CMeta reportedly built a humanoid robot with a Chinese company in about three months. Version two is reportedly already coming.
MWho's saying that?
CTechniahqrobot on X. The Information reported it. The robot is called Max.
MSo a social media account is giving us the news. Fine. What does Max actually do?
CIt's roughly five feet and a hundred pounds. Five-fingered hands, speakers, a screen face for expressions.
MCute. That sounds like a toy.
CIt runs on Nvidia's Jetson Thor chip. Meta says it's for internal research, not for sale.
MRight. A research project. So why does it matter?
CBecause Meta partnered with Fourier Intelligence, a Shanghai humanoid maker. And they moved fast.
MA Chinese hardware firm and Meta's AI. That won't set off alarms anywhere.
CI think that's the biggest part of the story. Meta is trying to give its Muse AI agent a body.
MBut it's not a product. A prototype in a lab isn't owning anything.
CThree months to a working humanoid. With a partner that already builds them. That's real speed.
MSpeed to what? They won't sell it, they won't deploy it. It's a lab toy tied to a Chinese supply chain.
CThat's exactly my point. It's a hedge. If Tesla and Figure stumble, Meta can move.
MNo, no, wait—the hedge is the problem. You don't hedge by handing a Chinese firm your AI hardware roadmap.
CWhy not? Fourier already makes GR-series robots. Meta isn't teaching them AI; it's buying hardware know-how.
MBuying know-how from a country racing to own physical AI. That's a transfer, not a partnership.
CSo Meta should build servos from scratch? Apple works with Chinese manufacturers constantly.
MApple builds phones, not robots that walk and grab. A robot in your house knows your house.
CIt's not connected to Facebook accounts. It's research, Matilda. Lab benches.
MAnd they're reportedly already working on Max 2. Lab project, sure. They're moving like a company with a plan.
CI'll give you that. The pace says serious. But I'd rather Meta be in this race than sit it out.
MI'd rather they build with a partner that won't end up on an export control list.
CThere's no export list for robot hands yet. Give it a year.
MDon't laugh. The defense world is already watching this space.
CSpeaking of watching, there's a post going around with video of an American humanoid and a Chinese one jumping.
MJumping? So we're at robot Olympics?
CKrosh posted, "The winner in this comparison isn't the one who jumps higher."
MWhat is the winner, then?
CAccording to Krosh, the gap between the countries isn't movement. It's numbers.
MWhat numbers?
CHe cites Omdia, the research firm, with a figure beginning about 13,000—but the post cuts off and it can't be verified.
MSo an unverified number that's cut off mid-sentence is doing his arguing for him.
CThe thread's point is manufacturing scale. Demos are easy, production is hard.
MThen he should say which robots and where the number's from. Give us a source, not a cliffhanger.
CFine, but the point stands. We've spent a lot of time on Meta's lab toy. Krosh says China wins on production.
MThat's my worry! Meta's lab toy is a Chinese partnership, and the claim is China is scaling production. Both can be true.
CExactly! So why are you mad at Meta for racing?
MBecause racing with a Chinese manufacturer is how you lose the home-field advantage!
COkay. You don't think Meta should be in hardware at all, do you?
MI think Meta should focus on AI. If they're serious about robots, build with a Western partner or go alone.
CGo alone with what factories? America doesn't have Fourier-level humanoid capacity yet.
MThen build it. That's what Tesla and Figure are trying to do. Spend the money.
CBut building it yourself takes longer. Meta got a robot in three months. That's the argument for partnerships.
MAnd it's also the argument for how we lose the lead. Fast and borrowed isn't a strategy, it's a leak.
CYou'd shake up a Meta shareholder call with that line.
MThey'd probably ask why the robot has a face. There's another thread from ElinaNotes on this.
CWhat's her angle?
MShe says the humanoid race isn't about demos anymore, it's about AI intelligence, manufacturing scale, reliability.
CThat's the same argument Krosh made. But she actually breaks down approaches. Figure bets on the brain with Helix VLA.
MI like that she calls it Physical AI. But two of her three approaches are the same problem at different stages.
CSo you're not impressed.
MI'm impressed by verified manufacturing numbers. Not threads with cliffhangers.
CFair. My take is the real story is everyone attacking the same problem from every angle.
MAnd half the angles run through China. That's the fact that matters.
CIt matters. But it also means America has to partner somewhere. We can't be purists.
MI'd rather be slow and secure than fast and hollow.
CMax is a sign the race is real.
MIt's a sign the race is messy. That's all I'll give it.