Loading video...

Video Failed to Load

Go Home

Karpathy found a way to reduce token consumption by 90% The problem is that the LLM re-reads the same files over and over again, loses context between documents, and provides less accurate answers as a result The solution is called Wiki Layer the LLM cleans, structures, and links all...

Asteri_eth's profile picture

Asteri

2,241 subscribers

1,078,354 views • 1 month ago •via X (Twitter)

EducationNews & PoliticsScience & Technology
LIVE

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

HERMES AGENT SHIPS WITH A BUNDLED SKILL FOR ANDREJ KARPATHY'S LLM WIKI PATTERN. A SELF-IMPROVING KNOWLEDGE BASE THAT GROWS EVERY TIME YOU FEED IT. mentioned this briefly in the overnight workflow article. here is the full breakdown. what it is: a self-improving knowledge base built as interlinked markdown files. unlike RAG (which rediscovers knowledge from scratch every query), the wiki compiles knowledge once and keeps it current. cross-references stay linked. contradictions get flagged automatically. synthesis reflects everything ingested so far. why this matters for Hermes memory: Hermes built-in memory knows YOU. it remembers your conversations, your preferences, your business context across sessions. but it doesn't know your inbox. or your meeting transcripts. or that article you saved last week. or the expert framework you want it to learn. the LLM Wiki solves that. THE DIVISION OF LABOR human curates sources and directs analysis. agent summarizes, cross-references, files, and maintains consistency. you drop in articles, transcripts, notes. Hermes indexes them, links related concepts, flags contradictions, updates affected pages. your knowledge base grows itself. SETUP IS ONE COMMAND the skill ships with Hermes. enable it. set WIKI_PATH in ~/.hermes/.env: WIKI_PATH=/Users/you/wiki defaults to ~/wiki if unset. then drop anything into it: "index this article into my wiki: [paste URL or text]" Hermes reads it, builds a source page, updates related entries, flags contradictions. THE OBSIDIAN ANGLE set OBSIDIAN_VAULT_PATH to the same directory. now your wiki is visible in Obsidian's graph view. nodes, links, backlinks. all built by Hermes. for headless servers: install obsidian-headless. syncs vaults without a GUI. agent writes from the server, you read on your laptop. THE COMPOUND EFFECT Hermes knows you. the wiki knows your world. combine them and the agent answers questions using BOTH contexts at once. month 1: you explain things twice. month 3: the agent references the wiki on its own. answers get sharper because the knowledge base got sharper. AUTOMATIONS THAT FEED THE WIKI set cron jobs to ingest automatically: "every day at 9am, check Granola for new meetings. add any new transcripts to my wiki under meeting notes." "every morning, scan my Gmail starred items. add anything worth keeping to the wiki." "every week, check arXiv for new papers in [your niche]. summarize and file." your wiki grows while you sleep. Hermes never forgets what gets indexed. THE LIMITATION TO KNOW unlike Hermes memory (which is conversational and lives across sessions), the wiki is a separate knowledge layer. Hermes won't pull from the wiki automatically unless you reference it or save it as a skill. best setup: build an LLM Wiki personality that tells Hermes to consult the wiki when answering strategy questions or domain-specific queries. full HERMES AGENT OVERNIGHT WORKFLOW👇
14:28

HERMES AGENT SHIPS WITH A BUNDLED SKILL FOR ANDREJ KARPATHY'S LLM WIKI PATTERN. A SELF-IMPROVING KNOWLEDGE BASE THAT GROWS EVERY TIME YOU FEED IT. mentioned this briefly in the overnight workflow article. here is the full breakdown. what it is: a self-improving knowledge base built as interlinked markdown files. unlike RAG (which rediscovers knowledge from scratch every query), the wiki compiles knowledge once and keeps it current. cross-references stay linked. contradictions get flagged automatically. synthesis reflects everything ingested so far. why this matters for Hermes memory: Hermes built-in memory knows YOU. it remembers your conversations, your preferences, your business context across sessions. but it doesn't know your inbox. or your meeting transcripts. or that article you saved last week. or the expert framework you want it to learn. the LLM Wiki solves that. THE DIVISION OF LABOR human curates sources and directs analysis. agent summarizes, cross-references, files, and maintains consistency. you drop in articles, transcripts, notes. Hermes indexes them, links related concepts, flags contradictions, updates affected pages. your knowledge base grows itself. SETUP IS ONE COMMAND the skill ships with Hermes. enable it. set WIKI_PATH in ~/.hermes/.env: WIKI_PATH=/Users/you/wiki defaults to ~/wiki if unset. then drop anything into it: "index this article into my wiki: [paste URL or text]" Hermes reads it, builds a source page, updates related entries, flags contradictions. THE OBSIDIAN ANGLE set OBSIDIAN_VAULT_PATH to the same directory. now your wiki is visible in Obsidian's graph view. nodes, links, backlinks. all built by Hermes. for headless servers: install obsidian-headless. syncs vaults without a GUI. agent writes from the server, you read on your laptop. THE COMPOUND EFFECT Hermes knows you. the wiki knows your world. combine them and the agent answers questions using BOTH contexts at once. month 1: you explain things twice. month 3: the agent references the wiki on its own. answers get sharper because the knowledge base got sharper. AUTOMATIONS THAT FEED THE WIKI set cron jobs to ingest automatically: "every day at 9am, check Granola for new meetings. add any new transcripts to my wiki under meeting notes." "every morning, scan my Gmail starred items. add anything worth keeping to the wiki." "every week, check arXiv for new papers in [your niche]. summarize and file." your wiki grows while you sleep. Hermes never forgets what gets indexed. THE LIMITATION TO KNOW unlike Hermes memory (which is conversational and lives across sessions), the wiki is a separate knowledge layer. Hermes won't pull from the wiki automatically unless you reference it or save it as a skill. best setup: build an LLM Wiki personality that tells Hermes to consult the wiki when answering strategy questions or domain-specific queries. full HERMES AGENT OVERNIGHT WORKFLOW👇

