FREE LIVE MASTERCLASS · BEGINNER · 60 MINUTES

Build Your Personal Wiki in 60 Minutes.

Saturday, June 2010:00 AM PST

Tired of re-explaining your context to Claude every session? Or losing track of insights because your notes are scattered across Notion, Slack, and PDFs? This session fixes that — you'll build a personal wiki that feeds Claude exactly what it needs, every time.

Claim Your Free Spot →Free · Online · Open to all

David Prorok

AI Systems Architect & Founder

Ex-Software Engineer · Facebook

Former Facebook engineer (Learning Products, Music & Rights, Events). Builds AI-powered systems across legal ops, bookkeeping, music, and education. CS, UIUC.

Background

Facebook

Uber

UIUC

WHAT YOU SHIP

Six artifacts. Plus a certificate.

David experimented with Karpathy's Wiki and in this session he'll share what has worked out for him — and how you could too build a useful app with your own corpus, on your own URL, in one sitting.

This will help you stop losing insights across scattered notes, transcripts, and docs — and give you a personal thinking partner that actually knows your context, grounded in sources you've chosen, queryable on demand.

A live, deployed wiki

A real URL with your data, your schema, and your auth — owned by you.

The Karpathy template

Forked, documented, and tuned to your domain. Remix it for your team on Monday.

An ingest + lint pipeline

Edge functions that turn raw sources into interlinked pages with citations and a self-healing index.

A query agent that cites

Grounded answers from your own corpus — saveable back into the wiki as new synthesis pages.

A schema you can evolve

The CLAUDE.md pattern: edit the rules in chat, let the agent migrate the rest.

A weekly habit, not a project

The 10-minute ritual that turns scattered inputs into compounding knowledge.

THE AGENDA

Five tight sections. Ship by the end.

SECTION 018 min

The Wiki Mental Model

Why Karpathy's pattern works — three layers (raw sources, LLM-owned pages, index/log), the Programmer/User dual role, and why markdown beats Notion for AI workflows.

SECTION 0215 min

Stand Up Your Wiki in 15 Minutes

Spin up the database, storage bucket, and edge functions live with Lovable Cloud. Wire Gemini 2.5 Pro for structured ingest. Render markdown + wikilinks with backlinks and a graph view.

15m

SECTION 0317 min

Ingest, Process, Refine

Drop in your own transcripts, PDFs, URLs, and chat logs. Use the agent's request_pages / upsert_pages / append_log tools. Lint for orphans, contradictions, and stale claims.

17m

SECTION 0412 min

Query, Synthesize, Ship

Streaming Q&A grounded in your pages with citations. Save synthesized answers back as new pages — the compounding loop. Export to briefs, slide outlines, and PRDs.

12m

SECTION 058 min

Make It Yours — Forever

Edit the schema in chat, not the UI — bulk-migrate pages safely. Remix the template for your team or domain. The 10-minute weekly ritual that compounds.

Prereqs: Comfort with markdown · A free Lovable account · 5–10 real sources to ingest (transcripts, PDFs, links)

WHO IT'S FOR

Stop re-explaining yourself to Claude. Build context that persists.

Tired of pasting the same context every session

Your wiki becomes Claude's permanent memory — feed it once, query it forever. No more copying and pasting background into every new chat.

Losing insights because your notes are scattered

Transcripts in Notion, PDFs in Drive, links in Slack — all ingested into one interlinked, queryable wiki that actually surfaces what you need.

Want an LLM that knows your work, not the internet

Answers grounded in your own corpus, not hallucinated from thin air. Your context, your schema, your pages.

Done studying AI — ready to ship something

60 minutes. Deployed URL. A habit you can sustain in 10 minutes a week. David built this — you will too.