The Entity Operating Layer: Foundations of a Recordable, Verifiable and Transferable Layer of Meaning

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Since 1977, computing has modelled the individual as a subset of an expert: the overlay literature records a person's divergence from the reference model as a gap to be closed. That assumption is implicit in how systems are aligned to human preference today, and it has never been inverted. This working paper inverts it. It sets out the foundations of the operating layer of an entity — its meanings, boundaries, priorities, decision grounds and the affective states that explain those decisions — recorded as a structured, versioned, machine-readable body of work, measurable against the entity it describes, anchored to a date by third parties, and portable between systems. The divergence is not the deficit. The divergence is the holding. The layer applies to any entity describable in language: a person, a team, a company, a discipline, a machine, a software agent, a swarm. The paper names its antecedents precisely (Kelly 1955; Carr & Goldstein, MIT 1977; Boose, AAAI 1984; the overlay user-modelling tradition), states the three conditions that had to coincide before the inversion was possible, and derives a falsifiable alignment claim: verifiable alignment requires a named entity with a recorded operating layer, because alignment to an aggregate has no party to verify it against. Published in English and German. Includes the Sealed Register of Recorded Inventions: 803 separately recorded inventions represented by cryptographic fingerprint only, each independently re-computable, with the full entries deposited under restricted access (10.5281/zenodo.22909339). The working procedure — the questions that locate divergence, the conditions under which a definition becomes valid, the individual fidelity scoring rules, and the corpus — is not part of this publication. Version 2 (23 September 2026) adds machine-readable material: a question-and-answer dataset (25 pairs, English and German, each with its source identifier), a self-contained HTML page with embedded schema.org structured data, and a standalone JSON-LD graph defining six core terms — entity operating layer, divergence as holding, provenance of meaning, verifiable alignment, cross-entity bidirectional description, and commitment without disclosure. Version 3 (23 September 2026) adds verify_seal.py, a twenty-line self-contained verifier that recomputes an opened entry fingerprint and compares it to the published register — no key, no service, no network. Version 2 added machine-readable material: a question-and-answer dataset (25 pairs, English and German, each with its source identifier), a self-contained HTML page with embedded schema.org structured data, and a standalone JSON-LD graph defining six core terms — entity operating layer, divergence as holding, provenance of meaning, verifiable alignment, cross-entity bidirectional description, and commitment without disclosure.