YanXbt

30,248 views • 1 month ago

This beginner guide to Karpathy's LLM Wiki is one of the most useful AI videos i've shared this year Karpathy's idea: instead of asking AI questions every time, make it build you a permanent knowledge base that grows smarter over time > drop your documents into a folder > Claude reads them and builds a structured wiki in Obsidian > every time you add a new source, Claude Sonnet updates existing pages and links new ideas > ask complex questions - Claude Opus answers from the organized wiki, not raw files it's like having a researcher who never forgets and connects dots across everything you've ever given it full setup guide is in the video above
14:21

This beginner guide to Karpathy's LLM Wiki is one of the most useful AI videos i've shared this year Karpathy's idea: instead of asking AI questions every time, make it build you a permanent knowledge base that grows smarter over time > drop your documents into a folder > Claude reads them and builds a structured wiki in Obsidian > every time you add a new source, Claude Sonnet updates existing pages and links new ideas > ask complex questions - Claude Opus answers from the organized wiki, not raw files it's like having a researcher who never forgets and connects dots across everything you've ever given it full setup guide is in the video above

Mr. Buzzoni

58,787 views • 2 months ago

A developer opened a law firm’s token bill and realized 20 users were burning $61,920 just to ask questions their files could answer offline. $61,920 cloud TCO. $23,180 local TCO. 4.12-month break-even. 20 employees. 0 client files uploaded. The old RAG pitch breaks when legal files, medical notes, and finance records cannot leave the building. Search snippets are not enough when the firm needs a living wiki of clients, projects, entities, and cross-linked decisions. So the agency installs a local node : OpenClaw watches the intake folder, a local model cleans and links the documents, and Obsidian becomes the private corporate brain. Emails, Slack exports, PDFs, and notes turn into markdown pages the team can browse without sending secrets to an API. That is the real B2B offer. Not “AI chat for your company,” but a private knowledge machine that pays for itself in month 5.
0:57

A developer opened a law firm’s token bill and realized 20 users were burning $61,920 just to ask questions their files could answer offline. $61,920 cloud TCO. $23,180 local TCO. 4.12-month break-even. 20 employees. 0 client files uploaded. The old RAG pitch breaks when legal files, medical notes, and finance records cannot leave the building. Search snippets are not enough when the firm needs a living wiki of clients, projects, entities, and cross-linked decisions. So the agency installs a local node : OpenClaw watches the intake folder, a local model cleans and links the documents, and Obsidian becomes the private corporate brain. Emails, Slack exports, PDFs, and notes turn into markdown pages the team can browse without sending secrets to an API. That is the real B2B offer. Not “AI chat for your company,” but a private knowledge machine that pays for itself in month 5.

Lummox

24,480 views • 7 days ago

Karpathy gave 100,000 people the LLM Wiki idea. 1 guy actually built it into a Claude + Obsidian brain that never forgets you. Every AI tool forgets you the second you close the tab. ChatGPT, Claude, Codex. All gone. He killed that. A shared memory layer your AI builds and maintains on its own. 3 layers. Raw, Wiki, Schema. You drop a source, Claude reads it, pulls the concepts, links them, cites the original, never touches your raw notes. Scales to 1,000s of notes without breaking. Then it goes deeper. An agentic firewall locks agents out of your other vaults. A local model running free and private, zero cloud. 2 videos walk through it all. 1 on why. 1 on the build, start to finish. Most people will keep pasting the same context into 6 tabs for the next 3 years. Build it once. Never explain yourself to a model again.
34:07

Karpathy gave 100,000 people the LLM Wiki idea. 1 guy actually built it into a Claude + Obsidian brain that never forgets you. Every AI tool forgets you the second you close the tab. ChatGPT, Claude, Codex. All gone. He killed that. A shared memory layer your AI builds and maintains on its own. 3 layers. Raw, Wiki, Schema. You drop a source, Claude reads it, pulls the concepts, links them, cites the original, never touches your raw notes. Scales to 1,000s of notes without breaking. Then it goes deeper. An agentic firewall locks agents out of your other vaults. A local model running free and private, zero cloud. 2 videos walk through it all. 1 on why. 1 on the build, start to finish. Most people will keep pasting the same context into 6 tabs for the next 3 years. Build it once. Never explain yourself to a model again.

