Nico

We built an autonomous robot in 4 weeks! Solar is the cheapest way to generate electricity, but operations do not scale. We bring automation to the solar market to solve this problem. Come see us this afternoon at Founders Inc Fort Mason for a live demo!

@RemiCadene @UMA_Robots Hardware in Paris is the best hardware

Starting with the fundamentals Prototype Version 0 AI, Software, Hardware A small team, 9 months Designed and assembled in Paris at @UMA_Robots

@johnfurr538500 Are you printing PLA or PLEC? Have you maximized the % infill?

No pain no gain Cracks in the battery tray. New part printing as I type. I think the current wheel boxes are a design flaw. Too much stress in the center. Thinking it needs to be single connected box vs disjointed. Anyone else have thoughts or suggestions?

@eptmpa Finally your own space! Enjoy!

Moved to a new location this week, printing will commence shortly

@johnfurr538500 Awesome congrats!

The B2 left the lab today. Pure Neural Networks RL trained balance bot. Powers through hills, grass, and maneuvers around patio. Still prototyped and ugly, but proof of platform is coming along nicely.

@sirwart Interesting thank you!

@NicoRobotics The LeRobot format is fundamentally columnar, whereas this format is row-based. I believe row-based is better for training (you want to lookup all data for a sample at the same time, not a single column across a wide range of time)

There's a lot to like about this paper, but my favorite part is the data format used for training. I adopted it and it's a big improvement over what I was doing before. If you want to try it out, I pulled it out into it's own library here: github.com/tetra-dynamics…

Tried the Franka Robotics dual arm teleoperation at Europe Embodied AI in Munich! Expensive but smooth as butter

@dAmineKharrat In my opinion KUKA is the easiest as you can find a lot of open accesses or easy to use tools, then ABB and then Fanuc

@NicoRobotics How complicated was it to get it running on ABB? We had complicated setup to do similar on Fanuc and Kuka.

Teleoperating gentle giants Definitely not built to be controlled by a SO101 but life finds a way

@dAmineKharrat We use EGM (externally guided motion). Theoretically 250 Hz but with Inverse/Forward Kinematics loop we are slower than this

@NicoRobotics Wow! What is running on the ABB controller ? What frequency is this running on ?

@alanchanguk Couldn’t agree more, great work! We should chat!

Today we're starting something new at Fuse. We're building robots. First, why this matters, because it's easy to get wrong. Skilled trades are ageing out faster than they're being replaced. The average electrician is over 45, the average plumber is over 50, and every year we lose more of them to retirement than we train. At the same time the world needs far more of them, not fewer, to build for the surge in energy demand ahead of us. You cannot build power plants, grid and AI data centres without them, and there are not enough of them. So we're attacking this from both ends. We're building a training centre in Birmingham, the first of many: untrained human in, skilled tradesperson out. We grow the workforce. And we're building robotics to amplify what our technicians can do. We don't have enough skilled people as it is. The goal is to make them more productive. Our technicians are at the core of Fuse's mission of acheiving low cost energy and energy abundance. Why we think we'll win where standalone robotics companies have struggled: 1. We already generate the data robots need. Robotics is bottlenecked on real-world physical data, and you cannot scrape it off the internet. Our technicians do real physical work every day across building power plants, electrical work and energy hardware installs. Only a handful of companies have real physical data. Rarer still to have it across this breadth of skilled trades, on sites you own. 2. We can deploy fast, and we learn faster. Standalone labs spend years convincing big customers to let them deploy on site, often against legacy resistance. We own the worksites, so we build, test and deploy on real Fuse jobs from day one. Continuous deployment means an extremely fast learning rate: more deployment, more data, faster improvement. This is greenfield. Everything is open for the new team to define, from initial use case to policy choice to the embodiment itself, including whether we build in-house or partner. We're open to working with the best robotics and world model companies out there. The team will work directly with me. Energy is the bottleneck for AI. Our goal is to unleash it, and robots are how we make sure it never becomes one again. We're hiring the founding team now. If you want to build robots that ship into the real world from week one and make skilled trades more productive, come talk to us.

