Ryven

I still don’t understand why people keep building a Claude second brain by hand when these 10 repos already built the whole system They turn your Claude sessions into a wiki, let Claude read and write your vault from your phone, and one of them comes straight from Obsidian’s own CEO 01 llm-wiki → your Claude Code and Codex sessions become a searchable wiki, with an MCP server to query it github.com/Pratiyush/llm-wik… 02 llm-wiki-claude-skill → one skill file teaches Claude to ingest sources, answer from them, lint the wiki and compile it into clean pages github.com/win4r/llm-wiki-cl… 03 obsidian-skills → Obsidian’s own CEO teaches Claude how Obsidian really works github.com/kepano/obsidian-s… 04 llm-wiki → a Claude Code plugin with init, ingest, compile, query and lint, where every change is committed to git github.com/ekadetov/llm-wiki 05 obsidian-dev-second-brain → your tickets, sprints and project docs live in the vault as notes, and Claude keeps them in sync with the code github.com/jesuslemusco/obsi… 06 obsidian-mcp-server → 14 tools to read, write, search and surgically edit notes, tags and frontmatter github.com/cyanheads/obsidia… 07 mcpvault → safe vault access with a read-only mode, and Obsidian doesn’t even need to be running github.com/bitbonsai/mcpvaul… 08 obsidian-claude-code-mcp → connect Claude Code to your open vault with one /ide command github.com/iansinnott/obsidi… 09 obsidian-web-mcp → reach your vault from your phone or any MCP client, with OAuth and a Cloudflare Tunnel github.com/jimprosser/obsidi… 10 obsidian-mcp → a git-backed vault with 18 tools, running locally or on AWS Lambda github.com/eddmann/obsidian-… Nobody is going to build your second brain for you, and every chat you close without one is knowledge thrown away. Bookmark this, pick one repo from 01 to 05 and ingest your first source tonight. A month from now you have a wiki, and everyone who skipped it is still opening a blank chat.

Andrej Karpathy co-founded OpenAI, and one post about one idea pulled 21M views: stop using AI only to write code, use it to build a second brain that never forgets what you feed it He built a wiki of about 100 articles and 400,000 words on a single topic, and the LLM wrote it, not him “The LLM is rediscovering knowledge from scratch on every question. There’s no accumulation.” I still don’t understand why people use AI every day and let every answer disappear the second they close the tab. His compounds. Yours resets 5 minutes to set up, no app to build, no database, no code to write: → brew install --cask obsidian, mkdir ~/vault, cd ~/vault → run your agent in that folder with CLAUDE.md as the rules, raw/ for sources and wiki/ for pages → drop an article, pdf or transcript into raw/ and tell it to ingest → it writes the page, links it to the ones already there and logs the change → from then on every question gets answered from your own sources, and the good answers get filed back as pages → one lint pass a week catches contradictions and orphan pages “Obsidian is the IDE; the LLM is the programmer; the wiki is the codebase.” A year from now one of you has a wiki that gets smarter every week. The other is opening a blank chat again. Karpathy already picked a side, and you can pick yours before you go to sleep The full setup is in the article below

This is f*cking brilliant. 13 free repos turn a folder of notes into a second brain an agent can actually use: it captures what you read, searches everything you saved and never starts from zero [01] capture obsidian-clipper: official Obsidian web clipper, saves any page as Markdown github.com/obsidianmd/obsidi… markitdown: turns PDF, Word, PowerPoint, Excel and audio into Markdown github.com/microsoft/markitd… yt-dlp: pulls subtitles from videos without downloading them github.com/yt-dlp/yt-dlp paperless-ngx: OCR and full-text search for every scan and PDF github.com/paperless-ngx/pap… [02] structure obsidian-dataview: turns your vault into a database you can query github.com/blacksmithgu/obsi… obsidian-git: version history and automatic backup for every note github.com/Vinzent03/obsidia… [03] search qmd: on-device search for your notes, combines keywords, vectors and re-ranking, has an MCP server github.com/tobi/qmd meilisearch: search as you type with typo tolerance and hybrid semantic mode github.com/meilisearch/meili… [04] agent memory letta: stateful agents that learn over time github.com/letta-ai/letta langmem: facts, behavior and past interactions for long-term agent memory github.com/langchain-ai/lang… MemOS: a memory layer for agents, stored as a graph you can inspect and edit github.com/MemTensor/MemOS [05] own it quartz: publishes your Markdown notes as a website github.com/jackyzha0/quartz syncthing: syncs your vault between devices, peer to peer and encrypted github.com/syncthing/syncthi… A pile of notes becomes a second brain only when an agent can search it and remember it. Every repo above is free, and most of it runs on your own machine.

