openai/gpt-chat-latest → B (50:1)

Side B adds substantial new functionality and infrastructure: a dedicated pair-selection module with bridge-aware voting logic and tests, a full `/vote` compare UI with in-place morph updates after voting, ID normalization via `ItemId::from_storage` to fix canonicalization bugs, and accompanying integration/end-to-end tests. Side A only removes a single test function, reducing test coverage without introducing lasting behavior or design improvements.