532156</guid>
    <pubDate>Wed, 31 Dec 2025 00:00:00 +0800</pubDate>
    <dc:creator>Tool Use - AI Conversations</dc:creator>
    <description>&lt;p&gt;回顧 2025 年 AI 領域的劇烈變化，涵蓋 GPT-5.2 等前沿模型的加速迭代、開源模型如 DeepSeek 的衝擊，以及資料中心能源瓶頸與物理機器人等新趨勢。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://podcasters.spotify.com/pod/show/tooluse/episodes/The-State-of-A-I--2025-ft-Richardson-Dackam-e3d05g5&quot;&gt;https://podcasters.spotify.com/pod/show/tooluse/episodes/The-State-of-A-I--2025-ft-Richardson-Dackam-e3d05g5&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>AGI 與長期趨勢</category>
    <category>開源模型</category>
    <category>模型發布與實測</category>
  </item>
  <item>
    <title>What the Freakiness of 2025 in AI Tells Us About 2026</title>
    <link>https://aiwulin.itsmygo.uk/c/2956e05f5a/</link>
    <guid isPermaLink="false">aiwulin-2956e05f5a</guid>
    <pubDate>Wed, 24 Dec 2025 00:00:00 +0800</pubDate>
    <dc:creator>AI Explained</dc:creator>
    <description>&lt;p&gt;回顧 2025 年 AI 領域的奇異進展與 2026 年預測，涵蓋推理模型、Genie 3 可玩世界、AI 垃圾內容主流化及 GPT-5 的誤解。觀眾能了解當前技術邊界、評估指標的缺陷，並掌握「側向生產力」與自動化資訊發現等未來趨勢。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.youtube.com/watch?v=FMMpUO1uAYk&quot;&gt;https://www.youtube.com/watch?v=FMMpUO1uAYk&lt;/a&gt;&lt;/p&gt;</description>
    <category>影片</category>
    <category>AGI 與長期趨勢</category>
    <category>模型發布與實測</category>
    <category>研究前沿</category>
  </item>
  <item>
    <title>GPT 5.2: OpenAI Strikes Back</title>
    <link>https://aiwulin.itsmygo.uk/c/45cb8cd363/</link>
    <guid isPermaLink="false">aiwulin-45cb8cd363</guid>
    <pubDate>Sat, 13 Dec 2025 00:00:00 +0800</pubDate>
    <dc:creator>AI Explained</dc:creator>
    <description>&lt;p&gt;深入解析 OpenAI 最新發布的 GPT 5.2 模型，探討其在專業任務表現、推理成本與不同評測標準下的優勢與劣勢。透過實際測試與多項 benchmark 比較，幫助讀者釐清模型真實能力與市場宣傳之間的差距。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.youtube.com/watch?v=4p73Uu_jZ10&quot;&gt;https://www.youtube.com/watch?v=4p73Uu_jZ10&lt;/a&gt;&lt;/p&gt;</description>
    <category>影片</category>
    <category>模型發布與實測</category>
  </item>
  <item>
    <title>Nano Banana + Gemini 3 = S-TIER UI DESIGNER</title>
    <link>https://aiwulin.itsmygo.uk/c/77a6ccf3b7/</link>
    <guid isPermaLink="false">aiwulin-77a6ccf3b7</guid>
    <pubDate>Fri, 12 Dec 2025 00:00:00 +0800</pubDate>
    <dc:creator>AI Jason</dc:creator>
    <description>&lt;p&gt;示範如何利用 Nano Banana 與 Gemini 3 結合，透過四步驟工作流快速生成創意十足的 UI 設計。觀眾能學習如何將設計規劃、影像生成與程式實作串聯，大幅提升介面設計的創意與效率。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.youtube.com/watch?v=RYnxU_MTVvU&quot;&gt;https://www.youtube.com/watch?v=RYnxU_MTVvU&lt;/a&gt;&lt;/p&gt;</description>
    <category>影片</category>
    <category>圖像生成</category>
    <category>模型發布與實測</category>
  </item>
  <item>
    <title>Claude Opus 4.5: The Engineers' Model</title>
