Files Over Prompts: Build HTML Artifacts and LLM Wikis Your Claude Agent Actually Reads (AI Memory Playbooks Book 3)

Prompts plateaued. Files compound. The smart developers already know.

On reddit builders are posting setups where Claude reads from a structured wiki, edits it back, and ships HTML artifacts that hit 6.5 million views and 20,000+ bookmarks in 24 hours. The difference is not talent. It is one folder of markdown files the agent reads first, every session.

This is the first complete playbook for the era after prompts. By Saturday morning your wiki has five live entries the agent reads. By Saturday night your Claude session remembers you across every restart. By Sunday afternoon you ship five HTML artifacts the rest of your team actually opens. By Sunday night the entire system is yours, every file under your control, no SaaS dependency.

This is the first book on the shelf that puts the manifesto in writing. Five other AI Second Brain books on Amazon promise dashboards and prompt libraries. None of them teach the structured-files stance Andrej Karpathy named in his LLM Knowledge Bases gist. None of them ship the HTML Artifacts pillar 6.5 million developers saw and 20,000+ bookmarked. None of them give you the chunking-with-overlap retrieval mechanic that stops one bad summary from compounding into structural error.

What you walk away with this weekend, applicable to PKM, sales, support, code review, or research:

1. A personal LLM Wiki with five live entries (one per use case), structured the way the agent reads.
2. A persistent memory injection that fires every new session, no re-explaining.
3. A wiki lint script that catches stale entries before they compound.
4. A reusable design-system HTML file every future artifact pulls from.
5. Three production-ready editor HTMLs: triage board, feature flag editor, prompt tuner.
6. A two-agent verification workflow where one agent checks the other's plan.
7. The full system as a $5K to $20K consulting offer the next time a client asks "can you set this up for us?"

Every week, five more "AI Second Brain" books drop on Kindle and the shelf gets crowded. Two of them are 28-page micro-books. None ship HTML Artifacts. Peter Yang is interviewing the founder who built a Personal AI OS in Claude Code on his next podcast. The day that episode goes mainstream, "wiki over prompt" becomes table stakes and the early-mover window closes. This one is 176 dense pages, every chapter shipping a working artifact, and you can be on it by Sunday night. Scroll up and grab your copy.