How Accurate Are the LLM Wiki Claims?
I told Claude to grab my YouTube videos and ingest them into a wiki.
I didn't connect a single concept myself.
Now I have a living mind map where every video, tool, and technique links to the others, and my AI OS gets smarter every time I add to it.
The idea comes from Andrej Karpathy. He's been using LLMs to build personal knowledge bases and Obsidian as the front end.
Here's the whole thing:
1️⃣ Install Obsidian and create a vault
2️⃣ Open it in Claude Code and paste in Karpathy's LLM wiki gist
3️⃣ Tell it to build the schema, the index, the log, and the folder conventions
4️⃣ Drop sources into a raw folder (a PDF, a URL, a transcript) and say "ingest this"
5️⃣ Watch it split one source into 5, 10, sometimes 20 cross-linked wiki pages
In this tutorial I dropped in two model system cards, and it didn't just summarize them.
It found where the two sources referenced each other and flagged a detail I would have missed reading them separately.
The best part is what's under the hood. Every page is just a markdown file with routing.
So you're not locked into Claude Code. You can point Codex, your own agent, or anything else at the same wiki.
I just dropped a full YouTube video walking through the entire build step by step, from empty vault to a connected second brain.
do fact check
This post is mostly accurate, but it mixes verified concepts with subjective claims and workflow-specific outcomes.
Fact check
"Fable 5 + Karpathy's LLM Wiki is Basically Cheating."
🟡 Opinion.
This is promotional language, not a factual claim.
"I told Claude to grab my YouTube videos and ingest them into a wiki."
🟡 Mostly accurate.
Claude Code can perform this workflow when given access to transcripts or downloaded content and the appropriate tooling. However, Claude Code does not automatically fetch arbitrary YouTube videos without the necessary tools or user-provided data.
"I didn't connect a single concept myself."
🟡 Personal experience.
This is the author's own experience and cannot be independently verified.
"Now I have a living mind map where every video, tool, and technique links to the others."
🟡 Mostly accurate.
Karpathy's LLM Wiki methodology builds interconnected Markdown pages with backlinks and cross-references. However, whether every concept is correctly linked depends on the LLM's output and may require human review.
"My AI OS gets smarter every time I add to it."
🟡 Marketing language.
The knowledge base grows as new information is added, but the underlying LLM does not become smarter or retrain itself. It simply has access to a richer knowledge base.
"The idea comes from Andrej Karpathy."
✅ Accurate.
Andrej Karpathy publicly introduced the concept of using an LLM to maintain a personal wiki and demonstrated the workflow using Obsidian and Markdown.
"He's been using LLMs to build personal knowledge bases and Obsidian as the front end."
✅ Accurate.
This matches Karpathy's published LLM Wiki concept.
"Install Obsidian... paste in Karpathy's LLM wiki gist..."
✅ Accurate.
Karpathy published an LLM Wiki prompt/gist that users can use inside Claude Code or similar coding assistants to generate the wiki structure.
"Drop sources into a raw folder... say 'ingest this'."
🟡 Mostly accurate.
This is how Karpathy's workflow is designed. The exact command and folder structure depend on the specific prompt or implementation rather than being built into Claude Code itself.
"Watch it split one source into 5, 10, sometimes 20 cross-linked wiki pages."
🟡 Reasonable outcome.
The LLM Wiki prompt intentionally decomposes source material into multiple linked Markdown pages. The exact number of generated pages varies by source and prompt.
"It found where the two sources referenced each other and flagged a detail I would've missed."
🟡 Personal anecdote.
This is a plausible outcome but reflects the author's experience rather than a guaranteed capability.
"Every page is just a Markdown file with routing."
✅ Accurate.
The LLM Wiki stores knowledge as standard Markdown files connected through links and folder organization.
"You're not locked into Claude Code. You can point Codex, your own agent, or anything else at the same wiki."
✅ Accurate.
Since the knowledge base is plain Markdown, it can be used by other AI coding assistants and tools that support Markdown files.
"From empty vault to a connected second brain."
🟡 Marketing language.
The workflow can build a structured knowledge base, but whether it functions as a "second brain" depends on the user's content, maintenance, and usage.
Overall verdict
Accuracy: 9/10
The post accurately describes Karpathy's LLM Wiki workflow and its integration with Obsidian and Claude Code. The weaker parts are subjective claims about the AI becoming "smarter," personal anecdotes about discovering hidden insights, and promotional language such as "second brain."
Real vs Fluff
🟢 90% Real | 🟡 10% Fluff
Real (90%)
Karpathy introduced the LLM Wiki concept.
Uses Obsidian and Markdown files.
Claude Code can generate and maintain the wiki.
Sources can be ingested into linked knowledge pages.
Cross-linking and backlinks are core features.
The wiki remains tool-agnostic because it uses plain Markdown.
Fluff / Needs correction (10%)
"AI OS gets smarter" refers to a growing knowledge base, not improvements to the underlying model.
Cross-links are generated automatically but are not guaranteed to be complete or correct.
Discovering hidden relationships between documents is a possible outcome, not a guaranteed feature.
"Second brain" is a metaphor rather than a technical capability.