When Two Designs Both Work, Let the Players Break the Tie

The studio's founding design fork — do you look at the room, or stand inside it — was never settled by argument. The same puzzle was built both ways, shown to people who play escape rooms, and their preference decided it. What made that a safe way to decide, rather than a gamble, is that the artifacts were disposable: being wrong would have cost almost nothing.

The same June the flat MVP settled the technical questions, two files sit in a storage bucket, dated the 4th — the same Cold War operation, the same counting-cards puzzle, built two ways. cube.html: you float outside a cube in the dark, turning it with a finger like a puzzle box; one face is the keypad. And room.html:

Same operation, same lock — but now the walls are around you and the door is something you turn to face.

A fork you can't reason your way out of

On paper both are defensible, and that is the problem. No argument settles "look at it" against "stand in it" — the difference is entirely in how it feels, which is the kind of question theory answers badly and a person answers in a second. So the studio stopped arguing and put both in front of a few friends who love escape rooms.

The verdict was consistent: inside. Standing in the room read as an experience; spinning the cube read as a toy. It also matched the shape of the thing it was imitating — a real escape room is a place you are inside of, not an object you hold in your hand. That pull is almost certainly why the same testers kept pushing toward what rooms eventually became: full spaces you act in, with fixtures and inventory.

The road not taken is still a road

Choosing room.html did not make cube.html wrong — it made it a different business. A cube you spin like a puzzle box wants to become a physical object: a white-label puzzle box a buyer customises with AI and receives in the mail, for themselves or as a gift. That is a real idea and still a live one. It is also a company with warehouses, shipping, and returns — far more logistics than this studio can carry while it is still hunting for product-market fit. The bucket keeps that road on file.

What this leaves behind

The tie-break itself is not the lesson. Five or ten friends is not serious validation, and the record shouldn't dress it up as one. The lesson is why the studio could act on a weak signal at all: the same cheap-artifact method from the first MVP made both versions quick to build and cheap to discard. When reverting a decision is a cheap move, a directional preference is enough to commit on — you don't need certainty, because nothing has been locked in.

Taste picks the direction; disposable artifacts make it safe to follow.