This is f*cking gold. One repo from Obsidian’s CEO teaches agents how your vault works. 10 more give them memory, search and everything you save 01 obsidian-skills → teaches Claude Code, Codex and other agents to write notes, bases and canvases in your vault the way Obsidian expects github.com/kepano/obsidian-s… 02 obsidian-copilot → runs Claude Code, Codex and OpenCode inside your vault, local models included github.com/logancyang/obsidi… 03 supermemory → remembers facts about you, rewrites the ones you contradict and drops the ones that expire github.com/supermemoryai/sup… 04 khoj → ask your notes, PDFs and the web one question, on your own server github.com/khoj-ai/khoj 05 obsidian-smart-connections → finds the notes you forgot you wrote while you type, no API key needed github.com/brianpetro/obsidi… 06 karakeep → saves anything, tags it with AI and finds it later, on your own server github.com/karakeep-app/kara… 07 siyuan → you and your AI agents edit the same notes on your own server github.com/siyuan-note/siyua… 08 zk → a plain-text wiki you run from the terminal github.com/zk-org/zk 09 silverbullet → Markdown notes you can program with Lua github.com/silverbulletmd/si… 10 memos → quick notes on a timeline, found by search, tag or date github.com/usememos/memos 11 trilium → notes in a hierarchy big enough for a whole personal knowledge base github.com/TriliumNext/Trili…

it is getting f*cking dangerous. a 23-year-old junior engineer from China makes $10,000 a month with no team, no assistant and no agency. one folder pays him. the analyst teams he replaced need weeks for what he delivers in one run, and his clients never asked where the team went. he is not a genius. he built a vault that remembers everything he has ever read, catches him whenever he contradicts himself, and tells the AI who he is before he types a word. then he pointed 300 agents at it. each one writes, a cheaper model checks every claim against the source in 100 milliseconds, and nothing moves until it passes. 100 company profiles, three sweeps, nothing left to fix by the third. you spent years collecting knowledge and your AI still meets you as a stranger every single morning. you are not losing to someone smarter. you are losing to a 23-year-old whose AI remembers him, while yours has never met you.

This is getting f*cking scary. 15 GitHub repos now wrap REA in a desktop studio, a local-model plugin and a scanner that shows what it could pull out of your own app [01] wrappers and installs ERA: wraps REA with run records, reports and a desktop studio github.com/dav1dngy/era-agen… reverseport: turns findings into rebuild specs and ports between engines and languages github.com/laaielichter/reve… rea-mcp: REA inside LM Studio for local models github.com/Nowng/rea-mcp rea-flake: REA with Ghidra packaged for Nix github.com/Fractal-Tess/rea-… [02] check and defend reaproof: scans your own build for what REA could recover, like source maps and keys github.com/Junhan2/reaproof antirea: stop directives that tell agents not to run REA on your project github.com/MockirDev/antirea REA-demo: REA analyzing its own package, graded against the source github.com/az9713/REA-demo [03] worked examples OpenBSD_netcat_REA: a static analysis of netcat-openbsd, every claim linked to an evidence file github.com/NullLabTests/Open… hexgrab: a color picker written from REA’s study of PowerToys behavior, with no PowerToys code github.com/iammukeshm/hexgra… [04] reading rea-research: a deep write-up of REA’s architecture and trust model, in Chinese github.com/ringozzt/rea-rese… reatool: unofficial install guides and a comparison with GhidraMCP and ida-pro-mcp github.com/ZooHero500/reatoo… [05] maps Awesome-REA: a bilingual catalogue of REA tools, cases and papers github.com/reacher-z/Awesome… awesome-rea: every entry names its own acceptance test github.com/yibie/awesome-rea rea-case-rankings: cases from GitHub, HN, X and Reddit on a 1 to 5 score github.com/miles990/rea-case… rea-gems: an auto-updated gallery of REA cases github.com/sqwu/rea-gems how the loop runs: app -> REA -> findings -> spec -> rebuild -> check -> harden start with an app you own and watch it explain itself.

