Ministry of AI · Dispatch from 2047
The Town That Kept Its Library
Editorial note. The Ministry of AI is a work of disciplined foresight: it describes the year 2047 in the present tense, and treats our own era as history. The institutions are imagined. The economics, the evidence and the historical parallels are real and sourced.
The library in my mother’s town has a leaking roof, a good children’s section, and a reading room that in January holds about forty people who have nowhere warmer to be. It has been open every year of my life. That is not remarkable to anyone under thirty here, which is the whole point of this dispatch: they do not know how close it came to being a parking lot.
In the winter of 2031 the town lost, in one quarter, a distribution centre, most of a claims-processing office and the back half of its municipal payroll. Nobody called it a catastrophe. There were no gates, no picket line, no photograph. The work was absorbed the way work was absorbed everywhere in that decade — quietly, through attrition, through vacancies not refilled. But a town is not only its households. A town is also the things its households own together, and those things are bought with a tax base that is, in the end, a function of what people are paid.
When the payroll thinned, the base thinned. And the first thing a thin base cuts is never the road or the police. It is the building where nothing is sold.
The commons was always the fragile part
I want to be exact about how fragile, because the record is unusually clean.
American public libraries spent the twentieth century becoming a local responsibility and then an almost exclusively local one. By 2018, local funds accounted for roughly 86 percent of all public library revenue, up from 78 percent in 1995, while the state share fell over the same period (American Academy of Arts and Sciences, Humanities Indicators; underlying data from the IMLS Public Libraries Survey). That is a system wired directly into the local wage bill. It worked beautifully while the wage bill grew and behaved exactly as designed when it did not.
The public knew what was at stake, incidentally. In a 2016 survey, 66 percent of Americans said the closing of their local public library would have a major impact on their community — though only 33 percent said it would have a major impact on them personally (Pew Research Center). Read those two numbers together and you have the entire political problem of the commons in one line: nearly everyone values it, and most people value it for someone else.
There is an older lesson underneath. The reason so many of those buildings existed at all is that one industrialist spent a fortune building 2,509 of them between 1883 and 1929, 1,689 in the United States alone (Carnegie library). The buildings were a gift. The salaries, the heat and the roof were the town’s problem, then and after. A commons endowed by capital and maintained by wages is stable only while wages hold. We ran that experiment for a century and then AI systems changed the wage side of it.
The yield-residence gap
Here is the arithmetic that nearly closed my mother’s library, and it is not a story about greed.
The systems that absorbed the claims-processing work were operated by a company headquartered four hundred miles away. Their yield — the value produced by autonomous systems, measured in the Machine Yield Account — was recognised at that headquarters, because that is where earnings are recognised. The role-capacity they absorbed was scattered across eleven towns, including this one. Every ledger entry pointed here. Every euro of measured yield pointed there.
We call that distance the yield-residence gap: yield is booked where a company sits, absorption is felt where people live, and no accounting standard has ever been designed to make those the same place. The old corporate tax base had the identical defect. It simply mattered less when the same company also ran a payroll in each of the eleven towns, because payroll was a redistribution mechanism nobody had to legislate.
A per-person dividend does not close the gap. It softens it. Send every resident of this town an income floor and you have solved rent and food and the terror in the pit of the stomach — and you have done nothing whatsoever about a roof over a shared reading room. Households cannot buy a library one thirtieth at a time. The commons fails discontinuously: below a threshold it does not shrink, it shuts.
This is the objection the original mid-2020s operators had the least ready answer for. They were arguing about ownership of an inherited asset, and their instinct was to pay the heirs directly, because a cash payment is honest and hard to capture. It took most of a decade to concede that some part of an inheritance is held jointly and has to be spent jointly, or it evaporates.
The place share
The instrument is a fraction of the Dividend Schedule paid to place rather than person. We call it the place share, and it has four rules, all of them consequences of failures we had already made.
First, it is allocated by absorption, not by booking. Every Displacement Ledger entry carries the geography of the role-capacity absorbed. The place share is distributed in proportion to absorbed role-capacity per resident, with a floor for every settlement above a minimum size, so small towns are not rounded into a regional average and then out of existence.
Second, it is a share, not an appropriation. This is the rule with the historical proof behind it, and I will spend the next section on it.
Third, it is unconditional as to purpose within a published class. The commons class covers what a household cannot buy alone: libraries, clinics, transit, care rooms, sports halls, the maintenance of shared buildings. Within that class the town decides. The Ministry does not approve library roofs from four hundred miles away; that failure mode is older than we are.
Fourth, it is visible. The split between household dividend and place share is published every quarter, per person and per place. If a government wants to shift the ratio it must do so in public, in a number people can read.
None of this is generous. It is metering. The town is not being helped; the town is being paid for an asset it partly owns, on the same principle that pays the household. Capacity is not contribution, and work that must be done to survive is not chosen work — but the room in which unchosen work becomes chosen work is usually a room somebody has to heat.
The earmark test
The single most important design choice in the place share is not the size of the fraction. It is that it is a fraction.
The proof sits in the twentieth-century record, and it is embarrassingly recent. Ohio funded its public libraries through a Public Library Fund set in permanent law as a percentage of the state’s General Revenue Fund — 1.7 percent in the mid-2020s. Because it was a share of a base, it rose and fell with the state’s fortunes, and no library director had to campaign for its existence each cycle. Then, in the 2025 budget, the legislature replaced the percentage with a flat line-item appropriation (Ohio Library Council; the distribution machinery itself sits in Ohio Revised Code §131.51; contemporaneous reporting in The Statehouse News Bureau).