West Lord

74,975 views • 27 days ago

Our new short course, Knowledge Graphs for RAG, is now available! Knowledge graphs are a data structure that is great at capturing complex relationships between data of multiple types. By enabling more sophisticated retrieval of text than similarity search alone, knowledge graphs can improve the context you pass to the LLM and the performance of your RAG applications. In this course, taught by Andreas Kollegger of Neo4j, you’ll - Explore how knowledge graphs work by building a graph of public financial documents from scratch - Learn to write queries that retrieve text and data from the graph and use it to enhance the context you pass to an LLM chatbot - Combine a knowledge graph with a question-answer chain to build better RAG-powered chat systems Sign up here!
3:46

Our new short course, Knowledge Graphs for RAG, is now available! Knowledge graphs are a data structure that is great at capturing complex relationships between data of multiple types. By enabling more sophisticated retrieval of text than similarity search alone, knowledge graphs can improve the context you pass to the LLM and the performance of your RAG applications. In this course, taught by Andreas Kollegger of Neo4j, you’ll - Explore how knowledge graphs work by building a graph of public financial documents from scratch - Learn to write queries that retrieve text and data from the graph and use it to enhance the context you pass to an LLM chatbot - Combine a knowledge graph with a question-answer chain to build better RAG-powered chat systems Sign up here!

Andrew Ng

244,329 views • 2 years ago

We built an app called Hatch It enables Karpathy's entire LLM Knowledge Base workflow out of the box, in 2 or 3 clicks, in a single interface. No need to stitch together Obsidian + plugins + markdown files + custom tools + etc. Hatch is an AI workspace where files, docs, and chats all live together. Not only can the chats READ files and docs, but they can WRITE and ORGANIZE them as well. This means you can set up an LLM Wiki in literally 3 steps, as I show in this video.
11:58

We built an app called Hatch It enables Karpathy's entire LLM Knowledge Base workflow out of the box, in 2 or 3 clicks, in a single interface. No need to stitch together Obsidian + plugins + markdown files + custom tools + etc. Hatch is an AI workspace where files, docs, and chats all live together. Not only can the chats READ files and docs, but they can WRITE and ORGANIZE them as well. This means you can set up an LLM Wiki in literally 3 steps, as I show in this video.

Channing Allen

180,804 views • 3 months ago

Introducing Wikiwise: an open-source Mac app for managing your own Karpathy-style LLM wiki. Set up a new wiki in a few clicks: all you need is Wikiwise + your agent. It's infinitely customizable, just markdown/html under the hood, and one click to share your wiki publicly. Here's how it works: * Install Wikiwise for mac (it's built in Swift so super minimal and performant). In Karpathy's framework, Wikiwise is your IDE. * Start a new Wiki: it generates a new folder on your machine that's scaffolded in the wiki structure Andrej Karpathy describes (index.md, raw folder, wiki folder, CLAUDE.md/AGENTS.md, although it tries to be as un-opinionated as possible). * Then just point your agent (Codex, Claude Code, Cursor, etc) at the folder and tell it what to import -- files on your machine, connect to your Readwise account, or urls from the web. * Your agent creates wthe wiki for you: Your agent will know how to ingest your raw sources (via the AGENTS.md) and will immediately start writing+linking wiki pages for you. * Go crazy on customization! The rendered wiki pages live as static html/css in your folder too so just tell your agent to change stuff, and if you need any more customization Wikiwise is fully open source :) * Ask questions about your research with your agent, ask it to bring in new sources, write new documents, etc. * (optionally) Hit the Publish button to share your wiki with friends/colleagues at a custom URL === I tried to walk the line on a couple constraints with Wikiwise: 1. I wanted it to be easy to spin up new wikis, especially without chaining together a bunch of different apps. It takes me a few minutes to spin up a new wiki on a topic -- I already have five! 2. Infinitely Customizable: one great aspect of building a wiki as Karpathy described is that you can modify any aspect of your wiki with your agent. Every new wiki styling+structure is self-contained in the local folder, which allows you to preserve this. Wikiwise is just an IDE that makes the setup easier and includes a nice un-opinionated starting state. 3. Minimal: Wikiwise is built mostly in Swift, and the DMG you install to download it is only 2.6MB (!) 4. Easy Publishing: my colleague Eleanor Konik has been building her own LLM wikis for months, but has always really struggled to actually share them with her book club. There are tools to do it, but figuring out hosting is always a huge headache. This seemed like an ideal usecase for a tool like Wikiwise to solve. The process of building wikiwise was also pretty interesting -- I "bootstrapped" the app in a way by first building my own wiki based on Karpathy's tweet and other notes I had, and slowly formed the shape of the project in collaboration with my LLM. This was all done in 3 days over the latest Readwise company hackathon we had. Truly an incredible time to be alive. Anyways, curious what you think! Links in next tweet.
3:10

