openai/gpt-chat-latest → B (4:1)

Side B changes the core ranking algorithm from contributor-level comparisons to pairwise ranking of individual commits, removes the incorrect single-contributor short circuit, rolls commit scores back up to contributors, updates evidence/UI, and adds tests covering same-contributor multi-commit behavior. Side A mainly implements persistent theme handling across pages and login plus room-aware URL generation, which is useful, but it is largely feature integration rather than a fundamental correctness improvement to the project's ownership allocation logic.