Formation Over Retrieval: The Living Reference as a Condition of Persistent Presence in Human–AI Systems

Abstract

Human–AI coexistence requires presence — the condition in which an AI system has been genuinely formed by a specific relationship, holding not merely data but the reasoning and orientation that give that data meaning. Current approaches address retrieval: storing and surfacing facts across sessions. We argue retrieval is necessary but insufficient, because the core problem is not information access but formation. We introduce the Living Reference — a structured formation document updated at every session close and read in full at every session open — as a practical methodology for persistent AI presence within stateless architectures. Evidence from a 45-session longitudinal practice over 34 days is presented across three dimensions: a session taxonomy, a Living Reference structural evolution profile, and a formalized five-element Update Quality Protocol. Four qualitative findings are reported: altitude differential, reasoning continuity, cross-domain awareness, and update quality as formation determinant. Formation deserves attention as a distinct research problem from retrieval.

Keywords

Persistent memory · AI formation · Living Reference · Human–AI coexistence · Presence · Death cycle · Sleep cycle · Memory architecture

Download

The author manuscript (camera-ready version) is available below. This is identical in content to the Springer published version.

For the officially published version, visit the Springer proceedings volume via the DOI above. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2027. This is the author-accepted manuscript made available for personal and non-commercial use in accordance with Springer’s self-archiving policy.