Tier 3: Major Experiments

LIGO O4/O5 + multi-messenger follow-up  •  3–5 years

Prediction: Gravitational wave arrival time correlates with dark matter column density along line of sight

Kill: No correlation at 10⁻¹⁶ level across 50+ multi-messenger events (archive registration; sample needed for a 3σ detection stated there as 20–50 events). This card read "20+" with no σ until 2026-09-19 — a weaker criterion than the one registered, undisclosed. Reachability: the best existing bound is 10⁻¹⁵, so the criterion cannot fire on any current instrument

Current verdict: Unreachable / monitoring-only. The kill criterion (a null at 10⁻¹⁶) sits an order of magnitude below the best bound ever achieved (GW170817, |Δv/v| ≲ 10⁻¹⁵), so it can never fire. The framework predicts no positive signal, so GW170817 was passed vacuously. See Falsifiability and Decisive Tests. (corrected 2026-09-14, explorer finding 2026-09-12)

Quantum optics lab with precision environment control  •  2–3 years

Prediction: Decoherence shows discrete steps at MRH boundaries, not continuous decay

Kill: Decoherence is perfectly smooth at all measured timescales

Current verdict: Unrunnable as stated. The MRH boundary is under-determined: where it falls depends on a choice the framework never states, so there is no boundary to look for steps at. See MRH. (corrected 2026-09-14, explorer finding 2026-09-12)

X-ray + optical survey telescopes  •  2–4 years

Prediction: Cluster-scale γ shows characteristic profile: γ ≈ 2 at outskirts, γ < 1 in ICM cores

Kill: No radial γ gradient in cluster profiles

Current verdict: Closed upstream. Reading a γ(r) profile off cluster data needs a bridge from C(ρ) to apparent mass. All four bridges tried on Coma failed (2026-05-28): two overshoot by 10⁴, one collapses to Newtonian, and one is capped at ≤2 against an observed 4.6. See Cluster Scale. (corrected 2026-09-14, explorer finding 2026-09-12)

Materials science lab + cryogenics  •  2 years

Prediction: New materials with optimized η (pair-breaking efficiency) show enhanced T_c

Kill: η optimization produces no T_c improvement over random search

Distinguishing power: MODERATE — tests whether η is useful as design parameter

Neuropixels + high-density EEG + fMRI  •  3–5 years

Prediction: Neural coherence shows scale-free structure with phase transition at C ≈ 0.50

Kill: Neural coherence shows no scale-free structure; threshold varies >50% across subjects

Current verdict: Unrunnable as stated. Predictions keyed to C ≈ 0.50 inherit the untestable-as-stated verdict, because no measurement maps to C. See Consciousness Threshold. (corrected 2026-09-14, explorer finding 2026-09-12)

Radio telescope time (resolved HI rotation curves in voids); proposed at $1M–$3M, 2–3 years  •  MOND+EFE branch measured 2021

Prediction: Void galaxies show higher DM fraction than cluster galaxies at same M_bar

Kill: DM fraction independent of cosmic environment

Current verdict: Self-eliminating-or-tie, and already executed by others. The two frameworks do not predict the same thing. On the SPARC mass models, at the environmental contrast SPARC actually spans, the density-keyed framework predicts a void-to-overdense change in DM fraction of Δf_DM ≈ 1.3×10⁻⁵ (low-g median, best case over knees and γ). MOND+EFE predicts 3.2×10⁻³, about 255× larger. Under the framework's acceleration keying C(a), the prediction is identically zero, since g_bar is held fixed. A 3σ stacked detection of the framework's signal would need ~1.4×10⁸ resolved HI rotation curves, against ~2.1×10³ for MOND+EFE. No outcome selects the framework, the same class as TEST-02. The measurement has also been made: Chae et al. 2021 (arXiv:2109.04745) located the SPARC galaxies in the cosmic web and found the EFE at >4σ. That detection is contested: Paranjape & Sheth 2022 (MNRAS 517, 130) show an EFE-like signal is generically expected in ΛCDM; Freundlich et al. 2022 (A&A 658, A26) find no EFE in Coma-cluster ultra-diffuse galaxies, a different sample; a 2025 re-analysis (Sargent et al., arXiv:2511.03839) calls it inconclusive. This is not a seventh refutation, and the count stays at 6. See TEST-05 for the lever-magnitude adjudication. (corrected 2026-09-14, explorer finding 2026-09-12)

Euclid mission data  •  3–5 years

Prediction: BAO peak shows fine structure (sub-peaks) from coherence interference

Kill: BAO peak is smooth Gaussian to 10⁻⁵ level in Euclid data

Distinguishing power: EXPLORATORY — no derived amplitude exists. Without a predicted sub-peak amplitude, any non-detection is consistent with "amplitude too small." Demote to exploratory until amplitude is derived. (Same standard applied to TEST-07 — 500 Mpc oscillation demoted for missing amplitude.)