https://imitation-alpha.github.io/zh-Hant/blog/ 關於 agentic systems、ML infrastructure 與 tooling 的繁體中文筆記。 zh-Hant Wed, 10 Jun 2026 00:00:00 +0000 https://imitation-alpha.github.io/zh-Hant/blog/loop-engineering.html https://imitation-alpha.github.io/zh-Hant/blog/loop-engineering.html Loop engineering 是裝上時鐘的 harness：schedule、state、spawn 與 hard oracles。難點是選出 loop 無法 game 的 signal。 Wed, 10 Jun 2026 00:00:00 +0000 Arthur Yau https://imitation-alpha.github.io/zh-Hant/blog/recursive-self-improving-ai-rubicon.html https://imitation-alpha.github.io/zh-Hant/blog/recursive-self-improving-ai-rubicon.html Self-improving AI 的 spectrum map：pseudo-labels、rewards、AI judges、generated tasks、harness improvement，以及為什麼我們仍接近 Rubicon。 Tue, 19 May 2026 00:00:00 +0000 Arthur Yau https://imitation-alpha.github.io/zh-Hant/blog/ai-self-correction-decoding-workflow-reasoning.html https://imitation-alpha.github.io/zh-Hant/blog/ai-self-correction-decoding-workflow-reasoning.html AI self-correction 的實用地圖：decoding tricks、workflow feedback、trained reasoning 都需要真實 signal，而不只是 reflection。 Tue, 19 May 2026 00:00:00 +0000 Arthur Yau https://imitation-alpha.github.io/zh-Hant/blog/ai-agent-harness-engineering.html https://imitation-alpha.github.io/zh-Hant/blog/ai-agent-harness-engineering.html 面向 agent builders 的 harness engineering：tools、permissions、workflows、feedback、memory 與 verification loops 如何讓 LLM 變得可用。 Tue, 19 May 2026 00:00:00 +0000 Arthur Yau https://imitation-alpha.github.io/zh-Hant/blog/ai-agent-research-work.html https://imitation-alpha.github.io/zh-Hant/blog/ai-agent-research-work.html AI agents 如何改變 research work：first pass 變便宜、verification 變更重要，而 human judgment 移到 problem choice 與 evidence。 Tue, 19 May 2026 00:00:00 +0000 Arthur Yau https://imitation-alpha.github.io/zh-Hant/blog/ai-agent-multi-agent-interaction.html https://imitation-alpha.github.io/zh-Hant/blog/ai-agent-multi-agent-interaction.html 面向 builders 的 multi-agent topology 地圖：什麼時候該 branch、critique、compete、merge，什麼時候該保持獨立。 Tue, 19 May 2026 00:00:00 +0000 Arthur Yau https://imitation-alpha.github.io/zh-Hant/blog/ai-agent-context-engineering.html https://imitation-alpha.github.io/zh-Hant/blog/ai-agent-context-engineering.html 面向 agent builders 的 context taxonomy：prompt-visible workspace、external memory、compression、bounded observations、sub-agents 與 on-demand skills。 Tue, 19 May 2026 00:00:00 +0000 Arthur Yau https://imitation-alpha.github.io/zh-Hant/blog/ai-agent-systems-hung-yi-lee-summary.html https://imitation-alpha.github.io/zh-Hant/blog/ai-agent-systems-hung-yi-lee-summary.html 用 computer architecture 重新理解 AI agent：LLM 是 processor，context 是 working set，memory 是 storage，tools 是 I/O，harness 是 OS。 Tue, 19 May 2026 00:00:00 +0000 Arthur Yau https://imitation-alpha.github.io/zh-Hant/blog/orchestrating-coding-agents.html https://imitation-alpha.github.io/zh-Hant/blog/orchestrating-coding-agents.html 如何用 tmux、Tailscale、iOS 上的 Termius，加上一個小型 PWA 控制台，同時操控 10+ 個並行 CLI 程式設計 agent。 Tue, 28 Apr 2026 00:00:00 +0000 Arthur Yau