Introducing Wikiwise: an open-source Mac app for managing your own Karpathy-style LLM wiki. Set up a new wiki in a few clicks: all you need is Wikiwise + your agent. It's infinitely customizable, just markdown/html under the hood, and one click to share your wiki publicly. Here's how it works: * Install Wikiwise for mac (it's built in Swift so super minimal and performant). In Karpathy's framework, Wikiwise is your IDE. * Start a new Wiki: it generates a new folder on your machine that's scaffolded in the wiki structure Andrej Karpathy describes (index.md, raw folder, wiki folder, CLAUDE.md/AGENTS.md, although it tries to be as un-opinionated as possible). * Then just point your agent (Codex, Claude Code, Cursor, etc) at the folder and tell it what to import -- files on your machine, connect to your Readwise account, or urls from the web. * Your agent creates wthe wiki for you: Your agent will know how to ingest your raw sources (via the AGENTS.md) and will immediately start writing+linking wiki pages for you. * Go crazy on customization! The rendered wiki pages live as static html/css in your folder too so just tell your agent to change stuff, and if you need any more customization Wikiwise is fully open source :) * Ask questions about your research with your agent, ask it to bring in new sources, write new documents, etc. * (optionally) Hit the Publish button to share your wiki with friends/colleagues at a custom URL === I tried to walk the line on a couple constraints with Wikiwise: 1. I wanted it to be easy to spin up new wikis, especially without chaining together a bunch of different apps. It takes me a few minutes to spin up a new wiki on a topic -- I already have five! 2. Infinitely Customizable: one great aspect of building a wiki as Karpathy described is that you can modify any aspect of your wiki with your agent. Every new wiki styling+structure is self-contained in the local folder, which allows you to preserve this. Wikiwise is just an IDE that makes the setup easier and includes a nice un-opinionated starting state. 3. Minimal: Wikiwise is built mostly in Swift, and the DMG you install to download it is only 2.6MB (!) 4. Easy Publishing: my colleague Eleanor Konik has been building her own LLM wikis for months, but has always really struggled to actually share them with her book club. There are tools to do it, but figuring out hosting is always a huge headache. This seemed like an ideal usecase for a tool like Wikiwise to solve. The process of building wikiwise was also pretty interesting -- I "bootstrapped" the app in a way by first building my own wiki based on Karpathy's tweet and other notes I had, and slowly formed the shape of the project in collaboration with my LLM. This was all done in 3 days over the latest Readwise company hackathon we had. Truly an incredible time to be alive. Anyways, curious what you think! Links in next tweet.

Tristan

95,704 views • 3 months ago

THIS MIGHT BE THE #1 OPEN-SOURCE REPO FOR CLAUDE CODE RIGHT NOW. IT GIVES CLAUDE A MEMORY AND SLASHES YOUR TOKEN COST ON EVERY QUESTION The repo is safishamsi/graphify, a free open-source skill that turns any codebase into a knowledge graph Claude Code can read instantly. Instead of grepping through your files every session, Claude gets a map of how everything connects The problem it fixes: Every time you ask Claude Code about a big repo, it does the same thing, greps through dozens of files like a brute-force Ctrl+F, blows through your context window, and sometimes still misses the answer hiding in a file nobody searched. Claude Code has no memory of how your project is structured. Every session starts from zero What it does: It maps your entire codebase into a knowledge graph, capturing not just which files exist, but which functions depend on which, which modules are central, and which files cluster around the same concern. Claude queries the map instead of scanning files How it works, three passes: 1. Code structure, free and local. Tree-sitter parses your files and pulls out classes, functions, imports and call graphs. No LLM, no tokens, just your actual code mapped deterministically 2. Audio and video, if you have them. Transcribed locally and folded into the graph 3. Docs, papers, images. Here an LLM does semantic analysis, figuring out what each document means and where it fits. Only the meaning gets sent up, never your raw source It saves you money: Normally a question about a big repo makes Claude spawn explore agents that scan file after file, eating your context window and your token budget before you get an answer. With the graph already built, Claude queries the map instead of re-reading the codebase every time. Same answer, a fraction of the tokens. The graph only gets built once, then a hook rebuilds it after each commit for free, so you never pay that scanning cost again. The bigger the repo, the bigger the gap The best parts: it's a skill, so once installed Claude knows when to use it without you memorizing commands. It works on non-code folders too, point it at docs or notes and it can spin up an Obsidian vault How to add it to your Claude: 1. Install Claude Code if you haven't: npm install -g Paul Jankura-ai/claude-code 2. Add the skill: claude skill add safishamsi/graphify 3. Open your project folder and run /graphify . to build the graph 4. Optional, make it automatic: graphify hook install so the graph rebuilds after every commit That's it. Ask Claude about your repo and it reads the map instead of burning tokens on a file hunt Bookmark this
13:23

THIS MIGHT BE THE #1 OPEN-SOURCE REPO FOR CLAUDE CODE RIGHT NOW. IT GIVES CLAUDE A MEMORY AND SLASHES YOUR TOKEN COST ON EVERY QUESTION The repo is safishamsi/graphify, a free open-source skill that turns any codebase into a knowledge graph Claude Code can read instantly. Instead of grepping through your files every session, Claude gets a map of how everything connects The problem it fixes: Every time you ask Claude Code about a big repo, it does the same thing, greps through dozens of files like a brute-force Ctrl+F, blows through your context window, and sometimes still misses the answer hiding in a file nobody searched. Claude Code has no memory of how your project is structured. Every session starts from zero What it does: It maps your entire codebase into a knowledge graph, capturing not just which files exist, but which functions depend on which, which modules are central, and which files cluster around the same concern. Claude queries the map instead of scanning files How it works, three passes: 1. Code structure, free and local. Tree-sitter parses your files and pulls out classes, functions, imports and call graphs. No LLM, no tokens, just your actual code mapped deterministically 2. Audio and video, if you have them. Transcribed locally and folded into the graph 3. Docs, papers, images. Here an LLM does semantic analysis, figuring out what each document means and where it fits. Only the meaning gets sent up, never your raw source It saves you money: Normally a question about a big repo makes Claude spawn explore agents that scan file after file, eating your context window and your token budget before you get an answer. With the graph already built, Claude queries the map instead of re-reading the codebase every time. Same answer, a fraction of the tokens. The graph only gets built once, then a hook rebuilds it after each commit for free, so you never pay that scanning cost again. The bigger the repo, the bigger the gap The best parts: it's a skill, so once installed Claude knows when to use it without you memorizing commands. It works on non-code folders too, point it at docs or notes and it can spin up an Obsidian vault How to add it to your Claude: 1. Install Claude Code if you haven't: npm install -g Paul Jankura-ai/claude-code 2. Add the skill: claude skill add safishamsi/graphify 3. Open your project folder and run /graphify . to build the graph 4. Optional, make it automatic: graphify hook install so the graph rebuilds after every commit That's it. Ask Claude about your repo and it reads the map instead of burning tokens on a file hunt Bookmark this

