Put it on the stone.

Two endpoints go in: the one you doubt and the one you trust. A few minutes later you know whether they answer like the same model. First time? Two-minute start.