Case 13 — Markdown Memory: Machine-Readable Records and AI Memory Patterns

Overview

Item	Content
Period	2020s–2030s (markdown ecosystem, AI agent memory)
Technology	Markdown records, folder-based digital archives, AI session memory (MEMORY, checkpoints, notes)
Literacy	Structured records, searchable documents, archives shared by people and machines

Starting from the User’s Experience (Field Story)

“I used markdown even before AI.
It’s the easiest format for machines to record, so I use a lot of md files.
The experience of building digital archives
resembles how mimocode leaves memories.”

This case is not tool promotion — it starts from the question activists actually need: “How do I leave something so I can find it again later?”

How mimocode Leaves Memories (Reading the Structure)

In AI work environments like mimocode, memory is usually split across multiple layers of markdown files.

Layer	Role	Corresponding activist archive
Project `MEMORY.md`	Long-lasting rules and judgments	Project operating principles
Session `checkpoint.md`	Status summary of today’s work	Daily work log
`notes.md`	Notes not to forget	Notebook / jotter
Per-task records	Item-level progress	Activity / program-level files

Core principles:

One context per file — split rather than append endlessly
Titles, dates, tags — easy for people to read, easy for machines to search
Newest on top — keep old records but gather the flow upward
Sources in folders — each file becomes evidence and material

Why Markdown (Between Machine and Human)

Feature	Effect
Plain text	Openable even when programs change
Minimal syntax for headings, lists, tables	Structure becomes search and TOC
Easy to edit, diff, backup	Great for keeping history
Photos, links, tables together	The basic form of an archive

Where 2000s documents were “program-dependent files”, 2030s records converge on text archives read by both people and AI.

Literacy Improvement Mechanisms

Split recording: Separate agenda / decisions / to-dos / check results
Dates and sources: When, where, and who left it
Design for rediscovery: File names and titles as search terms
Handing to machines: md organization → AI summarizing, searching, reusing

Field Operation Tips (50 minutes)

Split one experience program into three md files (overview / operation log / materials list)
Rewrite five titles for “me finding this next month”
Sample activist folder:

paju-experience-archive/
  00-index.md
  2026-10-07_training-ops.md
  materials/
    photo-list.md
    faq.md

Output: my activity archive folder structure + one index md

Implications for Paju Activist Training

Records activists build become assets passed to the next activist and to AI
A routine of “leave it in my folder” before “posting to a blog” matters more
2030s collaboration = people and AI reading the same md materials
With digital archive experience, AI memory structures become intuitive

Notes

Markdown is not “pretty writing” but a language where machines and people work together
How we leave memory is the next stage of literacy — enabling search, summary, and transfer