I just watched GPT-6 Astra design a PCB directly inside KiCad. The model was shown taking an electronic schematic and turning it into an actual PCB layout — positioning components and routing the copper traces needed to build the board. And that’s more significant than it might sound. PCB design involves a lot of repetitive and precision-heavy work, from component placement and trace routing to clearances, design rules, and manufacturing constraints. If AI agents can interact directly with engineering software like KiCad, they could potentially automate a large portion of this process. But there’s a big difference between successfully completing a demo and reliably producing hardware that’s ready for manufacturing. The real challenge is making sure the final board meets electrical, mechanical, thermal, and manufacturing requirements. So maybe the future isn’t AI replacing hardware engineers. It could look more like: Engineer defines the requirements → AI handles the repetitive l...