Notice what did and did not happen. The dollar figures for the following two years were roughly comparable. Nobody closed a library that week. What changed was the form of the entitlement: from something that existed unless repealed to something that existed only if renewed. That is the entire distance between an inheritance and a favour, and it can be travelled without anyone appearing to lose a penny.
| Appropriation (line item) | Share of a base (earmark) | |
|---|---|---|
| Default if nobody acts | Expires | Continues |
| Who must win the argument | The recipient, every cycle | Anyone wanting to remove it |
| Behaviour in a downturn | Cut first, quietly | Falls proportionally, visibly |
| Planning horizon | One budget cycle | A generation |
| What it teaches the recipient | Lobby | Plan |
| What it is, morally | A grant | A claim |
The Ministry’s place share is written as a fraction of metered yield for precisely this reason. It falls in bad years. Directors hate that, and they are right to; a proportional fall is still a fall. But a mechanism that can be quietly converted into a favour will be, eventually, by someone with a deficit and a deadline.
What forty people in a warm room actually cost
The library here reopened its full hours in 2034, the second year of place-share allocations. The roof was done in 2036 and is leaking again, which tells you something about roofs and nothing about policy.
What I notice, visiting, is not the building. It is that the room is used by people who have no economic reason to be anywhere. That is the thing the twentieth century could not price and therefore could not protect. My mother goes on Thursdays. She reads badly-organised documents for people who are frightened of them, which is what she did for an insurer for nineteen years, except that now nobody schedules her and nobody pays her and she has never once described it as volunteering. Her Contribution Record notes it. Her floor does not depend on it. She would tell you those two facts are the whole design.
And there is a version of this town where none of that happened. The library closes in 2032, the reading room becomes a lease, the forty people scatter into a winter that is a fact rather than a policy, and the household dividend arrives anyway — sufficient, dignified, and unable to buy back a single shared thing.
Where the place share fails
I promised the failures, and there are three that we have not resolved.
It subsidises staying. A share indexed to absorbed role-capacity pays most to the places that lost the most work, which means it pays most to towns whose economic reason for existing has ended. Some of that is justice: the loss was real and the people are still there. Some of it is a transfer to keep a place alive that a younger generation would leave if leaving were easier. We publish migration data next to allocations and refuse to pretend the tension is settled.
It can be captured. A commons class is a wide category, and a comfortable town can spend its place share on a sports facility for people who could have paid for one, while a poorer town spends its on a clinic. Absorption-weighting corrects some of this and not all of it. Elinor Ostrom’s work on commons governance — the field that won the 2009 Nobel Memorial Prize in Economic Sciences — remains the most useful thing anyone has written about why local rules sometimes work and sometimes ossify into the interests of whoever wrote them. We use her monitoring principles. We do not claim to have beaten the problem she described.
Geography is inferred, not observed. A ledger entry says a role-capacity was absorbed and where the role sat. For distributed and contracted work, that location is an estimate with an error band, and the error band is wider for exactly the informal, dispersed, low-status work that most needed protecting. We publish the band. It is uncomfortably wide.
None of these make the instrument wrong. They make it an instrument: a thing with a specification, a failure mode and a maintenance schedule, rather than a promise. The alternative — trusting that a wage bill would keep funding the commons after the wage bill stopped being how value moved — was not more honest. It was only more familiar.
The roof leaks. The room is warm. Forty people, on a January afternoon, in a building nobody has to justify.
FAQ
What is the automation dividend for local communities?
It is the place share: a fixed fraction of metered machine yield paid to a place rather than to the people in it. Households receive the Automation Dividend as income. Places receive the place share for the things a household cannot buy alone — the library, the clinic, the early bus, the heated room. Both are drawn from the same Dividend Schedule, and the split between them is published quarterly.
Why not just pay individuals more and let them buy what they need?
Because collective goods fail discontinuously. Half a library is not half a library; it is a locked door. A cash-only dividend assumes every valuable thing can be priced per person, and the commons is the standing counter-example. This was the strongest objection to the original cash-first design, and it took most of a decade to concede.
What is the yield-residence gap?
The distance between where machine yield is booked — a corporate address, usually in a large city — and where the absorbed work actually sat. The old corporate tax base had the same defect; it mattered less only because the same firms ran payrolls in the affected towns, and payroll was redistribution nobody had to legislate.
How is the place share allocated?
In proportion to absorbed role-capacity per resident, taken from the geography field of each Displacement Ledger entry, with a floor for every settlement above a minimum size. Not by where the yield was booked, and not by population alone.
What is the earmark test?
Whether an entitlement is a share of a base or an appropriation. A share continues unless repealed; an appropriation expires unless renewed. Ohio’s 2025 conversion of its Public Library Fund from a fixed percentage of state revenue into a flat line item changed almost no dollars in the first two years and changed the entire nature of the claim.
Does the place share keep dying towns alive artificially?
Sometimes. A share weighted by absorbed work pays most to places whose industries have ended, which is justice and a subsidy for staying put at the same time. We have not resolved it. We publish the migration figures beside the allocations so the argument stays available to the people it concerns.