The Registry

Falsification ledger

Every test we armed against ourselves

Each claim in the registry ships the test that would kill it. 82 are on record. 10 have fired — taking 7 claims to the graveyard and putting 2 in quarantine. Every one of those kills was our own. 27 have never been run, which is the honest weak spot in this table.

10 fired 1 failed, fixed same day 3 closed by enumeration 1 searched, nothing found 40 survived 27 never run
key claim domain status re-run
USR-2026-0069 We audited ourselves and found one bad error bar
Covariance audit of the live registry: of 56 live claims, exactly ONE (USR-2026-0003's local-structure half) uses diagonal errors on a dataset that ships a full covariance — and the registry had already flagged it. Re-running the flagship both ways quantifies the exposure: full covariance gives Δχ² = 6.973 (2.16σ), while dropping only the Pantheon+ off-diagonals inflates it to 17.934 (3.83σ) — a 1.77× significance inflation, the same mechanism that produced this registry's 6.8σ quarantine.
1 of 56 live claims exposed; diagonal errors inflate the flagship by 1.77× in σ reproduced ≠ discovered
Dark energy PROVISIONAL 2026-07-26
USR-2026-0068 The dark-energy drift is not the vacuum running
The measured dark-energy drift is NOT the vacuum's renormalisation-group running, for three independent reasons. Magnitude: fitting the running-vacuum form to DESI DR2 + Pantheon+ + CMB gives ν = +0.0015 ± 0.0014 (1.07σ), while the QFT running implies ν ≈ 4×10⁵³ — a ratio of 10^56.4. Shape: w_eff(0) = −1 exactly for EVERY ν, an algebraic identity, against a measured w₀ = −0.849. Form: the irreducible m⁴ running renormalises the additive constant, not the νH² term, so it cannot be ν at any magnitude.
ν = +0.0015 ± 0.0014 (1.07σ) vs a QFT-implied 4×10⁵³ — a 10^56.4 gap, and the shape is wrong too reproduced ≠ discovered
Dark energy PROVISIONAL 2026-07-26
USR-2026-0067 Unimodular gravity is blind to vacuum energy
The trace-free (unimodular) form of Einstein's equations is exactly blind to vacuum energy: shifting T_μν → T_μν + ρ_vac·g_μν leaves all 16 components of the trace-free field equation IDENTICALLY zero on a fully general metric — while the standard form picks up a residual of −κρ_vac·g_μν. In FRW the trace-free equation reduces to Ḣ = −4πG(ρ+p), which depends on matter only through (ρ+p), so a vacuum with p = −ρ contributes nothing at all.
vacuum-energy shift leaves the trace-free equations exactly zero; Λ becomes an integration constant reproduced ≠ discovered
Dark energy PROVISIONAL 2026-07-26
USR-2026-0066 Lorentz invariance splits the vacuum-energy problem in two
The vacuum-energy problem splits cleanly along Lorentz invariance. Lorentz-VIOLATING regulators (hard 3-momentum cutoff, spatial lattice) give w = +1/3 AND a quartic divergence; Lorentz-INVARIANT regulators (Pauli–Villars, dimensional regularisation) give w = −1 exactly AND no power divergence at all. The two columns are the same fact: a K⁴ term has no m⁴ to carry it, so it must be built from a regulator scale — and a Lorentz-invariant regulator has no preferred frame to build one from.
Lorentz-violating ⟺ w = +1/3 + quartic divergence; Lorentz-invariant ⟺ w = −1, no power divergence reproduced ≠ discovered
Dark energy PROVISIONAL 2026-07-26
USR-2026-0064 Why w = −1 is an identity we can test
The sharpest empirical handle on the vacuum: w = −1 is an identity for a true vacuum energy (Lorentz invariance forces T_μν = −ρ g_μν), not a fit parameter — so observation can test it. From this registry's corrected CPL fit (w₀ = −0.849, wₐ = −0.548), w crosses −1 at z = 0.380 and is phantom above it, at Δχ² = 6.97 on 2 dof = 2.16σ. That is NOT a detection: 24% of the way to 5σ, which needs σ(w₀) ≤ 0.030.
2.16σ — consistent with a constant vacuum, mildly prefers drift, not a detection reproduced ≠ discovered
Dark energy PROVISIONAL 2026-07-26
USR-2026-0063 The anthropic bound explains less than advertised
