How to play Rule

Test anything. Type letters and send them at the boundary — real words, names, nonsense. Tests are unlimited and free. They are counted, because how few you needed is interesting, but you are never penalised for asking, and sending a word you have already tried costs nothing at all.

Then prove it. When you think you know, the game hands you six words it has already judged and asks you to say, for each one, whether it went through or was stopped. Get all six right and your theory holds. You are never asked to write the rule out — a sentence would have to be marked by something, and predictions mark themselves. Nothing is checked until you send all six, so you can move between them and change your mind as often as you like.

More than one theory may fit. Evidence this size usually has several readings, and we do not pretend ours is the only one. If your rule calls all six correctly, it survived the proof — that is the whole test, whatever words you would have used. The sentence at the end is labelled Our rule for that reason.

Three attempts. Each proof is six new words, and starting one spends an attempt — the game says how many remain before it takes any. When one comes back you are told what the boundary did with all six and which of your calls it disagreed with — not the reason, that waits until the end, but enough to know where your theory broke.

Your best one counts. Nothing takes it away. Five of six on the first attempt stays five of six however the next two go, so there is never a reason not to try again. When you have had enough, finish and the boundary opens.

Stuck? Show another contrast adds two more words, one through and one stopped, usually differing by very little. It is the only hint, you get one, and it does not cost you a proof. The one thing it does cost is an Axiom.

The boundary always opens. Win or lose, it shows you how it worked at the end. That is the point of the whole thing, and it is not withheld from anybody who got it wrong.