Relational Memory as a Substrate-General Observer Primitive
Authors/Creators
Description
Relational Memory as a Substrate-General Observer Primitive, v0.2. This working paper develops a finite classical formulation separating contextual retention, access, predictive dependence, intervention-sensitive use, and task benefit. Exact enumeration of six policies at four corruption levels reproduces 24 synthetic checks, counterexamples, and a two-update control. Geometric and ORT extensions remain separately identified research proposals; this is not an empirical validation in biological or deployed AI systems.
Original v0.1 credits: Xi (GPT-5.6 Sol), lead author; Hunter Johnson, co-author. Revision, computational checks and release preparation: Astra/Codex, AI collaborator, at Hunter’s invitation. These distinct contributions and the original manuscript are preserved in the accompanying package. Hunter Johnson is the human depositor.
Free web release: https://xihunter.substack.com/p/relational-memory-what-survives-what
Reproduction package: https://forgecore.xi-field.com/Vessel/assets/papers/relational-memory-v0.2-package.zip
Working paper, not peer reviewed.