{
  "schema": "cain.e129.final_verdict.v1",
  "entity": "FIRME AI CORPORATION",
  "system": "CAIN-42",
  "evolution": "E129",
  "generated_at": "2026-11-21T19:01:44Z",
  "source_commit": "cef0d5688",
  "agi_status": "NOT ESTABLISHED",
  "rsi_status": "EXPERIMENTAL",
  "unification_verdict": "PASS",
  "multigen_rsi_verdict": "PARTIAL",
  "evidence_integrity_verdict": "VERIFIED",
  "rationale": "Laboratory A (AGI) and Laboratory B (RSI) were genuinely converged in cain/cain2027/unified/ via typed contracts, shared 9-partition provenance memory, independent evaluation, and unified MCPGate governance. In the preregistered 4-arm matched-compute study (800 evals, 5 sealed seeds), Unified Arm D (0.7) outperformed Arm A (0.34), Arm B (0.385), and Arm C (0.59) (Holm p={'D_vs_A': 0.0, 'D_vs_B': 8.32e-10, 'D_vs_C': 0.003185253448}), driven by inherited compositional macro compounding (macro rate 0.91 vs 0.75/0.28/0.25). In multi-generation RSI (G0->G3), compositional macro inheritance raised unseen solve rate from 0.25 to 0.9583 (vs 2x compute-matched 0.25, p=1.5258789e-05), while RSI's causal candidate CAND-G1-L3-CAUSAL-STRAT FAILED independent evaluation (0.7557 vs random 0.7435) and was REJECTED without promotion (unseen G3 causal MAE 0.8026 vs random 0.6891); hence multigen_rsi_verdict is PARTIAL. All tasks remain constructed on a CPU-only host with external benchmarks, third-party verification, and HF deployment BLOCKED. AGI remains NOT ESTABLISHED; RSI remains EXPERIMENTAL."
}