I Couldn't Find Anything in My Claude Chats, So I Accidentally Built a Life OS

A month ago I had 30+ Claude web chats and couldn’t find anything. The search worked, technically. But which chat had the meal planning discussion? Which one had the business pricing analysis? Was that recipe idea in the health chat or the general brainstorm? I tried using ‘projects’, but those got filled with many chats too.

I started copying key conversations into markdown files. Just to have a reference. Then Claude Code got involved, and now I have 500+ files running my entire life.

Obsidian graph view showing 500+ interconnected markdown files The Brain after one month. Each dot is a file. Each line is a link between them.

The Problem That Started It

Claude web is great for conversations. Terrible for building knowledge.

Every good idea lives in a chat thread somewhere. You vaguely remember having the conversation. Search returns 12 results across 8 chats. You click through them trying to find the one where you actually made the decision.

I was burning time re-having conversations because I couldn’t find the original one. Or we’d hit context limits and need to start a chat again and feed it the context again.

The Fix: Just Dump It Into Markdown

January 4th. I created a folder called “Brain” and started summarizing.

That health chat where we figured out my protein targets and meal prep strategy? Summarized into Health/nutrition/preferences.md. The 3D printing business economics deep dive? Now ArcaneLayer/cogs-reference.md. The content strategy chat? Career/blog/content-plan.md.

Nothing fancy. Just getting ideas and decisions out of chat threads and into files I could actually find.

Then I pointed Claude Code at the folder.

What It Grew Into

I didn’t plan to build a “system.” I just kept dumping conversations and the structure emerged.

Health (~100 files): Started as my protein targets and allergy notes. Now has 50+ recipes, weekly meal plans, workout programs, daily tracking journals, a freezer inventory. All because one summarized chat spawned another.
ArcaneLayer (~100 files): My 3D printing business. That one chat about product pricing became a full COGS reference. Launch discussions became launch checklists. Now there’s team rates, creator performance data, weekly income reviews.
WendyOwensBooks (~40 files): My wife’s author business got its own section. Revenue splits, live stream analytics, affiliate tracking.
Career/Business (~50 files): Contract outreach, content strategy, this blog’s migration from Gatsby to Astro.

500+ files. One month. All from “let me just save this chat somewhere I can find it.”

The Structure That Makes It Work

Not just a pile of files. There’s a pattern.

Daily rhythms: Each domain has a daily-rhythm.md. Monday is planning. Friday is launch prep. Sunday is rest and review. /today pulls all of them into one list.
Templates: Launches, projects, recipes, daily logs. Every repeating thing has a template. You don’t think about format, you just fill it in.
Preferences as data: preferences.md in Health doesn’t just say what I like. It says what to never suggest (salmon, bell peppers, onion because my wife is allergic), equipment I own (crockpot, smoker, deep freeze, ice cream maker), and exact macro targets. Every recipe and meal plan reads this file first.

The Claude Code Unlock

Here’s the thing about markdown files: Claude Code can read them AND write to them.

When I pointed Claude Code at the Brain folder, it could suddenly see all my decisions, constraints, and context. It wasn’t starting from zero anymore. And when we made new decisions together, they went straight into the files.

Claude Code helped me build everything that came next.

CLAUDE.md: The Master Brief

“Hey, we should probably write down who I am and what matters so you don’t have to ask every time.”

One file at the root. My health constraints (gout, wife’s onion allergy). My defaults (50/50 chicken breast/thigh, casseroles always doubled). Pointers to important files. Claude Code helped write it, and now Claude Code reads it every session.

AGENTS.md: Domain Context

Different parts of my life have different rules. Claude Code and I built an AGENTS.md for each major area:

Health’s knows my cut, my rehab constraints, never to suggest bland fitness food
ArcaneLayer’s file knows margin targets, team pay rates, product economics
WendyOwensBooks knows which affiliate drives 75% of revenue

Claude reads the local AGENTS.md before touching anything in that domain. We built these together as edge cases came up.

Skills and Scripts

This is where it got fun. Claude Code can build its own tools.

“I’m tired of planning meals manually.” So we built /meal-planner. It reads my preferences, checks what I ate the last two weeks, respects the constraints, outputs a week of dinners with a prep checklist.

“I need a grocery list from this meal plan.” So we built /grocery-list. Pulls ingredients, rounds meat to 1lb increments (how Kroger sells it), groups by aisle.

“What should I focus on today?” So we built /today. Reads every domain’s daily rhythm file, matches today’s day of the week, consolidates tasks.

Each skill is just a markdown file with instructions that Claude Code follows. I describe what I want, Claude Code writes the skill, we iterate until it works.

The Shift

Before (Claude web): Great conversation. Good ideas. Then it disappears into a chat thread I’ll never find again. Next session starts from zero.

After (Claude Code + Brain): Every decision persists. Every constraint is remembered. Claude reads the context, does the work, updates the files. Next session picks up where we left off.

The chat became a codebase. The AI became a collaborator with memory.

How To Start

You don’t need 500 files. You need one folder and the willingness to dump your chats.

Create a folder. Call it Brain, Notes, whatever. Put it somewhere Claude Code can access.
Summarize one chat. That conversation where you made a decision or figured something out? Copy the key parts into a markdown file. Just the key info, not the whole thread.
Point Claude Code at it. Open the folder in Claude Code. Now it has context. Ask it to help you organize, expand, or build on what’s there.
Let it compound. Every good conversation adds to the files. Every file gives the next conversation more context. A month later you have a system you didn’t plan.

The magic isn’t the structure. It’s that decisions stop disappearing into chat threads.

I didn’t set out to build a “Life OS.” I just got frustrated that I couldn’t find my own conversations.

Markdown files in a folder. Claude Code with read/write access. One month of just… using it.

500 files later, I have meal planning, business ops, content strategy, and health tracking all in one place. Claude knows all of it. Every session picks up where the last one ended.

Start with one folder and one summarized chat. See where it goes.