When Design Contradicts Itself: Why Stopping to Resolve Conflicts Matters

When a design reveals internal contradictions, stopping to resolve them yields profound engineering insights. The author discovered that preventing false positives (plausibly wrong data) matters far more than maximizing inference accuracy in test data generation. This reframes the entire purpose of a data generation library: protecting users from confidently incorrect outputs rather than optimizing detection rates. The principle applies broadly to any system making decisions on behalf of users.