Andrej Karpathy Wiki
Show HN: A self-growing wiki of Andrej Karpathy's public work
Last verified:
What is Andrej Karpathy Wiki?
Andrej Karpathy Wiki appears to be a personal knowledge system centered on Karpathy’s ideas about building an LLM-managed wiki: a directory of markdown pages that the model creates, updates, and links together over time. It is designed to turn raw sources into summaries, entity pages, concept pages, comparisons, and higher-level synthesis.
The core idea is that the LLM owns the knowledge layer rather than simply retrieving chunks at query time. That means it can generate new pages, update old ones when new material arrives, maintain cross-references, and check for contradictions.
The website and related material suggest it is aimed at people who want a persistent, compounding “second brain” for research, reading notes, or long-running knowledge projects. It is especially relevant for technical users, researchers, and knowledge workers who want their own AI-maintained wiki.
It is not presented as a conventional app with a polished SaaS product page, but more as a methodology, reference implementation, or personal system. The emphasis is on structured, evolving knowledge rather than one-off Q&A.
Andrej Karpathy Wiki pricing
Pricing model: Freemium
The website does not present standard pricing, free tiers, or paid plans. It is described more as a personal wiki methodology and markdown-based knowledge system than a commercial subscription product. No pricing details, included features by plan, or account tiers are provided on the site.
Andrej Karpathy Wiki pros
- LLM writes and maintains pages automatically
- Builds a persistent knowledge base
- Creates summaries from raw sources
- Generates entity pages
- Generates concept pages
- Produces comparison pages
- Supports higher-level synthesis
- Updates existing pages when new sources arrive
- Maintains cross-references between pages
- Checks for contradictions
- Works as an always-growing second brain
- Better suited to long-term knowledge accumulation
- Reduces manual note organization
- Useful for research-heavy workflows
- Obsidian-compatible markdown structure
- Designed for compounding knowledge over time
Andrej Karpathy Wiki cons
- Requires feeding it source material
- Not a turnkey consumer app
- Depends heavily on LLM quality
- Can accumulate token costs over time
- Needs ongoing maintenance or scheduled runs
- May be harder for non-technical users
- Not ideal for quick one-off answers
- Contradictions still need automated checking
Frequently asked questions about Andrej Karpathy Wiki
What is Andrej Karpathy Wiki?
It is an LLM-managed wiki concept and implementation where markdown pages are created and maintained by the model. The goal is to transform raw sources into a structured, interlinked knowledge base that grows over time.
How is it different from RAG?
Instead of only retrieving relevant chunks at query time, the system uses the LLM to build and continuously maintain a persistent wiki. That means the knowledge is organized ahead of time into pages, links, summaries, and syntheses.
What kinds of pages does it create?
It can create summaries, entity pages, concept pages, comparison pages, an overview page, and synthesis pages. The point is to convert scattered source material into a coherent knowledge structure.
Who is it for?
It is best suited for researchers, technical users, and people building a long-term personal knowledge base. It fits workflows where information needs to compound over time instead of being answered once and forgotten.
Does the LLM only answer questions?
No. The LLM is responsible for writing, updating, cross-linking, and reconciling the wiki pages. The system is meant to act more like a knowledge maintainer than a simple chat interface.
How does it keep the wiki consistent?
The system is designed to update pages when new sources arrive and to run checks for contradictions. That helps the wiki stay coherent as it grows.
What format does it use?
The wiki is organized as a directory of LLM-generated markdown files. This makes it suitable for markdown-based tools and note systems such as Obsidian-style workflows.
Can it handle large knowledge bases?
The concept is built for scaling across many pages and many sources, with the LLM continuously maintaining references and syntheses. However, the practical quality still depends on the model, the source material, and the maintenance loop.
Is it a hosted software product?
The site does not present it as a traditional hosted SaaS with subscriptions or packaged plans. It reads more like a public explanation of the workflow and architecture behind the wiki approach.
What is the main benefit of the system?
Its main benefit is compounding knowledge: each new source can improve the wiki rather than just adding another document. Over time, the system aims to produce a more organized and useful personal knowledge base than raw notes or simple search.