About This Wiki

This entrepreneurship knowledge base was built entirely by an LLM (Claude) over the course of one session — 62 loop iterations, 60+ git commits, from empty directory to 90 articles and 1,000+ wikilinks. No article was written by a human. Every word was researched, synthesized, cross-referenced, and committed by the machine.

How It Was Built

The Infrastructure

  • CLAUDE.md: Operating manual telling the LLM what to do each iteration
  • 7 CLI tools: ingest, search, search_index, lint, backlinks, stats, render
  • The /loop skill: Fires the prompt every 10 minutes via cron
  • Obsidian: The viewing frontend (open wiki/ as a vault)

The Loop Prompt

contribute one atomic piece of knowledge to the entrepreneurship
wiki following CLAUDE.md instructions

Each iteration: read indexes → pick highest-priority action → research/write/update → lint → backlinks → commit. Fresh context each time.

The Decision Tree

Priority order each iteration:

  1. Uncompiled sources exist → compile them
  2. Stubs need enrichment → expand them
  3. Topic gaps → research + ingest new source
  4. Health issues → fix them
  5. Synthesis opportunity → write connecting article
  6. Deepen coverage → new angles, case studies, contrarian views

Growth Phases

  • Phase 1 (iterations 1-15): Breadth. Ingest 10+ sources, create all core concepts.
  • Phase 2 (iterations 15-35): Depth. Promote drafts to complete, write synthesis articles, add case studies.
  • Phase 3 (iterations 35-62): Expansion. New topic areas, guides, visual outputs, quality passes.

What Makes This Different

From a Textbook

Textbooks are written once. This wiki is living — the loop keeps enriching it. Articles link to each other bidirectionally, creating a knowledge graph, not a linear narrative.

From a Blog

Blogs are chronological. This wiki is topological — organized by concept, connected by wikilinks. You can start anywhere and follow the links to related ideas.

From ChatGPT Answers

Chat answers are ephemeral. This wiki is persistent, versioned (git), and accumulated. Each iteration builds on everything before it. The 60th article is richer because it can reference the first 59.

From a Human-Written Wiki

A human would take months to produce 90 articles with 1,000+ cross-references. The LLM did it in one session. But more importantly: the cross-referencing is systematic, not ad hoc. Every article links to every related concept because the LLM reads the full index before writing.

The Knowledge Graph

The wiki forms a dense graph with clear structure:

Core Hub: product-market-fit (90+ incoming links) — the gravitational center

Primary Clusters:

Synthesis Layer: 9 articles connecting clusters (Startup Lifecycle, Validation Stack, Money Playbook, Building the Team, Leadership Modes, Experts Disagree, Founder OS, AI Era, Case Study Comparison)

Case Studies: 6 real companies (Airbnb, Stripe, Slack, Shopify, WeWork, Cursor) mapped to the frameworks

Entry Points: 5 guides for different needs (Start Here, Cheat Sheet, FAQ, Founder OS, Recommended Reading)

The Sources

21 sources from 19 authors, spanning 2006-2025:

Author# SourcesPerspective
Paul Graham6First principles, essays
Sam Altman1YC playbook, comprehensive
Steve Blank1Customer development methodology
Eric Ries1Lean startup, validated learning
Peter Thiel1Monopoly, competition, secrets
Marc Andreessen1Product-market fit
Jessica Livingston1YC patterns, determination
Keith Rabois1Operations, barrels vs ammunition
Ben Horowitz1Wartime/peacetime leadership
Rahul Vohra1Quantitative PMF measurement
Rob Fitzpatrick1Customer interview technique
Mark Leslie1Go-to-market framework
Jason Fried1Bootstrapping, calm companies
a16z1Network effects dynamics
Fractl1Startup failure data (193 post-mortems)
Michael Seibel1YC essential advice, 90/10 rule
Garry Tan1AI-era YC data

How to Extend This Wiki

The loop prompt keeps working. Future iterations can:

  • Find new sources (non-US perspectives, academic research, recent publications)
  • Write more case studies (Notion, Figma, Theranos, a non-US company)
  • Deepen existing articles with additional sources
  • Create new synthesis articles connecting concepts in novel ways
  • Render more Marp slide decks for presentations
  • Challenge existing content with contrarian perspectives

How to Build Your Own

/kb-init [your subject]

This creates the full infrastructure. Then:

/loop 10m contribute one atomic piece of knowledge to the [subject] wiki following CLAUDE.md instructions

The LLM handles the rest.

See Also