Artificial individual - technical specification

This technical specification follows on from "Architecture of an Artificial Individual - version 2, informed by Human Psychology and Physiology" (Orta, 2026, doi:10.5281/zenodo.23003276), which opens a series of dated deposits in the public domain. It describes the infrastructure of an "artificial individual": components, interfaces, data and verification protocol. The internal algorithms of each module are left to the following documents in the series. The specification tests a single postulate: a temperament and a history are enough to make an individual, whatever equipment of modules it receives. It separates three layers: the equipment (pre-trained modules, common to all individuals and frozen in the minimum viable product), the temperament (five parameters fixed at birth) and the history (what the individual lives through and learns). Identity is carried by temperament and history, not by the equipment. The specification tests neither general intelligence nor consciousness. Specialised modules emit contributions in a common envelope; a deterministic gatekeeper scores them and admits a few into a bounded global state, broadcast to all modules; the decision emerges from a competition between candidate actions whose thresholds depend on temperament. A direct path handles reflexes, veto and habits. Five rewritable memories sit alongside an immutable, hash-chained log that records every contribution and decision, and consolidation runs offline. A worked scenario follows two individuals with the same equipment who begin to diverge. The postulate is broken down into six testable hypotheses (coherence, history, stability, invariance to equipment, bifurcation, architecture), with an explicit refutation criterion. The minimum viable product (MVP) places several individuals with the same equipment in a small grid world with a simulated body. It uses a crossed design of temperaments, worlds and random seeds, and a trait battery of five fixed situations. The MVP is stage 1 of a six-stage roadmap (stages 0 to 5), in which each stage opens only after the previous stage's gate. The document also covers evaluation metrics, safety risks and countermeasures, limitations, a pre-registered protocol, a register of 18 architecture decisions, a traceability matrix for the 62 modules of v2, data schemas and a glossary. First complete draft: nothing in it has been implemented or tested. Written with the assistance of Claude (Anthropic), under the author's direction; not peer reviewed. New references are cited from memory and flagged as such in the document. Dedicated to the public domain (CC0).