Livestream Build Track

Build Your Own Second Brain

This guide walks through the full EJ's Palace Wiki stack: long-term memory, curated wiki structure, agent identity files, and the HARVEST loop that keeps the whole thing learning after every session.

5core components
43K+drawers in production memory
4wiki operations to keep it alive
1system that compounds over time

How This System Works

Each layer fixes a different failure mode

MemPalace remembers raw reality. The wiki curates what matters. SOUL.md and MEMORY.md orient the agent. HARVEST closes the loop so work does not evaporate.

Install just what you need

Use the individual step pages if you are adding one capability at a time or demoing each concept separately.

Install the whole operating system

Use the full setup walkthrough when you want the complete chain working together as one memory architecture.

Components

Install Individually

Each card opens a dedicated page with an overview, deep expandable notes, and two installation paths.

Resources

Grab every building block from one place

Use these links if you want the fast path: install the package, read the pattern, then wire the whole stack together.

1. MemPalace

Get MemPalace →

Install the memory engine first so your system can remember real work.

2. LLM Wiki Pattern

Read the LLM Wiki Pattern →

Start with Karpathy's original concept, then adapt the structure inside your own WIKI.md.

3. EJ's Palace Wiki

Browse EJ's GitHub Org →

Use EJ's public projects as your reference point for the wiki-centered variation on the pattern.

4. SOUL.md + MEMORY.md

Create Your Context Files →

These are files you create yourself, using existing agent context conventions as the starting template. OpenClaw and Hermes are both useful references for agent-hosted workflows.

5. HARVEST

Read the HARVEST Pattern →

HARVEST is part of the Palace Wiki pattern, not a separate OpenClaw download. Start with the public pattern, then run the loop on whatever agent host you use.