I Added Obsidian Last. That Was the Right Order.
I added Obsidian last. That was the right order.
Most people building a personal knowledge system do it backwards. They install Obsidian first, configure a vault, explore the plugins, build a folder structure for notes they have not written yet. Then they discover that the system needs content to be useful, and content is harder to generate than the system was to build. The tool just sits there.
I avoided this by accident, not by design. I built my system for months before Obsidian entered the picture. By the time I installed it, there was something real to point it at.
The sequence matters. You need critical mass first.
My system runs on a folder of plain text files. When I installed Obsidian, I pointed it at that folder as a vault. That was the entire setup. Obsidian read every file instantly, rendered all the markdown, and showed me my system in a visual interface I had never had before. The sidebar organized my directories. The graph view showed how the files connected. Search worked across everything at once.
Before Obsidian, I was navigating these files through Finder, clicking through subdirectories, reading raw markdown in a text editor. Obsidian did not change a single file or move anything. It just showed me what was already there. That was enough.
The Dataview plugin was the one addition worth making early. It lets you query your files like a database. I have a dashboard note that pulls my recent briefings, active projects, and writing drafts automatically, from files Claude Cowork writes every day. The dashboard updates itself. I did not build something static and then maintain it. I built one query and the content runs through it.
I did not install Obsidian on my phone. There is a mobile app. But on my phone, all I actually need is to read a couple of files occasionally, and the Dropbox app does that natively. Installing Obsidian on my phone would have meant configuring sync, paying for a subscription, and solving a problem I do not have.
Don’t add friction where none is needed just because the option is there.
I also tried the Tasks plugin, which manages checkboxes across your entire vault as a unified task list. I tested it, concluded that maintaining its syntax costs more than it returns, and moved on. When I want a sweep of everything open, I ask Claude Cowork. It reads the files, groups the items, and returns a clean list. No schema to maintain. No plugin to configure.
The Tasks plugin is a good tool. It is not necessary when you have an AI that can do the reading.
Right now I am using Obsidian for one thing: rendering the markdown files so I can actually see them. When I want to browse the system, I open Obsidian. When I want to know what is happening inside it, I ask Claude.
The tool renders. The AI thinks.
I used to use Roam Research for this kind of work. Roam is a serious tool, built for people who want to own the architecture of their own thinking. But Roam is cloud-based, which means an AI cannot read the vault before it acts. Everything I put into Roam stayed locked inside it. My markdown files are plain text in a folder on my computer. Any application can read them. Any session can pick them up and have full context before it responds.
I have not yet built out the backlinks and connections I used to maintain in Roam. That is the next layer. For now, the rendering alone was worth the five minutes the setup took.
If I were starting over, the sequence would be this.
Start with Claude Cowork and two files. A file that describes who you are. A file that tracks what you are working on. Let those generate the first outputs. Let the outputs accumulate. Once the folder has real content, markdown files that actually contain something, point Obsidian at it.
Obsidian without content is a good-looking empty folder. Pointed at a folder that has been generating content for months, it becomes a window into something real.
Build the system first. Install the viewer when there is something worth viewing.
