Skip to content
AI-Native PM
Field Notes

Field Note

How We Used Preview Labs to Design This Site

Almost nothing here went straight to a single answer. Each piece started as a set of options in a preview lab, because when a tool hands you ten finished-looking versions in a minute, the work is no longer making one, it is choosing the right one.

· 5 min read

Almost nothing on this site was built by going straight to a single answer. The home hero, the About page, the value map, dark mode, the field-notes landing you are reading from, the small control that switches the theme: each one started as a set of options laid out in a preview lab, side by side on real data, so we could look at them together and choose. The making was fast and cheap. The choosing was the part that took the judgment.

Many options arrive cheap; choosing one is the workOn the left, four generated options stacked together, all plausible. Arrows lead to the right, where one is kept and a quiet branch reads, or none. Caption: the options were free; the choice was the work.EVERY FEATURE WENT THROUGH A PREVIEW LABGENERATED, ALL PLAUSIBLEOption AOption BOption COption DTHE ONE WE KEPT…or none of themThe options were free. The choice was the work.

That order, options first and a decision second, is the natural pattern of building with these tools, and it changes where the difficulty sits.

When a tool can hand you ten finished-looking options in a minute, building one of them stops being the hard part, and choosing the right one becomes the work.

Choosing well, it turned out, was rarely a matter of picking one option whole. More often the right answer was a mix: this layout, with the background from that one; this hero direction, with the title from another; this map treatment, made more striking the way a different option had been. Seeing that a piece of option B belonged inside option A is a finer judgment than ranking A above B, and it is the judgment that actually shaped the site. The labs did not hand us the answer. They gave us a clear field to make the answer from.

And sometimes the right choice was none of them. A whole round of options would go up, get a careful look, and lose, and the decision was to keep what we already had or to start over. Declining the entire set is a real choice, not a failure to choose, and treating "none of these" as available is part of why the labs worked. The ones we sent back are their own note, What We Built for This Site and Then Deleted.

This is the argument of Taste Is a Muscle, Not a Gift, lived across a real build rather than stated. The essay says the bottleneck has moved from making to choosing, and that taste is the muscle that does the choosing. Building option-first is what that looks like at the desk: generate freely, then spend the real effort on selection, combination, and the nerve to keep none of it when none of it is right.

Where this goes next

The full case for why choosing is now the scarce skill, and how to train the judgment that does it, is in Taste Is a Muscle, Not a Gift and our essay The Discipline. The companion habit, being willing to delete the options that lose even after you have built them, is What We Built for This Site and Then Deleted.

Sources