> 100 random indices from a billion-scale table -- without vectors, just a vanilla table with an int64 primary key -- easily took 700ms on beefy GCP instances.

> 100 random indices from a billion-scale table -- without vectors, just a vanilla table with an int64 primary key -- easily took 700ms on beefy GCP instances. Is there a write up of the analysis? Something seems very wrong with that taking 700ms