Argument
Capacity to reason compounds. The rate at which reality emits evidence does not
The ratio between what can be computed and what can be observed widens every year, and the gap is now the binding constraint on industrial decisions.
Argument
Warrant: Derivation
This problem existed long before artificial intelligence and will outlast every model currently in production. What has changed is not the problem. What has changed is that it has become addressable.
01 · What has become cheap
Computation is cheap. Storage is cheap. Software is cheap and becoming cheaper. Reasoning, until recently the scarcest input in any technical undertaking, is now bought by the hour at a price that keeps falling.
Whatever remains scarce determines outcomes instead. So the question worth asking is not what can now be computed. It is what cannot.
02 · The asymmetry
A refinery shuts down for a turnaround once. For a number of weeks, thousands of people who have never worked together execute tens of thousands of tasks under conditions that will not recur. Welds are made. Vessels are opened. Judgements are taken at three in the morning by people whose competence has never been measured under those conditions and will never be measured under them again.
Then the plant restarts and the opportunity is gone. You may reason about that turnaround for a thousand years and generate no new information about it. The information existed only while the event was occurring, and only for those positioned to capture it.
As the cost of reasoning falls toward nothing, the binding constraint on knowing anything about the industrial world moves entirely to the acquisition of observations from it. Time spent computing becomes cheap. Time spent in contact with reality becomes the expensive thing.
The turnaround is one member of a class. Nuclear outages, offshore hook-ups, shipyard erections, commissioning windows, pipeline tie-ins, unplanned repairs, first-of-class fabrication: each is a window that opens once, emits more about real capability in days than routine operation emits in years, and closes for good.
Almost every organisation is allocated the other way around.
Argument
The ratio between what can be computed and what can be observed widens every year, and the gap is now the binding constraint on industrial decisions.
Observation
A refinery shuts down once, and no quantity of subsequent analysis reproduces the observation that closed with it.
03 · The unmeasured quantity
Not what they are certified to do. Not what they say they have done. What they can do, under load, in a specific plant, to a specific standard, on a specific day.
The industrial economy has no direct measurement of this. It has never had one. It has five indirect ones.
Observation
A CV is a proxy. A certification is a proxy. An interview is a proxy. An employment history is a proxy. A practical assessment is a proxy. None of these is capability itself. Each is an observation of something much deeper, taken through an instrument that nobody has characterised. The five are taken apart, one at a time, below.
Everything a person can do, narrowed to the one channel industry actually records.
Written by the subject. Describes work that cannot be verified by the party relying on it.
A threshold crossed on one morning, under conditions the certificate does not describe and does not carry.
Measures how a person performs in an interview. The correlation with execution is assumed rather than established.
Records that someone was retained. A statement about an employer's tolerance as much as about competence.
The closest of the five, and still a single rehearsed sample taken away from the conditions that decide the work.
Five noisy instruments, pointed at a quantity no one has named.
04 · The five instruments
Observation
Warrant: Foundation
A document written by the person being assessed, describing work that cannot be verified, in language chosen to succeed.
Records that a threshold was crossed on a particular morning, sometimes years ago, under conditions the certificate does not describe.
Measures how a person performs in an interview.
Records that someone was retained, which is a statement about an employer's tolerance as much as about a worker's competence.
The closest of the five. Still a single sample, taken in a rehearsed setting, of a quantity that varies with fatigue, supervision, equipment, weather, and the standard being worked to.
The industrial economy holds five noisy instruments pointed at a quantity it has never named, and reads them one at a time.
05 · The dimensions of description
Definition
Warrant: Foundation
Any description of what a person can do resolves onto eight dimensions. Industry currently records one or two of them and discards the rest, which is why capability claims from different sources cannot be compared, weighted, or checked against what actually happened.
Eight dimensions, of which most industrial records carry two. That is the measurement problem, stated exactly.
06 · Who is actually affected
A contractor bidding a project does not know, with any rigour, what its own workforce can execute. An operator planning a shutdown does not know which of the people arriving on site will perform under the conditions that will actually occur. A board committing capital to that work is committing it against an unmeasured quantity. A training provider designing a course is correcting a deficiency nobody has located.
Every one of them is reasoning carefully from observations too thin to support the decision.
07 · One condition
Recruitment is difficult. Deployment is uncertain. Training is inefficient. Planning is guesswork. Project execution is unpredictable. These are treated as separate industries with separate solutions.
An economy cannot allocate what it cannot see. Every downstream inefficiency in industrial labour is a consequence of the same missing measurement.
No improvement to any single one of them reaches the cause. The problem is not that industry recruits badly. The problem is that industrial capability has never been made observable.
08 · Bayswater
Not to recruit better, assess better, or predict better. Those are consequences. Bayswater constructs a representation of industrial capability, a computational description of what the industrial workforce can actually do, against which such questions become answerable rather than estimated.
The representation is the work. Everything else is a projection of it.
How it is constructed is not described here, and will not be.