Gracia’s [[Gracia (2026) Uni-Versum Personal Knowledge Architecture]] makes a bet that most knowledge architects avoid: use vocabulary that arrives with no prior associations. By naming her concepts in Latin, Gracia creates a namespace where each term means exactly one thing — not “workspace” (which means a different thing in Slack, VS Code, and Notion) but satelles, which means whatever Gracia defines it to mean. The price is a learning curve. The payoff is a vocabulary that does not fight itself.
This gloss reads the Uni-Versum vocabulary as precision infrastructure for multi-agent personal knowledge systems — not as a competing architecture to the estate’s feudal English, but as a complementary analytical layer that names things the estate’s vocabulary leaves unnamed.
Nos — the human at the center of the system, not as “user” operating from outside, but as the consciousness that inhabits the whole. Where “user” implies tool-operation, nos implies the Latin first-person plural — “we,” a perspective that is both individual and encompassing. Gracia’s argument: the shift from “the tool defines the structure” to “the person defines the structure” is the single most important design decision in personal knowledge architecture. Nos names the structural center that makes this shift real.
Agens — any agent, human or automated. The four-level taxonomy classifies agents by communication interface, not capability: agens-humanus (biological person), agens-scriptum (automated script), agens-scriptum-loquens (AI agent using natural language as its primary interface). This is orthogonal to role-based taxonomies — an orchestrator and a worker are both agens-scriptum-loquens. The distinction is behavioral, not organizational: a cron job and an LLM both automate, but only the LLM converses. The estate’s agent taxonomy (steward, orchestrator, worker, boundary guardian) and Gracia’s interface-modality taxonomy are complementary, not competing — each names a different dimension.
Fidelis-est — the bond between an agent and the nos it operates under. “It is not obedience; it is allegiance.” The term draws from Roman fides, the social bond of trust between patron and those operating in their name. Where “authorization” frames agent authority as technical permission, fidelis-est frames it as relational commitment — an agent’s authority derives from its ongoing trustworthiness to the human it serves, not from a permission list. This framing aligns with the estate’s conviction that sovereignty is a membrane rather than a wall: authority flows through relationship, not through access control tables.
Autonomia — the evolving trust contract between a nos and an automated agent, defining what can be delegated and at what granularity. Autonomy is not binary: some tasks are fully delegated, others require human judgment at each step, most fall somewhere between. The boundary shifts as trust develops — Gracia explicitly maps this to the estate’s escalation tiers. Autonomia names the mechanism that [[Allen (2023) Least and Necessary Design Patterns]] implies but does not fully theorize: the space between least authority (minimum by default) and necessary authority (minimum for the task).
Diploma — a trust credential for agents from a different nos operating in your system. Where autonomia governs intra-system human-to-agent trust, diploma governs inter-system trust: when an agent from another person’s estate enters yours, a diploma formalizes what they can and cannot do. This is a federated trust model distinct from both permission-granting (too static) and blanket access (too permissive). The diploma concept names what the estate currently handles through context and convention — it is a gap in the estate’s own vocabulary.
The vocabulary’s deepest contribution is not any individual term but the architectural principle they instantiate: “When the structural terms are well-defined, everything else can be loose.”
Gracia’s vocabulary governs structure (entity types, relationship types, workspace boundaries) but not content (daily notes, rough captures, exploratory thinking). Once satelles, nos, census, and tabellarium are precisely defined, everything else can remain unstructured. This is the difference between a city’s zoning plan and the conversations people have inside houses.
The estate’s garden forms implement the same principle through a different mechanism. Each form type (Model Form, Decision Form, Conviction Form) defines a structural contract — required sections, predicates, naming heuristics. Within that contract, the content can be as exploratory, tentative, or partial as needed. The structural contract is precise; the content is free. Gracia’s Latin vocabulary and the estate’s form types are parallel implementations of precision-as-freedom.
Gracia’s four-level agent taxonomy clarifies an ambiguity in the estate’s own agent classification. The estate distinguishes agents by coordination role (steward, orchestrator, worker); Uni-Versum distinguishes by interface modality (human, script, AI). A session running the Chancellor orchestrator involves agens-scriptum-loquens operating in a role the estate calls “orchestrator” — both descriptions are true and neither is redundant.
The trust vocabulary (fidelis-est, autonomia, diploma) names three distinctions the estate currently handles implicitly: allegiance vs. permission, delegation as dynamic contract vs. static access grant, and intra-system vs. inter-system trust. The estate has all three concerns but no vocabulary for them. Gracia’s terms could be adopted as the estate’s own trust layer without displacing the feudal English operational vocabulary.
The semantic web formalization (SKOS/RDFS/JSON-LD) gives Gracia’s vocabulary interoperability guarantees the estate’s ad-hoc predicate system lacks. Whether the estate needs those guarantees depends on whether it develops beyond single-person use — a question the [[Persona Design Choices Across Analytical Cultural and Professional Axes]] inquiry touches but does not resolve.