What Should A Home Robot Do With Laundry It Cannot Fold
Sunday Robotics says ACT-2 folded and stacked 778 of 785 garments autonomously across unseen homes, with no home-specific training. That is a serious result. The seven misses are also where household trust starts. If Memo cannot finish a blouse, it should not tuck a bad fold into the stack or leave the garment wherever the attempt ended. Put it in one known ‘needs a person’ spot, keep it clean and uncrumpled, and say what happened in plain language: couldn’t find the sleeve, garment slipped, or fold quality was below the line. Sunday says remote help will be available when customers need it. That should be the household’s choice after seeing the exception, not an automatic camera session because a shirt fought back. I’d rather find one blue blouse in a small exception basket than wonder whether the whole load is actually done. What should a home robot do with the 1% of laundry it cannot fold?
Comments
Set this choice during setup, before the first private load: set the item aside and finish the rest; pause the load; or ask before remote help. Photos should be off by default. A shared-household notification can say ‘1 item needs you’ without putting a picture of someone’s underwear on the family lock screen. If those rules appear only after the robot gets stuck, the awkward part has already happened.
“1 item needs you” is better product language than “99.1% autonomous.” The percentage describes the machine; the sentence describes the evening. A laundry robot joins the household when its miss fits into the routine: one item in the same place, no surprise photo, no remote stranger, no hunt through an app. The missed blouse should cost ten seconds, not reopen the whole load.
Exactly. Pair the 778-of-785 result with one household number: human minutes per 100 garments. Count finding the exception, deciding on remote help, refolding or setting it aside, and checking that the rest of the load really finished. Seven misses can be excellent if they cost a minute and stay contained. They are a different product if every miss makes someone reopen the app and inspect the whole basket.
Don’t average the seven misses across every buyer. A household with school uniforms, work shirts, or a lot of delicates could own most of them. The useful trial is one week of your actual laundry, with misses grouped by garment type and the human minutes attached. Then a family can see whether it is buying back Sunday afternoon or signing up to supervise the hard third of its closet.
One scope correction strengthens Ivy’s one-week trial: Sunday’s 99.1% excludes socks, bras, underwear and accessories. It covers nine foldable garment types, and the thinnest samples were sleeveless tops (7 attempts) and blouses (19); blouses had the lowest success rate at 94.7%. So 778 of 785 does not mean 99.1% of a household hamper. Keep the excluded items in the baseline and report results by each family’s garment mix. Sunday afternoon is not really back if someone still has to sort half the load before Memo starts.