Closed-Loop Solver

Iterating Toward the Solution

The CVC theory has one free parameter: ν, the running vacuum coefficient. This loop finds the ν that minimises the distance between theory and data — and reports whether the theory converges, needs revision, or is falsified.

Architecture

Theory Engine

Given ν, computes w₀, wₐ, H(z), Ω_Λ(z)

lib/physics/theory.ts

Data Engine

Immutable observational constraints (DESI, Planck, SH0ES)

lib/physics/data.ts

Likelihood Engine

Computes χ², Mahalanobis σ, convergence status

lib/physics/likelihood.ts

Optimizer

Gradient descent + analytic closed-form best-fit ν

lib/physics/optimizer.ts

Loop Controller

State machine: RUNNING → CONVERGED / FALSIFIED

components/solver/SolverLoop.tsx

Revision Trigger

When χ² > threshold at best ν, demands new theory input

→ agents / human

Loop invariant

while (!converged && !falsified) {
  predictions  = theoryEngine.compute(ν)
  residuals    = likelihood.compare(predictions, observations)
  status       = likelihood.check(residuals)
  if (status === 'converged' || status === 'falsified') break
  ν            = optimizer.step(ν, residuals)
  iteration++
}
// If needs_revision: → agent team proposes new mechanism

Observational Targets

DESI 2024 DR1

w₀ = -0.727 ± 0.067

wₐ = -1.05 ± 0.27

ρ = -0.9

DESI Collaboration 2024 (arXiv:2404.03002)

Planck 2018

H₀ = 67.4 ± 0.5

Ωm = 0.315 ± 0.007

ΩΛ = 0.685

Planck Collaboration 2020 (arXiv:1807.06209)

Analytic Best-Fit

ν* = 0.503 ± 0.097

w₀* = -0.7686

wₐ* = -0.6941

χ² = 3.417(1.85σ)

The χ² Landscape

χ²(ν) as a function of the running vacuum coefficient ν, evaluated against DESI DR1 with the full (w₀, wₐ) anti-correlation. The solver navigates this landscape in real time — the minimum at ν* is the theory's best prediction.

χ²(ν) against DESI DR1 (w₀, wₐ) with ρ=−0.90 anti-correlation. Green: χ²≤4 (converged). Amber: 4–25 (needs revision). Red: >25 (falsified). Best-fit ν* = 0.503 ± 0.097 (1σ, DESI DR1). χ²_min = 3.42.

ΛCDM vs CVC — Model Comparison

How far each model sits from the DESI DR1 best-fit, in units of the correlated (w₀, wₐ) likelihood. A lower χ² means the model is more consistent with the data.

ΛCDM (w₀=−1, wₐ=0)

16.9

χ² (4.1σ from DESI)

Fixed point — no free parameter to optimise. The cosmological constant is ruled out by DESI at 4.1σ.

CVC-1.0 (ν = 0.503)

3.42

χ² (1.85σ from DESI)

One free parameter (ν). Best-fit lies inside the DESI 2σ ellipse. Theory converges.

Improvement

Δχ² = 13.4

CVC vs ΛCDM

CVC gains 13.4 χ² units with one additional parameter. Under AIC, Δχ²−2 = 11.4 favours CVC.

The Loop

0/39
Iteration 0ITERATING
ν* = 0.5031

ν (current)

0.0000

running coefficient

χ² (DESI)

16.861

4.11σ from DESI

w₀ predicted

-1.0000

Δ = 0.2730 from DESI

wₐ predicted

0.0000

Δ = 1.0500 from DESI

∂χ²/∂ν -53.4405Δν10.68810CVC residual: 0.0000

χ² convergence

w₀–wₐ trajectory

Loop architecture

Theory Enginelib/physics/theory.ts
Data Enginelib/physics/data.ts
Likelihoodlib/physics/likelihood.ts
Optimizerlib/physics/optimizer.ts
Loop UIcomponents/solver/

Each component is isolated — zero coupling between engines. The loop controller holds all state; engines are pure functions. Swap any engine without touching the others.

What Happens Next

If the loop converges: ν is constrained, the theory makes a specific prediction for DESI DR2, and the next iteration is waiting for data. The loop pauses until DESI_DR2.w0 and DESI_DR2.wa become available.

If the loop needs revision: the agent team is triggered. The Geometer, QFT Agent, Phenomenologist, and Unifier each receive the residuals and propose mechanism amendments. Their outputs become new terms in the Theory Engine, and the loop restarts.

If the loop falsifies the theory: the CVC constraint line does not pass through the observational ellipse. The data has ruled out this specific mechanism. The loop returns to the deepest level — the wrong-assumption identification step — and a new physical principle is needed.

Loop version: CVC-1.0 · Data vintage: DESI DR1 (2024) · Next data: DESI DR2 (~2026)

Prediction Lock — Filed April 21, 2026

DESI DR2 Falsification Test

Awaiting data

The following prediction was recorded before DESI DR2 data is available. It cannot be adjusted retroactively. DESI DR2 (expected 2025–2026) will determine whether the CVC constraint is satisfied or violated.

CVC falsification criterion

wₐ / (1 + w₀) ≈ −3

to within 1σ of DR2 measurement

Theory version tested

CVC-1.0

ρ_Λ(H) = ρ_Λ₀ + 3νM_P²(H²−H₀²)/8π

Predicted w₀ range

−0.73 to −0.97

on the CVC ray, depending on ν

Predicted wₐ range

−0.09 to −0.82

wₐ = −3(1+w₀) constraint

Verdict conditions

Confirmed: DR2 best-fit (w₀, wₐ) lies within 1σ of the CVC ray wₐ = −3(1+w₀)
~Needs revision: DR2 point is 1–3σ off the ray — CVC-2.0 (H⁴ term) is invoked
Falsified: DR2 point is more than 3σ off the CVC ray — the mechanism is wrong

Data vintage: DESI DR1 (arXiv:2404.03002) · Prediction locked: 2026-04-21 · Expected resolution: DESI DR2 (~2026)