Verifiable Proof Systems (VPS) helps design partners track and measure pilot success by establishing a fixed-fee, four-week assessment that closes with a letter of intent if agreed criteria are met. This guide covers the core architecture of VPS, the specific metrics used to evaluate pilot outcomes, and the audit trail documentation required to verify authority boundaries in autonomous infrastructure.

Verifiable Proof Systems Architecture

Verifiable Proof Systems is a research initiative focused on developing infrastructure for verifiable authority. The core problem it addresses is that software now proposes consequential actions faster than anyone can review them. A proposal is not permission, and permission is not the effect. VPS builds the boundary where a proposed action either becomes a consequence or is refused.

The Three Planes of Authority

The architecture separates computation, authority, and consequence at one admission boundary. This separation is critical because models now propose tool calls, database writes, payments, and setpoint changes. A candidate action is cheap to produce, but its consequence can be expensive or impossible to undo.

• Plane A, Proposal: Untrusted. Models, planners, tools, and people propose. It holds no authority over production state or a plant.
• Plane B, Admission: A small, deterministic check of the attempt and its predicted effect, under a fixed policy.
• Plane C, Execution: Minimal. It carries out only what Plane B authorized, and records the outcome.

Admission requires all five conditions to hold together. These include independent evidence not produced by the proposer, bounded single-use authority scoped to the predicted effect, a verified workload identity and a separate human sponsor, state and sequence continuity, and declared constraints satisfied. Commit also requires a durable decision record before any effect. Rejected attempts are recorded too.

Custom Tools and Boundary Definition

Pilot Project Metrics and Baselines

For design partners, the pilot is a four-week, fixed-fee assessment on one of your systems. The success of this pilot is not measured by vague qualitative feedback but by specific, pre-agreed criteria. The pilot closes with a letter of intent if the criteria agreed in week one are met. This structure ensures that both the partner and VPS are aligned on what constitutes a successful outcome from the start.

Establishing the Week One Baseline

ROI is measured with each partner against a baseline agreed in week one. This baseline typically includes the current state of the system's authority boundaries, the frequency of proposed actions that fail admission checks, and the documentation gaps identified by the AgenticX-DYE(TM) tool. By establishing this baseline, partners can track improvements in boundary clarity and evidence quality over the four-week period.

Measuring Boundary Integrity

The primary metric for pilot success is the reduction in weak, drifted, or non-existent boundaries. VPS uses its custom tools to surface these points, and the pilot tracks how many are resolved or documented. Additionally, the pilot measures the completeness of the decision record. Every decision at the boundary, including refusals, must be recorded. A successful pilot demonstrates that the system can consistently produce this durable decision record without manual intervention.

Evidence Quality and Falsifiers

VPS maintains a public claim ledger containing 115 entries that document stated assumptions, explicit non-claims, and their respective falsifiers and statuses. During the pilot, partners can use this ledger to verify that the claims made about the system's capabilities are supported by evidence. The quality of the evidence is a key metric. Weak evidence, such as model confidence or a passing unit test, is testimony about a proposal but does not admit the transition. Strong evidence is independent and bounded.

Audit Trail Documentation and Evidence

Audit trail documentation is the backbone of verifiable authority. It provides a durable record of every decision made at the admission boundary. This record is essential for post-incident analysis, regulatory compliance, and continuous improvement. VPS emphasizes that outputs are not proof. A fluent explanation or a signed artifact from the proposing system is not sufficient. The audit trail must capture the independent evidence, the bounded authority, and the verified identity that allowed the action to proceed.

The Durable Decision Record

The durable decision record is a log of every attempt to cross the admission boundary. It includes the proposed action, the evidence presented, the authority granted, and the final decision. Rejected attempts are recorded too. This comprehensive logging ensures that teams can reconstruct the state of the system at any point in time. It also provides a clear history of how authority was exercised and constrained.

Documentation of Refusals

One of the most critical aspects of the audit trail is the documentation of refusals. When an action is refused, the record explains why. This could be due to insufficient evidence, lack of bounded authority, or a violation of declared constraints. By documenting refusals, teams can identify patterns in failed proposals and adjust their policies or evidence generation processes accordingly. This feedback loop is essential for improving the system's reliability over time.

Integration with Smart Contract Tools

Key Takeaways

• Verifiable Proof Systems separates computation, authority, and consequence at one admission boundary.
• The pilot is a four-week, fixed-fee assessment that closes with a letter of intent if week one criteria are met.
• ROI is measured against a baseline agreed in week one, focusing on boundary integrity and evidence quality.
• Admission requires five conditions: independent evidence, bounded authority, verified identity, state continuity, and declared constraints.
• The audit trail must record every decision, including refusals, to provide a durable decision record.
• AgenticX-DYE(TM) surfaces points where boundaries are weak, drifted, or non-existent.
• VPS is a research initiative, and CEAK-PRO 1.0 is a conceptual working paper with no model-checking or proof results yet.
• Smart contract tools can be used to add transparency and immutability to the audit trail.

Frequently Asked Questions

What is the primary goal of the VPS pilot?

The primary goal of the VPS pilot is to assess one of your systems over four weeks to determine if the criteria agreed in week one are met. If they are, the pilot closes with a letter of intent.

How is ROI measured in the pilot?

ROI is measured with each partner against a baseline agreed in week one. This baseline includes the current state of authority boundaries and documentation gaps.

What does the AgenticX-DYE(TM) tool do?

AgenticX-DYE(TM) surfaces the points where boundaries are weak, drifted, or non-existent. It is written in DARKc, a formally verified fail-closed compiler and syntax.

Does VPS guarantee that the system is secure?

No. VPS finds and documents gaps. It records each decision, including refusals. It makes changes detectable. It shows where the evidence stops. It does not guarantee security or prevent all unauthorized actions.

What is the difference between a proposal and permission?

A proposal is a candidate action generated by a model or planner. It carries no permission and no effect. Permission is bounded authority over a named object, under explicit limits. A good proposal does not create permission.

Is CEAK-PRO 1.0 a formally verified system?

No. CEAK-PRO 1.0 is a conceptual working paper. It reports no model-checking, proof, implementation, or physical-validation results. Formal models and an implementation are in development for a companion paper.

How does VPS handle rejected attempts?

Rejected attempts are recorded in the durable decision record. This ensures that teams can analyze why actions were refused and adjust their policies or evidence generation processes accordingly.

Can VPS help with regulatory compliance?

VPS produces evidence for requirements such as AIUC-1 B006, D003, and D004. It helps teams prepare for an outside test or audit. VPS is not a certifier or an accredited auditor, and it is not affiliated with AIUC.

Conclusion