brain

English · 한국어

A personal knowledge base of everything I learn. Not a repo I fill by hand — three learning agent skills populate and keep it up to date automatically. It’s a living store.

Learning runs as a cycle called PTR (Place → Teach → Review):

place-me  →  teach-me  →  review-me
 (assess)     (learn)      (review)
   └──────────  levels.md  ──────────┘

PTR is the name for this three-skill loop. Run it once per topic: assess where you stand, learn piece by piece, then review until it sticks. Repeat as the level rises.

The three skills

This repo is built to work with the learning skills from the kidow/skills repository.

• place-me — Before learning a topic, a short adaptive diagnostic gauges how much I already know. Reads any existing levels.md entry for the topic as a starting guess. One question at a time until the level is clear, then it writes a prose level summary to levels.md and hands off to teach-me calibrated to the result.
• teach-me — Explains one piece of knowledge at a time, always progressing from foundational to advanced. Reads levels.md first to calibrate depth. Advances only when I say “next”, never re-teaches covered pieces, and flags imprecise vocabulary for correction. Complex concepts are grounded in analogies to my native language and culture. A visual aid is drawn when something is hard to grasp, and an audio aid plays pronunciation when learning a foreign language. When all pieces are covered, suggests 2–3 related topics. Each piece is committed to notes/topic_name.md; notes auto-organize into domain subfolders as they grow.
• review-me — Quizzes me on the saved notes one item at a time using spaced repetition (Again / Hard / Good / Easy). With no context given, it prioritizes the weakest topics from levels.md; offers to run place-me first if levels.md is missing. When the session ends, it updates the level in levels.md and deletes the ephemeral .review/ folder.

The three skills are tied together by a shared topic_name key: the notes/topic_name.md filename and the ## topic_name section in levels.md point to the same topic.

Repository layout

brain/
├── notes/            # learned knowledge (written by teach-me, read by review-me)
│   ├── topic_name.md #  - one file per topic, no date (git history tracks time)
│   └── <domain>/     #  - notes auto-organize into domain subfolders as they grow
├── levels.md         # per-topic level memory (Blank/Glimpsed/Grounded/Fluent)
└── README.md

• notes/ — knowledge notes, content only, no frontmatter.
• levels.md — each topic’s level as a four-stage label plus a prose summary.
  • Blank — no prior exposure to the concept
  • Glimpsed — seen it before but cannot reproduce or explain it reliably
  • Grounded — explains it independently and applies it in new contexts
  • Fluent — handles exceptions, nuance, and edge cases without hesitation
• .review/ — an ephemeral folder that exists only during a review session and is deleted when it ends (excluded from git).

Installing the skills

npx skills@latest add kidow/skills/place-me
npx skills@latest add kidow/skills/teach-me
npx skills@latest add kidow/skills/review-me

Usage

Run the learning skills from inside this repository.

/place-me Japanese writing system   # assess level → record in levels.md
/teach-me Japanese writing system   # learn piece by piece → commit to notes/
/review-me                          # review weakest topics → update levels.md

Start your own brain

Copy the prompt below and paste it to your AI assistant (Claude Code recommended):

Set up a personal knowledge base using the PTR learning system (Place → Teach → Review).

1. Create a new git repository and open it in Claude Code.
2. Install the three learning skills:
   npx skills@latest add kidow/skills/place-me
   npx skills@latest add kidow/skills/teach-me
   npx skills@latest add kidow/skills/review-me
3. Create this file structure:
   - notes/      (empty folder — add a .gitkeep file so git tracks it)
   - levels.md   (empty file)
   - README.md   (copy from https://github.com/kidow/brain)

Once done, start with /place-me <topic> to assess your first topic.