Meet the robots learning to survive real warehouse chaos
Nomagic's North America chief explains how AI-guided robot arms are moving from controlled demos to the messy reality of order fulfilment, and what that shift means for the warehouses that pack your online orders.
Key points
- Nomagic, a robotics company specialising in AI-driven order fulfilment, is expanding into North America under general manager Josh Cloer.
- Cloer was a founding team member of Mujin's North American operation before joining Nomagic.
- Google DeepMind's Gemini Robotics 2 can now coordinate a robot's full body, not just a single arm.
- US defence contractor HII signed a deal worth up to $900 million with Path Robotics and GrayMatter Robotics for automated welding.
- The University of Florida opened a robotics lab focused on construction automation.
Picking the right item off a shelf sounds simple. For a robot arm, it's been one of the hardest problems in engineering.
That gap between the tidy lab and the chaotic warehouse floor is exactly what Nomagic, a European robotics company, is trying to close. The company builds AI-guided picking systems: robot arms that can identify and sort items in a fulfilment centre, the kind of facility that packs your online orders. Josh Cloer, Nomagic's general manager for North America, laid out the company's approach in Episode 256 of The Robot Report podcast.
What makes warehouse robots so hard to get right?
Warehouses defeat simple robots because the real world won't stay tidy. Boxes arrive crumpled. Products shift in transit. Lighting changes. A system that works on a test bench can fail the moment a squashed parcel hits the conveyor.
Nomagic's answer is what the industry calls physical AI, meaning software that learns from physical experience rather than following fixed rules. The robot adjusts and improves each time it handles a new object. Cloer describes the goal as bringing a "new level of intelligence and adaptability" to picking, not just raw speed.
He isn't a researcher. He's a commercial operator who spent years scaling Mujin, another industrial robotics firm, across North America as a founding team member before joining Nomagic. His job is turning the technology into contracts.
What else happened in robotics this week?
Three other stories from the same episode are worth a quick look.
| Story | Who | What | Scale |
|---|---|---|---|
| Welding automation deal | HII, Path Robotics, GrayMatter Robotics | Agreement to deploy automated welding systems | Up to $900 million |
| Full-body robot control | Google DeepMind | Gemini Robotics 2 coordinates whole-body movement | Research announcement |
| Construction robotics lab | University of Florida | New lab dedicated to building-site automation | Single facility opening |
The HII deal is the largest of the three by a wide margin. We reported it the day before this podcast aired, in our story US Navy shipbuilder commits up to $900 million to robot welding and sanding firms. HII is a US defence and shipbuilding contractor, and the agreement signals that heavy industrial welding, one of the most physically demanding trades, is now a serious target for automation.
On the Gemini Robotics 2 announcement: earlier robot AI typically controlled a single arm or gripper. Full-body coordination opens the door to machines that can work in spaces built for people. We covered the details on 30 July in Google DeepMind's New Robot AI Can Now Control a Full Body, From Feet to Fingertips.
Common questions
Will warehouse robots replace warehouse workers?
The industry's honest answer is: some jobs, not all. Picking and sorting roles are the most exposed; roles involving maintenance and judgment about edge cases are far less so. The pace of change depends on how quickly the technology proves reliable outside controlled conditions.
Does any of this affect what I pay for online deliveries?
Not directly or soon. Automation can lower fulfilment costs over time, but those savings have historically gone to shareholders or subsidised faster shipping rather than lower prices for shoppers.
