AIMM Session — November 6, 2025: Your Expertise Is a Product

30-Second Summary

Dense with wins and landmark live demo. Dirk crossed a major milestone — self-hosted N8N in Europe. Don shaved five days off a university grant project using Claude deep research. Lou walked through his actual production RAG stack: a Qdrant vector database powering custom ChatGPT action and Claude MCP for a law firm client.

1. The Dip Has a Winner: Dirk’s Infrastructure Milestone

Dirk Ohlmeier has a self-hosted N8N instance running on a European server — fully set up, secure, saving real money monthly. Next move: swap Airtable for Supabase.

Lou: “Nothing is ever as easy as people say it’s going to be, and so a lot of people give up at the first block. You’re a study in persistence.”

2. Claude Deep Research Just Saved Someone a Work Week

Don was ghostwriting a grant for U of Alberta Department of Chemistry — normally 5 full days. Used Claude deep research surgically: exact question, specific sources, exact format match.

Result: Five days became hours.

3. Your Expertise Is a Product

Rich Schefren has “Rich’s Brain” — journals, frameworks, decades of content synthesized into premium AI product.

Jay Abraham with Michael Simmons: took Jay’s entire content library into prompts that reproduce his thinking. Launching via Substack; SaaS planned.

The model applies to everyone: identify the most valuable thing you deliver, capture it, package into RAG database, offer standalone or as a complement to live work.

4. Live Demo: RAG Database → ChatGPT Custom GPT → Claude MCP

Lou’s production stack for law firm (Janssen legal team):

Vector database (Qdrant, free tier) — upload files, chunk, embed, store
Retriever endpoint — URL that ChatGPT/Claude can query
Custom GPT action — create action, paste schema, add API key as bearer token
Claude MCP — find server code, compile, paste config into claude_desktop_config.json

Queried “Nash Johnson” live, got full AI-generated dossier from deposition transcripts directly in ChatGPT.

The economics: this replaces tools priced at $22,000–$35,000/year. Lou’s cost: ~$100 in API usage over 6 months of building and embedding hundreds of thousands of pages.

Community Corner

Dirk: self-hosted N8N live in Europe after months of pushing through the dip.

Don: Claude deep research turned a 5-day grant project into hours.

Bally: present and taking notes on the RAG architecture for potential healthcare applications.

Donald: building his client-facing tools one Claude-generated HTML page at a time.