How I Turned Obsidian into an AI-Powered Knowledge System

Part 1: How I Got Here

This is Part 1 of a 6-part series on building an AI-powered knowledge system with Obsidian and Claude Code.

  1. How I Got Here – you are here
  2. The Architecture That Fixed It
  3. Teaching an AI Agent Your Note-Taking Rules
  4. Building Workflows That Run Themselves
  5. Searching 3,300 Notes in Seconds
  6. Connecting the Vault to the Outside World

I’m a network and security engineer. On any given week I’m in 15-20 meetings across multiple clients, juggling firewall migrations, SD-WAN deployments, FedRAMP authorization tracks (let’s not talk about FedRAMP), Prisma Access management, and vendor escalations. Every one of those threads generates notes, decisions, action items, and reference material that I need to find again later.

For years, “later” meant “never.” I’d write something down in Apple Notes, or a black book, or even a sticky note during a call. I’d paste a config snippet into a text file on my desktop in a random folder and save a vendor doc to a Downloads folder I never opened again. The information existed somewhere on my machine – but it was scattered, unsearchable, and disconnected across a mess of a file system.

The breaking point came when I was working at a different company and thinking about a change. I’d been talking to my wife about it when I got a random call from a contact I hadn’t talked to in years. It was a short call – just “hey, what are you doing these days? Are you happy? Have you thought about a change? We could use some help with a project and thought about you.”

In less than a week I had turned in my notice and was working on my exit docs. That’s when it hit me: I have no system. I need to learn to build a better way to organize information, notes, everything. Somebody has got to have had this problem before and shared their thoughts and systems for handling it. I made a promise to myself that I was going to go back to the school of hard knocks and learn everything I could.

I started looking at different apps. I tried Evernote – it felt dated. I checked out Notion, watched YouTube videos, tried different templates, but it just didn’t stick to how I think or how I wanted it to work for me. Not that I knew exactly what I wanted at that time, but something felt off. And I was not – and am NOT – a fan of anything where you can’t import and export your data. I knew I wanted something I could use now and move later. Little did I know how important that would turn out to be.

After spending time with Notion, I was after something that offered more flexibility and more control over my notes. It was about this time that I found Obsidian and the power of plain text. Right around the same time, Tiago Forte’s Building a Second Brain was released, and it opened up a world of options and ideas. That book led me to others:

Now I had what I needed. I started building something different – something that was me, but built on methods proven by people far crazier, smarter, and busier than I’ll ever be. This blog series covers that journey.

Why Obsidian (For Now)

I want to be clear: this isn’t an Obsidian sales pitch. I use Obsidian because it stays out of my way. The day it doesn’t, I’ll move on – and because everything is plain text, that move takes a minuscule amount of time compared to the friction any app could throw at me.

Before settling here, I evaluated several options – Notion, Roam Research, Logseq – installed some, browsed the websites of others. None of them clicked.

What clicked was plain text.

The markdown files are the system. The app is just a convenience – it gives me beautiful rendering, quick linking between notes, and excellent syncing across devices. But if Obsidian disappeared tomorrow, every note I’ve written is still a .md file I can open in any text editor on any operating system. That’s the whole point.

Plain text is king because:

  • Portability. My notes aren’t locked in a database or a proprietary format. They’re files in folders. I can move them anywhere.
  • Durability. Plain text files are readable today and will be readable in 20 years. No app has to survive for my notes to survive.
  • Toolability. I can grep through my vault, git commit it, diff changes, and point an AI agent at it – because every tool ever written knows how to read text files.
  • Recoverability. Version control has saved me more than once. I’ve had sync conflicts where my phone and computer were editing different versions of the same note – git let me untangle the mess. Another time, a different application reached into my vault directory and deleted several folders. Because the vault was under version control, I recovered everything in minutes.

That last point became even more important once I started letting AI agents read and write my notes. When an automated process can modify your files, you need the safety net of version control. Plain text makes that trivial – git init and you’re covered.

I do really like markdown specifically – it’s readable as raw text, widely supported, and the syntax is simple enough that you stop noticing it. I haven’t spent serious time with alternatives like AsciiDoc or Org-mode. Markdown works, and the ecosystem around it (Obsidian, static site generators, documentation tools) is massive.

If something better comes along, I’ll look at it. The system is built to move.

What You’ll Build in This Series

By the end of these six posts, you’ll have:

  • A structured vault with PARA folders, folder-note dashboards, and YAML schemas that make every note queryable (Part 2)
  • A CLAUDE.md file that teaches an AI agent your vault’s conventions – naming rules, frontmatter schemas, linking style, and plugin config (Part 3)
  • Automated workflows for inbox processing, meeting transcript extraction, daily/weekly reviews, and git version control (Part 4)
  • Three layers of search – Bases queries for live dashboards, the Obsidian CLI for programmatic access, and QMD for semantic search across thousands of notes (Part 5)
  • External integrations that connect your vault to Slack, email, project boards, and document conversion tools (Part 6)

You don’t need to build all of it. Each part is designed to stand alone – if all you want is a well-structured Obsidian vault with PARA and Bases dashboards, Parts 1 and 2 get you there. If you want the full AI-powered system, follow the series through.

Next up: Part 2 - The Architecture That Fixed It – where we build the PARA folder structure, set up live dashboards, and turn notes into queryable records.