openai/gpt-chat-latest → B (4:1)

Side B fixes a core ranking algorithm bug by changing Rank Centrality to use degree-based d_max instead of summed edge weights, eliminating oscillation in star-topology graphs and producing correct stable rankings. It also adds focused regression tests (Rust and Clojure fixtures) that directly reproduce and guard against the failure, whereas Side A introduces substantial invite and audit functionality but also leaves newly added invite events unused in favor of ephemeral in-memory state, making the design less durable.