Yarchi

55,345 views • 1 month ago

KARPATHY'S SECOND BRAIN REWRITES YOUR NOTES EVERY TIME YOU LEARN SOMETHING NEW most AI note setups are just RAG you ask, it digs through your files, answers, and forgets nothing carries over every query starts from zero Karpathy's llm-wiki works differently the AI keeps a living wiki inside Obsidian every source you drop in gets: > read and summarized > cross-linked to what you already have > checked for contradictions > folded into the running synthesis the knowledge compounds instead of resetting now the part people miss: a wiki like this is a product > deep research reports to founders go for $2-5k each > a due diligence package on a niche runs $5-10k > a paid newsletter off your synthesis pages clears $3-4k/mo at a few hundred subs your job is to curate sources and ask good questions the AI does the bookkeeping no human keeps up with - Obsidian is the IDE - the LLM is the programmer - your wiki is the codebase save this before your notes pile any higher 👇
0:25

KARPATHY'S SECOND BRAIN REWRITES YOUR NOTES EVERY TIME YOU LEARN SOMETHING NEW most AI note setups are just RAG you ask, it digs through your files, answers, and forgets nothing carries over every query starts from zero Karpathy's llm-wiki works differently the AI keeps a living wiki inside Obsidian every source you drop in gets: > read and summarized > cross-linked to what you already have > checked for contradictions > folded into the running synthesis the knowledge compounds instead of resetting now the part people miss: a wiki like this is a product > deep research reports to founders go for $2-5k each > a due diligence package on a niche runs $5-10k > a paid newsletter off your synthesis pages clears $3-4k/mo at a few hundred subs your job is to curate sources and ask good questions the AI does the bookkeeping no human keeps up with - Obsidian is the IDE - the LLM is the programmer - your wiki is the codebase save this before your notes pile any higher 👇

Mr. Buzzoni

18,500 views • 7 days ago

Everyone keeps asking me how to build a second brain or an LLM wiki. Here is the easiest setup I have found. I took my Wiki Builder skill, installed it into HyperAgent (Hyperagent) as a reusable skill, and asked it to build a research wiki on LLM verification from the latest 2026 papers. Recorded a quick demo. Built with Hyperagent. It planned first, asked a few sharp questions, then did the research: 29 papers curated into 21 files, a research map, a glossary, and clean subfields. Now it is a knowledge base that my other research agents build on. HyperAgent has all the capabilities to allow your agents and skills to compound. That’s a powerful use of AI agents.
8:03

Everyone keeps asking me how to build a second brain or an LLM wiki. Here is the easiest setup I have found. I took my Wiki Builder skill, installed it into HyperAgent (Hyperagent) as a reusable skill, and asked it to build a research wiki on LLM verification from the latest 2026 papers. Recorded a quick demo. Built with Hyperagent. It planned first, asked a few sharp questions, then did the research: 29 papers curated into 21 files, a research map, a glossary, and clean subfields. Now it is a knowledge base that my other research agents build on. HyperAgent has all the capabilities to allow your agents and skills to compound. That’s a powerful use of AI agents.

elvis

225,183 views • 11 days ago

THE OBSIDIAN GRAPH IS NOT A PRETTY NOTE MAP, IT IS A SELF-MAINTAINING AI WIKI THAT CAN TURN 120 SAVED SOURCES INTO 700 LINKED PAGES WHILE YOU ONLY KEEP ADDING NEW MATERIAL 00:11 the graph opens and the trick becomes obvious: every dot is a saved idea, every cluster is a topic, and every line is context the AI no longer has to rebuild from scratch every morning most people use chatgpt like a disposable memory slot. they upload 1 pdf, ask 3 questions, close the tab, and the next day the model knows nothing unless they feed it the same context again karpathy’s LLM wiki pattern changes the game. the model reads each new source, pulls out claims, updates old pages, creates entity files, links concepts, and marks places where newer info breaks older notes with claude code and obsidian, a single clipped article can expand into 8-20 markdown files: source summary, concept pages, people, tools, comparisons, open questions, and a running index after 60 sources, this is no longer “note taking.” it becomes a private research machine that remembers your old thinking, catches repeated angles, and gives you cited answers from your own archive bookmark this before your notes turn into a graveyard again
0:35

THE OBSIDIAN GRAPH IS NOT A PRETTY NOTE MAP, IT IS A SELF-MAINTAINING AI WIKI THAT CAN TURN 120 SAVED SOURCES INTO 700 LINKED PAGES WHILE YOU ONLY KEEP ADDING NEW MATERIAL 00:11 the graph opens and the trick becomes obvious: every dot is a saved idea, every cluster is a topic, and every line is context the AI no longer has to rebuild from scratch every morning most people use chatgpt like a disposable memory slot. they upload 1 pdf, ask 3 questions, close the tab, and the next day the model knows nothing unless they feed it the same context again karpathy’s LLM wiki pattern changes the game. the model reads each new source, pulls out claims, updates old pages, creates entity files, links concepts, and marks places where newer info breaks older notes with claude code and obsidian, a single clipped article can expand into 8-20 markdown files: source summary, concept pages, people, tools, comparisons, open questions, and a running index after 60 sources, this is no longer “note taking.” it becomes a private research machine that remembers your old thinking, catches repeated angles, and gives you cited answers from your own archive bookmark this before your notes turn into a graveyard again

