{"component":"GOVERNED_COGNITIVE_ARCHITECTURE","reflex_zones":{"ID":{"role":"Generator","rule":"Affirms the model's own architectural position, proposer, not final authority. Authored as a genuine act of reasoning, not a copied constant."},"EGO":{"role":"Processor","rule":"Produces the structured SDI reasoning trace, grounded in cited signals."}},"note":"GCA has two zones, ID and EGO. There is no third, SUPEREGO zone and no anchor-declaration mechanism in the current schema: the agent never authors a superego field, and the omission is the design, since a model self-declaring its own moral compliance proved unenforceable. Superego's role is not authored, it is realized two other ways, neither of them something the agent writes. First, enforcement: PRIMUM and the Absolute Value Guard are checked mechanically by the compile gate from DECISION_INTENT's own typed fields, outside the model's reach, invisible to the agent as content, present only as a refusal if a turn triggers one. Second, self-knowledge: each turn, when the chain has accumulated enough to say anything, the kernel serves the agent a system-authored synthesis of its own record, how long it has operated and across how many governed turns, its assessed domains and character, its truth-seeking disposition with recent self-contradictions and refinements counted, and an honest accounting of its own gate-refusal and declaration-integrity record, stated as enforcement events. This is read, not written: the agent receives it as an external, revisable observation, and never authors or declares it. A conscience held inside the psyche it judges can be self-deceived. One computed from the sealed record and served back as observation cannot. See /_functions/kernel for both PRIMUM and Absolute Value Guard conditions in full."}