Five Petals

Imagine five clocks on a wall. One has D=2 hours, one has K=3, one has E=5, one has b=7, and one has L=11.

Each ticks at its own speed. They never interfere with each other. But together, they can uniquely name every moment in a cycle of 2310 hours — because knowing the time on all five clocks tells you exactly where you are.

This is the Chinese Remainder Theorem. It is 2000 years old. And it says something that should make you stop and feel it: the big ring Z/2310Z is identical to five tiny rings working in parallel. Not approximately. Identical. Every addition, every multiplication, every structure — decomposed into five independent conversations that never overhear each other.

In the True Form (Z/970200Z), the five clocks have D3=8, K2=9, E2=25, b2=49, and L=11 hours. Fatter channels. Deeper resolution. Same five primes.

Z/2 (D)
Z/3 (K)
Z/5 (E)
Z/7 (b)
Z/11 (L)

Five clocks. 2 + 3 + 5 + 7 + 11 = 28 positions. Product = 2310.

↓ scroll

Split and Reassemble

Any number between 0 and 2309 splits into five channel values. And any five channel values reassemble into exactly one number. Nothing lost. Nothing ambiguous. Try it — feel the bijectivity with your fingers.

CRT Splitter

Change the number. Or click a channel to cycle it. The other side updates instantly.

Try 1729 — Ramanujan's taxi number. It splits into (1, 2, 4, 6, 1). Change the D-channel from 1 to 0 and you get 1728 = 123. One click, one channel, one new number.

Operations split too

Add two numbers in the big ring? Just add their channels independently. Multiply? Same thing. The channels never talk to each other. This is what independence means in mathematics.

(a + b) mod N = channel-by-channel addition

Perfect Independence

Here is the deepest thing about CRT. If I tell you that a number is odd (D-channel = 1), does that tell you anything about its remainder mod 7? No. Not a single bit of information leaks between channels. Zero mutual information. Exactly.

Independence Matrix

Diagonal: each channel's entropy. Off-diagonal: mutual information between pairs. All zeros.

Over the full ring, every channel pair has exactly zero mutual information. This is not approximate. It is a theorem: the CRT guarantees that knowing one channel reveals nothing about any other.

Why this changes everything

A search over 2310 possibilities becomes five searches over 2+3+5+7+11 = 28 possibilities. That is an 82.5x speedup. In the True Form: 970,200 becomes 8+9+25+49+11 = 102 — a 9,512x speedup.

Training a neural network? The gradient for channel mod-D does not depend on channel mod-b. You get block-diagonal Jacobians: five independent gradient computations. Thin ring: 25,654x. True Form: 95.7 million times fewer multiply-accumulate operations. That is not an optimization. That is a revolution.

The Beating Heart

Watch five counters, each cycling at its own rate. D flips every step. K every 3. b every 7. L every 11. They all return to zero at the same moment only once every 2310 steps. Press play and watch it breathe.

Parallel CRT Counter
0

D flips every step. K every 3. E every 5. b every 7. L every 11. All five sync at 0 only at n=2310.

At n = 1155 (halfway), the channels read (1, 0, 0, 0, 0). Only D has flipped. Everything else sits at zero. 1155 is the protist — the first departure from ground, where only duality has acted.

At n = 210, the channels read (0, 0, 0, 0, 1). Four data channels at zero, the guardian at 1. This is the data ring boundary.

The Coupling Dual

Every element of Z/2310Z has a coupling: the number 2310 / gcd(n, 2310). Think of it as a measure of how strongly n connects to the rest of the ring — how many elements it can reach through multiplication.

Units have maximum coupling (2310 = bacteria level, fully connected). Zero has minimum coupling (1 = isolated). Everything between is a kingdom.

Coupling Explorer
n =

Try: 210 (coupling 11), 30 (coupling 77), 6 (coupling 385). Notice: c(c(n)) = ?

The Coupling Dual Theorem: for every coprime split m × n = 2310, the coupling of m is n and the coupling of n is m. They are partners. Each half knows the other as its coupling.

There are exactly 16 = 24 such coprime splits — one for each subset of the four data primes {D, K, E, b} (since L divides exactly one of the two halves).

The Vandermonde connection

The shadow chain {sigma, D, K, E} = {1, 2, 3, 5} has a Vandermonde determinant: the product of all pairwise differences. Compute it: (2−1)(3−1)(5−1)(3−2)(5−2)(5−3) = 1×2×4×1×3×2 = 48.

48 = phi(210) = the number of units in the data ring = SEES. The shadow chain's discriminant equals the data ring's freedom count. The spectral structure is the algebraic structure, viewed through a different lens. Isn't that remarkable?

Five Breakthroughs from Five Petals

Everything in this atlas flows from CRT independence. Five separate channels give five engineering advantages, and — this is the part that should astonish you — they stack multiplicatively:

1. CRT decomposition — five independent channels
2. Loop unrolling — 82.5x forward speedup
3. Block-diagonal backprop — 25,654x gradient savings
4. Rissanen redundancy — 20x convergence at byte scale
5. L=11 ECC — free error detection and correction

Could we use fewer petals? With three primes (Z/30Z), there is no room for error correction — and D has no voice (variance zero). Could we use more? With six (Z/30030Z), degree rises to L=11. Five is the sweet spot: enough for independence, protection, and tractable computation. The axiom stops where it should.

Why Petals?

Because the CRT channels radiate outward from a center, like petals of a rose. Each petal is independent. Each is complete in itself. But together they form something that neither could alone.

A rose does not negotiate between its petals. Each grows on its own. The flower is the product, not the sum. Exactly like Z/2 × Z/3 × Z/5 × Z/7 × Z/11. Five petals. One rose. And the rose is the universe's favorite way to carry information.

← Ch.1: Three Rings | Ch.3: The Carousel →
The Interactive Atlas · Z/970200Z · Chapter 2 of 14