Rosie the Robot made it look effortless in 1960s TV reruns. Sixty years later, folding a single T-shirt remains one of the hardest problems in robotics.
It seems absurd. We fold clothes without thinking—grab a sleeve, find the collar, align the edges. But that intuitive geometry, learned through years of handling different fabrics, is precisely what robots can't replicate. A shirt that lands in your laundry basket one way will land differently the next time. A robot trained on flat, wrinkle-free images of clothing sees each crumpled variation as a completely new puzzle.
"It's not the fabric itself that is the challenge," says David Held, a robotics researcher at Carnegie Mellon University. "It's the amount of variations that can be created by the way fabric can be crumpled, and all the different kinds of clothing items that exist."
We're a new kind of news feed.
Regular news is designed to drain you. We're a non-profit built to restore you. Every story we publish is scored for impact, progress, and hope.
Start Your News DetoxFor years, robots relied on what's called "pick and place"—a predetermined sequence of moves to manipulate fabric. The problem: soft fabric doesn't follow a predetermined plan. It crumples. It distorts. It refuses to cooperate.
A New Strategy Takes Shape
That's where AdaFold changes things. Developed by researchers including Alberta Longhini (now starting postdoctoral work at Stanford), this newer algorithm doesn't commit to a single folding path. Instead, it watches the fabric as it folds, constantly adjusting its approach based on how the cloth actually behaves—its elasticity, its shape, the way it responds to each movement.
It's the difference between following a recipe exactly and cooking like someone who tastes as they go. The robot monitors progress at each step, recognizes when its intuition was wrong, and corrects course. For humans, this adaptive thinking is automatic. For robots, it's still genuinely complex.
We're not at the point where robots are folding your laundry tomorrow. But the shift from rigid "pick and place" to adaptive algorithms suggests the gap is narrowing. The real breakthrough isn't a faster robot—it's one that can think more like we do: flexibly, responsively, learning from what it observes rather than what it was programmed to expect.
When that happens, Rosie finally gets to do her job.
