I have been slowly pulling my life off Google.

Not all at once. Just piece by piece, every time I notice I am dependent on something I do not actually control.

Google Docs was one of the harder ones to let go of.

Not because I needed the storage or the syncing. I have Obsidian for that. I just needed one specific thing Google Docs does really well: the ability to leave comments on a piece of writing while you are revising it. Highlight a sentence, leave a note, keep moving. Come back to it later. That back-and-forth with yourself is a real part of the revision process, and I had not found anything in Obsidian that replicated it.

So my workflow was ugly.

Write in Obsidian. Hit a point where I needed to revise. Copy the whole thing, paste it into Google Docs, do my revision process there with comments and highlights, then figure out how to get it back. Every time.

That is two tools, two contexts, and a copy-paste step right in the middle of the work.

What I actually needed

I kept waiting for a plugin that solved this. Nothing quite did. So I built one.

Writing Annotations lets you highlight text inside a note and leave a comment directly on it. The annotation lives in the YAML front matter, which means when you are reading the note normally it renders clean. No clutter. But the data is structured and attached to the exact passage you flagged.

The revision process stays inside Obsidian. Google Docs stays out of it.

The part I did not expect to matter

That last detail, storing annotations in structured YAML, turned out to be more useful than I originally designed for.

If you are working with a local LLM, something like Qwen or DeepSeek running on your own machine, and you hand it a note that has embedded annotations, it can actually see what you want changed and where. You are not saying look at paragraph three and see if it flows. You are saying: here is the specific line, here is my concern about it, now respond to that.

The annotations become a precise interface between your thinking and the model output. Instead of sending a full draft and hoping the LLM guesses what you care about, you are targeting the exact areas you already know need work.

That is a different kind of collaboration than most people are doing with AI writing tools right now.

One less dependency

I am still improving it. There are things I want to add: better filtering on the export, a way to mark annotations resolved without losing them, a proper review mode. It is at version 1.2.6 and stable, but not finished.

If you use Obsidian and you write things you actually care about, it is free. I built it for my own workflow, but I figured someone else might find it useful too.

One less reason to open Google Docs. That is enough for me.