AI Parallels to Hardware Prototyping

Years ago, building hardware meant a garage full of equipment, a supplier relationship, and a tolerance for $10k mistakes. Now? A $4 ESP32 and a 3D printer in my bedroom take me from idea to working prototype in a weekend. It's genuinely incredible. The barrier to 0→1 has basically collapsed. But here's what cheap ESP32s and accessible 3D printing didn't do: → They didn't make it easier to maintain a product over years — to handle the edge cases, the firmware updates, the part that goes EOL six months after launch. → They didn't make it easier to decide what to actually build. A prototype answers "can I?" It says nothing about "should I?" → They didn't solve distribution. Printing one enclosure is trivial. Manufacturing 10,000, certifying them, shipping them, supporting them — that's still the hard part. I keep thinking about this as I watch the conversation around AI replacing software engineers. AI has done to software what cheap hardware did to prototyping: it crushed the cost of going from 0→1. Anyone can ship a working prototype app from their bedroom now. But maintaining a system over years, deciding what's worth building, and actually getting it into people's hands at scale? Those were always the hard parts. And they still are. 0→1 was never the whole job. It was the fun part.