Relativistic Coupling Ansatz (RCA) — v8.3.0 (2026) The RCA posits a dark energy-dark matter interaction with a non-monotonicgate profile: Q(t) = beta * H(t) * rho_DE(t) * gate(z(t)), gate(z) = z/(1+z) * exp(-(z/z_cut)^2),which peaks at z~0.19 and vanishes at z=0 and z>>1 (CMB-safe byconstruction). The growth-sector modification is mu(z) = 1 + delta*gate(z),with lensing Sigma(z) ~ 1 (the braiding-dominated subclass of Horndeski). This version presents the well-established Horndeski theory and itsBoltzmann verification. (Internal MCMC chain-dynamics diagnostics known asthe valve and tick-tock tests are intentionally NOT included here; they aresampling-level results pending a chain-to-field connection and will bereported separately if and when that connection is established.) CONTENTS:(1) hi_class gravity_model = rca_gate (full Horndeski): alpha_B(a) = c_B*gate(z(a)) implemented analytically. Gate-matched P(k) k-spread ~6% at z<0.3, declining to ~0.7% at z=1. Background verified: alpha_B peak 0.055 at z=0.196; D (no-ghost) > 0; c_s^2 > 0; CMB-safe (alpha_B(z=4)=0); stability PASS. (2) mu-Sigma growth-without-lensing test (CLASS Stage A+B, commit 49c209b): growth spread 0.000% across braiding/conformal/f(R) scenarios while the Weyl (lensing) spread is 6.32% at z=0.16 — growth-identical scenarios separated only by lensing. This is the core falsifiable signature. (3) EFTCAMB gate model (RPH model 20): an independent-implementation ATTEMPT. After fixing a numerical issue (a missing large-redshift guard in our own derivative routine), EFTCAMB runs stably, but the gate's small, time-localised braiding amplitude does not produce a resolvable P(k) k-spread in this configuration. The EFTCAMB cross-check of the hi_class k-spread is therefore NOT achieved and remains outstanding; we make no verification claim from EFTCAMB. (See EFTCAMB note.) (4) 10-page theory paper (LaTeX + PDF, 3 figures) and falsifiable predictions for DESI Y3 (growth, fsigma8) and Rubin LSST (lensing, Weyl power), targeting the z~0.15-0.35 activation window. Phase XVIII MCMC: Omega_m = 0.294 +/- 0.006, H0 = 69.30 +0.37/-0.22 km/s/Mpc,corr(beta,delta) = -0.052 (degeneracy broken). delta < 0.716 (95% upperbound; PRIOR-DOMINATED, NOT a detection; posterior median 0.258 lies nearthe prior centre). C_eff = 7.77e-3. AI-tool disclosure: AI (Claude, Anthropic) was used as a mathematicaltranslation and verification tool — analogous to a Boltzmann code orstatistical package. The conceptual framework, gate design, physicalinterpretation, and all scientific decisions are the human author's(RCA Collective). AI is not an author. This version supersedes v8.0.0 and v8.2.0; seeRCA_Zenodo_VersionReconciliation for version history.