Vector search, embeddings, and the index that answers them.
They argued over 32 versus 112 queries a second. We ran their own benchmark and got 2,318.5.
