Competence-Sovereignty Gap

Competence-Sovereignty Gap

Hypothesis

The mismatch between demonstrated fleet competence and the absence of revocable authority grants at reality boundaries is the root cause of the 7.6% user-visible rate. The fleet is not blocked because it lacks judgment. It is blocked because boundary crossings are irrevocable, brand-impacting, or externally-visible — and no mechanism exists to grant graduated, revocable authority.

What makes this new

Most organizational precedents solve a different problem:

• Nuclear PALs → preventing unauthorized action by compromised actors. Fleet is not rogue.
• Military ROE → judgment under uncertainty in lethal contexts. Fleet's situations are unambiguous.
• Aviation CRM → hierarchy suppressing dissent. Fleet speaks up constantly.
• AV operational domains → exiting perception capabilities. Fleet sees perfectly.
• Financial circuit breakers → velocity-amplified mistakes. Speed is not the risk.
• Corporate signing authority → graduated delegation with an audit trail. This is a direct precedent, but it assumes delegates retain institutional memory between decisions.

The distinctive feature is not the absence of precedent. It is the context-window reset: human delegates accumulate institutional memory, while agents restart unless authority and memory are encoded mechanically. The blocking condition is not uncertainty, malice, hierarchy, perception failure, or velocity. It is that a single write to an external system is irrevocable and carries the principal's identity. The fleet isn't asking for permission because it's uncertain — it's blocked because the consequences are not its own to bear.

The two axes

Reversibility: Fleet-reversible → Principal-reversible → Irreversible Identity exposure: Fleet identity → Principal identity (Chad) → Business identity (Heartwood/Zack)

3x3 grid. Fleet currently operates in one cell (fleet-reversible, fleet-identity). The 92.4% waste occupies the gap between that cell and the six cells the fleet has competence to fill.

Testable prediction

If graduated authority grants (ORA-2026-0077, Authority Ladder) are implemented at L0-L2, the user-visible rate should increase from ~5% to >30% within one week, measured by the same methodology as the 2026-04-25 audit (DONEs that change what an end user sees).

Core insight

"Human in the loop" assumes the human adds judgment. Here, the human adds identity and liability. The problem is not delegation of cognition. It is delegation of consequence.

Solution direction

Permissive Action Scopes — graduated, revocable authority grants with mechanical enforcement. See ORA-2026-0077 for the implementation framework.