pal (formerly self-org) is my second brain that organizes itself: a headless, local-first knowledge engine where free-form thoughts are classified by an LLM chain into a markdown wiki whose taxonomy grows organically. Storage uses ULID-stable IDs and atomic writes with one git commit per operation, so the brain directory’s history is a full time machine.
The engine now ships triage and link ingestion (web articles are fetched, summarized, deduplicated, and filed into the reading list automatically), a pal lint reorganization pipeline, and FastMCP tools exposing the brain to AI clients — all engine-first, so CLI, MCP, and any future app share the same API. The suite has grown to 290 tests with hermetic live-test isolation.