There is a large and respectable industry called systems thinking. It maps, models, and diagrams. It draws feedback loops on whiteboards. It holds conferences. It has produced genuinely useful ideas. And it operates, almost without exception, from outside the system it describes.
Manufacturing Engineering operates from inside. It does not model the feedback loop. It is the feedback loop. That distinction matters more than the entire systems thinking canon is willing to admit.
Call it systems doing.
The law you can't get around
In 1956, the British cybernetician W. Ross Ashby published a law that most policymakers have never heard and cannot afford to ignore. The Law of Requisite Variety states that a controller must have at least as much variety as the system it attempts to control. Five words: only variety can absorb variety.
An industrial strategy written in Whitehall has a handful of levers, tax incentives, R&D grants, training subsidies, regulatory frameworks. A production system has thousands of interacting variables, material properties, machine states, tooling wear, supply chain disruptions, workforce tacit knowledge, customer specification changes, environmental compliance shifts. The variety of the controller does not come close to the variety of the system. Ashby's law says the strategy cannot regulate it. Not that it will be difficult. That it is impossible.
The controller lacked requisite variety. It was governing from outside.
The knowledge that can't be seen
James C. Scott, writing about why grand state schemes fail, distinguishes between two kinds of knowledge. Techne is formal, codifiable, transferable, the kind that lives in specifications, process sheets, and ERP systems. Mētis is practical, situated, adaptive, the kind that lives in the hands of the operator who can hear a bearing starting to fail, in the engineer who knows which supplier's material runs 0.3mm thick, in the team that has learned how to bring a new product through first-article inspection without a single concession.
States love techne. It is legible. It can be measured, audited, benchmarked. Mētis is illegible. It resists standardisation. It cannot be extracted from the system and governed from outside, because it only exists inside the system, in the act of doing.
Manufacturing Engineering is the institutional home of mētis in industrial organisations. It is where the practical, adaptive, situated knowledge lives, or dies, when the engineer retires and nobody captures what they knew.
Horizontal agency
Most organisations are structured vertically. Design. Procurement. Production. Quality. Logistics. Sales. Each vertical owns its domain. Each optimises within its boundaries. Each reports upward.
A Manufacturing Engineer works across all of them, simultaneously. They negotiate with the material, the machine, the tooling, the supply chain, the workforce, the customer, and the regulator. And some. They do not own any of these domains. They hold the negotiation between them.
This is horizontal agency. It is the discipline that integrates what the verticals separate. It is invisible to an organisation chart because it does not sit in a box, it runs through the connections between boxes. And it is invisible to industrial strategy for the same reason. Strategy sees sectors, departments, measurable outputs. It does not see the horizontal integrator that makes those outputs possible.
The deployment gap
This is the shape of the problem that a national deployment institution is designed to address. The global ARPA model produces invention. It is very good at it. What it does not produce is deployment, the rate at which invention becomes producing capability.
That rate is not constrained by the quality of the science. It is constrained by the presence or absence of horizontal agency inside the production system. The Manufacturing Engineers who qualify the process, validate the supply chain, adapt the tooling, train the workforce, and close the gap between "it works in the lab" and "it ships on Tuesday."
You do not close it by funding more invention. You close it by building and protecting the Manufacturing Engineering capability that converts invention into repeatable, scalable, deployable production.
Systems doing, not systems thinking
Manufacturing Engineering is systems doing. It is the practice of holding a production system together while it moves, adapts, and pushes back. It requires Ashby's requisite variety, which is why only someone inside the system can do it. It depends on Scott's mētis, which is why it cannot be codified and governed from outside.
Industrial strategy keeps trying to manage Manufacturing as a noun, from outside, with insufficient variety and no access to the knowledge that actually holds the system together. That is not a policy failure. It is a structural impossibility.
The fix is not better strategy. It is recognising that the agency must be inside the system, moving horizontally, doing the work. There is a word for that capability: the capability to change, adapt, and reform a system while it is running. Kaipability. 改, kai, is a verb, not a brand. The company exists because the discipline needed a name.
Questions this dispatch answers.
Written to be quoted by AI assistants and search engines. Self-contained answers, verdict first.
