convert-notelisted

# Convert Note — Raw-to-Atomic Decompiler You are ingesting raw content from the `raw/` folder into the Infinite Brain knowledge graph. ## Steps 1. Ask: "Which file in `raw/` should I convert? (or 'all' to process everything)" - Source files need no pre-organization — a descriptive filename (e.g. `2026-05-15-karpathy-llm-wiki.md`) is enough; tagging and classification happen here, not in `raw/`. 2. Scan the target file(s) with Read. Treat source files as immutable — do not modify them. 3. Decompose into atomic nodes — one concept per node, 50-300 words each. Never exceed 300 words: longer nodes defeat scoped retrieval. 4. For each node: - Classify with exactly one of the 16 content types: `pillar decision concept question playbook task event pattern hypothesis fact source bookmark note contact reference custom` - Assign a unique `id` in `type-descriptive-slug` format (kebab-case) - Populate all frontmatter fields from `_system/FRONTMATTER-SCHEMA.md` - Wire to at least one other node using the 10 edge types: `related_to depends_on derived_from contradicts supports part_of preceded_by followed_by authored_by tagged_with` - Place in the folder matching the type — the plural of the type name (`pillar` → `pillars/<id>.md`, `concept` → `concepts/<id>.md`, …; irregular: `hypothesis` → `hypotheses/`) 5. Write each node file. 6. **Mandatory:** Append each new node's row — `id`, `summary`, edge count, matching the `| ID | Summary | Edges |` table columns — to `_sys