A NEW LINE IS NOT A GUILTY VERDICT 🏭
Fremont went from a few dozen Optimus units a week to several hundred, with talk of a thousand by year-end. Most stay inside Tesla, on scripted tasks, and the AI still fails when the task changes. By January 2026 none were doing useful work. That part holds. 🤖
It does not mean the job was never the point. Early units teach the line. Tesla already did this with the Model 3. The lease fits the same fact: the V3 hand still has over a hundred small parts, some assembled by hand. Sell it and every fix is a recall. Lease it and you can pull the machine and change the hand. The logs come home either way. 🔧
Half a million hours are the cost of a field the internet cannot train, not proof the robots are decoys. A recording project does not need a thousand-a-week target. The October memory cut was how to get chips at all. And if actuators are up to 60 percent of the cost and scarce, extra robots are a bad way to buy video. 📹
The sharp part is the timeline. Thousands of useful robots by the end of 2025 became zero. A lease customer and an unrehearsed task are still the tests. ⏱️
THEY’RE TAKING SIDE GIGS FROM 18-YEAR-OLDS TOO 😱
I’m watching a video where a metal hand scoops popcorn and suddenly shapes it into a heart — and I feel a slight shock. Not from the gesture itself. But from how quickly everyone around is ready to shove money into the slot. “So cute, let’s tip him!” people write, as if there’s not a mechanism standing in front of them, but someone who’ll go home after the shift with heavy legs. 💸
Something tightens inside me from this. Because tips have always been about fatigue, about patience, about someone actually spending their day on you. And here it’s just a precise movement, a little spilled popcorn, and a neat heart in response. No emotion, no need. And yet we’re already reaching for our wallets. 🤯
If this robot were standing right next to me, I’d probably freeze for a moment and smile — the gesture was genuinely pleasant. But money? No. Because in that moment it gets a little creepy: we’re starting to pay algorithms for successfully imitating us. 🤖
THEY PROMISED THIS ROBOT WOULD TAKE YOUR JOB. INSTEAD IT TOOK YOUR ORDER 🍿
Optimus at the popcorn machine, scooping again and again until the box is full to the top, like a guy on his first shift who really wants to keep this job.
Now look at the counter. Kernels everywhere. That is how you know he is a real employee.
Then stay for the last three seconds. He hands the box over, the guest holds up half a heart with their fingers, and he finishes it with his hand. A robot doing a finger heart with a customer.
I worked a summer behind a café counter and nobody ever made me a heart. Mostly they told me the coffee was cold.
So here is where I land. The article says the robots already showed up. Fine by me, as long as the first one on shift is the one making hearts at the popcorn machine. Would you tip him?
YOUR NEXT COWORKER JUST CLOCKED IN FOR $30,000 - AND IT DOESN’T EAT LUNCH 😱
A humanoid used to be a clip you close. It matters when the machine shows up as a product. Figure’s robot at BMW didn’t do a demo. It worked 10-hour shifts, logged over 1,250 hours, and sat in a line tied to more than 30,000 cars.
The price moved too. Build costs fell fast enough that $30,000 no longer sounds absurd, and simpler models have appeared near $16,000. Not a finished worker. Just hardware leaving the toy shelf.
Factories don’t pay for tricks. They pay for one move repeated without stopping the line. Buy the robot once, pull hours out of it. If a useful hour gets cheaper than hands, nobody waits for a revolution. They place an order.
Homes can wait. Warehouses won’t. Cool not because they’ll outthink us -because the metal can already finish a shift, and the price no longer sounds like a house.
$38B market. $16K humanoids. 30,000 BMWs. Robots rolling off production lines by the hour.
The humanoid race isn't about who builds the coolest robot anymore.
It's about who makes human labor economics stop making sense. https://t.co/gF86ESM751