BURHAN is an independent verification plane for bounded AI coding tasks. It seals a contract before execution, qualifies validators against correct and incorrect controls, captures candidate state, and verifies it in a fresh workspace before issuing its own verdict.
The payment-idempotency demo includes real historical live Codex Validator Architect and Executor evidence. BURHAN captured the empty candidate state and independently issued REJECTED; AgentExecutionClaim never determines the verdict.
The repair sequence is explicitly labeled DETERMINISTIC REPAIR DEMO, not live Codex repair. The original live run reports REPAIR_CONTEXT_UNAVAILABLE for same-thread repair. The deterministic repair uses the same sealed validator standard, SamePackProof, fresh verification, linked local-artifact-integrity receipts, and tampering detection to reach VERIFIED.
This submission does not claim successful live GPT-5.6 Platform API inference because quota was unavailable.
Built With
- git
- next.js
- node.js
- openai-codex-cli
- openai-responses-structured-outputs
- typescript
- zod
Log in or sign up for Devpost to join the conversation.