Plug-and-play for the physical world

Two stacks

A developer today can build an application that serves users on the other side of the world, takes payments and sends messages, without understanding any of the infrastructure beneath it. Decades of shared standards made the digital world composable: almost anything can talk to almost anything else.

The physical world never got that layer. The factories, water works, buildings and power networks we depend on are instrumented with a patchwork of protocols that accumulated over a century, each solving a local problem in its own decade. Walk through a typical plant and you'll find current loops from the 1950s, serial buses from the 1980s, several incompatible industrial networks from the 2000s, and a building system above it all that can't speak to any of them. It isn't a sensing stack. It's a sensing museum.

The tragedy isn't that the old protocols exist. It's that they all exist at once, in the same building, and connecting any two of them is a project.

The cost of a sensor that doesn't know what it is

Today, adding a sensor to a site is mostly configuration. Someone has to tell the system what the device is, what it measures, what units it reports in, what its range is, how it was calibrated, and where it has been fitted. Every one of those steps is a chance for a mistake that quietly corrupts the data — a pressure reading in the wrong units, a temperature probe registered as the wrong channel, a calibration that expired a year ago.

Multiply that by thousands of sensors and the result is predictable: integration costs more than the hardware, deployments stall, and the data nobody fully trusts is the data nobody acts on.

A probe that introduces itself

The fix is old and simple: let the device carry its own description. When a Vigil probe is plugged into a node, it says what it is — what it measures, in which physical quantity and unit, its range, its calibration and who made it. The node passes that on; the platform files the reading in the right place, in the right units, with the right history. Plugging it in is the set-up.

That small idea has large consequences:

• No configuration screens. A technician fits a probe and walks away; the platform already knows what it is.
• No wrong units. Every reading arrives as a physical quantity, so a pascal is never mistaken for a percent.
• Probes are portable. The same probe works on any node on any Vigil site, from any maker who designs to the standard.
• Calibration travels with the device. When a probe is swapped, its history and calibration come with it, and the memory records the change.

Actuators too

The same idea applies to the things a node can move. An actuator that describes itself — a valve driver, a relay, a damper motor — tells the platform what it can do and within what limits. That's what lets programmes, reflexes and an AI assistant propose actions safely: the platform knows not just that there's an output on a node, but what's attached to it and what moving it would mean.

An open registry

Self-describing devices only pay off if many makers build them. That's why the registry of probes and actuators, Forge, is open: what each device measures, its interfaces and ratings, in one place the platform itself uses. Partners design to it once and their hardware works on every site — and as our assistant learns to recommend probes for a site's purpose, the registry is the catalogue it draws on.

Where it leads

When sensing becomes plug-and-play, the economics flip. Adding a measurement stops being a project and becomes a purchase. Coverage stops being rationed. And the data becomes trustworthy enough to act on — which is exactly what a facility needs before it can safely let a platform move its valves.

The Forge partner programme →