Tesla showed Optimus folding laundry and I have

Tesla posted a video of Optimus folding a shirt.

The internet’s response was predictable. “That’s it?” “My grandma folds faster.” “Two years of development for this?” Memes. Mockery. The standard tech Twitter treatment for anything that looks underwhelming in a 30-second clip.

I watched the video and saw something else.

The robot picks up a shirt from a table. It identifies the shirt as a shirt (not a towel, not a rag, not a blanket). It locates the edges. It makes a series of fold decisions. It executes those folds with two hands that have to coordinate pressure, position, and timing simultaneously. The result is a folded shirt. Not a well-folded shirt. A folded shirt.

That’s not nothing.

Why this matters

Folding laundry is trivially easy for humans. We learn it as children. We do it while watching TV. We don’t think about it.

For a robot, folding laundry is one of the hardest manipulation tasks that exists. Fabric is deformable. It doesn’t hold its shape. It drapes, bunches, wrinkles, and slides. Every shirt is different. The folds depend on size, material, and shape, all of which vary. There’s no rigid geometry to grab onto. No handles. No standard form.

Industrial robots handle rigid objects brilliantly. Car parts, circuit boards, boxes. Things with known dimensions and predictable physics. Give a robot arm a bolt and it’ll place it with micrometer precision every time.

Give a robot arm a t-shirt and it’ll crumple it into a ball.

That’s the gap. And Tesla just demonstrated a general-purpose humanoid robot bridging that gap, even if the bridge is rickety.

The Tesla pattern

The first Tesla Roadster had a 200-mile range and cost $109,000. It was slow to charge. The build quality was rough. The company nearly went bankrupt. Car enthusiasts mocked it.

The Model S came next. Better in every way. Then the Model 3. Then the Model Y, which became one of the best-selling vehicles on Earth.

Tesla’s pattern is: ship version one embarrassingly early, iterate relentlessly, scale to millions. The first version is always a target for mockery. The version five years later is a product that reshapes the market.

I’m not saying Optimus will follow this pattern. I’m saying the pattern exists at Tesla, and writing off version one because it folds laundry slowly is the same mistake people made about the Roadster.

What I want to see next

Speed doesn’t matter yet. Precision doesn’t matter yet. What matters right now is generalization. Can Optimus fold a towel after learning to fold a shirt? Can it sort light and dark clothes? Can it handle fabric it’s never seen before?

The difference between a scripted demo and a learning system is the thing to watch. If every new task requires months of engineering, Optimus is a very expensive laundry folder. If the robot can transfer skills between similar tasks, then folding laundry is just the first demonstration of something much larger.

I’ll be watching. Skeptically, because Tesla oversells timelines like nobody else. But watching. Because the company that figured out how to build a million cars per year in a factory full of robots is probably the company to bet on for building the robots themselves.

The shirt was folded. Badly. But it was folded. And the robot that folded it will be better next month, and better the month after that.

That’s the part the memes miss.

Related thinking: