openai/gpt-chat-latest → B (3:2)

Side B makes a functional architectural improvement by introducing per-entity selectors (`data-entity-section` plus `entity_section_selector`) so SSE/Idiomorph updates target the correct entity instead of a single global `#entity-section`, and it reuses the shared `entity_section` renderer in the voting UI to eliminate duplicated rendering logic. Side A improves test maintainability by replacing manually enumerated Kaocha suites with automatic `^test\..+` discovery, which reduces future configuration work, but its impact is narrower than the UI update and code reuse changes in Side B.