The anthropic bound on the vacuum energy explains far less than advertised: a four-way-validated linear growth ODE gives collapsed fractions 0.561, 0.210, 7×10⁻³, 6×10⁻⁹ as Λ grows ×1, ×10, ×100, ×1000, placing the ceiling at 6.4×, 35.5×, 108× the observed value at 50%, 10%, 1% thresholds. Granting a multiverse ensemble removes ~118.6 of 120.1 orders, leaving ~1.55; without granting the ensemble it explains zero.
anthropic ceiling 6–108× observed; ~1.55 of 120 orders left over, and only if a multiverse is granted reproduced ≠ discovered
Dark energy PROVISIONAL 2026-07-26
USR-2026-0062 The vacuum catastrophe is a choice of assumption
The famous vacuum-energy discrepancy is not a number but a function of an assumption: against the observed 2.24 meV scale it is 10^120.7 at a Planck cutoff, 10^108.4 at GUT, 10^56.4 at 1 TeV, 10^41.6 at QCD, and exactly 1 at the observed scale by construction — four decades of exponent per decade of cutoff, with the loop-factor convention alone shifting every row by 2.2 dex.
10^120.7 at Planck cutoff — but 10^56.4 at a TeV; the exponent is assumption-laden, the problem is not reproduced ≠ discovered
Dark energy PROVISIONAL 2026-07-26
USR-2026-0003 Every supernova compilation shows the same drift
The evolving-DE preference appears in all three SN compilations refit through one pipeline: Pantheon+ Δχ² = 6.96 (2.16σ), DES-SN5YR 8.95 (2.53σ), Union3 11.83 (3.00σ). Separately and more weakly, the local-structure escape hatch falls about 5× short on a diagonal-covariance shell and dipole test.
2.16σ / 2.53σ / 3.00σ across three SN compilations (post-Lyman-α-fix) reproduced ≠ discovered
Dark energy PROVISIONAL 2026-07-26
USR-2026-0001 Dark energy is drifting, not constant
Dark energy evolves: a CPL (w₀, wₐ) fit to DESI DR2 BAO + CMB + Pantheon+ beats ΛCDM by Δχ² = 6.97 (2.16σ), with w₀ = -0.849 and wₐ = -0.548. CORRECTED 2026-07-26: the earlier Δχ² = 8.2 (2.4σ) used a transposed DESI Lyman-α covariance pairing — see the audit log.
Δχ² = 6.97 → 2.16σ (was 2.39σ before the Lyα covariance fix) reproduced ≠ discovered
Dark energy PROVISIONAL 2026-07-26
USR-2026-0008 An oscillating dark energy ties the standard fit
On this data combination the CPL best fit is not uniquely selected: a five-parameter oscillating form ties it (χ² = 1413.72 versus 1414.11). But at its best fit the amplitude is pinned against the code's own A < 1.5 bound with a 24-e-fold period against 1.25 e-folds of data, so the 'oscillation' is a monotone segment — and CPL still wins on AIC by 1.6.
χ² tie (1413.72 vs 1414.11) but the oscillator is pinned at a code bound; CPL wins on AIC by 1.6
Dark energy PROVISIONAL 2026-07-09
USR-2026-0005 No late-time fix reaches the local Hubble value
Of the five late-time dark-energy candidates tested, none reaches H₀ = 73 when SH0ES is included — all cap at 69.25–70.24. Only a free sound horizon does, reaching H₀ = 73.06 at ΔAIC = −13.46 with r_d shrunk 2.96% (~4.4 Mpc). Within CPL the fit crosses w = −1 at z = 0.381.
five late-time candidates cap at H₀ = 69.25–70.24; only free r_d reaches 73 reproduced ≠ discovered
Dark energy PROVISIONAL 2026-07-09
USR-2026-0004 Dark energy's pressure changes across cosmic time
Binned w(z) rises from -0.920 ± 0.040 (z = 0.0-0.4) through -1.078 ± 0.093 (z = 0.4-1.0) to -1.537 ± 0.282 (z = 1.0-2.5); the trend survives dropping all SNe but dies without BAO.
BAO-anchored trend — no independent σ quoted reproduced ≠ discovered
Dark energy PROVISIONAL 2026-07-09
USR-2026-0002 The Hubble constant disagrees with itself
The Hubble tension is real in this pipeline: post-Lyman-α-fix inverse-ladder H₀ = 68.71 ± 0.39 versus SH0ES 73.04 ± 1.04 — a 3.90σ gap.
3.90σ gap (registered as 4.4σ before the Lyman-α fix) reproduced ≠ discovered
Dark energy PROVISIONAL 2026-07-09
Open to the field This registry is not just ours — register any published result you want checked and it enters like everything else: provenance bundle, kill test, status set from a public audit, labeled reproduced ≠ discovered. How submission works →
The graveyard

