The LLM Wiki Directory for Humans, Not Just Hackers
Curated tools, schemas, and guides based on Andrej Karpathy's LLM Wiki pattern — from no-code platforms to advanced Claude Code setups.
Whether you're a researcher, writer, or developer, start building your AI second brain today.
Featured in HN discussions · Curated by humans · Free schemas to copy
What is LLM Wiki?
Andrej Karpathy introduced the LLM Wiki pattern in April 2026 — a persistent, compounding knowledge base maintained by an LLM. Instead of re-deriving answers through RAG every time, your model writes a wiki that gets richer every time you use it. The idea caught on so fast that within a week there were already open-source implementations, YouTube explainers, and a dozen hot-take blog posts. What was missing was a single place that turns the spec into things you can actually download and apply without another weekend of trial and error. That is the gap this site fills.
Persistent Knowledge
Unlike RAG, your knowledge compounds. New sources ripple through existing pages, contradictions get flagged, and everything stays interlinked. Add a paper about transformer scaling and the pages for attention, optimization, and GPU economics all get updated — not just retrieved from a chunk cache.
Markdown-First
Plain markdown files in a folder. No vector DB, no embeddings, no infrastructure to maintain. Works with Obsidian, Notion, or any editor. You can grep your LLM Wiki, diff it in git, or read it on a plane. The knowledge base outlives any tool you use to build it.
LLM-Maintained
Let Claude, Gemini, or any capable LLM compile, update, and cross-reference your raw sources into a structured wiki — automatically. The model follows a schema you define, so the output is predictable and auditable. You stay in the loop without doing the typing yourself.
Token-Efficient
Compiled knowledge fits in context. A 400K-word wiki beats RAG on small-to-mid knowledge bases for both latency and accuracy. For most personal knowledge bases a compiled LLM Wiki answers faster than a RAG pipeline and stays consistent between sessions and devices.
Why Choose llmwiki
We turn Karpathy's spec into things you can actually download and use — templates, schemas, and step-by-step guides for three kinds of users. Our readers tell us they come here for the same reason: the rest of the internet explains the pattern, we show you how to run it on your own machine, your own vault, your own papers. Everything on the site is free to read, and the paid kits exist so you can skip the trial-and-error weekend if your time is worth more than the ticket price.
Build Your LLM Wiki in 3 Steps
No terminal required. Works with Claude, Gemini, or any capable LLM. The same three steps hold whether you are a developer shipping a pipeline or a writer who just wants a better second brain. The hardest part is picking a schema that matches how you actually think — once that is locked in, every new source compiles itself in the same predictable shape.
Pick Your Schema
Start from one of our five battle-tested schema.md templates — general, research, engineering, product, or SEO. Each one is a ready-to-paste CLAUDE.md companion that covers the most common LLM Wiki use cases. Edit it to fit your project, or start from scratch once you have seen the pattern a few times.
Drop in Your Sources
Put your raw files (papers, notes, screenshots, links) in a folder. Point Claude or Gemini at it with your schema.md and say 'compile'. Raw files live in a raw/ folder — PDFs, screenshots, meeting notes, research papers, whatever. Claude reads the schema and writes a structured wiki alongside in wiki/ without touching the sources.
Query & Compound
Ask questions, add sources, fix pages. Every new source ripples through — your wiki gets smarter every week, not every session. As new sources arrive, the LLM updates existing pages rather than creating orphans. After a few weeks your wiki becomes the highest-signal draft of everything you have learned in the space.
Everything You Need to Build an LLM Wiki
From zero-code setups to advanced Claude Code pipelines — curated for how you actually work. Every item on this list exists because a reader asked for it and we wrote it. Nothing is speculative, nothing is generated; every guide has been tested against a real knowledge base before it ships. If we haven't tried it yet, it doesn't go on the site.
Plain-English Primer
Our anchor guide to Karpathy's LLM Wiki pattern — 5 minutes, no jargon, written for non-technical readers first. A 10-minute read that replaces the hour of hot takes you would otherwise wade through on X and YouTube. No jargon, just the LLM Wiki pattern in plain prose.
Schema Library
5 battle-tested schema.md templates (general, research, engineering, product, SEO) with real-world examples and full source. Each schema is battle-tested against real projects before it ships — no wishful examples, only the shapes that actually produce clean LLM Wiki output.
Open-Source Directory
Curated list of LLM Wiki implementations on GitHub — ranked by difficulty, stack, and 'time to first useful page'. We install every repo before reviewing it, score the time-to-first-useful-page, and flag the ones that abandon the LLM Wiki semantics halfway through.
No-Code Tools Directory
For writers, researchers, and consultants who don't want to touch a terminal. Side-by-side comparison of no-code LLM Wiki tools. For readers who never want to see a terminal, we compare native apps, Obsidian plugins, and hosted services side by side with honest pros and cons.
Academic Use Cases
Deep-dive guides for managing research papers, literature reviews, and citation graphs with LLM Wiki + Zotero. Managing 200+ papers is a different problem than managing 20 notes. Our research guides cover Zotero integration, citation graphs, and end-to-end literature review workflows.
Curated Newsletter
Free newsletter with the best LLM Wiki resources, schema patterns, and community discussions. We send when we have something worth your time — no fixed schedule, no ads, no sponsored fluff — just the highest-signal LLM Wiki content we have found.
Frequently Asked Questions About LLM Wiki
Still confused? Read the free primer or skim the answers below. If your question isn't covered, email us — we read every message.
Schema patterns and LLM Wiki news, sent occasionally.
No fixed schedule, no upsell. Unsubscribe in one click.
Drop your email
One-click unsubscribe. We never share or sell.