Shipping in days · early-bird pricing live

Stop teaching Claude Code the same lesson every Tuesday.

A battle-tested memory architecture for solo operators who use Claude Code as their pair — so it actually remembers what you taught it, and stops blowing things up the same way twice.

Get notified at launch See what's inside

40 markdown files. One philosophy. Built running a real production system, refined over months of incidents.

The problem

You've been here, haven't you?

You explained your deploy quirk to Claude last Wednesday. Today it's gone.
You told it to stop summarizing every diff. It summarized. Again.
Your CLAUDE.md is 600 lines and somehow still misses the one rule that matters.
You tried the memory feature. It saved your shoe size and forgot the migration plan.
You're not sure if a rule belongs in a doc, a comment, a memory, or a hook.

The problem isn't Claude. It's that nobody told you how to structure what it should remember.

The kit

40 files. 2 hours to read. Pays back for 12 months.

A 4-type taxonomy, a why-line discipline, a verification step before acting, and a set of templates you can drop into ~/.claude/ today.

A

ARCHITECTURE.md

The 25-minute read that frames everything. Not theory — lived through months of incidents.
T

8 templates

Empty + filled versions of each memory type. Copy, rename, edit, ship.
E

20 sanitized examples

Real memory files from homelab clusters, indie SaaS, and multi-VPS deploys.
P

5 operational playbooks

Destructive-op checklist · backfill protocol · upgrade ritual · postmortem template · audit ritual.
C

6 starter CLAUDE.md files

Next.js, FastAPI, Docker Compose, k8s, Homelab, Generic SaaS — drop-in starting points.
!

ANTI_PATTERNS.md

The 17 things I learned not to memorize the hard way.
▶

30-min walkthrough video

Me using the system on a live deploy, in real time. Watch the architecture in motion.
∞

Lifetime updates

Private GitHub repo access. Every revision lands here. Forever.

The discipline

What a memory entry actually looks like.

The format is small. The discipline is everything. Every entry answers why and how to apply — without those, future Claude over-applies the rule to edge cases.

---
name:           Integration tests must hit a real DB
description:    why mocking the database failed us in 2024-Q3
type:           feedback
---

Never mock the database in integration tests.

**Why:** Q3 2024 incident. Mocked tests passed; the production
migration dropped a column the worker still read. Two-hour outage.

**How to apply:** any test marked @integration. Spin up the test
container, run the migration, assert state. Mocks are for unit only.

That's the format. That's the discipline. Multiply by 40.

Self-filter

Who this is — and isn't — for.

This is for you if

You use Claude Code weekly+ for real work, not just experiments
You're a solo operator, indie hacker, homelab tinkerer, or small-team SRE
You've felt the pain of Claude forgetting, over-applying, or drifting
You'd rather buy 12 months of lessons than learn them yourself

This is not for you if

You've never used Claude Code (start with Anthropic's docs first)
You want a SaaS, plugin, or CLI tool — this is markdown + philosophy
You're looking for prompt engineering tricks — this is about persistence
You work in a team of 4+ — protocols this kit doesn't cover

Pricing

One-time. No subscription.

Early-bird · first 50 buyers

$19$39

one-time · lifetime updates

ARCHITECTURE.md + 40 markdown files
30-minute walkthrough video
Private GitHub repo, updates forever
MIT license for templates
Email support, 30 days
14-day refund, no questions

Get notified when it ships

Pre-launch list

One email when it ships. That's it.

No drip campaign. No upsells. You get one email the day the kit goes live, with the early-bird link.

FAQ

Honest answers.

Anthropic is improving memory. Will this be obsolete in 18 months?

Maybe. The file format will change. The architecture — taxonomy, why-lines, verification step — outlives the format. You'll port it.

Will this work for teams?

Designed for solo. Two people OK with discipline. Three+ needs protocols this kit doesn't cover.

Can I share with a colleague?

Use it together at work, yes. Re-distribute publicly, no.

Is this prompt engineering?

No. It's about what persists between sessions. Orthogonal to prompts.

Refund?

14-day, no questions. Email hi@solosre.dev.

Is the source case study public?

No — it's a real production system serving real users. Every example in the kit is sanitized to homelab / indie SaaS / multi-VPS scenarios.