Differential QA Evidence Lab

Compiler defects can alter state and events even when return values look correct. Clean differential results are unreliable without a calibrated detector.

Solution

Built a Vyper differential workflow that froze observable behavior, used a historically fixed defect as a positive control, and compared REVM, Py-EVM, and optimizer modes.

Deliverables

Harness; eight campaign families; test plan; execution matrix; positive-control report; machine-readable results; documented exclusions and limits.