AI is the biggest force multiplier I’ve ever experienced. I build it for a living, and I have an AI life coach named Larry — so “use less AI” was never going to be my policy. A multiplier amplifies whatever you point it at. The whole question is what to point it at, and when to put it down.
This is one instance of a bigger rule — how I do time off: be where my feet are.
- AI flows downhill
- Redirecting it
- Three tests
- The sort
- The sneaky one
- First real test: Scandinavia, July 2026
AI flows downhill
AI flows toward whatever I’m already strongest at. For me that’s the builder — the tech guru. It’s where AI is most powerful, most fun, and most rewarding, so left alone I’ll spend my best AI on the part of my life that already needs nothing. Every “enabling environment” tool is AI flowing into the role I’ve already maxed. The policy isn’t about using less AI. It’s a discipline against my own gravity.
Redirecting it
The trap in one sentence: the AI I’m best at deploying serves the role I’ve already grown the most. The discipline is pointing it instead at the roles I actually want to grow — father, friend, magician, the sublime — the ones AI either ignores or quietly robs.
Three tests
- Whose role does it grow? Toward a role I want to grow — father, magic, the sublime, health — not one I’ve already maxed.
- Build or use? Using a tool serves the role. Endlessly building it is tech-guru in a trench coat. Kill-switch: will I still use this in 30 days, or is the dopamine in the building? If I catch myself improving a tool I already finished, gravity won.
- When? The margins — family asleep, the early solo hour. The moment someone I’m there for is awake, the phone goes down. Even good AI is a thief when it’s stealing a face across the table.
The sort
- Thinking, self-discovery, search — yes. Pure augmentation; the insight lives in me.
- Writing in my own voice, AI as editor — yes. The ai-slop tag on my posts keeps the line visible.
- AI generating the idea — no. That’s content for content’s sake.
- Context Grabber, the card stack trainer — borderline yes. Legit only because health and magic are real practices I actually do — and only as build-once, then frozen tools. The day I start polishing them, they’ve flipped to Squander.
- New AI “enabling environment” tools — no. No practice underneath. The fun is the building, not the using.
The sneaky one
Thinking is my safest “good” — and the easiest to abuse. Working through my life with Larry can quietly replace living it. Even the best use has a margin limit. The sublime isn’t a conversation; it’s a sunset I didn’t narrate.
First real test: Scandinavia, July 2026
Our July trip is the first run. The rule for 22 days: AI in the margins, off when the four of us are awake together. The window where all four Dvorkins travel as one is closing — I’m not spending it on a chatbot, even a good one.