If Apple Built a Factory, It Wouldn't Use a PLC (physical-ai.ghost.io)

The piece argues that modern factories should be redesigned like Apple’s ecosystem: machines, sensors, robots and workers acting as peers that discover each other, share intent, and adapt in soft real time—rather than being siloed behind vendor-specific PLCs and brittle central controllers. The comparison isn’t about consumer features but about treating coordination as a first‑class protocol: secure identities, local mesh discovery, context sharing and semantic handshakes (the analogue of Handoff, Bonjour, UWB/Bluetooth) that make interoperability invisible. That shift matters for AI/ML because it reframes automation problems as distributed state‑estimation, intent negotiation and real‑time decision‑making tasks—exactly where learning, perception and multi‑agent coordination can add value. Technically, the article highlights obstacles—safety and determinism requirements, 30‑year hardware lifecycles, and incumbents’ lock‑in business models—that keep factories on isolated control stacks. The proposed remedy is “protocol literacy”: build common, secure primitives for discovery, authentication, intent exchange and feedback so autonomous agents can cooperate without brittle integrations. For ML practitioners this implies new focuses: certified, verifiable policies for safety‑critical control; lightweight semantic protocols for intent and anomaly signals; decentralized state fusion and learning that tolerate legacy nodes. The payoff is not smarter robots per se but far richer, evolvable conversations between them—enabling resilience, faster integration and emergent system‑level behaviour.