Feasibility of stateless architectures passing the NCT

Ascertain whether stateless inference-time architectures—systems that generate responses without a persistent identity-bearing state across turns—can satisfy the Narrative Continuity Test and qualify as continuous interlocutors.

Background

A core claim of the paper is that stateless inference and local plausibility objectives hinder narrative continuity because commitments do not persist across turns. Nonetheless, the authors stop short of declaring stateless designs incapable in principle and explicitly leave open whether such architectures could ever meet NCT criteria.

This question targets the architectural necessity of persistent state for identity continuity, asking if purely stateless approaches can ever achieve the joint, diachronic properties required by the NCT.