起點：串聯的野心

設立了個人網站的部落格功能後，我開始思考著如何串聯 mysecondbrain，以及將這樣的部落格貼文納入到未來 mysecondbrain 的編譯整合。為了實現這個願景，我依然透過 Claude Code 來串接起這些聯想，並自動化整個過程。

問題浮現：跨 Session 的記憶斷裂

但我逐漸發現一個根本的問題——當不同 session 因為主題或 context 超過限制而被迫分支出去時，跨 session 的記憶無法同步或連續。這導致我花費越來越多時間在向 Claude Code 重複說明、重新設定已經完成過的東西。

每當我表達不耐煩時，它才會乖乖去查看我在本機端筆記型電腦上到底設定了哪些東西。然後它道歉，然後我赫然發現——原來我早就和 Claude Code 一起協作過了。

情緒的漣漪

很有趣，也很累人。我覺得我的脾氣越來越不好，好像要對 AI 亂撒一氣。這種循環帶來的挫折感，來自於期待與現實之間的落差——我期待的是連貫的、有上下文記憶的協作者，但得到的是每次都需要重新喚醒的工具。

實驗：Gmail 自動化發佈流程

好吧！我決定用這封信件做一個測試。我使用 Gmail 寄給自己，加上「部落格」標籤，以及「mysecondbrain」標籤，看看 GitHub 會不會自動去抓到，然後自動到個人網站去貼文發表部落格文章。

更大膽的目標是：自動將我的部落格文章納入 mysecondbrain 的大腦編譯流程。

後記

這次實驗其實是在探索一個更深層的問題：我們該如何設計人機互動的工作流程，才能真正減輕認知負擔，而不是替換成另一種形式的疲勞？ 答案或許在於接受 AI 工具的限制，並聰慧地在自動化和手動控制之間找到平衡點。

English version below

The Beginning: Ambition to Connect

After setting up a blog feature on my personal website, I started thinking about how to integrate it with mysecondbrain, and how to incorporate these blog posts into the future compilation and integration of mysecondbrain. To realize this vision, I continued to leverage Claude Code to bridge these ideas and automate the process.

The Problem Emerges: Memory Fragmentation Across Sessions

But I gradually discovered a fundamental issue—when different sessions fork due to topic or context overflow, memories cannot sync or continue across sessions. This led me to spend increasingly more time re-explaining and re-configuring things I had already set up with Claude Code.

Only when I express frustration does it diligently go back to check what I had actually configured on my local notebook. Then it apologizes, and I suddenly realize—I had already collaborated with Claude Code on this!

The Ripple of Emotion

It’s both amusing and exhausting. I feel my temper growing worse, as if I want to vent my frustration at AI. This cycle of frustration stems from the gap between expectations and reality—I expect a coherent, context-aware collaborator, but what I get is a tool that needs to be reawakened every session.

Experiment: Gmail-Driven Automated Publishing

Alright! I decided to run a test with this email. I sent it to myself via Gmail with a “blog” label and a “mysecondbrain” label, to see if GitHub automatically detects it and publishes it to my personal website as a blog post.

Even more ambitiously, I hope to automatically integrate my blog articles into the mysecondbrain compilation process.

Reflection

This experiment ultimately explores a deeper question: How should we design human-AI interaction workflows to genuinely reduce cognitive burden, rather than simply replacing it with another form of fatigue? The answer perhaps lies in accepting AI tool limitations and intelligently finding balance between automation and manual control.