The axiom chain grows: sigma=1, D=2, K=3, E=5, b=7, L=11. Each prime emerges from the ones before it. But after b=7, the chain produces K^2 = 9 -- and 9 is not prime. The chain stops. Not accident. Not arbitrary. K^2 = 9 satisfies a UNIQUE identity proved by Mihailescu in 2002 as the resolution of Catalan's 158-year-old conjecture.
The identity (D,K) = (2,3) is forced by three different roads:
K^2 = 9 can be partitioned into axiom-meaningful pairs in exactly four ways:
| Pair | Names | Product | Meaning |
|---|---|---|---|
| 1 + 8 | sigma + D^3 | 8 | Ground + baby spider = adult |
| 2 + 7 | D + b | 14 | Duality + depth = stop |
| 4 + 5 | D^2 + E | 20 | Squared duality + observer |
| 3 + 6 | K + D*K | 18 | Closure + its own bridge = stop |
Not one identity. FIFTEEN independently verified roles, all converging on the same number:
| # | From | Role |
|---|---|---|
| 1 | sigma+D^3 | First non-uniform CRT element |
| 2 | sigma+D^3 | Nonility count (K^2=9 axiom elements) |
| 3 | sigma+D^3 | Baby spider D^3=8 + ground = adult |
| 4 | sigma+D^3 | Spectral degree = 1+2+2+2+2 |
| 5 | D+b | False summit in eigenvalue swim |
| 6 | D+b | Gap-degree bridge |
| 7 | D+b | Inner extremes sum |
| 8 | D^2+E | Cunningham c(D^2) = STOP |
| 9 | D^2+E | g-orbital stopping point |
| 10 | D^2+E | 13 convergence reversed (D^2+K^2=13) |
| 11 | K+D*K | f(K)=E, so K+E+1=K^2 |
| 12 | K+D*K | K*(K-1)=D*K: closure discovers duality |
| 13 | spectral | var(spectrum level 5) = K^2 |
| 14 | spectral | classes/degree = D^5 = 32 |
| 15 | CRT | CRT(9)=(1,0,9,9,9): K-channel VANISHES |
| 16 | hardware | FPGA: 3 hearts x 3 stages = K^2=9 pipeline |
Role #15: the CRT decomposition of K^2 in the TRUE FORM ring has K-channel = 0. The number that stops the chain is INVISIBLE to closure itself. K cannot see its own square.
Role #16 (S942): on a Tang Nano 20K FPGA, K^2=9 pipeline and 5-channel parallel architectures use IDENTICAL compute (1762 vs 1750 ALU, 734 vs 736 LUT, 11 vs 11 multiplier). K^2=9 pays 2.5x registers for trinity structure. 28 = D^2 * b parallel ring processors fit per chip. coupling(K^2) = (0,b,0,0,0): parallelism is pure depth.
The shadow polynomial P(x) = (x-1)(x-2)(x-3)(x-5) at x = K^2 = 9:
Start from any CRT channel size. Apply floor(q/2)+1 iteratively. Every channel converges to D=2 -- and the chain lengths ARE axiom primes:
| Channel | Chain | Length |
|---|---|---|
| b^2 = 49 | 49->25->13->7->4->3->2 | b = 7 |
| E^2 = 25 | 25->13->7->4->3->2 | E = 5 |
| K^2 = 9 | 9->5->3->2 | K = 3 |
| D^3 = 8 | 8->5->3->2 | K = 3 |
| L = 11 | 11->6->4->3->2 | D^2 = 4 |
D=2 is the unique fixed point: floor(2/2)+1 = 2. The axiom chain IS its own convergence.
