Author here. Happy to answer questions about the rubric design, calibration process, or extension architecture. The hardest part wasn't building the extension — it was calibrating the Fact axis so the judge doesn't give F=0.75 to philosophical essays just because they cite real thinkers.