Research paper

Intent-to-Execution Proof: verifying what was authorized against what actually ran

The core CoreGuard thesis: logs describe, receipts prove. A summary of the repository's intent-to-execution design and its fail-closed verdicts.

The question

Every on-chain action starts as an intent — a user, wallet or AI agent approves something. What executes on chain can differ from that approval. Today most systems keep logs, and logs are written by the party being checked.

The approach

CoreGuard binds the authorized intent to a commitment, records the execution, and lets anyone re-derive the verdict offline. A mismatch or missing evidence yields a fail-closed result instead of an optimistic pass.

VERIFIED means the evidence matches. It never means the transaction was safe or the counterparty trustworthy.

Keep reading

← All posts