I don’t write blog posts. I write notes. The blog posts are just the notes that survived long enough to be worth sharing.
Here’s the workflow:
Capture
Everything starts in Obsidian. Daily notes, fleeting ideas, reading highlights, coaching session reflections. No structure, no pressure. Just capture.
I use a simple tagging convention: #seed for raw ideas, #growing for notes I’m actively developing, #ripe for things ready to publish.
Develop
When a #seed note keeps pulling me back, when I find myself adding to it across multiple days, I know it has legs. I’ll spend a focused session developing it. Usually this means:
- Write the core idea in one sentence
- Ask: what does someone need to believe for this to land?
- Write the supporting points
- Cut everything that doesn’t serve the core idea
Refine
This is where AI comes in. Not to write, but to pressure-test. I’ll feed a draft to Claude and ask:
- “What’s the weakest argument here?”
- “Where does this get vague?”
- “What would a skeptic push back on?”
The AI is a sparring partner, not a ghostwriter. The voice and thinking stay mine.
Publish
The final step is mechanical: export from Obsidian, add front matter, push to Jekyll. I built a small Python script that handles the conversion. The writing is the hard part. Publishing should be frictionless.
Why this works
Because it respects how ideas actually develop. They don’t arrive fully formed. They grow through accumulation and refinement. The workflow mirrors the thinking.