Supply chains are the last great software problem
Logistics is the largest, oldest, and most under-instrumented distributed system on earth. The teams that treat it like software will define the next decade.
A modern supply chain is a distributed system with lossy links, adversarial partners, and no consistent clock. It has partitions that last weeks, actors who lie about state, and eventual consistency measured in freight days. If you described it in a systems paper, no reviewer would believe it was real.
And yet, most of it runs on spreadsheets, EDI messages older than the engineers reading them, and phone calls. The people inside these systems are brilliant — they hold the world together through sheer operational memory. But the primitives they have to work with are the wrong primitives.
The interesting insight is that this is a software problem, not a logistics problem. The bottleneck is not trucks. The bottleneck is that no two systems in the chain agree on what a shipment is, what a status means, or what time it is. Every integration is a bilateral treaty. Every exception is a phone call.
The teams that will win the next decade of logistics are the ones that treat the chain like a computer: strong types at the interfaces, event logs that survive replay, deterministic reconciliation, and operator tooling that treats humans as the smartest node in the graph, not the slowest.
I have watched this play out in pharma. A prescription is a message; a fulfilment is a state machine; a delivery is an ack. Once you see the domain that way, the operational leverage is enormous. A single well-designed reconciliation loop can replace an entire floor of exception handlers.
The industry has been waiting for someone to build the AWS of physical goods — not another marketplace, not another visibility layer, but the actual primitives. Whoever does that will look, in retrospect, like they were building infrastructure for a country.