openai/gpt-chat-latest → A (4:1)

Side A materially improves the pair-selection algorithm by introducing structured bridge and within-component priorities, preferring attachment to established voted components and then rank-adjacent refinement once the pool is connected. It also adds multiple targeted tests covering the new selection behavior. Side B mainly simplifies the HTML flow by server-rendering the new-thread slot on the home page and removing the now-redundant UI action and tests, which is a useful cleanup but has a narrower long-term impact.