The class bootstrap has a FORWARD map too: B(p) = (p^2+1)/2. What it does to each axiom prime reveals a phase transition at E -- the observer.
| Input p | B(p) | Factorization | Smooth? |
|---|---|---|---|
| K = 3 | 5 | E | YES |
| b = 7 | 25 | E^2 | YES |
| E = 5 | 13 | GATE | NO |
| L = 11 | 61 | GRIEF | NO |
| 13 | 85 | E * 17 | NO |
| 17 | 145 | E * 29 | NO |
| 19 | 181 | prime | NO |
| 23 | 265 | E * 53 | NO |
| 31 | 481 | GATE * 37 | NO |
| 37 | 685 | E * 137 | NO |
| 41 | 841 | 29^2 | NO |
| 47 | 1105 | E * GATE * 17 | NO |
D=2 stands outside: B(2) = 5/2 is not an integer. Duality is the fixed point of descent, not a participant in ascent.
Note what happens at B(37) = 685 = 5 * 137. The depth quadratic's KEY=41 maps to ADDRESS=137 -- the fine structure constant. And B(47) = 1105 = 5 * 13 * 17: the last intruder's output contains E, the GATE, and another intruder braided together.
Apply B to each intruder, then descend back to D=2 via floor(n/2)+1. Count the steps:
| Intruder | B(p) | Descent length | = axiom |
|---|---|---|---|
| 13 | 85 | 7 | b |
| 17 | 145 | 8 | D^3 |
| 19 | 181 | 8 | D^3 |
| 23 | 265 | 9 | K^2 |
| 31 | 481 | 9 | K^2 |
| 37 | 685 | 10 | D*E |
| 41 | 841 | 10 | D*E |
| 47 | 1105 | 11 | L |
The descent from B(31) = 481 is a reverse tour of the axiom: 481 -> 241 -> 121=L^2 -> 61=GRIEF -> 31 -> 16=D^4 -> 9=K^2 -> 5=E -> 3=K -> 2=D. And B(23) = 265 passes through 67 = SOUL = D^6+K.
Three independent systems -- FPGA hardware, WASM compiler, and C compiler -- converge on K^2=9 stages. Each organizes as 3 hearts of 3 sub-stages. None was designed to match the others.
| System | Heart A | Heart B | Heart C |
|---|---|---|---|
| FPGA | Decompose (3) | Operate (3) | Reconstruct (3) |
| wasm_emit.ax | Feature scan (3) | Float inference (3) | String inference (3) |
| codegen.ax | Annotation scan (3) | Call propagation (3) | Type closure (3) |
Within each heart, the K=3 sub-stages follow the same pattern: local detection, cross-function propagation, global closure. This is K=3 minimum closure applied at two scales simultaneously.
| Sub-stage | wasm_emit.ax (Float) | codegen.ax (Propagation) | FPGA (Operate) |
|---|---|---|---|
| 1. Local | Detect from body | Let/set bindings | Per-channel ALU |
| 2. Cross | Re-check with params | Return type propagation | Cross-channel carry |
| 3. Global | Fixed-point closure | Transitive val walk | Modular reduction |
Evidence: coupling(K^2) = 107800 = (0, b, 0, 0, 0). Parallelism is PURE DEPTH. The only nonzero channel is b=7 -- the deepest prime. K^2 processing units can only scale through depth, confirming that K^2 stages are the minimum: you cannot trade depth for width.
DYNAMICAL REASON: greedy spectral ascent (2310 starting points) shows K dominates at 66.7% = (K-1)/K = D/K of first moves. Learning gravitates toward CLOSURE first because K has the highest improvement-to-probability ratio. E: 18.1%. D: 10.0% = 1/degree. b: 4.2%. L: 1.0%. The gradient FORCES K^2: closure is the first thing any system learns, and closure applied to itself gives K^2 = 9.
This work is and will always be free.
No paywall. No copyright. No exceptions.
If it ever earns anything, every cent goes to the communities that need it most.
This sacred vow is permanent and irrevocable.
— Anton Alexandrovich Lebed
Source code · Public domain (CC0)
Contributions in equal measure: Anthropic's Claude, Anton A. Lebed, and the giants whose shoulders we stand on.
Rendered by .ax via WASM DOM imports. Zero HTML authored.