Machenta taught a 3-robot assembly line in two minutes
Edward Johns says his team set up a three-robot packaging line in two minutes using one demonstration per robot, with a model trained entirely in simulation that had never seen the objects. The same day, a depth-only foundation model for robots landed with weights and code.
Shown this week
Machenta taught three bimanual robots a packaging line in two minutes
A week after coming out of stealth, Machenta's Edward Johns posted a three-robot assembly line he says took two minutes to teach, where integration normally takes weeks.
None of this reactivity and recovery needed to be shown in the demonstration: it all comes from the intelligence of the pre-trained model.
DeFM is a foundation model that sees only depth, and it shipped
Manthan Patel released DeFM, a foundation model (a big pre-trained backbone other systems build on) built for depth images instead of RGB, with weights, code and a demo out the same day.
The only input is a depth image. Yet stairs, railings, walls and car parts each get their own features.
Out in the world
Lukas Ziegler says Europe's Micro-AGI reached $100M ARR
Lukas Ziegler says Micro-AGI is the fastest company ever to reach $100m ARR, and that it is a European physical AI company. He spoke with Artjem Weissbeck at NVIDIA GTC, and says the company started in data collection and now does models, hardware, data and research end to end, while most of the field specialises.
They’re buying a working deployment, with safety and accuracy someone has to sign off on.