A different category of institution.

The conventional taxonomy sorts firms by output: steel, software, finance, construction. There is a deeper classification, and it is the one that will come to matter: the region of reality with which a firm maintains privileged, repeatable, economically meaningful contact, and from which it can therefore learn at a rate no one else can.

As reasoning becomes abundant, this second classification becomes the operative one. What a firm produces turns out to be a consequence of what it can observe. Production is downstream of measurement.

Under that classification Bayswater's answer is exact: industrial human capability, observed where it is exercised, not where it is documented.

A scientific method, run as an institution, embedded inside production.

Continuously: identify the largest remaining uncertainty in the representation. Design the contact with reality that resolves the most of it. Acquire the observation. Incorporate it. Choose the next contact better than the last was chosen.

An institution built this way is not defined by what it knows today. It is defined by how much every week of contact with the industrial world improves what it knows. That property, not headcount, not output, is what it is optimised for, and the state variables it publishes exist to make the claim checkable rather than rhetorical.

This is unlike consulting, where knowledge is a byproduct of engagements. Here the engagement is the byproduct. Knowledge generation is the production function. The function, stated exactly: the interface between industrial human capability and industrial execution.

Four of those five moves point at the world. The fifth points back at the operation itself. An earlier pass made a dated statement about something that has since either happened or not, so the answer already exists in the field; it is looked up, and it is recorded whichever way it went. Without that move the other four are a pipeline with a good name rather than a loop, and the word recursive would be doing work it had not earned.

Nobody purchases coordinates. Everyone positions by them.

The metric system. Coordinated time. Double-entry bookkeeping. Satellite positioning. None of these are products, and none of them is noticed in daily use. They are representations that became the terms in which coordination itself happens, so fundamental that their existence is simply assumed.

That is the layer Bayswater is built toward: the representation in which industrial capability is stated, compared, priced, regulated, trained and planned. Not a vendor within the industrial labour market: part of the coordinate system the market runs on.

Capability is the first region of that description, chosen because it is where the leverage is greatest. It is not the last.

Ambition of that order is cheap to declare and slow to earn. The earning is years of calibration against real outcomes, which is why every page of this site is marked provisional.

An ambition of that shape carries one serious objection, and it is stated here rather than left for somebody else to make: a shared measurement is not paid per use, and earns nothing at the moment it is consulted, which is precisely why everybody is free to consult it. Standing on its own, the reach therefore describes an institution with no means of surviving long enough to become one. The half on the right of the rule above is the answer to that, and it is argued where it belongs rather than a second time here.

What the institution becomes is one question. How it earns is another.