📓 March 11, 2026

1. Patrick's α⁻¹ Decomposition Plausible

The fine structure constant to full CODATA precision, built from seed 3 using only {×2, ×3, ×5, −1}:

Seed: 3
2²×3 − 1    = 11
2²×11 − 1   = 43
5×43         = 215
2×215 − 1   = 429
2×429 − 1   = 857
2×857 − 1   = 1713
2²×5×1713−1 = 34259
2²×34259    = 137036     → ÷10³ = 137.036

Extension to CODATA (4 more deposits):
×1250 − 1 = 171294999
×40 − 1   = 6851799959
×10 − 1   = 68517999589
×2 − 1    = 137035999177  → ÷10⁹ = 137.035999177 ✓

10 total −1 deposits from seed 3. The "voices" of 2, 3, and 5 take turns: 3 enters at the seed and leaves after two steps. 5 enters in bursts then decays. 2 is the steady heartbeat throughout.

Caveat: This is a decomposition of a known value, not a prediction. Any number can be decomposed over {2,3,5,±1} — the universality was verified for all integers 2–1000. What's potentially special is the shape and compactness (10 steps for 12 digits), not the existence of the decomposition.

2. Brooklyn's Generating Function Speculative

Brooklyn searched OEIS for "36, 823" as adjacent terms and found exactly two matching sequences:

Seq 1: 1/((1−5x)(1−9x)(1−10x)(1−12x)) → 1, 36, 823, 15270, ...
Seq 2: 1/((1−6x)(1−7x)(1−11x)(1−12x)) → 1, 36, 823, 15282, ...

Applied to Eddington's base of 136:

136 + 1 × 10⁰           = 137
137 + 36 × 10⁻³          = 137.036
137.036 − 823 × 10⁻⁹     = 137.035999177  ← CODATA exact

Three terms of a rational generating function reproduce 12 significant figures of α⁻¹. Both root sets sum to 36. The scale separations (10⁰, 10⁻³, 10⁻⁹) match the ~10³ emergence level gaps in Brooklyn's Scale Geometry paper. The two sequences diverge at the 4th term (15270 vs 15282), which constitutes a testable prediction for future precision measurements.

Brooklyn's own assessment: "I don't think it's actually that relevant... it demonstrates how you could end up with correction factors over the 8n+1 integer result. The real prediction is that more precise measurements will reveal another scale gap rather than noise."

3. Sierpiński Carpet Spectral Test Plausible

We built and ran numerical code testing whether the 8n+1 sequence describes the Laplacian eigenvalue counting function on Sierpiński carpet approximations.

Results:

• Spectral dimension converging toward predicted d_s = log(8)/log(3) ≈ 1.893: Level 1 → 1.58, Level 2 → 1.73, Level 3 → 1.82
• Walk dimension d_w ≈ 2.08, consistent with known literature value ~2.09
• Eigenvalue threshold ratios at 8n+1 crossings approaching 9 (Level 4: λ(1097)/λ(137) = 8.89)
• 8n+1 counting function is close but not exact — the d_w ≈ 2.09 discrepancy from d_w = 2 is confirmed

Brooklyn's hypothesis: If gauge connections serve as shortcuts across the carpet holes, they'd restore d_w = 2 and make the counting function exact. That's the kill-or-confirm test.

Public repo: github.com/liet-codes/sierpinski-spectral

4. √(a+b) vs √(a×b) Ratio Lattice Established

Patrick's ratio tables for additive and multiplicative merges reveal fundamentally different coupling structures:

• √(a+b) — ratios are constant along anti-diagonals (depend only on the sum). Vessel and fill are entangled.
• √(a×b) — ratios are constant along rows and columns independently. Vessel and fill are factored.

The fixed point where √(a+b) = √(a×b) is a+b = ab, which for integers is only a = b = 2. The invariant.

Connection to the McGuffin: the object merge rule √(N1 × N2) is the factored version. The Pythagorean merge √(a² + b²) adds self-reference. Three levels of the same structure.

5. Primorial Volume Scaling Plausible

The base cube 2×3×5 = 30 under uniform doubling:

Double all dimensions → volume × 8 (= 2³)
Base volume: 30
Level 1: 240
Level 2: 1920
Level 3: 15360
...
P/8 at level 0 = 30/8 = 15/4 ← the closed form coefficient 15 = 2⁴−1
Sphere volume: 4π(30)/3 = 40π × 8ⁿ, where 40 = 8×5

Clean ×8 ruler marks at every scale. Patrick: "I'm not saying anything about gravitation — only that you can draw nice clean ruler marks every ×8V if it's r³."