Decisions

Three directions, settled

Every batch that ended with a question left both answers standing in the lab rather than picking one. These three were answered together, on one screen — which mattered, because two of them are the same question at different distances: how much a surface should say.

The losing direction is not deleted. It stays built, and it stays here, because a decision you cannot see the alternative to is not a decision anyone can revisit — the same reason eight button directions are still live at /lab/buttons.

D-04

Telemetry readout

Is a live distance a number, or a shape?

The device dock now opens on the measurement, above the board's static particulars. `Last serial output` stays below it — after a run that line reads `Barrier: closed`, not a distance.

A · The number, largeChosen

Distance

18cm

Borrows A-15's big mono verbatim rather than inventing a size. The cost is honest: at this weight a continuous reading starts to look like a gauge, and this is not an instrument panel.

B · The reading, in the dock's rhythm

Distance18 cm

The same shape as a key-value row, so it belongs to the Device tab's family instead of arriving as a widget. The trace carries no fact the number does not, which is what lets it be hidden from assistive tech without losing anything.

P-01

Project card

How much can a card say?

S-02 and S-03 are both grids of these, and both get the full inventory. The card answers what is in the box.

The full comparisonRead the grids, not the cards. Every card looks fine on its own.
W-06

Mock camera frame

How much should the camera look like a camera?

The inspection frame carries no camera furniture. `Demo vision result` and the capture time sit under the image, in the interface's own voice.

A · The capture

HC-SR04VCCTRIGECHOGNDSIG5VGNDOpenClosedEchoD6D7
Demo vision result14:32

The frame is a photograph. Viewfinder corners, the label burned into the top left, the time in its corner, a scan you feel rather than see. It says this is what the camera saw, and the vision result is drawn on top of an image. The cost is exactly the product's hardest promise: the more it reads as a real frame, the more work Demo vision result has to do on its own.

B · The plateChosen

HC-SR04VCCTRIGECHOGNDSIG5VGNDOpenClosedEchoD6D7
Demo vision resultCaptured14:32

The frame is evidence, not a photograph. Nothing is printed on the image; the label and the time move underneath it, into the interface's own voice. What keeps it from being a second canvas is that it is fixed — no controls, no view switch — and that it carries annotations the canvas never draws. The cost is the other one: if it is only a still canvas with boxes on it, the modal has to justify itself some other way.