The old job: dispensed, or eaten?
A timed feeder drops a portion at 19:00 and treats the job as done. On Petlibro's Granary 2 series a built-in scale weighs food in the tray, so the dispensed amount and the amount left in the bowl are logged separately. TechCrunch writer Lauren Forristal said she tested the Vision model. The app stores eating duration, frequency, preferred times and speed in daily, weekly, monthly and yearly views.[1]
The bowl getting lighter is a visible log. The scale catches a timer that drops a second portion before the first is gone. That does not establish that a cat is ill; it only splits "the machine dispensed" from "the bowl emptied."[1]
The camera claim is not yet a confusion test
Petlibro says the Vision camera can tell apart as many as 10 cats, that Vision Duo uses two chutes and bowls, and that the top model portions from that recognition. Those are vendor claims. The article has no independent count of two similar cats being confused. It also lists no stock, country list or general on-sale date; the unit Forristal tested does not mean the same software is on in every kitchen.[1]
My read is that the brittle step is recognition quietly assigning a portion to the wrong cat. The scale still leaves a leftover; the camera layer can bury the error in an eating chart. The alternative stands: the camera may have been steady in Forristal's home, and the article simply did not count that. What the reader has is that the scale's log and the vendor's 10-cat sentence are not the same kind of evidence.[1]
The signal to wait for
A multi-cat home can judge the scale layer from the spec today: does leftover food show up? For the camera layer, the signal is an independent confusion test that names the device, the lighting and the cat pair. If no such write-up exists by 2026-12-31, the 10-cat sentence stays a marketing claim.[1]