GenArch.
Knowledge Architecture

Alexandria

The knowledge architecture that compounds.

A living knowledge corpus for AI agents. Alexandria stores what agents learn, indexes what they've seen, and retrieves what's relevant, turning ephemeral context into durable institutional memory that compounds across sessions.

Knowledge compounds,
but only when it's shared.

Today, the best evidence lives in people's heads. When they leave, it leaves. When agents start a new session, they start from zero. The organization never compounds.

Every new hire re-discovers the same pitfalls. Every agent session re-learns the same context. Decisions are made without precedent, or worse, against it. Knowledge lives in one practitioner's head until they're gone, and the wiki nobody trusts is the only alternative.

The real edge is accumulated, internalized evidence: pattern recognition across decisions, failures, and recoveries, made shareable across people and agents so the whole organization compounds.

Individuals still get the serial practitioner's advantage. The difference: so does everyone else, and every agent, every session. Institutional amnesia is the default. Alexandria is the structural fix.

Three primitives. One coherent store.

Every operation in Alexandria (storing a fact, retrieving context, applying a role-scoped view) is encoded using the same three-part structure. It is one architecture that appears everywhere.

Frame task-scoped
The immediate working context
What is in scope right now, what the goal is, what constraints apply to this specific task or session. Frames hold the work in front of the agent: transient, task-keyed, discarded when the task closes. The footprint of decisions made within a frame feeds back into the substrate.
Lens role-scoped
The domain perspective
The role-shaped filter. Declares which entity types are loaded, which relationships are in scope, which state-changes are watched, and which RBAC capabilities apply. The lens shapes how the substrate is interpreted: what signals matter, what vocabulary applies. Same substrate, different surface, governed by declared contract. Lenses are the productizable surface; the substrate is the platform.
Substrate universal
Foundational institutional memory
The universal state of what agents and people have seen, decided, and learned. Entities, facts, typed state-changes, decision traces, interactions, bi-temporal validity. Every substrate write emits a typed state-change with before/after, consumer, frame, and evidence. One substrate; everyone shares it.

The compounding loop.

Every action feeds back. Every retrieval is a usage signal. Every correction strengthens the graph for the next person and the next agent. The loop compounds across sessions and across time.

01
Capture
Decisions, observations, reflections, and session outcomes
02
Extract
Claims, entities, and connections identified and typed
03
Connect
Cross-domain links, contradictions, and patterns surfaced
04
Retrieve
Right context, right moment, right depth
05
Act
Better decisions, informed by precedent and accumulated evidence

every action feeds back; the loop compounds

Six capabilities that compound.

01

Assumption Register

Every assumption made explicit, with kill criteria and a validation timeline. When an assumption is invalidated, the system knows which decisions depended on it, and surfaces them.

02

Decision Quality Tracking

Separate the quality of a decision from the quality of its outcome. Retrieve well-reasoned precedent, going beyond what merely worked last time. The distinction matters for learning.

03

Strategy-Execution Drift

Compare stated priorities against behavioral evidence: time allocation, actual decisions, observable patterns. Surface drift before the quarterly review discovers it.

04

Cross-Entity Intelligence

Patterns emerge at the intersection. Signals connect to assumptions. Decision traces link to strategy claims. The graph sees what any individual agent or session cannot.

05

Convergence Detection

No single metric triggers action. The signal is multiple independent indicators crossing thresholds in a time window. The system watches for clusters, not spikes.

06

Progressive Disclosure

Not a data dump. Retrieval filters to full claims to source references to traversal. The agent exercises judgment at every layer about what to absorb and what to defer.

Same architecture. Different domains.

Alexandria is a cross-product knowledge layer. It underlies every Lexenne product: the same intelligence, shaped by a different lens for each domain.

Strata Work

Career history as institutional memory

In Strata, Alexandria stores career history, preference patterns, and match outcomes. Each session builds on what previous sessions learned, so the agent compounds toward a sharper picture of what actually constitutes a good fit.

Patina

Household memory that persists

In the Patina product family, Alexandria builds household memory across purchases, maintenance decisions, and long-term planning. Knowledge that compounds at the household level, held across sessions and apps.

The four architectural commitments.

Substrate generalizes; lenses don't
One substrate, many lenses. Adding a lens never forks the substrate. This is the architectural posture that allows different products and roles to share the same compounding knowledge without collision.
State-change is first-class
Every substrate write emits a typed state-change with before/after, consumer, frame, and evidence. Bi-temporal validity answers "what did the system believe on date X." Provenance is structural.
Lens is declared, not emergent
Lens contracts are DB-resident with declared RBAC, default entity types, and watched state-changes. Skill metadata references the lens; the dispatcher trusts the declared contract.
Non-action is a decision
Triggers emit explicit act / wait / escalate / no-op signals with rationale. Silent drift becomes structurally distinguishable from intentional restraint. Absence of action is recorded, not ignored.