Stable core, adaptive edge: a model for manufacturing software

Every manufacturing system lives with two opposite demands.

The first is stability. Orders, routings, lots, quality results and production history must be consistent today, next year and during the next audit. A change in how a scrap declaration is stored can break traceability for every product the plant makes.

The second is adaptability. The way people work changes all the time: a new customer asks for an extra check, a line gets a new machine, a supervisor wants a different view of the shift. If every such change becomes an IT project, people stop asking — and start working around the system with spreadsheets and paper.

Traditional systems pick one side. Packaged MES products protect stability and make change expensive. Custom software makes change easy at first and stability expensive forever.

Two speeds, one model

The way out is to stop treating the system as one block.

The core holds what must be governed: the factory model, product and process definitions, execution records, quality and material genealogy, identity and access, planning logic and integrations. It changes carefully, through releases, because everything depends on it.

The edge is what people touch: screens, forms, dashboards, workflows, the order of steps on a terminal, which actions appear where. It should change as fast as the work does.

The crucial rule is that the edge is composed on the core, not built beside it. An operator screen does not have its own database; it reads the same work stage and writes the same declaration through the same action as every standard screen. A dashboard does not copy data; it queries the governed model.

What this changes in practice

• Changes become small. Adding a check to a station is a change to one experience, not to the data model.
• Rules never fork. Whether a quantity is declared from a standard terminal or a custom tablet app, the same action records it — with the same validations and the same history.
• Ownership is clear. The vendor carries the core and its lifecycle. The customer, or a partner, owns the experiences.
• Upgrades survive. Because experiences bind to stable entities and actions, the core can evolve underneath them.

Where AI fits

AI makes composition faster — describing an experience in plain language and getting a proposed screen bound to the right entities. It does not need to invent a database or a parallel application, and it should not. The value of AI in manufacturing grows with the quality of the model it works on.

That is the idea behind Metalya: stable core, adaptive edge.