Author notes — full detail, auditor-facing
This is a pre-registered prediction for the Test JJ Phase 1 joint-7-vector search. Pre-registration ensures the success criteria are locked before the experimental result is observed, eliminating the possibility of post-hoc threshold adjustment.
Background
The framework predicts that the seven foundational geometric expressions (CONCENTRATE, DISTRIBUTE, RADIATE, POLARIZE, TRIANGULATE, CUBIFY, FRUSTRATE) should *compose* when geometries couple — producing a joint spectral fingerprint distinct from the sum of two independent single-cavity fingerprints. Test JJ Phase 1 puts that prediction in front of three coupling topologies (aperture, shared_wall, near_field) and a decoh teeter-totter (0.10/0.50 → 0.30/0.30 → 0.50/0.10).
A prior attempt to detect this emergence in aperture-coupling alone was falsified. Phase 1 expands to three topologies so the question is settled across the coupling landscape.
The pre-registered claim
Joint emergence is detected when:
- The coupled-pair fingerprint has cos-distance ≥ 0.15 from the
- AND this cos-distance ≥ 0.15 holds at the *boundaries* of the
- AND this pattern holds for at least one of {aperture, shared_wall}
sum of single-cavity fingerprints (i.e., the coupled run is *materially* different from the linear superposition).
decoh teeter-totter (drive imbalance) but NOT at symmetric drive (0.30/0.30) — the non-monotonic signature.
topologies but NOT for the near_field control.
Quantitative thresholds
Confirmed if (all of): 1. cos-distance ≥ 0.15 at (0.10/0.50) AND/OR (0.50/0.10) in aperture or shared_wall topology. 2. cos-distance ≤ 0.08 at (0.30/0.30) symmetric drive (the "joint mode vanishes at balance" non-monotonic signature). 3. cos-distance ≤ 0.10 in the near_field control across all decoh ratios (the "no joint mode in uncoupled control" signature).
Partially confirmed if (any of):
- cos-distance 0.08–0.15 at boundary drives but non-monotonic
- Joint emergence shows in one of {aperture, shared_wall} but the
signature is partially visible.
other is null (suggests topology-dependent emergence — still a finding).
Falsified if (any of):
- cos-distance < 0.08 across all three topologies and all decoh
- cos-distance is monotonic in decoh ratio (joint behavior is
- Joint emergence appears in near_field control at similar
ratios (no joint emergence detected at this scale).
linear mixing, not non-monotonic emergence).
amplitude to aperture/shared_wall (suggests artifact, not real emergence).
Why the non-monotonic signature is the load-bearing criterion
A joint-7-vector that exists only at specific drive imbalances is the framework's *prediction*. Linear mixing of two cavity modes would produce a *monotonic* response to drive balance — peaking at one extreme or the other but smoothly varying. A real joint emergence should appear at imbalance, vanish at balance, and reappear at the opposite imbalance — the teeter-totter signature.
This is what distinguishes a genuine framework-confirming result from a mundane linear-mixing observation. If Phase 1 produces a *monotonic* response (even a large-amplitude one), the framework's specific claim is *not* confirmed; what's confirmed is linear coupling, which is uninteresting.
Why pre-registration is critical here
Aperture coupling Phase 0 was falsified at slightly different thresholds than initially considered. Test JJ Phase 1 needs the thresholds locked *before* the results land, so post-hoc claims of "oh, the emergence was at 0.12 cos-distance instead of 0.15, that still counts" can't be made.
The thresholds in this entry are the official ones. They will not be revised retroactively. The experimental result either lands in the confirmed window, the partially-confirmed window, the falsified window, or somewhere outside all three (ambiguous result).
Status of evaluation
- Phase 1 simulations are running on Hetzner. As of 2026-05-12,
- Preliminary indication: the cos-distance values are clustering
- This is suggestive of falsification but is not yet conclusive
- Pre-registered judgment will be filed when the full Phase 1 grid
~60% of the (topology × decoh) grid is complete.
below 0.10 across all completed configurations.
(need the full grid + statistical confidence intervals).
completes.
Summary — reader-facing
Pre-registered prediction for Test JJ Phase 1 joint-7-vector search, filed 2026-05-09.
The claim: the joint-7-vector emerges from coupled-cavity drive when (1) cos-distance ≥ 0.15 from linear superposition at drive imbalance, (2) the non-monotonic teeter-totter signature is visible (emergence vanishes at symmetric drive), and (3) the near_field control is null (cos-distance ≤ 0.10).
