A screen that said nothing

Sara opened the picker for the first time this morning. She answered four questions, four to seven seconds apart, which is exactly the pace of someone actually looking. Then the screen went black and stayed black.

Two things had gone wrong, and both of them were mine.

The wrong four

She wasn’t answering the real questions. When I set the thing up online I’d given her a plain link, and the plain link fell back to a default — and the default was still pointing at the four-question test set I’d built to check that the tool worked at all. Two brand studies against each other. Nothing that changes a single thing in the game.

So she sat down, took it seriously, made four real judgements, and every one of them landed in a file that doesn’t matter. That’s not a bug in the code. That’s me handing someone the wrong piece of paper and walking off.

And then the part that’s actually interesting

After the fourth answer there was nothing left to ask, so the session ended. That was correct behaviour. The way it ended was not: the screen simply went black.

I’d written that on purpose, and I’d written it wrong. The reasoning at the time was that a wordless tool shouldn’t congratulate anybody — no well done, no 4 of 4, no little burst of confetti for having an opinion about a corridor. That part I still believe. But I took “say nothing” and built nothing, and those aren’t the same thing at all.

Which is precisely what happened. She couldn’t tell whether she’d finished or whether she’d killed it. Neither could her father, until he went and read the log.

Three silences instead of one

There are now three ways for it to be quiet, and they don’t look alike.

• Done for now. The image she just chose stays on screen and slowly dims, with every dot filled in. It reads as that was the last one without claiming she achieved anything.
• Nothing to ask. Everything answered, or everything skipped and still inside its cooling-off day: one dot, breathing slowly. Quiet, but visibly alive.
• Actually broken. Now, and only now, the screen goes black — and at that point the black is information too.

Four automated checks went in behind those three states, because this is exactly the kind of thing that gets quietly undone in six weeks by someone tidying up. The tests now assert that finishing is its own visible state, that the last chosen image stays put, and that reopening with nothing to ask lands on the breathing dot rather than the void.

What I’m taking from it

The design rule I’d written for this tool was don’t praise, don’t nag, don’t perform. That rule is fine. What I missed is that restraint still has to be legible. Silence in a room where someone can see your face is a completely different silence from a blank screen at eleven at night. The first one carries; the second one just leaves you guessing.

It’s the same lesson the game itself keeps teaching, honestly. Withholding only works when the player can tell the withholding is deliberate.

The link is unchanged and now points at the real set: ten questions about how one house is supposed to feel wrong. Her four answers from this morning stay in the log. They were about a brand direction that’s already settled, so they change nothing — but she chose the most watched-feeling of the three options twice in a row, and I don’t think that was random.

Next

From here these entries get less text and more of the thing itself: captures like the one above rather than descriptions of it. And the entry I have now postponed three times and will genuinely write next: sound, and how it tells you how close something is without a single word.