Workshop Day / S4 · 4D

Your First Cairn From Your History.

The checkpoint that gives your system memory. You build an Agent Conversations Cairn from your own Claude Code history, keep the raw history local, turn it into L1 waypoints, a L2 Chain of Density Card Catalog Layer (Card Catalog Layer for short), and L3 source logs, then produce the feed your Morning Brief can read. Want the full picture of the memory layers and where Cairns came from? Read the Cairns Three Tiers and Mini-Cairns Build library cards. Here, you build the working Agent Conversations Cairn.

What you'll have

An Agent Conversations Cairn a local vault built from your own Claude Code thread history, on your machine.
L1-L3 memory layers L1 waypoints, a L2 Chain of Density Card Catalog Layer, and L3 raw source logs from about the last 14 days.
A Morning Brief feed a validated reports/Agent-Conversations-.json file that the S5 brief can consume.

Step 4 · 4D

4D Build

4D goal A local Agent Conversations Cairn from your Claude Code history

Done when The Cairn is written and the Morning Brief feed validates

You run /build-agent-conversations-cairn

Time ~20 min

The concept Cairns Three Tiers

One-click start

Build it with Claude.

Paste this into Claude Code. It runs the Agent Conversations Cairn skill against your own local Claude Code history, then validates the Morning Brief feed before you move on.

Your checkpoint.

This is the day-of execution path. Cairns is the foundation of the whole system, but your 4D move is concrete and small: build one local Agent Conversations Cairn from your own Claude Code history, validate the Morning Brief feed, and ask it a question. That is a working second brain.

4D goal

A searchable Agent Conversations Cairn, built from your own Claude Code history, on your machine.

Done when

Your Cairn writes L1-L3 markdown and a validated reports/Agent-Conversations-<date>.json feed.

You run

The /build-agent-conversations-cairn skill in 4D mode.

Want the how and why

This card is the checkpoint, kept tight for the day. For the memory layers (L1 waypoints, L2 Chain of Density Card Catalog Layer (Card Catalog Layer for short), L3 raw), the origin story, and how the capture pipeline works, read the Cairns Three Tiers and Mini-Cairns Build library cards.

Build it.

Step 01

Build your Agent Conversations Cairn

5 min

Use the starter prompt at the top of this card. Claude runs /build-agent-conversations-cairn, confirms your local vault path, and keeps the build personal and local.

Paste into Claude Code /build-agent-conversations-cairn

Worried about where your data goes? This vault is local and personal. It is not the shared YouTube demo vault and it does not write to any production system. Ask Claude to confirm the path before it creates anything.

Step 02

Produce L1-L3 and the feed

10 min

The skill scans about the last 14 days of your Claude Code JSONL history in ~/.claude/projects. It preserves source material in L3, writes L2 Chain of Density Card Catalog Layer summaries, writes L1 waypoints, and validates the Agent-Conversations Morning Brief feed.

Paste into Claude Code /build-agent-conversations-cairn --mode 4d

Step 03

Ask your Cairn a question

3 min

Ask it something only your history would know, like what you decided about a project last week. The skill also writes a small personalization harness from about the last week of history with no more than five questions, so your agent can check whether the memory actually fits you.

You have a searchable Agent Conversations Cairn on your machine. In S5 your brief recalls from the validated feed.

Going to the 8D room?

8D checkpoint

Merge memory across all your machines (Claude and Codex) live, and see the path to semantic plus graph search: one brain across your whole setup. Start at Merge Memory Across Machines.

Where to go next.

Workshop Day