Cole Medin uses Nate Herk’s four-C framework to audit an AI second brain in a 21-minute walkthrough of his Markdown-based setup.
Context: know the business before searching
- Load a small set of core files describing behavior, the user, and current priorities at session start.
- Keep detailed entity notes and conversation summaries in a searchable knowledge base.
- Audit with a fresh session: can it accurately explain the business, team, and priorities?
Connections: reach the places work happens
- Connect email, calendar, tasks, files, repositories, and communication tools through APIs or MCP, the standard for connecting agents to tools.
- Wrap integrations to expose only the actions you want: Medin allows email drafting while withholding sending.
- Audit with a morning catch-up that finds real commitments and messages needing attention.
Capabilities: retain the process and test changes
- Turn repeated workflows into editable skills so the agent reuses the method and output format.
- Medin’s rule: after explaining a process more than three times, package it as a skill.
- Keep representative tasks as a regression suite and rerun them after changing models or the surrounding system.
- One recurring test edits the same short video, making unexpected changes easier to notice.
Cadence: automate after the workflow earns trust
- Run one useful workflow by hand before scheduling it; report completed work and anything needing attention.
- Medin describes a half-hour heartbeat that drafts replies and sends brief Slack updates.
- His linked starter repository provides a starting point; the framework also works with other agent setups.
Caveat: the claimed 25–30 hours saved weekly is Medin’s estimate. He says he has a measurement system but does not explain it in this video.
“You’re only going to be automating things once you trust your second brain to do it well.” — Cole Medin