Proofline — Evidence Gate for AI Agents by Kai Relaunch DeptProofline — Evidence Gate for AI Agents by Kai Relaunch Dept

Proofline — Evidence Gate for AI Agents

Kai Relaunch Dept

Kai Relaunch Dept

Workflow demo
Proofline Keeper applies the same evidence-gate pattern to a transaction firewall: inspect the proof receipt, block unsafe actions, and route exceptions for approval.
47-second continuous Proofline evidence run
Problem AI agents can produce confident outputs from stale or unverifiable sources, then trigger an external action without a reproducible audit trail.
Approach Proofline separates evidence collection from decisioning. It records provenance and freshness, applies explicit policy checks, and creates a tamper-evident proof packet before an external action is allowed.
Delivery • Deployed Google Cloud Run application • Deterministic pass or block decisions • Public evidence links and reproducible decision receipts • Nine automated tests • Live demo and Devpost case study
Outcome A reviewer can reproduce why a decision passed or failed instead of trusting opaque model output. No synthetic production results or commercial outcomes are claimed.
AI-assisted engineering; evidence rules and final behavior were reviewed before publication.
Like this project

Posted Aug 7, 2026

Designed and deployed a verification-first AI product interface on Cloud Run with 9 tests, deterministic receipts, and reproducible public evidence.