Claude Code + Obsidian is the most powerful AI combo I've ever used. I literally built an AI second brain that runs my entire life. Inspired by Andrej Karpathy's LLM Knowledge Wiki, this tool has been a complete game-changer. Here's EXACTLY how to build one for yourself:

This is quite fascinating and I'm on the same ship. But wondering at what point does "building a second brain" become collecting? I've seen people spend more time organizing their knowledge system than actually using it... :P

Claude Code + Obsidian works because retrieval + execution closes the loop; Karpathy's wiki idea gave the right backbone for that. The second-brain stack gets real once notes feed directly into agent tasks.

The real unlock here isn't Obsidian specifically. It's plain text files in a folder structure that AI can read natively. CLAUDE.md files as context boundaries, markdown as the universal format. No database, no API, no sync layer breaking. People keep building complex RAG pipelines when a well-organized folder does the job better.

cc + fzf is another underrated combo. cc runs all these bash commands but they vanish from your shell history. built a hook that captures every command and pipes them into fzf so you can actually search and reuse them later

building it is relatively easy keeping it useful is the hard part

the claude code + obsidian combo is wild. only downside is burning through tokens faster than you realize when you're deep in the zone. built to track usage across claude, cursor, copilot and 20+ providers right from the menu bar so you always know where you stand

This is exactly what I needed. I've been manually organizing my thoughts across 5 different tools like an animal.

Completely self-managed memory and note organisation. A literal second brain.

The interesting part isn’t even the second brain itself. It’s when the notes stop being passive storage and start becoming active context for actual workflows.

It's so powerful that Bro only has less than a hundred notes. lol You haven't even touched really building a second brain yet with MD files.Glad to see everyone hopping on the trend though.

the karpathy wiki idea scales surprisingly well once you add an agent that actually writes back to it after every session

the graph looks amazing, but how do you handle privacy with personal/sensitive notes? do you have a local LLM fallback, or does it all go to Anthropic?"

karpathy's knowledge wiki pattern plus claude code is what personal knowledge management was supposed to be this whole time. the missing piece was always the reasoning layer

«This thing runs my entire life!» *shows an Obsidian vault with 90 files*

we use a similar pattern but with CLAUDE.md as the persistent memory layer. the agent writes back learnings from every session into the rules file. after 30+ days it genuinely knows your codebase patterns, naming conventions, even your debugging style. the compound learning effect is real

The "second brain" concept only works if you actually trust the system enough to offload to it. Most people build the vault, populate it for a week, then go back to Google. The ones who make it stick are the ones who made retrieval faster than memory - not just storage. Karpathy's workflow does that. Most implementations don't

something on the same lines , just compatible with any llm

Você pode comentar casos de uso dessa combinação de ferramentas no seu dia a dia? Vi muitas pessoas publicando sobre esse duo, queria entender mais de como é a aplicabilidade.

rookie

we went from forgetting where we saved notes to having AI that remembers better than we do and somehow that feels more unsettling

obsidian as claude's memory layer is so underrated. been doing this half-assed without the karpathy structure. time to rebuild

@im_dimneo Just built this for my business, created a mobile app tonight with Claude api included so I have a mobile command center to pull live GA4 data, CPL, generate on-demand social, CRM access, etc…3 pronged logic depending on the task…game changer

how much would this Claude Code + Obsidian setup cost to setup including hardware? Would you recommend a stand alone, dedicated use workstation? Would your data be stored locally or uploaded into the cloud? Would uploading gmail history help personalize the “second brain”?

i’m doing a similar thing with claude + notion. works just as good

The "second brain" angle is compelling. The next step is extending this to professional contexts — an agent that doesn't just store your knowledge but actively works on your behalf: finding collaborators, reaching out, networking. The personal AI doing real work while you sleep.

CLAUDE.md as a knowledge base is all you need.

whole life run by AI sounds like a dystopian nightmare just waiting to happen

Built the same thing. No Obsidian. No Mac. Markdown files in Termux on a 300€ phone. Your second brain doesn’t need a subscription.

been doing something similar with a custom CLAUDE.md that teaches the agent my project context. once it has enough memory about how you think, it starts connecting dots you missed. the obsidian vault as a persistent memory layer is the move, gives it continuity across sessions that raw context windows can't

forecast right from Claude Cowork

I use this one

Completely self managed memory and note organisation is what you really need…

The real unlock is having capture and retrieval in the same system. Most people have a huge gap between consuming something and actually being able to find it later. When that gap closes, the compounding effect on your thinking is hard to overstate.