Reading Agent · field method
Reading Agent · 手绘方法
Claude / GPT + Earth LinC MCP + one Reading Harness. You read. It keeps up.
Claude / GPT ＋ Earth LinC MCP ＋ 一张 Reading Harness。你读，它跟着你。
Connect
第一步
Add Earth LinC MCP inside Claude or GPT.
在 Claude 或 GPT 里接上 Earth LinC MCP。
Four actions on your own Stations. Nothing is written without your confirmation.
对你自己的 Station 的四个动作。没有你的确认，不会写入任何东西。
Copy a Harness
第二步
This is what keeps the agent steady for weeks.
它让这个 agent 连续几周保持一致。
book: reading_progress: spoiler_boundary: raw_notes: [] knowledge_branches: [] connections: [] station_url: last_sync: PERMANENT RULES 1. Human reading comes first. 2. Never treat unread pages as the reader's experience. 3. Preserve raw notes before rewriting. 4. Keep BOOK / YOU / KNOWLEDGE / AI CONNECTION separate. 5. Accumulate notes first; do not update the Station for every small note. 6. AI suggestions are questions, not conclusions.
book: reading_progress: spoiler_boundary: raw_notes: [] knowledge_branches: [] connections: [] station_url: last_sync: 永久规则 1. 人先读，AI 后到。 2. 我还没读到的部分，不算我的阅读经验。 3. 先原样保留原句，再谈整理。 4. BOOK / YOU / KNOWLEDGE / AI CONNECTION 分开存放。 5. 先累积笔记；不要每条小笔记都更新 Station。 6. AI 的建议是提问，不是结论。
Stay with me while I read this book. Keep this Harness updated. Do not summarize ahead of my reading. Preserve my raw notes, help me open relevant knowledge, notice connections, and only write back to Earth LinC when my understanding has genuinely changed.
陪我一起读这本书。持续维护这份 Harness。不要超出我的阅读进度去总结。原样保留我的原句，帮我打开相关的知识，留意可能的关联；只有当我的理解真的发生变化时，才写回 Earth LinC。
You read
第三步
At your pace. The AI never runs ahead of your page.
按你的节奏。AI 不会跑到你读到的页数前面。
Drop a note
第四步
Any shape, any hour. It is kept raw.
什么形式都行，什么时间都行。它原样存下。
他好像很浮躁，却又是一种通透。
He seems restless, and somehow clear.
Stored word for word, tagged with where you are in the book. No summary, no polish, no rewriting.
逐字存下，并记下你读到的位置。不总结，不润色，不替你改写。
AI opens knowledge
第五步
Background may widen the book. It never replaces it.
背景知识把世界变大，但不替代这本书。
AI finds connections
第六步
It asks. You decide whether it counts.
它只提问，算不算数由你决定。
尽情地活，却不留恋。
Live fully, hold on to nothing.
他好像很浮躁，却又是一种通透。
He seems restless, and somehow clear.
你前面也写过「尽情地活，却不留恋」。这两句会不会在问同一个问题？
You wrote “live fully, hold on to nothing” before. Are these two asking one question?
Understanding forms
第七步
Wait until one thought can stand on its own.
等到一个想法能自己站住，再说。
MCP → Station
第八步
A new Station is written in three moves.
新建一个 Station，只有三步。
The first Station can stay private.
第一个 Station 可以先设为私密。
Next time
第九步
Then the loop starts again, one page further.
然后这个循环再走一圈，往前多读一点。
The AI does the comparing. MCP only provides Read / Quote / Compose / Refine — and only refines when your understanding actually moved.
比较是 AI 做的。MCP 只提供 读取 / 引用 / 写入 / 修改 —— 而且只有理解真的变了才动笔。
The chat is not the memory. The Station is.
对话不是长期记忆，Station 才是。
Either one can continue — if it reads the same Harness and the same Station first.
哪一个都能接着往下走 —— 只要它先读同一张 Harness 和同一个 Station。
One book → many Stations.
一本书 → 很多个 Station。
Some questions leave the novel and keep growing on their own.
有些问题会从小说里脱离出来，自己继续长。
Unfinished, still moving.
还没读完，还在长。
See a real Reading Station ↗ 看看它真的长成什么样 ↗A Station is not the summary of a book. It is one formed unit of understanding.
Station 不是一本书的摘要，而是一段已经形成的理解。
The book stays yours. The reading stays yours. The Station helps your understanding keep growing.
书是你自己读的，理解也是你的。Station 只是让这些理解继续长下去。