And when they said they're moving towards knowledge graph like instant results, latency not gonna be pretty, lots of small fetchings after general queries from core search servers.