Video / Evidence / PactVerity technical series
How the PactVerity Proof Engine Works
A step-by-step explanation of measurable terms, evidence, deterministic checks and receipt verification.
Five-minute narrated explainer · original publication: 16 August 2026. Page and evidence links reviewed 27 September 2026.
Watch on the official YouTube channel ↗
Recheck the current evidence
Original narration transcript
The transcript preserves the historical recording. It is not rewritten to imply that later developments were already live when the video was recorded.
The PactVerity Proof Engine is designed for claims that can be expressed as measurable conditions. It does not ask a language model to guess whether a service was good. It takes structured terms, structured evidence and a deterministic policy, then returns a result and a receipt. That distinction matters because the same compatible inputs should produce the same evaluation path and make disagreement easier to locate.
Terms describe the service promise before evidence is evaluated. For an API, that might include a maximum response time, a required availability percentage, a freshness window or an expected output hash. Each condition should be specific enough to test. Ambiguous phrases such as fast, reliable or high quality are not enough until they are converted into thresholds, time windows and permitted evidence sources.
The evaluator supplies observations such as response timing, status samples, timestamps, content digests or signed records. Evidence should include enough context to understand where it came from and what period it covers. The engine does not independently observe the outside world. It checks the evidence submitted to it. Strong integrations therefore pay careful attention to provenance, access control and the possibility of missing or manipulated data.
A policy defines how each condition is calculated. The engine runs the checks, records whether each one passed and preserves the policy version used. Deterministic does not mean infallible. It means the path is explicit and reproducible. If parties disagree, they can inspect the term, input evidence or policy logic instead of debating an opaque score produced by a changing model.
The resulting receipt links the agreement, evidence summary, checks, outcome, engine version and timestamp. A digest helps detect later alteration. The receipt can be stored, exchanged and submitted to the verification endpoint. It is designed to answer a focused question: given these inputs and this policy, is this result internally consistent and reproducible?
Verification confirms the receipt schema, digest and expected relationships between its fields. A known-answer health test also checks that the deployed service can reproduce a predefined result. These are useful technical signals, but they are not an external security audit, uptime history or proof of adoption. PactVerity publishes those distinctions so evaluators can separate operational checks from broader assurances.
A receipt may be internally valid while the underlying evidence source is compromised. A poorly designed threshold may reward the wrong behaviour. Two parties may choose an incomplete measurement window. These failure modes are why the Proof Engine should begin with low-risk, measurable digital services and why independent review remains necessary before any future financial settlement layer is considered production-ready.
Open the Proof Engine at pactverity dot com and run a bounded example. Inspect the output, receipt digest and verification step. Developers can also review the OpenAPI specification and Receipt API beta. The goal is to let evaluators challenge the model directly. Contact PactVerity if you have an API or agent workflow suitable for a transparent technical pilot.
Continue the series
- What Is PactVerity? Verifiable Receipts for AI Agents →
- Why AI Agents Need Verifiable Receipts →
- Verifying APIs: Latency, Uptime, Freshness and Integrity →
- The AI Agent Trust Stack: Identity, Authority, Evidence and Receipts →
- PactVerity Receipt API Beta: Developer Walkthrough →
- PVTY Token Facts, Utility Proposals and Current Boundaries →
- PactVerity Roadmap: Simulations, Security Gates and Real-World Adoption →
- How to Evaluate an AI Agent Claim: A Practical Checklist →
- Building Accountable Agentic Commerce →