Dated, falsifiable bets, each with a kill condition fixed in public before the data — and the sheet that indexes them. Registered under the adversarial self-registration protocol: a bet that dies stays frozen rather than being refitted.
Working note
A Falsification Sheet for the Finite-QEC Substrate Programme
A deliberately adversarial map of the simplest empirical ways to kill, demote, or constrain the framework. Near-term kill switches: a confirmed primordial tensor signal at r ≳ 10⁻³; a dark-energy reconstruction excluding w(a) = −1 + a/28; a scalar tilt excluding n_s = 27/28; a secular drift in G or α; a neutron EDM far above 10⁻³¹ e·cm with no extra CP source; or dark matter shown to be a freely streaming particle gas. Several tests are branch-level (the 17.7 keV sterile neutrino, the R4/MOND line law, the K04 debris walls) and would retire a branch without collapsing the matter sector.
Prediction
A pre-registered galaxy-scale gravitational-slip prediction: screened metric response γ_eff(g_b) from a finite-QEC substrate
A frozen bet, published as made: a falsifiable galaxy-scale prediction fixed before the deciding data, with no-retune rules written into the registration. The substrate's dark-sector split gives massive tracers and photons different readings — photons additionally see a screened metric response g_M = a₀√g_b · exp(−g_b/(k·a₀)) — so lensing-plus-dynamics comparisons should read an effective γ_eff(g_b) above one, rising as acceleration falls. Neither GR+CDM nor MOND-class theories produce that shape; both give γ = 1 at every g_b. It sits within 2.2σ of the strongest published constraint, and one SLACS-class reanalysis binned in g_b, or a single resolved lens at g_b ≈ 4 × 10⁻¹⁰ m s⁻², settles it. Registered 10 July 2026 and posted unaltered.
Prediction
A Rational Target for Dynamical Dark Energy: w₀ = −27/28, wₐ = −1/28
A registered dark-energy branch preserved as the public record of a prediction retired under its own rules, because pre-registration requires failed predictions to stay visible. In the CPL parameterisation the framework fixed a single rational point, w₀ = −27/28, wₐ = −1/28 — the one-line law w(a) = −1 + a/28 — a mild thawing-like departure from ΛCDM produced not by a fit to survey data but by the same 28-channel service clock that sets n_s = 27/28. The dated outcome: the branch is internally killed by the pinned-H₀/acoustic-clock consistency test and kept only as a frozen reopen condition. Surveys may still test the point, but it failed internally before the external posterior became decisive. An honest negative, retained rather than refitted.
Prediction
A Primordial-Tensor Null from Boundary Printing: Pre-registering r_linear = 0
A short, dated prediction from the boundary-printing branch: a primordial-tensor null, r_linear = 0. Inflation here is not a smooth stretching of an existing metric but the printing of fresh boundary cells by a scalar counting process, and a scalar source has no transverse-traceless part, so the spin-2 projector annihilates it at linear order. The often-quoted r ∼ 2×10⁻⁹ is only the second-order scalar-induced floor, far below any near-term reach. The near-term claim is therefore a null: no primordial B-mode detection at r ≳ 10⁻³, a clean kill switch, since a robust LiteBIRD or CMB-S4 signal would refute the branch. The post-hoc rules are pre-committed: the printer premise and A_s = (3/4)α₀⁴ cannot be retuned after the data arrive.
Prediction
A Sterile-Neutrino X-ray Target from the Finite-QEC Dark Sector: m_nuR = 17.68 keV, E_gamma = 8.84 keV
A staged, pre-registered X-ray target for the framework's sterile-neutrino dark-sector state, with a flux audit that grades its own claims. The hard claim: the passive ν_R state has mass m_νR = α₀²Λ_QCD = 17.68 keV, so its radiative decay gives a photon line at 8.84 keV, distinct from the much-discussed 3.5 keV line, with a conditional abundance Ω_νR h² = 0.02418. The flux tier is downgraded: the registered κ = 1 branch gives a deliberately faint line below XRISM-class sensitivity, but current canon does not derive κ = 1, and the flux scales as κ². The robust content is the line position, the relic-density target and no bright line at this energy: a much brighter 8.84 keV line refutes the branch; a deep non-detection only bounds κ.
