I've been sitting with something uncomfortable.

A few weeks ago, I finished a 1,200-word piece in under an hour. Solid structure, right angle. Nothing technically wrong with it. I closed the file feeling efficient.

Then I noticed I didn't feel anything else.

Not the particular tiredness that comes after real thinking. Not the small satisfaction that arrives when a paragraph finally sounds like yourself (aka dopamine). Just done. Like I'd processed something rather than made something.

I've been trying to figure out what happened since.

The AI didn't do anything wrong. The output was genuinely good. That's not the problem.

The problem is simpler: I skipped the rep.

Csikszentmihalyi found that flow requires a specific balance: the challenge has to match the skill. Too easy, and the brain disengages. Too hard, and it shuts down.

When AI takes the hard part, you don't enter flow. You exit the challenge-skill balance entirely.

Mueller and Oppenheimer's research on note-taking pointed at the same thing. Laptop notes produced more content than handwritten ones. But handwriting produced better learning — because the act of paraphrasing in real time forces synthesis. The constraint was the work.

AI gives us the laptop version of thinking. Fluent. Efficient. Complete.

And sometimes, not always, but sometimes — that's too easy.

There's a quiet trade happening.

Every time you reach for AI before you've wrestled with the problem yourself, you trade the rep for the output. The output arrives faster. The thinking that would have happened in the gap — the synthesis, the struggle, the moment where the idea becomes yours — doesn't.

You don't notice it in one session.

You notice it in a hundred.

The edge softens. The work feels faster, and something else feels further away.

That's not a reason to stop using AI. It's a reason to use it differently.

This week: The Pre-Prompt Journal.

Before your next significant AI-assisted task, open a blank page: a notebook, blank doc, anything without a cursor blinking at you.

Set 5 minutes.

Write three sentences:

Then prompt with your constraints as the brief, not the request. Include sentence 3 explicitly: tell the AI what it's not doing for you in this session. This is the design choice that separates a prompt from a brief.

When the output comes back, hold it against sentence 1. Did it extend your thinking, or assume it? If it assumed it — if the output started where you started — you handed over too much. Redirect. Go again.

This is iteration, not generation. The three sentences aren't warm-up. They're the spec. The gap between what you knew going in and what the output revealed is where judgment compounds.

The friction is the workout. And flow follows friction.

But for now: when was the last time you noticed the work felt too easy?

– Michael
Founder, The Drop In
& Author of 'Human Traits — a novel exploring humanity's relationship with AI'