why a tiny rule, rooted down, closes into a cycle — and the real math under the toy
The whole engine is a single recurrence: start with two numbers, and forever after, each term is the sum of the two before it. From 0, 1 that's Fibonacci — 0, 1, 1, 2, 3, 5, 8, 13… — but any two seeds will do. The numbers race off to infinity, and the ratio of neighbours settles onto the golden ratio φ ≈ 1.618: an irrational constant emerging from nothing but integer addition. That's the first emergence, and it's real.
Now "root" each giant term down by taking it mod M — for M = 9 this is the classic digital root (sum the digits until one remains). The unbounded sequence collapses onto the small wheel of residues 0 … M−1. Connect each residue to the next and you draw a path on that wheel.
The numbers are infinite; their shadows on the wheel are not.
Here's the part that feels like magic but is a theorem. Mod M, only finitely many pairs of consecutive residues exist — at most M². Since each pair determines the next, the sequence must eventually revisit a pair, and from there it repeats exactly. That repeat length is the Pisano period, π(M). For Fibonacci mod 9 it is 24, and the sequence touches 0 every 12 steps — two beats of twelve inside one cycle of twenty-four. (The same 24 from the gravity dossier shows up here through a completely separate door; see the colophon for how much to read into that.)
The cycle a word produces depends only on its seed reduced mod M — the residue-pair (a mod M, b mod M). The raw size of the seed is irrelevant once you're on the wheel. Two consequences follow, and the tool turns both into features:
First, collisions: different words whose seeds reduce to the same residue-pair (or to another pair on the same orbit) close into the same cycle. Second, decoding is many-to-one: from a signature you can recover the cycle-class and the handful of residue-seeds on it, but never the unique word — infinitely many map there. Raise M and the M² pairs spread across more, longer orbits, so collisions thin out: the cipher's resolution rises with the modulus.
Read as a cipher: the base is the key, the closing cycle is the signature, and the modulus is a resolution dial. Encode a word, export its signature as a code, compare two words for a collision, or hunt the smallest modulus that finally separates them. It's a faithful, deterministic encoding — and, as the next note insists, a toy one.
The math is real. The recurrence, the golden-ratio limit, digital roots, and Pisano periods are textbook number theory; Fibonacci mod 9 genuinely has period 24 with zeros every 12.
The 24 is a real coincidence of structure, not a bridge. That this 24 matches the bosonic-string transverse count from the gravity thread is a true and pretty echo — both are real facts — but nothing here derives one from the other. Two different rooms, same house number. Enjoy it as resonance; don't build a causal claim on it.
Not cryptography. "Cypher" is the framing, not a security claim — this encoding is trivially many-to-one and easily inverted to its class. It's an instrument for seeing emergence and closure, not for keeping secrets.