The Easy Problem — Summary

Gil superintends a water plant serving nine thousand people in the Sandhills. Two of the 1968 pumps still run. The monitoring model that watches turbidity and dosing went in three years ago, cost less than a valve, and was adopted in an afternoon on his own authority. The proposal for a closed-loop dosing skid, the plant’s first physical AI decision, has sat on his desk for seven months, because it is a different kind of decision. The cognitive tool asked for one competent person. The skid asks for a technician who is currently a hundred and forty miles away, a parts shelf, a plan for when the distributor consolidates, and a named person answerable for decades, because wrong dosing carries a docket number. A physical deployment is not a purchase with a maintenance cost; it is a maintenance obligation with a purchase attached, and that inversion raises the corpus’s serving-unit floor from an absorption test to a custody test. The layers with deflation curves are the ones places like his already have. The layers holding him at seven months have no deflation curve and have thinned for two generations.

Inside the same subject sits a movement the corpus has been slow to write: the first melting asset that is a population. The largest gap in the world economy is the price difference of an hour of physical work between places, and whole regional economies stand on that spread. Physical AI is aimed at it directly. The merciful finding is the rate: the spread closes at the speed of capital, task by task across replacement cycles, formation-scale time a country can use. The unmerciful finding is the holder: a population holds its position with no ledger and no sell order, and the proceeds accrue to whichever account the closing is billed to. What a population can do that a firm cannot is act through the bodies that answer for it, which returns the question to serving institutions, formation, and the floor.

Then the turn the whole three-essay set has been waiting for. The corpus is pessimistic about a cheap thing: cognitive deployment costs nothing and stalls for years on identity, because the obstacle is people. It must now be optimistic about an expensive thing: every difficulty in the physical ledger, depreciation schedules, actuarial tables, parts contracts, capital programs, is a difficulty institutions were built to eat, and the numbers will clear one replacement cycle at a time without anyone having an identity crisis first. The hard problem is us, and we were never the expensive part. A dated claim enters the register: through September 2033, physical adoption in existing facilities tracks replacement cycles, not capability releases. Gil approves the skid in October with one handwritten clause, training for two people, repeated at year five. Above his desk sits the plant’s green commissioning logbook, five superintendents’ hands across fifty-eight years, and what he is really deciding is whether the book will have a sixth.