Rule Out
An old sorter keeps some trays and returns the rest. Someone set it and left no note.
What it is
The machine has a setting. You build an arrangement of up to three coloured pieces, run it, and the machine either keeps it or returns it. Nothing else. No hints, no warmer-colder, no partial credit.
You get eight tests to work out what the setting is.
You never say the rule. You prove it.
When you think you have it, you press I've got it, and the machine asks you to demonstrate. Four arrangements in order: one it will keep, one it will return, one it will keep, one it will return.
No feedback until all four are in. You cannot feel your way through the proof the way you felt your way through the tests.
That is the whole design. Believing you know the rule and being able to produce a case it rejects are different things, and only the second one is knowledge.
Spending tests on being wrong
Eight is not many, and the instinct is to spend them confirming what you already suspect. Arrangements that pass feel like progress. They are usually the least informative thing you can do.
The tests that teach you something are the ones you expect to fail. That is uncomfortable, and it is the point of the title.
How to play
- Put a piece on the trayTap a piece
- Take one offTap it on the tray
- Test the arrangementRun it
- Start the proofI've got it
- Hand in each proof trayCall tray
- Sound on or offThe speaker, top right
It starts silent. The speaker turns the sound on, and the game remembers your choice next time.
On a desktop, click instead of tapping, or play from the keyboard. Tab moves between the piece grid and the buttons, the arrow keys move around the grid, Enter or Space puts a piece on the tray or presses a button, Backspace takes the last piece off, and M turns the sound on and off.
Nothing in the game is longer than a sentence: it is built to be read by people whose first language is not English.
What isn't settled yet
The proof works and the machine gives away nothing it shouldn't. What has not been tested is whether someone who has never met a puzzle like this understands, before their tests run out, that they are supposed to go looking for a no.
Build Log
-
Build 15
About one setting in thirty at depth 2 was solved before you touched it: its two free examples ruled out every other possibility, so your tests went on a question already answered. The examples are now redrawn until at least one other setting still fits them. Losing also drops you one depth instead of back to the first, so a long run no longer ends in a long climb through easy settings.
-
Build 14
The game could not be played from a keyboard. The pieces were not buttons, so a keyboard could run a tray but never load one. They are buttons now: Tab reaches them, the arrow keys move among them, Enter places one and Backspace takes the last one off. One family of settings went too. Its rules could differ from another setting only on one-piece trays, so a player holding the wrong one passed the proof about 39 times in 40.
-
Build 10
Gave the machine an owner. It read as cold and arbitrary until one line fixed it: someone set it, and left no note. That makes the two starting examples answers already on the board, and makes a limited number of tests a physical fact rather than a game being stingy.
-
Build 6
The same setting turned up twice running, past a guard that was working. It compared names. Two rules can carry different names and do the same thing: "every large piece is hollow" is "every solid piece is small". Generated content must be deduplicated by what it does, not by what it is called.
Credits and licences
No third-party engine and no embedded typefaces: it uses whichever fonts your device already has. Cover drawn from the game’s own piece code and colours.
Rule Out is original work by Launchway Games.