Codette: Multi-Perspective Reasoning as aConvergentDynamical System with Meta-Cognitive Strategy Evolution

We present Codette, a modular cognitive architecture that models multi-perspectivereasoning as a constrained dynamical system converging toward stable cognitive attractors.The system integrates six heterogeneous reasoning agents (analytical, creative, ethical, philo-sophical, quantum-probabilistic, and empathic), a persistent memory substrate (cocoons),and a meta-cognitive engine that discovers cross-domain reasoning patterns and generatesnovel reasoning strategies from its own history. The RC+ξ (Recursive Convergence + Epis-temic Tension) formalism provides a dynamical-systems-inspired lens for describing cognitivestate evolution via agent-weighted updates with coherence and ethical constraint terms; wetreat the convergence discussion as conditional on explicit modeling assumptions ratherthan as a general guarantee. We evaluate Codette through a benchmark suite of 17 prob-lems across six categories (multi-step reasoning, ethical dilemmas, creative synthesis, meta-cognition, adversarial robustness, and Turing naturalness) under four experimental condi-tions: single-agent baseline, multi-perspective synthesis, memory-augmented reasoning, andfull Codette with strategy evolution. On this benchmark (timestamp: 2026-04-08), the fullsystem achieves a 93.5% higher mean composite score than the single-agent baseline (0.356→ 0.689). Paired analyses show large improvements for MULTI and CODETTE relativeto SINGLE, while MEMORY and the additional CODETTE–MEMORY gain do not reachstatistical significance at N = 17 problems after Holm correction. The architecture runs onconsumer hardware (Llama 3.1 8B with LoRA adapters) and is open-source.