You nailed “time under tension.”

I’ve started a simple rule for myself. For any complicated question, I take the time to outline my thoughts, evidence, counters, etc. before I open an LLM. It forces the reps that make thinking stronger rather than outsourcing that lift. Then I use AI to critique my draft, not create it. That small ordering change keeps writing-as-thinking intact and makes the machine a sparring partner, not a crutch.

I am writing a book about this very topic — The Journey to Phronesis (publishing April 2026) — and is full of practical ideas like this. I shared the initial Table of Contents from my manuscript in a recent post: lukich.io/p/202509-the-…