Evidence & Predictions

What the model predicts, how confident we are in each prediction, and what would kill it. For the derivations behind these numbers, see the Framework page.

Current Evidence

ObservableΛCDM saysWe predictData showsStatus
σ₈ (clustering amplitude)0.811≈ 0.76 ± 0.030.76–0.79 Consistent
S₈ (weak lensing)0.83≈ 0.78 ± 0.030.76–0.79 Consistent
Lyman-α k-cutoffnonek_c ≈ 0.75 h/Mpcsuppression observed Consistent
G_eff during leakageG_N≈ 0.75 G_Nconsistent Consistent
CMB acoustic peaksfitunchangedPlanck fit Consistent
Damping-tail suppression0%1–3%pending CMB-S4 Pending

Predictions by Confidence Tier

75%Tier 1 — Core mechanism
  • Effective gravity is transiently suppressed near z ≈ 50,000.
  • Small-scale power is suppressed relative to ΛCDM.
  • CMB acoustic peaks preserved.

These follow directly from the derived Bricks 1–3 and are the most robust part of the framework.

60%Tier 2 — Strong directional predictions
  • Power-spectrum cutoff at k_c ≈ 0.75 h/Mpc.
  • S₈ ≈ 0.78 ± 0.03.
  • Damping-tail suppression of 1–3%.

Requires the phenomenological Brick 4 mapping to hold; validated only if Boltzmann code reproduces σ₈.

45%Tier 3 — Specific numerical predictions
  • σ₈ = 0.76 ± 0.03.
  • Radion equilibrium r₀ ≈ 0.75.
  • Potential steepness β ≈ 8.5.

Parameters are currently fitted with zero degrees of freedom; Stage 3 must derive them from first principles.

Falsification Criteria

What kills this model
  • ◇ Precise linear P(k) at k ≈ 0.5–1 h/Mpc shows standard ΛCDM with no cutoff.
  • ◇ σ₈ measured higher than ΛCDM (wrong sign of the effect).
  • ◇ Lyman-α forest shows no small-scale suppression.
  • ◇ Growth rate f(z) matches ΛCDM exactly.
  • ◇ CMB damping-tail suppression > 5% (too large).
  • ◇ PIXIE-class spectral-distortion measurement rules out any H(z) deviation near z ≈ 50,000.

All are testable within 2–3 years.

Four-Stage Validation Roadmap

  1. 1
    August 2026Active
    Stage 1: Boltzmann code validation

    Patch CAMB/CLASS with modified G_eff(z). Run with fixed parameters, extract σ₈, verify against target 0.76.

  2. 2
    2027
    Stage 2: Alternative-model comparison

    Show the leakage mechanism is distinguishable from Warm Dark Matter, Fuzzy Dark Matter, and other small-scale suppressors.

  3. 3
    2027–2028
    Stage 3: First-principles 5D derivation

    Derive Brick 4 (scale selection) from the 5D Einstein–Gauss-Bonnet equations rather than fitting.

  4. 4
    2028–2029
    Stage 4: Observational tests

    DESI, Euclid, Roman, and CMB-S4 measurements of P(k), growth rate, and damping tail.

Future Tests (2027–2029)

  • CMB-S4: 1–3% damping-tail suppression prediction.
  • Euclid: weak-lensing S₈ to 1% precision.
  • DESI Lyman-α: k_c cutoff extraction from the forest.
  • Roman: high-z galaxy-cluster growth history.