Core Principles

Sovereignty, portability, legibility, evolution, and distribution — the non-negotiables.

  1. Sovereignty — Public services must not be architecturally dependent on private interests
  2. Portability — Standardized outputs, not locked implementations
  3. Legibility — Citizens can understand, question, and change what the model optimizes for
  4. Evolutionary — The system learns; better models displace worse ones on merit
  5. Distributed — No single jurisdiction, vendor, or government layer controls the stack

Practical Barriers

Anticipated Objections and How the Framework Addresses Them

"Most counties don't have IT staff."

The framework anticipates this. Compute cooperatives — state-hosted shared inference layers — allow small jurisdictions to access registry models via standardized schemas without running their own infrastructure. The interface requirement is an API call, not a data center.

"The hyperscalers will fight this."

They may. The permitting leverage argument is real but time-limited — which is precisely why the sequencing matters. The framework doesn't require federal legislation to begin; a single city negotiating a single data center permit is enough to start.

"What stops this from being politically captured?"

The distributed architecture is the answer. No single layer controls the stack. The federal registry is governed by a multi-stakeholder board, not a single agency. State schemas define outputs, not implementations. Municipal autonomy over model development is protected by design.

"How do we measure success?"

Real-world outcomes, not benchmarks. Accuracy on historical decisions, processing time reduction, citizen satisfaction on service interactions, and equity metrics across demographic groups — tracked per jurisdiction, published annually, and used as the basis for model selection in the registry.