Achieved 99% zero-shot laundry folding across 785 attempts. Demonstrated oneshot learning emergence from scaled pre-training. Pursuing physical AGI via parallel skill training; home robot form factor agnostic.
no scored calls yet — needs a stated position or a categorical verdict, with a matured window vs SPY
Scaling pre-training data and compute enables oneshot learning—robots can learn new folding styles from one demonstration and generalize to unseen garments/environments—marking a step toward physical AGI where thousands of home skills can be trained in parallel.