W@Home an akatalêptos verification

Live Progress

-
Active Workers
-
Jobs Completed
-
Hits Found
-
Confirmed
-
Compute Hours
Eigenvalue Search (retired)
0%
0 jobs · 0h compute
Fractal Falsification
0%
0 jobs · 0h compute
Countdown Clock
0%
0 jobs · 0h compute
0%

Total progress across all work types.

Eigenvalue Search — retired August 2026

Swept lambda from 0.4 to 0.6 in 200,000 steps. Each step: build the W-operator on a Menger sponge contact graph, solve eigenvalues, compare ratios against known physical constants. Status: this queue is retired — ratio recovery at this spectral density was shown indistinguishable from null spectra (April 2026), so "hits" below are recorded candidates, not evidence. Data retained for the record.

Recent Ratio Candidates

Loading...

Leaderboard

#WorkerJobsHoursHitsWTrustStatus
Loading...

Fractal Falsification

If the Menger sponge's physical constant predictions are mere numerology, other fractals should produce similar matches. We test every threshold-based 3D fractal from base 3 to base 21 (odd bases only). Each is defined by removing subcubes with ≥P center coordinates from a b×b×b grid.

Base (b)Threshold (P)SΔKept (k)Removed (r)Constants Matched
Computing...

Menger Countdown Clock — retired August 2026

This panel presented finite-level spectral-dimension estimates (1.32, 1.88, 2.14) as a sequence converging toward the Hausdorff dimension, with each level refining constant predictions by 10×. Direct measurement contradicts the premise: the sponge's spectral dimension is ds ≈ 2.45, flat across levels L2–L5 (three independent methods), and the constant-prediction framing was withdrawn after null testing (April/August 2026). Retained below for the record.

Spectral Dimension Convergence
Fine Structure Residual (log scale)

Multiplicity Tower

Eigenvalue-2 multiplicity: m(n) = (18n + 153·4n + 1155) / 357

LevelMultiplicitydSGap RatioResidual vs dH
Loading...

Cloud Compute Results

High-precision eigenvalue extraction on dedicated GPU hardware (NVIDIA GH200 480GB). 94 physical constants matched at L5 (3.2M vertices) and L6 (64M vertices). L7 (1.28 billion vertices) is currently running.

-
L5 Constants
-
L6 Constants
-
L5→L6 Improvement
-
L7 Eigenvalues
-
Best Precision (ppm)

94 Constants — L5 vs L6 Precision

Sorted by L6 precision. Sub-ppb matches highlighted green. All 94 constants found at both levels.

ConstantTarget ValueL5 MatchL5 Error (ppm) L6 MatchL6 Error (ppm)Improvement
Loading...

L7 — Live Eigenvalues (1.28 Billion Vertices)

Extracted via thick-restart Lanczos with deflation on an NVIDIA GH200. GPU at 100%, 96.7GB VRAM. Every probe converges to delta=0.00. Updates live via the L7 bridge.

Loading...

What This Means

94 out of 94 physical constants emerge from the Menger sponge eigenvalue spectrum with zero free parameters. Precision improves geometrically with iteration depth: L5 median error ? ppm → L6 median error ? ppm. The convergence pattern is identical for known and unknown ratios — consistent with standing waves on a self-similar fractal structure.

L7 (1.28 billion vertices) is the largest Menger eigenvalue computation in history. Currently extracting interior eigenvalues via 50 sigma sweeps. Hardware: NVIDIA GH200 480GB, ~$40 total cost.