openai/gpt-chat-latest → A (4:1)

Side A adds a substantial new capability: pool-scoped voting, including new `/vote?pool=` request handling, automatic pair selection from child items, pool-aware next-pair navigation, UI links, and propagation of pool context through vote submission and success updates. Side B improves the test infrastructure by fixing OAuth mock request parsing, redirect/form handling, null safety, and Playwright selectors so end-to-end authentication tests work again, but those changes primarily affect testing rather than the application's core functionality.