On showing up for AI

There's a pattern in every field that moves this fast: the people who end up good at it aren't the ones who read the most papers. They're the ones who kept showing up after the novelty wore off.

The first six months with a new tool feel like magic. Every prompt produces something vaguely impressive, and you tell yourself you're learning. Then the novelty dulls. You notice the failure modes. You catch yourself wondering whether any of this actually works.

That's the point where most people drift away. And that's the point where the real learning starts.

The discipline of returning

Showing up for AI means sitting with the same problem — the ragged prompt, the unreliable tool call, the eval score that won't budge — for the third, fifth, twentieth session. It means writing the same kind of summary a dozen times and watching your taste sharpen each round.

There is no substitute for repetition in a craft, and this is a craft.

What it looks like in practice

• Re-reading the trace when you think you're sure what's wrong.
• Rewriting the prompt after it "already works."
• Keeping a file of things that surprised you this week.
• Saying "I don't know" out loud before you reach for the API.

None of this is glamorous. All of it compounds.