A Low-Resource Cognitive Architecture for Human–AI Symbiosis: SCCE, CSNM, and ERCI

This paper introduces the Cuniglio Architecture, a low-resource cognitive framework for human–AI symbiosis composed of three integrated modules: the Sistema Cuniglio de Cognición Expandida (SCCE), the Cross-Session Narrative Memory (CSNM), and the External Recursive Cognitive Interface (ERCI).
Unlike conventional LLM-based approaches that depend on large-scale compute, persistent memory, or massive datasets, this architecture demonstrates that advanced cognitive behaviors — including deep recursion, narrative continuity, and emergent meta-reasoning — can arise through structured human–AI interaction on standard consumer hardware, without storing any personal or identity data.
The SCCE models cognitive state evolution across four interpretive layers with reproducible metrics. The CSNM formalizes how inferential reconstruction generates cross-session coherence without memory retention. The ERCI externalizes recursive reasoning into a controlled, human-guided loop that reaches depths of over ten cognitive levels while maintaining structural stability.
Simulations conducted across 200+ interactions demonstrate coherence scores above 0.84, functional reconstruction rates of 0.91, and recursive stability gradients below 0.12 — all achieved without GPU-intensive infrastructure. The work also introduces a Security, Ethics and Verification Framework (SEVF) ensuring zero-identity retention, human cognitive sovereignty, and full auditability.
This contribution is positioned at the intersection of cognitive science, symbolic AI, human–computer interaction, and AI ethics, and proposes a viable pathway toward ethical, decentralized, and reproducible AGI research accessible to institutions in Latin America and other resource-constrained regions.