In the 1950s, a German sociologist named Niklas Luhmann started writing ideas on index cards. One idea per card. Each one numbered, tagged, cross-referenced. Over four decades he pumped out 50+ books and 550 articles from this system. He called it a Zettelkasten — literally a slip-box. The man didn't just store information. He built something that thought alongside him.
I had no idea about Luhmann when I started using Obsidian. I just needed somewhere to dump things — meeting notes, project specs, random technical rabbit holes, daily journals. But after a while it hit me: the way Obsidian works — plain Markdown, linked together, sitting on your filesystem — is basically a digital Zettelkasten. And that turns out to have a very unexpected superpower in 2026.
AI agents can read it too.
The Second Brain That Actually Works
Most "second brain" setups die because they're too much work. You spend more time filing and tagging than actually thinking. Obsidian hits a sweet spot:
- One note, one idea. A presale process. A workshop outline. A debugging trick. Each gets its own .md file.
- Links over folders. Forget rigid hierarchies. You connect notes with [[wikilinks]]. A web of knowledge grows on its own.
- Local, plain-text. No vendor lock-in. No proprietary nonsense. Your knowledge is just files on disk.
I throw everything in there — daily journals, finances, project docs, technical research. It's messy and imperfect and that's fine. The point is: everything goes in, and the links make it findable later.
Obvious, But Almost Nobody Does It
Here's what bugs me. The exact properties that make Obsidian good for humans — Markdown, local filesystem, structured folders — make it a perfect knowledge base for AI agents.
When I work with Claude Code, it reads my Obsidian vault directly. No export step. No copy-paste marathon. No API integration to build. The agent just... reads the files. Because they're files.
I know almost no one who actually does this. Everyone's talking about AI-assisted workflows, but most developers still re-explain their entire context from scratch every single conversation. Meanwhile:
- Point the agent at a folder. Your presale notes, workshop materials, technical specs — they're already there, structured and linked.
- Build on past thinking. Instead of giving the agent a wall of context every time, it reads what you already wrote. Your notes become its context window.
- Human stays in the loop. You write notes for yourself. The AI benefits as a side effect. You never twist your notes to please the machine — the format just happens to work for both.
Why Markdown Is the Universal Interface
The tech industry keeps building elaborate knowledge platforms with databases, custom APIs, proprietary formats. Meanwhile, a folder of .md files turns out to be the most interoperable format you can get:
- Obsidian reads it.
- AI agents read it.
- Git versions it.
- grep searches it.
- Any text editor opens it.
Luhmann figured out that the connections between notes matter more than the notes themselves. Markdown with wikilinks keeps exactly that — in a format both humans and machines handle without friction.
So What
You don't need a fancy tool. You need a habit: write things down, link them, store them in a format that'll outlast whatever app is trendy this year.
If you already use Obsidian or any Markdown-based system, you have something most people don't: a knowledge base that works for you and for AI agents at the same time. In a world where AI is becoming a real collaborator and not just a chatbot — that's not a nice-to-have. It compounds.