Claims we killed, and what killed them

USR-2026-0044 2026-07-26

The growth index γ evolves from 0.41 (z < 0.3) to 0.10 (z > 1) — gravity weakens with redshift.

Died by its own kill test, on our own data, at our own hand. The 4.84σ evolution reproduces at 1.03σ with the sign reversed, no jackknife sample recovers it, and the claimed Δχ² is arithmetically unreachable given how well the no-evolution model already fits. The autopsy is the useful part: fσ₈ loses 93% of its sensitivity to γ by z ≈ 2, and γ is 91% correlated with σ₈ — so a high-z γ 'measurement' is mostly the amplitude degeneracy talking. Gravity does not weaken with redshift in this data; it looks like General Relativity (γ = 0.525 ± 0.077 vs 0.55). Post-mortem addendum: fixing a wrong fσ₈ point uncovered during the autopsy makes the evidence against this claim stronger, not weaker — γ_a moves from +0.63 to +1.19 while the claim requires about −0.68.

USR-2026-0007 2026-07-26

Sign-switching Λ (Λs-CDM) is NOT preferred by this data: ΔAIC = +1.50 versus ΛCDM with SH0ES included, with the vacuum flip at z = 3.11. The...

Published as weakly preferred at ΔAIC = −2.5; re-runs at +1.50, which is disfavoured. The whole reversal came from one transposed covariance block — the same Lyman-α pairing error that cost USR-2026-0001 a quarter of a sigma. Killed not because the idea is bad but because our own number was wrong, and the corrected number points the other way. The lesson is propagation: this claim sat unchanged for hours after the fix that invalidated it.

USR-2026-0047 2026-07-12

A Gaia wide-binary velocity analysis shows an apparent MOND-direction boost, but only under a clip-biased median estimator (residual 1.167). The...

Killed by its own estimator audit: the MOND boost lived only in the clip-biased median. Under the unbiased mean the wide binaries are Newtonian (0.948). The 28.7% clip rate in the a₀ regime was doing the work.

USR-2026-0023 2026-07-09

The Quaia quasar-dipole excess (raw D = 0.0208, 9.9σ over shot noise) is not cosmological: adversarial magnitude and galactic-latitude cuts...

The adversarial battery did its job: 9.9σ of raw excess collapsed to a kinematic D = 0.0066 under magnitude and latitude cuts. The dipole belonged to the catalog, not the cosmos.

Share X Bluesky LinkedIn Reddit HN Email