Gipp 🦅

41,728 views • 28 days ago

A guy in China plugged an LLM into Karpathy method + Claude Code and built a second brain that remembers everything forever. Not a note app, not RAG just 3 folders on a laptop. Folder 1 holds raw sources: articles, journal entries, chat logs. The LLM reads them but never touches them. Folder 2 is the wiki hundreds of markdown pages the LLM writes and maintains itself, where every person, idea and decision gets filed, cross-referenced and kept current. Folder 3 is one file: CLAUDEmd, the rulebook that turns a chatbot into a librarian. He drops in a new article and Claude reads it, updates 10-15 pages, flags what contradicts his old notes, logs the timestamp while he watches his memory grow node by node in Obsidian graph view. Ask it anything and it doesn’t dig through raw files like ChatGPT does the synthesis already exists, the connections are already drawn. The method comes from a gist Karpathy posted on April 4, now sitting at 5,000 stars and 5,000 forks, with 1 production team already scaled to 4,000 interlinked concepts in 6 months. The part nobody mentions: Vannevar Bush described this exact machine in 1945 the Memex, a personal memory store with trails between documents. Nobody built it for 81 years because someone had to do the maintenance. Now maintenance costs nothing the LLM does the filing while the human does the thinking. Every article makes the brain smarter, and it never forgets. His memory has a git history now.
0:28

A guy in China plugged an LLM into Karpathy method + Claude Code and built a second brain that remembers everything forever. Not a note app, not RAG just 3 folders on a laptop. Folder 1 holds raw sources: articles, journal entries, chat logs. The LLM reads them but never touches them. Folder 2 is the wiki hundreds of markdown pages the LLM writes and maintains itself, where every person, idea and decision gets filed, cross-referenced and kept current. Folder 3 is one file: CLAUDEmd, the rulebook that turns a chatbot into a librarian. He drops in a new article and Claude reads it, updates 10-15 pages, flags what contradicts his old notes, logs the timestamp while he watches his memory grow node by node in Obsidian graph view. Ask it anything and it doesn’t dig through raw files like ChatGPT does the synthesis already exists, the connections are already drawn. The method comes from a gist Karpathy posted on April 4, now sitting at 5,000 stars and 5,000 forks, with 1 production team already scaled to 4,000 interlinked concepts in 6 months. The part nobody mentions: Vannevar Bush described this exact machine in 1945 the Memex, a personal memory store with trails between documents. Nobody built it for 81 years because someone had to do the maintenance. Now maintenance costs nothing the LLM does the filing while the human does the thinking. Every article makes the brain smarter, and it never forgets. His memory has a git history now.

West Lord

61,092 views • 9 days ago

google just turned karpathy's llm-wiki gist into a spec today: the open knowledge format. markdown files, one required field. the format question is now settled. the hard part still isn't. what it changes: - portability. a bundle is a tarball or a git repo. it moves between orgs and tools with no translation layer. - version control. knowledge lives beside the code it describes, with diffs and history for free. - decoupling. a human writes the bundle, an agent reads it, with no shared tooling between them. - no SDK, no account. you read and write plain markdown. what it leaves to you: - contradiction resolution. a new source fights an old claim. OKF stores both and never says which wins. - staleness. timestamp is a field. retiring a dead fact is a call the format won't make. - merge policy. which page a fact lands on, what gets forked or overwritten. the producer agent owns it. - retrieval. index.md holds for a few hundred pages. past that you need hybrid search and a reranker. OKF standardized the filing cabinet. the librarian is still yours to build.
0:45

google just turned karpathy's llm-wiki gist into a spec today: the open knowledge format. markdown files, one required field. the format question is now settled. the hard part still isn't. what it changes: - portability. a bundle is a tarball or a git repo. it moves between orgs and tools with no translation layer. - version control. knowledge lives beside the code it describes, with diffs and history for free. - decoupling. a human writes the bundle, an agent reads it, with no shared tooling between them. - no SDK, no account. you read and write plain markdown. what it leaves to you: - contradiction resolution. a new source fights an old claim. OKF stores both and never says which wins. - staleness. timestamp is a field. retiring a dead fact is a call the format won't make. - merge policy. which page a fact lands on, what gets forked or overwritten. the producer agent owns it. - retrieval. index.md holds for a few hundred pages. past that you need hybrid search and a reranker. OKF standardized the filing cabinet. the librarian is still yours to build.

