Author notes — full detail, auditor-facing
Patent 7 ("Hierarchical multi-scale cavity composition / onion architecture") claimed that N cavities arranged in a polyhedral spatial pattern with aperture coupling would form a Level-1 meta-cavity that itself resonates and computes. The geometric intuition was reasonable: small openings in PEC walls *do* allow some EM transmission, and if cavity-to-cavity coupling is strong enough, joint modes should emerge.
The claim was tested three times under increasingly favorable conditions, and rejected each time:
Test O (HPC-025 era, 2026): two octahedral cavities separated along x with an explicit PEC septum and a drilled circular aperture. Aperture radius swept from r=0 (closed control) to r=8 cells (large opening). Closed control verified septum isolation — cavity 2 received zero signal from cavity 1's drive. Open apertures (r=4, 6, 8) produced energy transmission ratios of ~10⁻⁶ to 10⁻⁵ — essentially noise floor. Coupled-mode prediction had hoped for ≥1% transmission. The result was 100,000× too weak for meaningful coupling.
Test D (cipher chain test, 2026): a chain of cavities tested for signal propagation along the chain. Result: 14 orders of magnitude signal drop between successive cavities. Chain configurations cannot propagate signal usefully via aperture geometry.
Test JJ Phase 1 (2026-05-11/12, current): the most rigorous test. Same octahedron pair, aperture coupling at the LARGEST setting (r=8 cells), with a full 5×5 sweep of decoherence teeter-totter AND frequency mismatch. The framework specifically anticipated that mismatched per-cell drive at certain (decoh, frequency) combinations might unlock coupling that uniform drive missed. Across all 25 sweep configurations: every single one produced cos_joint_vs_r1 ≤ 5×10⁻¹⁵ — bit-exact floating-point equality between the joint fingerprint and the singles. Zero emergence. Zero coupling. At ANY drive setting.
Three independent tests, three null results, three different geometries (pair / chain / pair-with-mismatch). The aperture-coupling mechanism doesn't work at this physical scale (~30-cell cavities, ~10⁻¹ wavelength apertures).
What this means for the framework:
- Patent 7's aperture-mediated meta-cavity claim is falsified at scales currently tested
- Multi-cavity coupling (Layer 2 of the circuit language, the joint 7-vector operator GWC is testing) requires a DIFFERENT mechanism — shared walls, near-field, or some structural connection beyond just a hole
- GWC Test JJ Phase 1 was specifically designed to extend the aperture test to mismatched drive (the user's hypothesis was that this might unlock coupling). It didn't. The mismatched-drive hypothesis is independently falsified for the aperture geometry — though it may still produce signal under shared-wall or near-field topologies, which are next-batch tests in the auto-chain
What this means for the broader posture: the framework is willing to publish negative results on its own predictions. Patent 7 was held in good faith; the data falsified it three times; the response is to abandon that specific mechanism, not to add correction terms that paper over the null. Same posture as the cipher-corrections finding.
Summary — reader-facing
Patent 7's hierarchical-aperture-cavity mechanism was tested three times across three years of progressively more rigorous experiments:
| Test | Setup | Result |
|---|---|---|
| Test O | Octahedron pair with septum-aperture, r=0..8 cells | Energy transmission 10⁻⁶ to 10⁻⁵ (100,000× too weak) |
| Test D | Cipher chain, multi-cavity sequence | 14 orders of magnitude signal drop |
| Test JJ Phase 1 | Octa pair, aperture r=8, full decoh×frequency mismatch sweep | All 25 configurations: bit-exact floating-point zero coupling |
Aperture geometry does not produce useful cavity-to-cavity coupling. The mechanism is falsified.
What this tells us: multi-cavity compute can't rely on small-hole coupling. Alternative mechanisms (shared walls, near-field) are being tested next under Test JJ topology_compare + full_phase1. The framework abandoned the aperture mechanism rather than patching it.
Status: Confirmed negative finding. Auto-chain continues to test shared-wall and near-field topologies; if those ALSO show null results, Layer 2 of the circuit language may not exist at the geometric scales we can currently simulate, and the framework would need to retreat to single-shape compute claims only.
See also: Cipher corrections hurt accuracy — same posture (publish failures, don't patch around them).