Building faster than you can think

Watch a painter work and you'll notice something obvious. They don't just paint. They step in, make a stroke, step back. Look at the whole canvas. Think. Step in again.

The doing and the thinking are both part of the work. Neither one is the interruption.

A software developer's loop is the same. Write code, step back, evaluate, think, write more code. It looks like a slow compile, a walk to get coffee, a bug you couldn't squash until you slept on it. Was it friction slowing you down, or was it an invisible pause that made the whole process work?

The doing is getting faster. Disorienting, can't-look-away faster. Bugs, features, whims, all gone before you've had time to think about them.

The thinking hasn't sped up. You still need the same time to figure out whether you're building the right thing. Whether the approach is sound. Whether what the agent just handed you is actually good or just looks good at first glance.

Judgment doesn't scale with tooling.

The doing was always the slow part. Now the doing is running ahead of the thinking. And when you can build faster than you can think, something breaks.

You build something that works. Tests pass. And you think: I could polish that last bit, or I could ship it and move on. Isn't that what being agile means? Ship the MVP, iterate later.

So you ship it. And maybe that's fine. But also maybe what you skipped was the moment where you would have realized the 20% that's wrong is the important 20%. That the approach feels incomplete and you're building the right feature for the wrong problem.

We can add an intentional pause by writing a spec before you turn the agents loose. Sit with the problem, figure out what the pieces and assumptions are. It sounds obvious but most people skip it because writing feels slower than building. It isn't. A spec is a forcing function for thinking. It's also what lets you evaluate the output against something concrete instead of vibing your way through a review.

Prototyping cheaply is another version. When you can look at multiple approaches to the same problem side by side, choosing between them forces you to articulate what you actually want. That's the pause, but it feels productive rather than like lost time.

And there's just the plain fact of limited energy. Three, maybe four hours of deep AI-assisted work before you're cooked. Treat those sessions as a limited resource, not an infinite one.

The painter doesn't step back because they ran out of paint. They step back because looking at the whole canvas is part of how good work gets made.

The pace of your tools doesn't have to set the pace of your thinking. The loop still needs both steps. You just have to protect the second one on purpose now.