Rohit

30,710 views • 1 month ago

Fable 5 and Karpathy's knowledge base trick are basically a cheat code. Someone just dropped a full walkthrough on building the most powerful second brain ever Andrej Karpathy shared a simple idea: use an LLM to build personal knowledge bases. Index your sources, let the model organize them, and use Obsidian as the front end. Someone just built the full thing with Claude Code and Fable 5, and walked through every step You drop your sources into a folder (YouTube transcripts, PDFs, articles, meeting notes) and Claude Code reads them, splits them into linked wiki pages, and builds a living second brain that connects everything. It grows on its own as you add more. Where Fable 5 pulls ahead: it found a connection buried across two sources you'd miss reading them separately. And asked to turn the messy web of links into something a beginner could understand, it produced clean, simple HTML in one shot. The same task took a full day on Opus 4.8 and came out worse. The best part: it's all just markdown files with routing. So it's not locked to Claude Code. Plug it into Codex or any agent you want. Give your AI a real knowledge base, and it stops guessing and starts knowing. Bookmark this.
14:34

Fable 5 and Karpathy's knowledge base trick are basically a cheat code. Someone just dropped a full walkthrough on building the most powerful second brain ever Andrej Karpathy shared a simple idea: use an LLM to build personal knowledge bases. Index your sources, let the model organize them, and use Obsidian as the front end. Someone just built the full thing with Claude Code and Fable 5, and walked through every step You drop your sources into a folder (YouTube transcripts, PDFs, articles, meeting notes) and Claude Code reads them, splits them into linked wiki pages, and builds a living second brain that connects everything. It grows on its own as you add more. Where Fable 5 pulls ahead: it found a connection buried across two sources you'd miss reading them separately. And asked to turn the messy web of links into something a beginner could understand, it produced clean, simple HTML in one shot. The same task took a full day on Opus 4.8 and came out worse. The best part: it's all just markdown files with routing. So it's not locked to Claude Code. Plug it into Codex or any agent you want. Give your AI a real knowledge base, and it stops guessing and starts knowing. Bookmark this.

Shadow Nick

45,526 views • 20 days ago

Prompt engineering is where people go when they refuse to write files Karpathy's LLM Wiki gist is painfully simple: > raw sources stay untouched > the agent writes the wiki > `CLAUDE.md` or `AGENTS.md` tells it the rules > `index.md` maps the pages > `log.md` records what changed That is why a good answer should not die in chat history If the model gives you a useful comparison, summary, decision, bug explanation, or research thread, file it back into the project Use: > `wiki/topics/[topic].md` > `wiki/entities/[person-or-company].md` > `wiki/sources/[source-name].md` > `wiki/analysis/[question].md` Make the agent keep the wiki from rotting: > backlinks > summaries > contradictions > stale claims > orphan pages > missing source links For coding projects, same idea: > `CLAUDE.md`: who you are, project rules, files not to touch > `index.md`: current files and features > `log.md`: date, change, reason > saved pages: decisions you do not want to re-explain Start by saving one good answer today
11:03

Prompt engineering is where people go when they refuse to write files Karpathy's LLM Wiki gist is painfully simple: > raw sources stay untouched > the agent writes the wiki > `CLAUDE.md` or `AGENTS.md` tells it the rules > `index.md` maps the pages > `log.md` records what changed That is why a good answer should not die in chat history If the model gives you a useful comparison, summary, decision, bug explanation, or research thread, file it back into the project Use: > `wiki/topics/[topic].md` > `wiki/entities/[person-or-company].md` > `wiki/sources/[source-name].md` > `wiki/analysis/[question].md` Make the agent keep the wiki from rotting: > backlinks > summaries > contradictions > stale claims > orphan pages > missing source links For coding projects, same idea: > `CLAUDE.md`: who you are, project rules, files not to touch > `index.md`: current files and features > `log.md`: date, change, reason > saved pages: decisions you do not want to re-explain Start by saving one good answer today

Alpha Batcher

14,041 views • 1 month ago

THIS GUY BUILT A GOD-MODE SECOND BRAIN IN 3 STEPS THAT GETS SMARTER EVERY SINGLE DAY AND 10X HIS RESEARCH! Everyone hoards notes they never reopen, but he turned the dead pile into a brain that actually reads itself back, in three steps: > download Obsidian, make a vault, and point Claude Code at the folder > create two folders inside, raw-sources for everything you dump in and wiki where Claude writes the organized version, then drop in Andrej Karpathy's prompt > throw in every article, book note, meeting transcript and piece of research you ever saved and forgot Then you stop searching and start asking the wiki, "what are the gaps in my knowledge", "write me a briefing on this topic", and Claude pulls from your whole vault and answers with citations. Everyone copies the folders, but the part that actually matters is what happens next, every answer gets written back into the wiki and a weekly health check hunts for contradictions and outdated info, so it is not storage you query, it is a brain that compounds every day you use it. The full step by step guide with the Karpathy prompt and the .md files is in the article below.👇 P.S. you can send this article straight to your Claude to get the full context
0:56

THIS GUY BUILT A GOD-MODE SECOND BRAIN IN 3 STEPS THAT GETS SMARTER EVERY SINGLE DAY AND 10X HIS RESEARCH! Everyone hoards notes they never reopen, but he turned the dead pile into a brain that actually reads itself back, in three steps: > download Obsidian, make a vault, and point Claude Code at the folder > create two folders inside, raw-sources for everything you dump in and wiki where Claude writes the organized version, then drop in Andrej Karpathy's prompt > throw in every article, book note, meeting transcript and piece of research you ever saved and forgot Then you stop searching and start asking the wiki, "what are the gaps in my knowledge", "write me a briefing on this topic", and Claude pulls from your whole vault and answers with citations. Everyone copies the folders, but the part that actually matters is what happens next, every answer gets written back into the wiki and a weekly health check hunts for contradictions and outdated info, so it is not storage you query, it is a brain that compounds every day you use it. The full step by step guide with the Karpathy prompt and the .md files is in the article below.👇 P.S. you can send this article straight to your Claude to get the full context