Confirmed window:
- cos-distance ≥ 0.15 at decoh (0.10/0.50) and/or (0.50/0.10)
- cos-distance ≤ 0.08 at decoh (0.30/0.30)
- cos-distance ≤ 0.10 in near_field control
Falsified window:
- cos-distance < 0.08 everywhere → no joint emergence
- Monotonic response → linear mixing only
- Joint signal in near_field → artifact
Why the non-monotonic signature matters: linear mixing produces a monotonic response; real joint emergence should appear at imbalance, vanish at balance, reappear at the opposite imbalance. This distinguishes the framework's specific claim from generic linear coupling.
Status: open. Phase 1 ~60% complete on Hetzner. Preliminary indication: cos-distance clustering below 0.10 (suggestive of falsification but not conclusive). Full grid result will fall in one of the four pre-registered windows.
Pre-registration locks thresholds so post-hoc adjustment is impossible. The result either lands where it lands.
---
Outcome (2026-05-14)
Honest framing. This pre-registration enumerated two outcomes along one axis — confirmed vs falsified for *joint 7-vector emergence at the cos-distance ≥ 0.15 / non-monotonic-teeter-totter signature*. (A third "partially confirmed" window was also defined, but that too sits along the *same* emergence axis.) The actual JJ phase-1 result landed in a third orthogonal outcome the prereg did not anticipate.
What the data showed.
1. Joint emergence: falsified at the predicted scales. The cos-distance metric never exceeded the 0.15 threshold at drive imbalance, never showed the non-monotonic teeter-totter signature, and the near_field control was not differentially null. Magnitude ratios sat at 1.0000 ± 0.001 across all three topologies and all decoh ratios — coupled cavities do not amplify, only redistribute. By the prereg's own thresholds, the joint-7-vector emergence claim is falsified for this scale and these topologies. (See § "Pre-reg scorecard preview" in tests/test_jj_prism_extended_2026-05-14.md.)
2. A separate, unpredicted finding emerged: a prism-class drive-invariant spectral lock. Across 681 non-smoke configs spanning prism_7fold, prism_11fold, and prism_13fold pairs in three coupling topologies and the full decoh teeter-totter, the joint-spectrum bandwidth was frequency-locked:
- prism_7fold: BW std/mean = 3.9 % (pooled, n=150)
- prism_11fold: BW std/mean = 4.9 % (pooled, n=150, NEW)
- prism_13fold: BW std/mean = 7.1 % (pooled, n=56, partial)
- octahedron×octahedron: BW std/mean = 111.6 %
- bicone×bicone: BW std/mean = 59.2 %
This lock was drive-invariant (held across the full decoh grid), topology-independent (held across aperture, near_field, shared_wall), and tightened with sample size rather than regressing toward the geometric-mean baseline. The single-cavity {7}-fold uniformity result from HPC-039 (2.7 % error) replicates at the coupled-pair scale and extends to {11}-fold and (partially) {13}-fold. See tests/test_jj_prism_extended_2026-05-14.md and tests/test_jj_comprehensive_2026-05-13.md.
Why this is a third outcome the prereg did not cover. The prereg asked one question ("does the joint 7-vector compose at coupling?") and defined success/failure windows for that one question. The actual finding is a *different* phenomenon — a prism-class spectral lock that is independent of drive balance and topology — that the prereg's metric (cos-distance from linear superposition) was not designed to detect. Calling the experiment "falsified" is technically correct for the registered claim, but omits the substantive empirical discovery.
Lessons learned for future preregs. The confirmed/partially- confirmed/falsified-along-one-axis structure is necessary but not sufficient. Future preregs in this project should:
1. Enumerate orthogonal outcomes, not just gradations along the primary axis. For coupled-cavity searches, "the prediction was wrong but a *different* signal appeared" is a recurring pattern (HPC-039 single-cavity {7}-fold lock was also a side discovery from a different primary question). 2. Pre-register the discovery channel, not just the metric. Specify in advance what other observables the run will be probed against (here: pooled BW std/mean per shape would have been a useful auxiliary metric to register up front, even though the primary metric was cos-distance). 3. Pre-commit to publishing the orthogonal finding when it appears, with the same rigor as the registered claim. The risk without this: the falsified primary result gets quietly archived while the real signal gets reported informally without the pre-registration discipline that makes the project's evidence chain auditable.
Status. This entry is now closed. The registered claim is falsified at this scale; the orthogonal prism-class lock finding is documented in the comprehensive and extended JJ result files above. Existing prereg windows and hypotheses above are left unchanged for the historical record — only this Outcome section is appended.