← Work
Case study · 2026 · Architect + operator
Morris Brain
A second brain that maintains itself — I read and decide; an AI does the bookkeeping.
The vault itself is private — it's my actual second brain. Want a walkthrough of how it's built?
What it is
A personal knowledge system run as a partnership with an AI collaborator: I supply documents and questions; it files, cross-references, and maintains everything against a formal schema, with nightly automation.
The problem
A fast-growing body of career, market, and learning context becomes unnavigable and inconsistent when upkeep is manual. I wanted it to compound without becoming a chore — and an AI collaborator whose judgment stays reliable over months, not just one session.
What I built
An Obsidian vault operated by AI agents in the terminal (Claude, Codex) against written governance rules: typed pages for people, projects, concepts and sources; a written protocol suite defining how the AI should delegate, verify, and escalate; and an unattended daily automation chain — price fetch, portfolio derivation, nightly research, morning and evening briefings.
Stack
Claude Code (agents, skills, hooks)CodexObsidianMarkdown + YAMLPythonmacOS launchdgit
Results
- Months of continuous, unattended operation
- Daily automation pipeline that runs itself
- A governance protocol suite, adversarially reviewed before adoption
What it shows
Systems thinking applied to myself: designing the information architecture and writing the rules an AI collaborator actually runs on — sustained as a real pipeline, not a one-off demo.