Think by hand, synthesize with AI

I worked in AI for a living. I also keep a fountain pen within arm's reach at all times, and most of my real thinking still happens on paper. People assume that's a contradiction. It isn't. It's my whole method. For two years, I wrote here about productivity for techies, and somewhere along the way, I went quiet. The honest reason is that I got tired of adding to the noise. The internet does not need another post about a to-do list app. What I kept coming back to in my own work was a quieter and more useful question: In a world where AI can generate infinite output, how do you keep your own thinking sharp? In this article, I share my answer. It's the system I actually use, and it's what this site is about now.

The trap everyone's walking into

The dominant advice right now is some version of put everything into the AI. Dump your notes into a model. Summarize the meeting. Generate the first draft. Let it remember for you. It feels like leverage. For a while it even is. But there's a cost that shows up slowly: when you outsource the act of thinking, you stop to build the thing that thinking creates: judgment, connections, a point of view that's yours. You end up faster and shallower. More output, but less of it genuinely yours. A hard drive full of AI summaries you never reread, ideas you never actually formed. The fix isn't to reject the tools. I'd be a hypocrite. The fix is to be deliberate about which part of the work is yours and which part you hand off.

Two jobs, two tools

Knowledge work is really two jobs, and they want different tools. The first job is thinking. Wrestling with an idea, noticing what connects, forming a view. This is slow, non-linear, and it benefits from friction. This is the part you should not hand to a machine, because the friction is where the value is. For me, this happens by hand. A notebook, a reMarkable, the slowness of writing a sentence out forces me to actually have the thought. The second job is processing. Summarizing, tagging, retrieving, connecting across hundreds of notes, resurfacing the thing you wrote six months ago. This is tedious, mechanical, and at scale, impossible for a human to do well. This is exactly what AI is good at. Most "AI productivity" setups let the machine do the thinking and you do the filing. That's backwards. A calm second brain flips it.

Capture by hand. Synthesize with AI. Keep what matters in a system you own and trust.

What the system actually looks like

Here's the shape of it. Not the apps (those change). The structure.

1. Capture deliberately. Quick capture for the fleeting stuff: a thought on a walk, a quote from a book, something you heard. For anything that might matter, I write it by hand. The friction is real. If an idea isn't worth the time to write out, it usually isn't worth keeping.

2. Funnel into one trusted home. Notes scattered across five apps means you trust none of them. Everything goes into one place (Obsidian, for me) that I own, that's plain text, that no company can turn off or paywall. One place you trust beats ten places you don't. The goal is an container for all your original rough notes and thoughts.

3. Let AI do the synthesis. This is where the model actually helps. Summarizing long captures. Suggesting tags as well as connections I'd miss manually. Answering questions across my whole vault. Finding old notes that connect to what I'm doing today. The AI isn't generating my ideas. It's surfacing the links I didn't see.

4. Review, so it stays alive. A weekly pass to prune, link, and resurface. A system without a review habit is just a graveyard for notes. Fifteen minutes a week is the difference between a second brain and a digital landfill.

Why "by hand" isn't nostalgia

I'm not romanticizing paper for its own sake. There's a practical reason. Writing by hand is slower. That's why it works. You can't type faster than you think, so you're forced to compress, to pick what matters, to actually form the thought instead of just letting it spill out. Try writing a difficult idea by hand, then typing the same idea. Notice which one forces you to think harder. Research backs this, but you'll feel it.The hand-writing step isn't anti-technology. It's a quality filter. Garbage in, garbage out applies to second brains too. Your handwriting is the gate. The AI then handles the tedious part: speed and scale. You get thinking that's actually deep because it was slow, plus the leverage of fast processing. Most people pick one and skip the other, and end up with either nothing useful or noise they never read.

Where to start this week

Don't build the whole system at once.

• Pick one surface for capture. A notebook, a tablet, anything you'll actually reach for. Use it for one week and see what happens.
• Pick one home. One app. Move your most important notes there. Stop spreading them across five places.
• Add one AI task. Mine to start was simple: "Summarize this note in three bullets. What does it connect to?" That's it.
• Review on Friday for fifteen minutes. Prune, link, pull up old notes that matter. Keep it short and regular.

That's a working second brain. Build on it from there.

What's next here

This is the foundation. From here I'll go more deeply into the specifics: my exact eink or handwriting-to-Obsidian workflow, the prompts I actually use, how to structure a vault you won't abandon, and how to use AI without letting it quietly take over your thinking. If that's useful to you, join the weekly note. One short essay a week on thinking clearly in the age of AI. No hype, no app reviews, the occasional ink drawing. And if you want to build your own system properly rather than piece it together from blog posts, that's what the Second Brain Sprint is for. I built it with you, one-on-one, in four weeks. The first few spots are open at founding rates. Think by hand. Let AI do the rest.

~ Regina