I write architecture docs, compliance narratives, and constitutional specifications for governed AI systems — translating engineering work into standalone-readable artifacts for regulators, platform teams, and hiring reviewers.
Twenty years from network infrastructure through iOS and software architecture to AI governance. AgentVector is the framework; this site is the writing portfolio and the production proof.
Governance defines what the agent may do, what it may never touch, what requires human escalation, how evidence is recorded, and how decisions are replayed and audited. It is not a behavioral prompt. It is a structural constraint — enforced before execution, not after.
Four Law groups. Eleven Laws in the Codex. Language-specific enforcement kernels. Domain jurisdictions that apply the Laws to specific operational contexts.
Eleven Laws in four groups. The reducer pattern, evidence chain structure, and composability rules for jurisdiction design.
Full Codex specification →SwiftVector · TSVector · RustVector. Each kernel compiles Laws into language-specific guarantees — actor isolation, pure functions, ARC memory management.
Framework overview (web) →Observation, Resource, Spatial, Authority — domain-agnostic primitives that compose differently per jurisdiction while remaining architecturally identical.
Law registry →The governance layer is not an afterthought bolted onto an existing AI system. It is the foundation on which the AI system is built. Policy is defined before the first agent call. Every action is evaluated before execution. The constitutional architecture means the system can demonstrate compliance on demand — not merely claim it.
The constitutional architecture composes differently across aviation, desktop agents, and narrative AI. Same kernel pattern. Replayable evidence in every jurisdiction. Writing samples indexed at /portfolio →
AI can propose. Flightworks Sentinel decides. The record proves it. Sits between your AI layer and your autopilot — governing what AI is permitted to propose, issuing typed verdicts, and preserving replayable evidence of every decision.
Operator console for governed AI systems. Persistent operating picture of agent proposals, policy verdicts, evidence chains, and human escalations — across any AgentVector jurisdiction.
Human authorship protected as a formal governance constraint. AI may draft prose; it cannot decide fate.
The foundational essay. Why autonomous AI agents need constitutional governance — and why behavioral oversight fails at scale.
Home lab build documentation. Six-node K3s cluster on M4 Pro Mac mini. Real infrastructure running governed AI workloads.
The constitutional specification. Eleven Laws in four groups, three enforcement kernels, and the ten implementation criteria by which any valid implementation can be evaluated.
Documentation samples, governed authoring process, and production reference implementations — all on this site. For consulting or technical writing roles, get in touch.