I still don’t understand why a second brain only stores notes when these 10 repos make it quiz you, retract old claims and run offline Your agent can critique your notes, pull in fresh meeting transcripts, and one repo compares 15 second-brain tools for you 01 llm-wiki → drop in a book, a PDF or a meeting transcript and Claude writes linked Obsidian pages with citations github.com/phyllissong136/ll… 02 claude-wiki-verbs → nine commands for your vault, including synthesize, critique, compare and eli5 github.com/daniel8824-del/cl… 03 obsidian-wiki-mcp → an MCP server for your vault with BM25 search and a diff on every write, and it never calls a model itself github.com/aud1ence/obsidian… 04 llm-wiki-md-cli → a CLI with supersede and retract for outdated claims, and a lint that can block a commit github.com/ankishraj/llm-wik… 05 WikiSpec → spec-driven changes in Claude: propose, plan, apply, validate, archive, and every archived change feeds the wiki github.com/christianb93/Wiki… 06 anansi → quizzes you on your own wiki and grades your answers with written feedback github.com/darthblanc/anansi 07 silt → pulls fresh material from Read.ai and Google Docs into raw/, and your agent merges it into the wiki (Mac) github.com/psuijk/silt 08 Personal-wiki-using-Gemma-4 → drop files in a folder and a local Gemma model on Ollama writes the wiki, with a chat UI on top github.com/aswin-pn/Personal… 09 karpathy-llm-wiki-boilerplate → a template repo with four skill files, a CLAUDE.md and a small example wiki already inside github.com/xiaoyuze88/karpat… 10 awesome-brains → 15 second-brain tools compared by cost, setup and self-hosting github.com/lakshyabuilds/awe…

I still don’t understand why everyone isn’t using this yet. Three folders cut my costs by $1,000. Karpathy described the system in April 2026, and his post passed 21 million views. It’s a second brain that knows everything about you. It’s not a plugin or an app. It’s three folders of plain text files that Claude reads and writes itself. raw/ holds your sources, read-only for Claude. wiki/ holds the pages it writes and keeps current. CLAUDE.md holds the rules. Add a source, ask against the wiki, run a check, and every new session starts where the last one stopped. Setup is ten minutes: mkdir -p brain/{raw,wiki,output,.brain} cd brain && git init claude Tell Claude to interview you and write its own rules file. It asks about your work, your goals, your strengths and weaknesses, and writes it all down. Then I tested it. I planted one false sentence in the wiki, and a 95-line script with no AI inside caught it in milliseconds, for zero tokens. The full guide has the folders, the prompts, the script, the hooks, the nightly run, and where the whole idea breaks.

I still don’t understand why people keep building a Claude second brain from scratch when these 10 repos already built the whole system They take in your sources, link them into a wiki, check every claim, and one of them runs fully on your own machine 01 obsidian-skills → Obsidian’s own CEO teaches Claude how Obsidian really works github.com/kepano/obsidian-s… 02 claude-obsidian → drop in any source and Claude reads it, links it and files it into one connected graph of plain Markdown you own github.com/AgriciDaniel/clau… 03 obsidian-second-brain → Claude Code remembers you between sessions, with 48 commands, all inside your Obsidian vault github.com/eugeniughelbur/ob… 04 karpathy-llm-wiki → one install line, then ingest, query with citations and lint, in Claude Code, Cursor, Codex or OpenCode github.com/Astro-Han/karpath… 05 second-brain → raw sources in, linked wiki pages out, browsed in Obsidian github.com/NicholasSpisak/se… 06 second-brain-os → the full kit: a 77-page guide, a starter vault, 18 skills, 72 commands and 6 subagents, four of them read-only github.com/undefined-ui/seco… 07 llm-wiki → two-layer memory: 10 to 20 rules Claude loads every session, 50 to 200 wiki pages pulled only when needed github.com/MehmetGoekce/llm-… 08 vouch → agents propose changes, you approve them, and a claim without a source is an error github.com/vouchdev/vouch 09 ai-second-brain → your ChatGPT and Claude history becomes a linked Obsidian vault, then a wiki github.com/charlie947/ai-sec… 10 LLM_Wiki_SecondBrain → the whole pipeline on a local model, no cloud API. 32 GB of RAM recommended github.com/DimitrisLianos/LL…

I still don't understand why everyone isn't using this yet. Five minutes of setup saved me thousands of dollars Andrej Karpathy, a co-founder of OpenAI, shared a simple idea that reached 21 million views: stop pointing AI only at code, point it at everything you know. Drop any article, transcript or PDF into a folder. Claude files it into a living wiki, a second brain linked to everything you already saved. Every new source makes the next answer sharper, and you stop pasting the same context into every chat. The whole setup: > Create an Obsidian vault and open it with Claude Code > Hand Claude Karpathy's wiki idea file and let it build the structure > raw holds your sources, wiki holds its pages, CLAUDE.md holds the rules > Drop a source into raw and type "ingest this" > Then ask anything, and it answers from all of it Set it up once, and Claude stops forgetting you between chats. The full step-by-step guide for Claude and Obsidian is in the article below. Bookmark this and set it up tonight