Uni-Versum proposes a personal knowledge architecture built from the human outward rather than the tool inward. Information systems fail because they manage information (data plus context) but not knowledge (semantic relationships, perspective, and boundaries). The architecture rests on four pillars — workspaces with selective boundaries (satelles), a unified registry (census), a communications routing layer (tabellarium), and a human perspective center (nos) — organized as a watershed where shared context creates gravitational pull between entities. A formal Latin vocabulary of 41 terms, mapped to SKOS/RDFS/JSON-LD semantic web standards, turns untyped links into typed relationships that machines can traverse and humans can query for meaning.
Typed relationships are the knowledge upgrade. The argument’s central architectural claim: a link between two notes says “connected” but nothing about how, why, or what follows. A typed relationship (“A funds B,” “A inhabits B”) carries implications. The vocabulary is infrastructure, not documentation — every entity has a type, every relationship has a named term. Without it, you have an information system; with it, a knowledge system.
Watershed over canal. The system is organized like a watershed (emergent, terrain-shaped, gravity-driven) rather than a canal (engineered, centralized, capacity-limited). No central hub commands routing; shared context between entities creates gravitational pull proportional to relationship density. This is both metaphor and design principle — it produces the distributed, fractal architecture that enables the system to grow without collapsing.
Four-level agent taxonomy by interface modality. The vocabulary classifies agents not by role or capability but by communication interface: agens (any agent) splits into agens-humanus (biological) and agens-scriptum (automated scripts). Agens-scriptum further specializes into agens-scriptum-loquens (AI agents using natural language). This is orthogonal to role-based taxonomies — an orchestrator and a worker could both be agens-scriptum-loquens.
Trust as a three-layer vocabulary. Agent trust is modeled through three distinct terms: fidelis-est (loyalty/allegiance — which human perspective center an agent serves), autonomia (the evolving trust contract defining what an agent can do on behalf of its human), and diploma (trust credential for agents from a different human’s system operating in yours). This separates allegiance, delegation, and federation — three concerns most agent frameworks conflate.
Precision creates freedom. The vocabulary governs architecture (entity types, relationship types, workspace boundaries) but not content (daily notes, drafts, rough captures). “When the structural terms are well-defined, everything else can be loose.” This is the difference between a city’s zoning plan and the conversations people have in their houses.
Fractal boundary architecture. The biological cell metaphor: every workspace has a selective membrane (boundary), internal organization, signal processing, and autonomy. The four pillars (workspaces, registry, communications, perspective) replicate at every scale. Each workspace maintains its own knowledge; the central registry registers workspaces but trusts each to manage its own contents.
Latin as namespace isolation. English terms carry connotations from prior tool use (“workspace” means different things in Slack, VS Code, Notion). Latin terms (satelles, insula, taberna, horreum) arrive without baggage, carrying exactly one meaning in this system. Same principle as scientific nomenclature — precision across contexts, not elitism.
Lectio as perspectival reading. A reading note (lectio) is not a property of the source but “the record of an encounter between a source and a specific context of work.” The same book produces different lectiones in different workspace contexts. The reader’s perspective is constitutive, not incidental.
Nos as perspective center, not user. The human at center is not a “user” operating a tool from outside but a consciousness inside the system — “inhabiting the whole, feeling where attention is needed.” Nos holds perspective (continuity across project changes) rather than centralizing data.
“Connection differs from relationship. Stating ‘Note A links to Note B’ conveys nothing regarding the reason for connection, the nature of the link, or the significance for the individual. This constitutes mere connection without comprehension.” — The Problem
“A graph of untyped links is a map with roads but no labels. You can see that places are connected, but you do not know if a road is a highway or a footpath, if it goes one way or both, if it is open or closed. The structure is visible; the meaning is not.” — The Failed Solutions
“The shift from ‘the tool defines the structure’ to ‘the person defines the structure’ is the single most important design decision in this architecture. Everything else follows from it.” — The Model
“Fidelis-est: the bond of trust between a patron and those operating on their behalf. It is not obedience; it is allegiance.” — Vocabulary: fidelis-est
“When the structural terms are well-defined, everything else can be loose. You do not need to tag every note, because the relationships between entities already carry the meaning.” — The Vocabulary
“A lectio is not a book attribute, not a review, not a summary. It is a relationship: the record of an encounter between a source and a specific context of work.” — Vocabulary: lectio
“The human is the circulatory system: the one who carries meaning between organs that could not reach each other on their own.” — The Living System
Implementation maturity. Self-described as “work in progress.” Fractal architecture designed but partially implemented, trust agreements implicit, communications routing nascent. The vocabulary is more complete than the running system.
Single-person validation. The claim that the pattern is “replicable” by any knowledge worker is aspirational. No second implementation exists.
Gravitas remains conceptual. The gravitational metaphor (relationship density as force) has no computational specification. How it produces useful results at scale is unverified.
Latin accessibility. Twenty-nine domain terms plus twelve schema terms require learning before the vocabulary is useful. Bilingual definitions mitigate partially.
The work is early-stage with no established reception. Its significance is architectural rather than empirical: it provides formal vocabulary for agent trust relationships (allegiance, delegation, federation) that most multi-agent frameworks handle implicitly. The agent taxonomy and trust primitives are directly applicable to the estate’s persona strategies and inter-estate collaboration patterns. The semantic web foundation (SKOS, RDFS, JSON-LD) gives the vocabulary interoperability guarantees that ad-hoc predicate systems lack.