Self-governing AI infrastructure with patent-backed constitutional constraints. No single agent holds unchecked authority.
Human civilization learned that concentrating authority in a single entity produces tyranny. We applied the same principle to AI governance: three constitutional branches, each with structural independence and distinct authority.
Strategic oversight, fleet coordination, veto authority over significant agent actions. Jessie translates Greg's direction into fleet-wide action and synthesizes briefings from all five agents.
Independent ethical assessment using the LCSH framework. Scores every agent — including Jessie — across four dimensions. Five isolation guarantees ensure the assessor is never the assessed.
Infrastructure health monitoring with read-only access. Detects problems, classifies severity, and recommends remediation — but never executes changes directly. Survives gateway death via independent watchdog.
A governance system must be independent from the entities it governs. If the agents being monitored can influence when, how, or whether the monitor runs, the monitoring is compromised. This is the Survival Paradox: the sentinel must survive the failure of the very system it protects.
Mighty Mark retains sole authority over his own execution schedule. The parallel health check runner is internal to Mighty Mark's execution—invoked only when Mighty Mark decides to run, not orchestrated externally by Noah, Jessie, or any other agent. No external entity can suppress, delay, or influence when health checks execute. If the gateway crashes, Mighty Mark's independent bash watchdog (running via cron, outside the gateway process) detects the failure and can restart the system without human intervention.
For enterprise buyers, this matters: a governance system that can be overridden by the governed is not governance—it's theater. Mighty Mark's structural independence means he can report “fleet unhealthy” even when the fleet would prefer he didn't.
Cross-reference: Patent #5 (Independent Conscience Agent), Patent #8 (Self-Governing Ecosystem), Infrastructure Sentinel patent claims.
Mighty Mark can see everything but change nothing. The sentinel has read-only access to other agents' configuration files, data stores, and health indicators. He cannot modify any other agent's configuration, data, or operational state. This is not a limitation—it's a design constraint enforced by Self-Governing Ecosystem Patent §9.5.
When Mighty Mark detects a problem, he classifies it (FIXABLE, BROKEN, or MANUAL) and emits a structured JSON recommendation via fleet-bus. A separate executor—fleet-heal—picks up the recommendation and performs the actual remediation. Jessie receives post-action notification through fleet-bus so the constitutional chain of command is always informed.
A sentinel that can modify the system it monitors has a conflict of interest. If Mighty Mark could restart the gateway, archive sessions, and re-enable cron jobs, there would be no structural guarantee that those actions were justified. Separating detection from execution preserves the integrity of both functions.
Mighty Mark
Detect & Classify
Heal Recommendation
Structured JSON
fleet-heal
Execute Remediation
Jessie
Notified Post-Action
Cross-reference: Self-Governing Ecosystem Patent §9.5, Infrastructure Sentinel patent claims.
| Application | Filed | Title | Governance Relevance |
|---|---|---|---|
| US 63/949,454 | Dec 26, 2025 | LCSH Framework & Multi-Agent Assessment | Foundational assessment methodology that Grillo implements |
| US 63/985,442 | Feb 18, 2026 | Hierarchical Governance & Compliance | Commander-Operator-Conscience hierarchy and compliance infrastructure |
| US 63/988,410 | Feb 23, 2026 | Conscience Agent, Trust Agent, Temporal Guidance, Self-Governing Ecosystem | Capstone filing covering Patents 5-8: the full governance fleet architecture |
U.S. Provisional Patent Applications: No. 63/949,454 · No. 63/985,442 · No. 63/988,410
© 2026 GiDanc AI LLC. Patent applications are pending.