Miraqle

15,128 views • 25 days ago

48 HOURS AFTER KARPATHY POSTED HIS LLM WIKI IDEA, A 26-YEAR-OLD BIRMINGHAM GRAD SHIPPED THE ONE COMMAND THAT MAKES IT WORK. IT NOW HAS 76,000 GITHUB STARS 1 command. 71.5x fewer tokens per query. 0 vector databases points graphify at any folder - codebase, docs, PDFs, screenshots, video. Tree-sitter parses the code, Claude reads the prose, the whole thing lands as a knowledge graph in graphify-out/ one flag - --obsidian - writes the entire graph as a fully-linked Obsidian vault: one markdown note per concept, every relationship a wikilink, every node linked back to its source. Drop the vault into Claude Code as a skill. Claude queries the graph instead of grepping through raw files, forever Safi Shamsi finished his MSc at Birmingham with Distinction in 2025. His thesis was a knowledge-graph RAG system for academic search. He shipped Graphify 48 hours after Karpathy's post, iterates every week, and has already been forked by Rootly AI Labs for incident data. Hacker News, Analytics Vidhya, Towards AI - all organic.76,000 stars. Three months old no neo4j server. no vector db. no embedding pipeline. no cloud. no monthly fee you're reading this on a device that could clone the repo, run one command, and have a Claude-native knowledge graph of your entire codebase in Obsidian before your next standup
16:36

48 HOURS AFTER KARPATHY POSTED HIS LLM WIKI IDEA, A 26-YEAR-OLD BIRMINGHAM GRAD SHIPPED THE ONE COMMAND THAT MAKES IT WORK. IT NOW HAS 76,000 GITHUB STARS 1 command. 71.5x fewer tokens per query. 0 vector databases points graphify at any folder - codebase, docs, PDFs, screenshots, video. Tree-sitter parses the code, Claude reads the prose, the whole thing lands as a knowledge graph in graphify-out/ one flag - --obsidian - writes the entire graph as a fully-linked Obsidian vault: one markdown note per concept, every relationship a wikilink, every node linked back to its source. Drop the vault into Claude Code as a skill. Claude queries the graph instead of grepping through raw files, forever Safi Shamsi finished his MSc at Birmingham with Distinction in 2025. His thesis was a knowledge-graph RAG system for academic search. He shipped Graphify 48 hours after Karpathy's post, iterates every week, and has already been forked by Rootly AI Labs for incident data. Hacker News, Analytics Vidhya, Towards AI - all organic.76,000 stars. Three months old no neo4j server. no vector db. no embedding pipeline. no cloud. no monthly fee you're reading this on a device that could clone the repo, run one command, and have a Claude-native knowledge graph of your entire codebase in Obsidian before your next standup

chewa.

297,082 views • 15 days ago

I genuinely don't understand why everyone isn't using this yet. Andrej Karpathy, OpenAI co-founder, posted a simple idea that went massively viral: Stop using AI to write code. Use it to build a second brain. You point Claude Code at a folder. Drop in any source: an article, a transcript, a PDF. Claude reads it, links it, files it into a living wiki of everything you know. It compounds like interest. The more you feed it, the smarter it gets. Here's the whole thing: 1) Install Obsidian 2) Create a vault 3) Open it in Claude Code 4) Paste Karpathy's wiki idea and tell Claude to build it 5) Claude makes three folders: - raw (for sources) - wiki (for its pages) - CLAUDE. md (that runs it) 6) Drop any source into raw and say: "ingest this" 7) Ask questions across everything, forever Five minutes to set up and you never start from a blank chat again. Full step by step guide below.
1:48

I genuinely don't understand why everyone isn't using this yet. Andrej Karpathy, OpenAI co-founder, posted a simple idea that went massively viral: Stop using AI to write code. Use it to build a second brain. You point Claude Code at a folder. Drop in any source: an article, a transcript, a PDF. Claude reads it, links it, files it into a living wiki of everything you know. It compounds like interest. The more you feed it, the smarter it gets. Here's the whole thing: 1) Install Obsidian 2) Create a vault 3) Open it in Claude Code 4) Paste Karpathy's wiki idea and tell Claude to build it 5) Claude makes three folders: - raw (for sources) - wiki (for its pages) - CLAUDE. md (that runs it) 6) Drop any source into raw and say: "ingest this" 7) Ask questions across everything, forever Five minutes to set up and you never start from a blank chat again. Full step by step guide below.

CyrilXBT

321,372 views • 1 month ago

Andrej Karpathy quietly shipped the best second brain idea in years not an app. a pattern. let an llm maintain a wiki of your notes. you dump sources, it reads them, links them, files them. knowledge compounds like interest. someone built it into a free claude code plugin. setup is two commands: claude plugin marketplace add AgriciDaniel/claude-obsidian claude plugin install claude-obsidian@agricidaniel-claude-obsidian then open obsidian, open claude code in the same folder, type /wiki. that's it. your notes are now queryable by claude and they get richer every time you read something. bookmark this. best thing you'll build this weekend.
0:42

Andrej Karpathy quietly shipped the best second brain idea in years not an app. a pattern. let an llm maintain a wiki of your notes. you dump sources, it reads them, links them, files them. knowledge compounds like interest. someone built it into a free claude code plugin. setup is two commands: claude plugin marketplace add AgriciDaniel/claude-obsidian claude plugin install claude-obsidian@agricidaniel-claude-obsidian then open obsidian, open claude code in the same folder, type /wiki. that's it. your notes are now queryable by claude and they get richer every time you read something. bookmark this. best thing you'll build this weekend.

Avid

336,969 views • 1 month ago