MOSS, the open source litter-picking robot, is taking shape 🤖 First prototype, first reality checks. Some components ended up elsewhere than planned in CAD, and the Jetson is now running on battery power. Plenty to improve in the next revision. That’s what prototypes are for.

welcome 🫡

The Rover moves continuously through the vineyard, recording changes with a basic, affordable camera and other sensors. Each pass adds dated observations to a shared record. We use it to build and check reconstructions like this one, for people and AI to work with. #robotics #PhysicalAI #AgTech #precisionAgriculture

There are a lot of partnership opportunities in embodied AI right now: hardware grants, cohorts, residencies. I love it. The synergies are real and they turn projects into real products faster. But please 😀, why does every one of them come with an arbitrary constraint? Use this specific hardware. Deploy X in Y weeks. Build in this specific way. Founders already know what their project needs to move forward. Ask them what the bottleneck is, and fund that. I'm saying this from inside a few of these programs, and happy to be there tho

Some rovers can even pick up trash! 😉 👊

Starting today, I’ll be sharing the partnerships and collaborations we’ve been working on over the past few weeks. First up: I’ve been selected to join the first @OpenAI Codex Physical Builds cohort, alongside builders from around the world! @OpenAIdevs will connect the 35 participants, provide a hardware kit and compute support, and help us create project videos and reach more people through their Substack, Instagram, X and LinkedIn channels. For MOSS, that means more support to keep building, testing and sharing the process, from simulation to the real robot.

Proud to announce I've been selected for the first @OpenAI Codex Physical Builds cohort, with 34 other builders from all over the world. What it means on the bench: some hardware, compute to keep the build going, and @OpenAIDevs helping us film and share what we make, on Substack, Instagram, X and LinkedIn. For MOSS, that's a more affordable Raspberry Pi version of the rover open-sourced by @SHOW_ROBOTICS and built in synergy in front of the other participants in the OpenAI program. More collabs and partnerships to be announced.

Robotics is starting to get really interesting in the best possible way. A hand with zero rotary joints. A robot learning from its own failures. Robots that could build robots. An open-source rover cleaning our streets, and @tnkrdotai We're shipping an entirely new experience for electronics in your repository. Featuring @Boyuan__Chen @dm_robots @DominiqueCAPaul @hiroindustries and @metrox_eth @tnkrdotai

astra rover update: determined that Onshape's API billing model was NOT going to be sustainable long-term, so i had codex download Fusion on my MBP and rebuild the head assembly this video shows abt 2.5hrs of work (astra MAX) and the end result

🎥 Day 4 Challenge: Dual RGB + Depth Build something inspired by dual RGB streaming alongside the full depth pipeline. Ideas: 3D scene understanding, object inspection, robotic picking, spatial perception, or mixed reality. 👁️📡 🎁 Best project video wins a @RealSenseai D421. realsenseai.com/perception-s… #NationalCodingWeek #RGBD #ComputerVision

By 2030, 1 in 6 people will be over 60 and providing nursing care will become challenging. The winners of Dimensional track at @HackMIT turned a wheelchair into a robotic caregiver that helps people bathe, eat, drink, manage medication and more. Open source code ⬇️

I have found that a few practical questions reveal a lot about a robotics team's simulation workflow. How long does it take to recreate a new customer site? Which physical properties are measured, predicted, or simply assumed? How much engineering work happens before policy training can begin? The answers usually show whether the bottleneck is asset creation, site specificity, or confidence in the physics. If you are deploying robots for manipulation, picking, or another physical interaction task, I would be glad to compare notes. You can send me the object, site, or workflow that is causing the most friction.

Yesterday I assembled the tracks on MOSS, my open-source litter-picking robot running @dimensionalos. It rolled for the first time! Jev looks promising in simulation, but I can’t wait to test the full loop on the real robot: detection → decision → actuation. Assembly revealed plenty of room for improvement. I’ll keep building and testing this prototype over the next few days, simplifying the design, choosing components better suited to the job, and trimming the BOM cost wherever I can. The really cool part? I can document changes as I go and hand them to Codex to work on the next version while I keep building. Here’s the V0.4 animation. We’re getting close to releasing the CAD files and BOM so you can start building yours. Build, break things, iterate. Now in parallel.

Big update, so many features will improve MOSS and VECTOR. Exciting !

dimOS v0.0.14 is out 🚀🦾 - A browser UI to control your robot: camera feed, 2D costmap, WASD driving, and a chat box for the agent. - Web SDK for building your own pages. - Zenoh is the default transport now. - A new evals framework to score agents against recordings and live sims. - 3D navigation and ray tracing for dynamic voxel clearing required for global maps. - Record every topic of any blueprint to replay, visualize, or share. - New robots: Galaxea R1 Pro, Deep Robotics M20, Boston Dynamics Spot (experimental), OpenYAM, A1Z. github.com/dimensionalOS/dim…

The last two days have been wild. My open-source litter-picking robot went a little viral, and my follower count tripled 🤯 Thanks for following along! It happened so fast I couldn’t keep up with the follow-backs. Drop a comment on any of my posts so I can find you and follow you back.