Read more from @alexrkonrad here.

Thank you to our friends and investors including @ericjang11, @xiao_ted, @radbackwards, @animesh_garg, @suzannexie, @apartovi, @conconconling, @mrjasonchoi, @pythianism, @rajivpoc, @Rewkang, @alyriadefi, @svangel, @A_StarVC, @BennettSiegel, @kevinhartz, @offlineventures,.

The inspiration for our name Mecka here are "mechas" from Japanese science fiction which are giant humanoid robots piloted by human fighters

Until robots enjoy the benefits of fleet-scale learning, with millions of robots deployed, we see the 8 billion humans as teachers to physical AI from how to walk this Earth to how to make an espresso!. By then, we'll be far along our goal of offering the pot and kitchen too!.

We've proven this out with great partners in the past year and have built several datasets in the tens of thousands of hours as our early bets become popular belief. But egocentric data is not enough by itself.

The naive common-sense hunch, along with wonderful research in the community, is that a humanoid and a human share a form factor. The question was whether we could transfer the knowledge from less noisy human demonstrations in a POV video to a robot at large-scale

The data soup in robotics is made up of many key ingredients, including tele-op, synthetic, and internet-scale data. Across high-quality but slow teleoperation data with fast yet lower-quality synthetic simulations, we made a bet on human data, starting with egocentric videos.

General-purpose robots will be the largest GDP impacting technology of our lifetime with a $42T injection of productive labor through humanoids and robots of many shapes and sizes!. But robot models, among many other challenges, face a data issue

Excited to announce our $8M seed round led by @neo to further our work as the data layer for robotics. Heres why we bet early on human data advancing autonomous robotics:

The teleoperation vs large-scale video learning argument in robotics will play out the same as LiDAR vs vision-only in autonomous cars.

no amount of money has ever bought a second of time.

There will robots of all types and shapes everywhere in the future.

bet on trajectory over position in time.

RT @elonmusk: @karpathy Andrej, my long lost brother, let us work together again!.

type nights

schools need their teachers to be expert ai chat users so they can teach about the experience, and considerations, of co-existing with AGI. this is likely not to happen in public schools for a few years. old dogs don’t learn new tricks and their jobs are protected. instead, the.

grok and gpt needs a student mode default on based on age that prompts the user to think from first principles like a teacher - the infinite answer machine is dangerous. i’m at an age where i’m out of school and i don’t have kids yet, so i am far removed, but it feels like we.

Japanese robots are coming

yes.

This feels like the modern Age of Exploration - “GPT-1 of robotics is already somewhere on arXiv”. We sail a vast, unmapped ocean, driven not by certainty, but by the thrill that something beautiful lies just beyond the horizon.