Hosting a research hackathon in two weeks in Paris Pull up :)

A must read! As a hardware founder, I'm fascinated how the Chinese economy managed to pull out such cheap EV at this quality. Curious to see if it will be the same with humanoids, and if the intelligence software will be Chinese as well theguardian.com/technology/202…

We'll organize our next event very soon: luma.com/Paris-hardware…!

6 months ago we started "Paris Hardware Meetup" after meeting the team behind the SF ones. 400 registrants at each event, the energy and desire to build is incredible. Today someone texted me "Found a job at your meetup, thanks for organizing it." What an incredible feeling!

@johnfurr538500 Would be a nice power upgrade! Might try it! Thanks!

@NicoRobotics These are the wheel specs I ordered: ebay.com/itm/4058645634…

B2 wheel test with new Vesc and new chassis. Prolly another week or so to first balance test.

@johnfurr538500 Nice, I've been using BY1016Z motors, 350W, would love to see how yours compare

@NicoRobotics I found them on ebay from a Chinese supplier. Took a few weeks ro arrive.

@fayezsalka Impressive, what model are you using?

I wonder what those arms are capable of 🤔 (Realtime and autonomous)

@PTrubey Thank you for your post! I was thinking around the rack to rack link. Current laser link bandwidth I think is around 100 Gbps. I think SpaceX is aiming for 1 Tbps. On ground we have 50-100 Tbps. But not sure it matters with inference. Any thoughts?

Oh my, my dunk of JerryRigs is going viral. Well, let's use this as a teaching moment. First, realize when people say "data centers in space" they aren't talking about lofting up giant Costco sized buildings. SpaceX and Starcloud are proposing satellites that each have the compute capacity of about one AI rack, or what the guy is pushing in the picture below. These individual sats won't be connected together in space to run large training jobs, they'll only be used for inference - answering people's questions, running agentic tasks, etc. So each satellite has relatively tractable power and cooling requirements. There will be a couple of largish solar panels attached to give it 24x7 cheap power (remember that you get like 5x more solar energy per panel in space than on Earth). And a smaller radiator that will radiate away waste heat into the vastness of space. Both the power and cooling technologies are simple, well tested and cost nothing to operate, unlike power and cooling on earth. In particular, cooling on earth requires extra power to run powerful water pumps to move fluid all over the place and then to dump the heat into a relatively hot atmosphere. Yes, space based cooling can only reject heat via radiative cooling, but it is doing it in the vacuum of space at -454 °F (-270 °C, 3 K) versus about 77 °F (25 °C, 298 K) on Earth, so that helps a lot. Point being that cooling in space has only a single upfront cost of building a passive radiator. But what about the overall cost, you ask? Well, think about all the things you don't need to build now. That rack the guy is pushing around weighs 1,400 pounds mostly because of all the metal required to support everything against Earth's gravity. Things can be built far more flimsy in space since they are in zero gravity. Also that rack has a bunch of power electronics and fans, neither of which are needed in space. Indeed, that entire building those racks sit in doesn't need to be built. All that fiber cabling isn't needed (lasers in space take the place, no need for cables). Giant utility transformers and a small army of step down transformers and battery packs don't need to be built. The land doesn't need to be bought. The permits don't need to be acquired. The supposedly huge amount of water used doesn't need to be provisioned (it's a tiny amount, but the detractors love to bring it up). There are in fact giant cost savings going into space. What about launch costs? That is small as well. Starship is fully reusable. The majority of launch costs are natural gas and liquified oxygen extracted from the air. That's it. Cheap access to space, really cheap I mean, is a huge unlock. I was initially shaking my head when I first heard about Elon's "crazy" idea of space based compute, but the more you look into it, it is far less crazy and more doable and practical. At least for SpaceX.