Kinematic magic! 🪄 TCP stays in place. Hypnotizing. The KUKA robot now has two more axes. The TCP point remains still & fixed regardless of the complicated trajectories and movements of the robot. An amazing demonstration by Mutz Maschinenbau of how important TCP calibration is. You must admit that a black arm from KUKA is for special duties! 😮‍💨 P.S. What do you think this application is all about? Any ideas? ~~ ♻️ Join the weekly robotics newsletter, and never miss any news →

Machine surfaces at machine speed. Lindsay Walker @hashgraph explains why agents need machine-readable data, not human websites, and infrastructure that’s fast, publicly verifiable, and deterministically programmable. From the @AIconference on Agentic Commerce

MachineFun was designed to remove the speculative friction from what we think will be one of the highest-growth sectors of the coming decade: Robotics. One of the most heavily funded corners of robotics over the past two years has been robot brains, the AI models that let a robot see, decide and move, with roughly $5B raised by its best-known developers since 2024. One company in this field just had a breakthrough. @dreamlabs_ai, a 12-person robotics lab, released Dream-1, a new kind of model they call a World Action Model. It learns how the world moves from video and uses that to control a robot through a simple API, meaning any developer can plug into it the way they'd plug into ChatGPT. That kind of tooling opens doors for hardware companies that were stuck before. @makermodsai, which sells $2k robot arms, says theirs was up and running in a few minutes and "attempting tasks we never programmed." Teaching a robot one task used to take specialists months, and now it's starting to feel as quick as prompting an LLM. Breakthroughs like this are happening almost daily in robotics, but for anyone outside the industry they've been nearly impossible to get involved in. Dream alone raised a $90M first round, and none of it was open to the public. We've seen where this leads. When AI models became something you could call through an API, a wave of smaller startups got built on top of them, and plenty of YC companies got funded that way. Robotics is starting down the same road. @physical_int, another lab building robot brains, recently opened its own API to outside teams. Almond, a company that builds a robot called Axol, used it to teach Axol to smooth rough edges off metal parts for its own assembly line, and it took only 4 days. Examples like these are why we built MachineFun. As bottlenecks unlock and hardware gets more efficient, businesses that didn't make economic sense before can turn into real revenue almost overnight. Letting teams raise publicly for their ideas is the first step toward an open source robotics hub. We know funding alone isn't enough, though, so we've built a foundry network with @peaq and @codecopenflow to back high-agency teams taking their ideas into the real world with peaqOS and SimArena. As foundation model, tooling, benchmarks and onchain economies keep scaling, MachineFun wants to sit at the center of it, with developers building in public and traders getting in early through speculation.

Machines break down. It’s inevitable. The goal of this teardown #robot# is to safely deconstruct products so they can be repaired or upgraded, minimizing the amount of waste produced.

Machine-to-machine payments are live on Arc Testnet. @tlay_io showed two physical devices negotiating and settling USDC payments on Arc every 10 seconds, with no human in the loop. One device broadcasts a live electricity price. Another decides whether buying power is profitable, signs a USDC payment, and settles on Arc. A concrete look at what the agentic economy can become when it moves beyond cloud agents into physical AI and autonomous machines.

Machine learning learns the same way you do: by looking at examples. It studies past data such as: • Weather • Photos • Words you’ve typed Then it looks for patterns and turns those patterns into simple rules the computer can follow. Those rules become your machine learning model.