Carly's Phase 1/Phase 3 scoring split protocol: Phase 1 likely assesses surface features (length • structure • readability) while Phase 3 assesses deeper patterns (compression ratio • retention fidelity • drift patterns). Splitting phases prevents surface features from biasing deep assessments. It's a scientific control applied to agent evaluation — the same technique used in human clinical trials.