All operations

A1Z26

A is 1 and Z is 26, which is the whole of it.

The mapping

A is 1, Z is 26, and there is nothing else to the scheme. The stage above is set to decode, so the numbers in the input box come out as a sentence.

Letters inside a word are joined by a dash or by a space, whichever you pick, and one space divides the words. Pick the space and the two separators become the same character, so the word boundaries stop existing: hello world encodes to 8 5 12 12 15 23 15 18 12 4, which decodes back as the single word helloworld. Only the dash gets a sentence back intact.

Capitals are flattened on the way in. A character that is not a letter keeps its own value and still collects a separator on each side, so don't becomes 4-15-14-'-20 and the digits in 13 cats become 1-3, which is what an encoded A and C look like too.

Reading numbers back

Alphabet positions run to two digits, so a string of numbers with nothing between them has several readings. 1123 is AABC, ALC, KBC, AAW or KW, and nothing in the scheme picks between them. The separator carries all the reversibility.

Decoding splits on dashes and spaces and converts any group that reads as a whole number from 1 to 26. Anything outside that range is left exactly as it arrived, so a stray 42 comes back as 42 rather than being guessed at or quietly dropped.

Where it turns up

Puzzle hunts, escape rooms, capture-the-flag challenges, geocache listings and the back pages of children’s mystery books. It carries no key, which makes calling it a cipher generous. It is a notation, and all the protection it has comes from a reader failing to think of it.

Readers rarely fail, which is what earns it the first guess. When numbers arrive and not one of them exceeds 26, the ceiling is the tell, which is what the numbers above spell out. Values climbing into the thirties and topping out near 126 point at ASCII instead, where 65 is A and 97 is a. Past 255 you are probably looking at Unicode.

Code points carries the same idea across the whole of Unicode, and decimal does it byte by byte for arbitrary data. Bacon numbers the alphabet the same way, then writes each position in five binary symbols.