AI-Augmented Note-Taking

Most knowledge workers fail at note-taking systems.

We start with elaborate tagging schemes, rigid hierarchies, and tight heuristics that promise to make everything findable. Then life happens.

Notes pile up. Tags become inconsistent. The system becomes a graveyard of good intentions.

This is the state of most personal knowledge management systems.

I experienced this for years. I relied on my ability to maintain perfect note-taking discipline so I could find what I needed later. It worked, but at great expense to my time and energy to find the things I'd written down.

The Shift: From Perfect Capture to Imperfect Collection

Claude Code (and Amp Code most recently) changed my relationship with notes entirely.

Not because it made me more organized, but because it made organization matter less upfront.

My current process looks like this:

Daily capture: I take 20-30 notes per day myself (I use a global quick capture workflow on my Macbkook and a global rapid logging workflow on my iPhone. Claude takes interstitial notes about my tasks automatically. The bar for "worth noting" dropped significantly because I'm no longer the only one who needs to make sense of it later.

Theme notes over tags: Instead of tags, I maintain theme notes (actual documents where I collect thoughts about specific topics). This lets me take notes about the themes themselves (and with Obsidian I use dataview to visualize connective tissue in the theme note, creating a meta-layer of sense-making.

Strategic synthesis: Claude does strategic sense-making based on my overall note structure and themes. It notices patterns I miss because I'm too close to the work.

Temporal reviews: Weekly and monthly roundups where Claude surfaces what it notices. These become springboards for new evergreen notes where I develop more polished thinking.

What I've Learned

This approach has taught me several lessons about how knowledge compounds. The insights emerged gradually as I stopped fighting the chaos and started working with it.

Writing It Down Creates Competitive Advantage

The most consistent feedback I get from others is about my ability to notice connections and see things from different angles. I don't think this happens if everything lives in my head.

Writing forces specificity. Having it written encourages nurturing and refining ideas. AI-assisted review forces synthesis. Together, they create a flywheel where quantity of notes leads to quality of insights.

The AI Doesn't Replace Thinking. It Scaffolds It

Claude isn't writing my thoughts for me. It's helping me structure the chaos I create through constant note-taking. Think of it like having an assistant who can:

• Read all your notes (something you'll never have time to do)
• Notice when three notes from different weeks are actually about the same thing
• Ask you questions about patterns it sees
• Draft synthesis that you can refine

The thinking is still mine. The scaffolding is collaborative.

This distinction matters because it fundamentally changed how I approach capturing information in the first place.

Lower the Bar for Capture, Raise the Bar for Synthesis

Traditional advice says "only write down what's important." This creates decision fatigue and leads to under-capturing.

My new philosophy: capture liberally, synthesize strategically. Claude helps with the second part, which makes the first part sustainable.

If I notice something about how my team communicates, I write it down. If I read something interesting about developer experience, I write it down. If I have a half-formed thought about content strategy, I write it down. I use wikilink references in my interstitial notes liberally.

Later (sometimes weeks later) Claude helps me see how these connect.

Theme Notes Beat Tags Every Time

Tags are metadata. Theme notes are content.

When I have a theme note for "Developer Experience," I can:

• Add observations over time
• Write about the theme itself, not just tag things with it
• See the evolution of my thinking
• Let Claude help me organize observations within the theme

This creates a living document instead of a static label.

The Weekly and Monthly Review Isn't Optional

Without the temporal reviews, notes just pile up. Claude's monthly and weekly roundups force a rhythm:

Weekly: What patterns emerged this week? What deserves to move from interstitial to evergreen? These often end up informing my weekly newsletter

Monthly: What themes are strengthening? What connections are appearing across weeks? What's ready to become an article or strategy doc?

This rhythm keeps notes active instead of letting them become write-only memory.

The Real Work Is Showing Up

Yes, this is a lot of work. Taking 20-30 notes per day is work. Reviewing Claude's roundups is work. Turning synthesis into evergreen notes is work.

But for me, it's the right work. The alternative (letting everything live in my head or in scattered docs) means doing the synthesis work under pressure when I actually need the insight.

This approach distributes the cognitive load across time and between human and AI.

Why This Matters Now

We're in a moment where tools like Claude Code, Amp Code (pick your local AI coding tool of choice) can be used on your local file system and read your entire corpus of notes, understand context across documents, and help you see patterns. This wasn't possible two years ago.

The question isn't whether AI-augmented note-taking is worth it. The question is whether consistently noticing connections in your field is worth it.

For me, that's been the superpower: seeing what others miss because I've been paying attention, writing it down, and letting AI help me connect the dots.

Getting Started

If you're curious about this approach:

1. Start capturing more, not less. Lower your bar for what's worth noting. (I prefer Obsidian for this)
2. Try theme notes instead of tags. Pick three topics you care about and create actual documents for them.
3. Ask your AI to do weekly reviews. Share your week's notes and ask what patterns it sees. (I prefer Amp Code for this lately, but Claude Code works just fine for it)
4. Move good synthesis to evergreen notes. Don't let insights live only in chat transcripts.

The system doesn't need to be perfect. It just needs to be used.

Changelog

• [2025-10-21] Initial draft created from Twitter thread discussion