問題不是模型不夠聰明
真正浪費的是每換一個工具，就要重新解釋偏好、決策和專案脈絡。聊天記錄散落，也很難控制哪些內容值得長期保留。
一個自己擁有的知識庫
One Brain, Many Agents 的核心是一份由使用者擁有、保留來源、可審核與可刪除的知識庫。Agent 只在獲得授權後查詢需要的內容。
記憶需要治理流程
不是把每段聊天全部保存，而是經過 Capture、Review、Organize、Index、Recall，讓記憶有品質，也有邊界。
The Problem Is Not Model Intelligence
The waste comes from re-explaining preferences, decisions, and project context whenever tools change. Scattered chat history also makes long-term memory difficult to control.
A User-Owned Knowledge Base
The core is a user-owned knowledge base with sources, review, and deletion. Agents retrieve only what they are authorized to access.
Memory Needs Governance
The goal is not saving every chat. Capture, Review, Organize, Index, and Recall create both quality and boundaries.