"What" is what matters

Programmers used to be valued for knowing how to build things. That’s fading fast.

AI agents can now do in minutes what used to take months. And this is AI at its worst. It only gets better from here.

The “how” — the syntax, the architecture, the implementation details — used to be the hard part. The expensive part. The part you hired for. Not anymore.

Some people think “why” is the safe ground. Understanding why something should exist, why it matters, why now. That lasts longer than “how.” But not as long as you’d think. Agents are getting better at reasoning too. They’ll connect dots. They’ll make cases. Give them enough context and they’ll argue for and against better than most people in the room.

What lasts is “what.”

What should we build? What problem is worth solving? What’s missing that people actually need?

No agent answers that for you. It’s not a question of logic or capability — it’s a question of taste, judgment, and lived experience. It’s noticing the gap before anyone describes it. It’s the instinct that says this matters before there’s data to prove it.

The “how” people built careers on execution. They were the hands. But hands are being automated. If all you know is how to build, you’re standing on a shrinking island.

The “what” people are standing somewhere else entirely. For them, agents aren’t a threat — they’re a lever. Every agent that gets better at execution makes the person who knows what to build more productive. Not less.

One group gets replaced. The other gets amplified. Same technology. Opposite outcomes.

Choose your side.