    <link>https://aiwulin.itsmygo.uk/c/e1fa1d5d77/</link>
    <guid isPermaLink="false">aiwulin-e1fa1d5d77</guid>
    <pubDate>Mon, 01 Dec 2025 00:00:00 +0800</pubDate>
    <dc:creator>IndyDevDan</dc:creator>
    <description>&lt;p&gt;示範 Claude Opus 4.5 如何透過多代理協作與代理沙盒，自動完成全端應用開發與瀏覽器測試。觀眾能學習如何部署多個子代理執行複雜任務，並掌握利用 Opus 4.5 進行高效代理程式設計與自動化測試的實際流程。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.youtube.com/watch?v=3kgx0YxCriM&quot;&gt;https://www.youtube.com/watch?v=3kgx0YxCriM&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>多代理系統</category>
  </item>
  <item>
    <title>EP116 - Google 挑戰 NVIDIA 霸權的開始？深度解析 TPU</title>
    <link>https://aiwulin.itsmygo.uk/c/0156c3dab3/</link>
    <guid isPermaLink="false">aiwulin-0156c3dab3</guid>
    <pubDate>Mon, 01 Dec 2025 00:00:00 +0800</pubDate>
    <dc:creator>科技浪 Tech.wav</dc:creator>
    <description>&lt;p&gt;Podcast深度解析Google TPU與NVIDIA GPU的競爭局勢，釐清Gemini 3全由TPU訓練的意義，並探討Meta等大廠買入TPU的動機。內容涵蓋硬體規格對比、軟體生態進化及未來市場走向，幫助讀者建立正確的技術認知。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://open.firstory.fm/story/cmimkhl9u023601y0a7che3vb&quot;&gt;https://open.firstory.fm/story/cmimkhl9u023601y0a7che3vb&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>AI 晶片與硬體</category>
  </item>
  <item>
    <title>Episode 63: Why Gemini 3 Will Change How You Build AI Agents with Ravin Kumar (Google DeepMind)</title>
    <link>https://aiwulin.itsmygo.uk/c/9e6d5b9d5e/</link>
    <guid isPermaLink="false">aiwulin-9e6d5b9d5e</guid>
    <pubDate>Sat, 22 Nov 2025 00:00:00 +0800</pubDate>
    <dc:creator>Vanishing Gradients</dc:creator>
    <description>&lt;p&gt;Ravin Kumar 與 Hugo 探討 Gemini 3 如何改變軟體開發，從簡單工具呼叫轉向自主「代理 harness」。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://hugobowne.substack.com/p/episode-63-why-gemini-3-will-change-2fd&quot;&gt;https://hugobowne.substack.com/p/episode-63-why-gemini-3-will-change-2fd&lt;/a&gt;&lt;/p&gt;</description>
    <category>Podcast</category>
    <category>AI Agent 基礎</category>
    <category>模型發布與實測</category>
    <category>Agent Skills 與 Harness</category>
  </item>
  <item>
    <title>Three Years from GPT-3 to Gemini 3</title>
    <link>https://aiwulin.itsmygo.uk/c/8f994faf80/</link>
    <guid isPermaLink="false">aiwulin-8f994faf80</guid>
    <pubDate>Wed, 19 Nov 2025 00:00:00 +0800</pubDate>
    <dc:creator>Ethan Mollick（部落格）</dc:creator>
    <description>&lt;p&gt;作者測試 Google 最新 Gemini 3 模型，透過實際任務展示 AI 從 2022 年寫詩到 2025 年編碼遊戲的巨大進步。&lt;/p&gt;&lt;p&gt;原站：&lt;a href=&quot;https://www.oneusefulthing.org/p/three-years-from-gpt-3-to-gemini&quot;&gt;https://www.oneusefulthing.org/p/three-years-from-gpt-3-to-gemini&lt;/a&gt;&lt;/p&gt;</description>
    <category>文章</category>
    <category>模型發布與實測</category>
    <category>機器人與具身智慧</category>
  </item>
</channel>
</rss>
