On John Collison's Cheeky Pint podcast (Oct 29), Sam Altman talks about robots building data centers and power plants as the bottleneck moves from chips to construction and power. Talk, not a deployment, but a clear read on where AI operators say the constraint sits.
Schneider Electric's Michael Martinez argues that structuring plant data comes before putting an agent on the path to autonomous operations. Vendor content, but the sequencing point matches what operators hit when agents meet messy plant data.
Neuron's Cortex controller lets engineers describe machine behavior in plain English, get readable control code, and test it in simulation before it runs. At CureWood it replaced a 1991 Siemens S5 controller with 189 hardwired I/O points in about three hours of engineering and eight hours of downtime, and the finger-jointing line ran on first power-up. The PLC programmer shortage is the bottleneck on new factories, and this is AI aimed at that work.
Toyota introduced an AI welding robot, jointly developed with Yaskawa, at its GR Kart plant in Gamagori. Trained on welding images and audio, it checks parts with microphones and sensors before welding and cut roll-cage fabrication from up to three weeks to three days. Teaching a robot the welder's ear, not just the path, is a skilled-trades story.
