I have used Microsoft OneNote as my daily journal and scratch pad for almost twenty years. It has been a faithful companion. Access on all my devices. Tabs, sections, web clippings, ink when I wanted it, and a rhythm I did not have to think about.
I am now leaving it behind.
Not because OneNote failed me as a note-taking app. It did not.
I am leaving because I finally found something I like better for the way I work now: Obsidian, a robust notes app that stores everything as simple text Markdown files in folders. My Obsidian "vault" is bundled into a single AI-enabled workspace where my journal, my projects, and my agents all live in the same place.
In a nutshell, this is a shared cloud folder, mapped to a local drive, with an intentional folder structure, containing mostly simple text files. (More on this concept here: https://github.com/scottheffield/results-studio )
Obsidian is the robust editor and front end for my journal work. A tool like Cursor is the robust interface for my agentic work. Both tools sit on top of the same shared workspace.
The shift is less about nostalgia and more about computing fundamentals. When your knowledge is plain text in a shared workspace, AI tools like Claude Cowork, Cursor, and Codex can engage with almost zero friction. When your knowledge lives in a proprietary format at scale, your options are limited and AI friction is high.
When I started using a shared workspace with AI agents I was running a split brain.
On one side: OneNote. Meeting notes, daily journal entries, quick ideas, the running log of how my weeks actually unfolded.
On the other side: a separate text-based workspace where I could use AI seriously. Client artifacts, frameworks, project folders, git-backed context files, the material I wanted agents to reason across.
I could use AI in the workspace. I could write in OneNote. What I could not do reliably was bridge the two.
My daily journal held the freshest signal: what I cared about today, what I promised someone, what I noticed in a call. My workspace held the durable structure: project artifacts, coaching session notes, blog drafts, research data, reference works, and archives.
The gap between them was not a discipline problem. It was an integration problem.
I tried to close it. The OneNote API was not robust enough for the sizable notebooks I had built over nearly two decades. At the scale I was operating, simple folder reads would fail, and retrieving specific documents wasn't reliable. I had high hopes trying an MCP connector someone had built (a standard way tools plug into AI apps), but it had the same problems because it used the same API under the hood.
I gave up on the idea of a robust OneNote to AI workspace integration. At least for now.
So AI helped me in the workspace, while my daily journal entries stayed elsewhere. Not where I wanted to be, but I accepted it.
And then I finally tried Obsidian.
Obsidian is not magic. What changes the game is that it is completely text based. No proprietary file formats. Just Markdown files in a normal folder tree, versioned in git, sitting next to everything else I care about.
An agent does not need a special connector to read a .md file. It does not need to query a clunky API to paginate through a notebook. It finds a file, reads the text, and just works.
My daily journal now lives under a "/Notebook" folder, in the same vault as my other workspace folders. When I write a - blog idea: or - LinkedIn idea: in today's daily note, AI driven scripts can sweep those lines into the right backlog. When I finish a coaching session, the notes I took in Obsidian can flow into the canonical session folder under Projects/Coaching. When I ask an agent to draft a post, it can read the journal entry where the idea started and the template in the blog pipeline and reference examples of how I have written before.
AI-native access, all in one place, without any fancy, but brittle connectors to bridge systems.
That is what I mean by zero friction.
The medium is in a format and structure that agents can read directly.
I want to be fair to OneNote, and honest about my own bar.
I had habits: daily pages, loose structure, quick capture, enough organization to find things without turning note-taking into a second job. I had developed some pretty specific ways of working in OneNote over the years. I wanted a journal that still felt like my journal.
Obsidian, with a small set of plugins and deliberate templates, got me there, and beyond. Daily notes land in dated folders. Coaching sessions, social media posts, and other repeated notes get consistent templates. The community is strong; the plugin ecosystem is mature. Lots of great YouTube content out there from passionate users.
Some of the capabilities are beyond where I had taken OneNote, especially for linking ideas across months and projects.
I do not feel like I am giving up a home. I feel like I moved into a cozy, and somewhat fancier house, where the doors open into every room.
The outcome I care about is not "I switched apps." It is one stitched-together world.
My work journal, client coaching artifacts, blog pipeline, thought-leadership drafts, and reference context can all sit in one repository that agents are instructed to read. Instructional files like AGENTS.md and CLAUDE.md in the key folders tell them how the vault works. Other files communicate urgency, reference, vision and purpose.
When AI is aware of both parts of my world, it can help me connect them. It can turn a bullet in today's note into a draft post. It can remind me what I committed to in a session last Tuesday while I am prepping for today's call. It can propose updates to a tracker because the transcript and the coach's notes are in paths it can actually see.
I am writing this post inside that very system.
The idea showed up in my daily note. The draft lives in Areas/blog-posts/drafts/. An agent read my instructions, my backlog, and my voice, and gave me a first version to react to. I used to have to connect all that manually, do a lot of copy and paste, then back and forth for ideation and editing.
Now it just flows.
I am only a few weeks into this migration, but I'm very happy so far.
All my old OneNote pages still exist, and I'll probably leave them there for now. I thought about doing a mass import, but I don't think that will add much value. I can always open up my old notebooks to find key artifacts from the past.
So I offer this question:
Where does your "thinking in progress" live relative to where your agents work?
Are they in different systems, or are they robustly connected? If not, then you will keep re-uploading context by hand and wondering why AI feels brilliant in demos and muddy on Tuesday afternoon.
I left OneNote because I got tired of the gap. Obsidian in a unified, git-backed, agent-aware workspace closed it.
If you are still straddling two worlds, I propose a very simple solution. Plain text on disk is probably the most boring answer I know. But the best solutions are often the most simple and straightforward. They are also the ones that stand the test of time.
If you are building an AI-enabled workspace for your daily work, I would like to hear how it's going. What lives outside your agents' reach today?