Today we’re out of stealth, and we’ve raised a $40m seed round co-led by @creandum and @usv, to bring Instant Robot Learning to life. For the past few months, we‘ve been busy building a new company. Machenta. This follows results from a new kind of robot foundation model we’ve built, which I’m so excited to share with you today! Usually, it takes hours/days of data collection and training to teach a robot. We’ve built a model that learns tasks instantly: one demonstration, no further training. A model trained entirely in simulation, without a single pixel of real-world data. A model where teaching robots is so quick and easy, it feels like physical vibe coding. A model for robots that learn like humans do, from someone like you. Welcome to the era of Instant Robot Learning. Thank you @Kathi_Wilhelm, @rebeccakaden, @hkanji, @jonno_evans and all of our angels for your support! @Machenta_ Learn more: machenta.com/blog/instant-ro… See below for what you can do with Instant Robot Learning!

Thanks @MTSlive for the chat last night about Instant Robot Learning! Here's a little clip about why we're excited at @Machenta_ to be building AI for robots that learn instantly.

London has a new Physical AI Lab. 🔥🇬🇧🇪🇺 And they have a legit claim to be soon a world leading one. What MACHENTA does is close to heretical. It's a completely different approach to building models, that doesn't require massive amounts of data, nor post-training, but can repeat a task after showing it only once. It's a complete narrative violation to everything we know in robotics right now. No absurdly large vision models. No hours of post-training needed. But it works instantly. You show it. It can repeat it. Not 20 video takes and then keeping the one that worked. You show it. And very very very likely it can repeat that new task after only that one demonstration. Their technical approach is also fundamentally different. Heavy focus on in-context learning, explicitly 3D, and tons more that i am unsure if it's public information. They is a non-zero chance that they will unlock autocomplete for the world. Do a task. Click a button. Get out of the way and the machine repeats it until it runs out of material. And yes, one more time. This is real. I have seen it multiple times myself (Investors since over a year). You demo it. And that's it. It can repeat it almost every time after only that one demonstration. 🤯 Kudos to @Ed__Johns and @vitalisvos19 for launching! PS: If you want to join a breakout Physical AI lab -> Machenta is likely they one to apply to right now. x.com/Ed__Johns/status/21150…

Fun times at ETN today — we made it to #1 on the funding leaderboard! (Small caveat: we’re the first company they’ve put up there 🤷‍♂️) @Machenta_

On the back of our @Machenta_ launch today, I'll be joining MTS live today at 8:45pm UK, 1:45pm PT. Join us at the link below to hear all about the world of Instant Robot Learning!

Thanks Ted! Yes, test-time steerability with a physical example is often essential to show what you really want a robot to do.

Thanks so much Michael! Happy memories talking about our early research on this with you on RoboPapers!

Big congrats to @Ed__Johns & team! Ed & his lab has done some of the most impressive work on super data efficient robot learning. Super bullish on them!

Thanks for the great summary — we're very excited to be bringing USV into the wider Machenta team, let's go!

Big congrats to @Ed__Johns and the entire team, in-context learning for robotics from sim only while adapting to real deployments, that's impressive! 🥳 Huge for UK and Europe too!

Glad to back @Machenta_ ! 🚀 Teach a robot a new task with a single demonstration, no further training. Powered by a foundation model trained entirely in simulation. “Vibe coding in the physical world” is an exciting vision. Now out of stealth!

Teach a robot with only a single demonstration. Key to making robots into useful tools for a wide range of work! Really cool launch.

A new AI lab has just emerged from London and it has raised a $40m seed. It is the largest ever robotics seed in the UK and the company has built a foundational model designed to achieve instant learning. i.e. show a robot something ONCE and it will learn how to do it. It was founded by a Professor of Robot Learning at Imperial College London (@Ed__Johns) Their model is trained entirely in simulation, so instead of learning specific tasks, it learns how to learn. Then when you give it one real demonstration, on objects it’s never seen, it picks the task up instantly. This is VERY COOL and it is amazing to see more and more frontier AI companies coming out of London. Huge congrats to the whole @Machenta_ team

Very exciting bet: in-context learning from one demonstration, pretrained entirely in simulation. The name “Vision-Demo-Action model (VDA)” is also a perfect fit.

Physical vibe coding is here. @Machenta_ brings instant learning to robots: show a robot a task once and it gets to work. You prompt it with a demonstration the way you prompt ChatGPT with text. While most of robotics races to collect more real-world data, @Ed__Johns and @vitalisvos19 spent the past years at @imperialcollege on the opposite bet: a model trained entirely in simulation that learns how to learn. That means it’s bounded by compute, not data, and it will only get better as it scales. We believe the hardest problems are often solved by people who bring a different lens. Ed is a true polymath: after his PhD, he spent a year as a singer-songwriter, playing gigs with a then still unknown Ed Sheeran. In a band, when someone shows you a chord, you rarely need to see it twice. Perhaps that’s why Ed asked a different question from most of his field: not how to make robots learn more, but how humans learn so fast. At @creandum, we're proud to co-lead Machenta’s $40M seed with @usv, backing Ed, Vitalis and the team.

Great little summary of what we're building here — thanks Boris!

Another world-class robotics lab coming out of stealth, exciting results in in-context learning! Ed is one of the leading UK robotics researchers, I remember seeing him demo Instant Policy at a tiny room in his Imperial lab early last year, amazing to see the progress since!

Love the work on few-shot manipulation learning from @Ed__Johns and co. (In particular, I really love InstantPolicy and DINOBot!) Looking forward to seeing where this goes! 👀

Vitalis, it’s been such an honour to continue this journey with you — from where we started together at Imperial to what we're now building at Machenta. 🚀 You are the master of in-context robot learning. Let’s go!

Rebecca, thank you for your support — we're so excited to be building Machenta with you! Let's go!

Here we go...! 🚀