GitHub Copilot Wiki: An AI-Powered Second Brain Template

A forkable template for building LLM-maintained personal knowledge bases using GitHub Copilot, implementing Andrej Karpathy's Compounding Knowledge Pattern.

View on GitHub

Template Customization Guide

This guide explains how to adapt the LLM Wiki Template to your specific domain.

1. Core Schema (.github/copilot-instructions.md)

This is the “brain” of your wiki. Copilot reads this on every request.

2. Page Formats

If your domain requires specific metadata (e.g., “Clinical Studies” might need “Methodology” and “Sample Size” fields), update the Page formats table in copilot-instructions.md.

3. The raw/ Directory

Keep the raw/ directory clean. It is meant to be an immutable record of where your knowledge came from.

4. Automation with Intake Scripts

If you have a large folder of existing markdown files:

  1. Copy them into raw/.
  2. Run scripts/intake.ps1 (Windows) or scripts/intake.sh (Mac/Linux).

The script will iterate through the files and call Copilot for each one, ensuring your wiki/index.md and wiki/log.md stay perfectly in sync.

5. Deployment (Optional)

Since the wiki is entirely markdown:


Follow the conventions. Trust the process. Build your knowledge.