Agent memory benchmarks have had a basic attribution problem. Every AI memory startup claims better recall, and until now there was no way to check. That is why this new shared benchmark caught my…