Prediction
A Registered Null Prediction for K04 Defects: Pinned Fossils, Not Mobile Halo Dark Matter
A registered negative prediction: the K04 crystallisation-defect sector is a substrate-pinned fossil, a gauge-blind relic of imperfect crystallisation, not the mobile collisionless halo. Topologically, the protected class is a ℤ₂ homology class, so finite misbonds heal while protected relics are extended strings or walls. Dynamically, every advancing move crosses a positive Peierls barrier, and both the astrophysical drive (∼10⁻⁴²) and thermal creep at today's CMB temperature are hopelessly small. The readings are nulls: no comoving with galaxies, no Bullet-cluster-like separated peaks, no terrestrial defect wind, no CDM-like subhalos. Using K04 as mobile dark matter is falsified unless new canon supplies a finite carrier or a zero-barrier move.
Prediction
A Pre-Registered Neutrino-Sector Prediction from the Finite-QEC Substrate: Normal Ordering at the Oscillation Floor, No Dirac Phase, and an Invisible m_ββ
The framework's complete neutrino-sector bet, fixed by substrate constants plus one declared scale, Δm²₃₁ = 2.515×10⁻³ eV². Four claims: no intrinsic leptonic Dirac CP violation (J_ℓ = 0, δ_CP ∈ {0, π}, because CP transport rides colour intermediates and the lepton register is colourless); normal ordering at the oscillation floor with m = (0.79, 8.72, 50.16) meV and Σm_ν = 59.7 meV; CP living in the Majorana sector, never in long-baseline δ_CP; and an invisible m_ββ = 1.76 or 3.07 meV, below the LEGEND-1000/nEXO band. At registration the CP-conserving point sits 0.16σ from the NuFit 6.0 best fit, with 2.2–3.7σ strains in the mixing angles openly declared. Grade: conditional.
Prediction
Leptonic CP is Majorana-only: A Registered Prediction Addendum
A dated addendum to the neutrino-sector registration (3 and 6 July 2026) with a machine-readable JSON attached. The upgrade is the reason for the Dirac-CP null: rephasing-invariant Dirac CP needs transport through the colour register, and the colourless lepton register carries only trivial phase, in the same walk engine that makes the quark sector complex. So leptonic CP is Majorana-only: the long-baseline phase is conserved (J_ℓ = 0, δ_CP ∈ {0, π}) while CP lives in a monitored Majorana recovery-orientation record, visible in neutrinoless double-beta decay and leptogenesis, not oscillations. Normal ordering with Σm_ν = 59.7 meV. Killed by intrinsic Dirac CP at high significance, an inverted ordering, Σm_ν far above 60 meV, or m_ββ ≳ 10 meV.
Prediction
Newton's Constant from the Proton Mass: A Locked Zero-Parameter Prediction
The sharpest single number in the programme: Newton's constant computed from the proton mass through a chain with no fitted parameters, G = ħc/M_P² with M_P² = 990α₀³Λ²/r₆ and Λ = m_p/(2√2(1+3α₀²)), where α₀ = 1/137 is an exact alphabet count, 990 = 2·9·55 an exact combinatoric, r₆ a computed queue current and (1+3α₀²) the junction-billing correction from the baryon's Y-junction. The result, G_pred = 6.674311×10⁻¹¹ m³ kg⁻¹ s⁻², lands at +0.07σ of CODATA-2022, with the framework side exact at the 10⁻⁹ level, so the whole uncertainty is experimental. The correction class was registered with a two-sided lock/kill rule before its landing was computed, and a big-G measurement at ≲15 ppm excluding G_pred kills the chain outright.
Prediction
Finite-Information QCD: A Pre-Registered Fingerprint Protocol at a Derived Lattice Spacing
In the substrate programme the lattice spacing is a derived physical constant, a₀ = 0.59494 fm, with no continuum limit to take, so deviations from continuum-extrapolated QCD at this spacing are predictions, not artefacts. This document freezes the deviation list first. It registers a closed four-item fingerprint list — an orientation-averaged ℓ = 4 potential scalar, the three-quark junction geometry, glueball mass ratios at a₀, and the roughening/Lüscher crossover — each with estimator, comparator and kill rule, plus a paired-lattice protocol and a two-stage freeze fixed before any production compute. A null across all four hardens the QCD-reproduction claim; the glueball clause gives an external kill.