Who holds the keys to public intelligence?

Sebastian Clark’s Plan for Public Intelligence, published by Isolarii this month, wants to do for artificial intelligence what rural electric co-operatives did for the American farm.¹ Its motto turns Google inside out: where Google built private systems on top of public information, Clark’s plan offers “a public architecture for private information.” He asks what it would take to build, in effect, a “Palantir for civil society” that genuinely belongs to it. His central question is one Counterfoil asks too: under what conditions can intelligence become common infrastructure?²

This is a serious, thoughtful plan, and it deserves a serious reading. In the present context, that means asking the questions Ivan Illich would have asked, which Clark never does. This isn’t quite a surprise. Illich is not among the social thinkers Clark credits (David Graeber, Édouard Glissant, Simon Fraser University’s Wendy Chun, and Jo Guldi are), and unlike them, Illich never settled into an academic career.³ Clark does use Illich’s favourite (and very common) image to frame the central problem: the enclosure of the commons.

What the plan proposes

The diagnosis properly comes first. Data has become a means of production, like land, and the firms that gather it are fencing it in as a resource they control. Google moved from organizing public information (search) to extracting private information: chat histories, medical records, and the working knowledge held in hospitals, schools, courts, and libraries. Clark calls that last kind of data “context” and argues it is what the AI labs now want most. Meanwhile, the institutions that hold “context” in trust have handed their computing to vendors like Google to escape the liability of holding data themselves. Along the way, they’ve lost their own judgment.

Clark points to the British Post Office’s Horizon scandal, Australia’s Robodebt, and the Dutch toeslagenaffaire (childcare-benefits affair), where public bodies believed flawed automated records over human evidence, much to the harm of the public these systems supposedly existed to serve. The Post Office relied on Fujitsu’s accounting software; Robodebt and the Dutch risk scoring were the governments’ own systems. In each case, people accused by a machine were left with the near-impossible task of disproving the accusations.⁴ Such cases are often filed under “algorithmic governance failure,” but the phrase flatters them.⁵ These were algorithms with no real governance at all, and the institutions that ran them unleashed what Illich called the “murderous logic” of technology run amok. The machine was believed over human testimony and evidence.⁶

Clark’s plan offers five remedies for making algorithmic machines governable by humans:

• A separation of powers for computation. Who holds data would be separated from who computes over it. Programmable cryptography (zero-knowledge proofs, multiparty computation, homomorphic encryption) enables computation based on records the user can’t see directly.
• Markets in computational rights. Clark’s instrument is a “computational warrant” — a priced, time-limited, machine-enforced licence to run one specified function over someone’s data, with a receipt. He compares it to an easement, the right to use land you don’t own, and expects a market to grow up around the warrants.
• The public university as anchor. Universities and their libraries would set standards, issue warrants over their own holdings, and pool patents on non-exclusive terms, while hospitals and other custodians keep authority over their own records. Clark’s first move is for a public university to refuse its records for AI training; after that, even proprietary frontier models could train and run on such records, but only under warrants that state what may be computed and released. Student movements would have to supply the political force to establish and sustain this role for universities by “forcing” them to use their “custodial leverage” against the AI labs. (At this point in the text, Clark shifts from talking about Stanford as the origin and guarantor of tech monopolies to note student protest movements in Dhaka, Kathmandu, Nairobi, Mexico City, “and now France,” but conspicuously not North America.)
• Open silicon. Inspectable chip designs and toolchains, designed for kinds of computation other than today’s language models, would break lock-in like Nvidia’s CUDA. Clark names three separate routes: the federal funders’ march-in rights under Bayh–Dole (never yet used), public patent pools among universities, and the precedent of the 1956 AT&T consent decree, which required royalty-free licensing of Bell’s patents.
• Co-operative infrastructure. Data centres owned by rural electric co-operatives, sited at their substations, and funded with low-interest federal loans would devolve the critical modern means of production to the American heartland.

The plan closes on two ideas. The first recasts privacy as a question of who may do what with computation, after Glissant’s right to opacity: the right of every person and culture to remain incompletely understood without losing dignity or the right to relation. Glissant was answering the Western habit of treating freedom as total transparency. Demanded by the powerful of everyone else, transparency is a means of control.⁷ The second is an old legal idea: knowledge belongs to the living in usufruct, the right to use and enjoy something without wasting it.⁸

What it gets right

These are provocative ideas, and Clark gets a lot right. The section on liability is counterfoil research in all but name. It documents institutions that bought a tool to do their work, then lost the ability to do the work themselves, and finally lost the ability to detect when and where the tool went wrong. Horizon’s accounting software sent subpostmasters to court and to prison under a legal presumption that computers work correctly. That is what Illich meant by counterproductivity: past a certain point, a tool defeats its intended purpose. (I’ve written about that pattern in “Why more makes less.”) Yet a system is what it does.

Clark is also right that “sovereign AI” — national models bought from the same few vendors — is dependence with a flag on it. He reaches for Machiavelli on “borrowed armies.” Illich had a word for it too: “radical monopoly” — the dominance not of one brand but of one kind of product, until people can no longer meet the need any other way.⁹ Clark looks for precedent in the right places: co-operatives, the Cooperative Extension System,¹⁰ and Adriano Olivetti’s factory town at Ivrea. Clark’s best idea is an analogy: the knot. Data tied together for one purpose, then untied, rather than chained together permanently can be governed.

Who owns it is not the question

Illich’s likely first objection would be to the plan’s frame. In Tools for Conviviality he argued that “certain tools are destructive no matter who owns them.”¹¹ Public ownership of a factory does not change what the assembly line does to the people on it. Clark’s plan spends most of its energy on ownership, custody, and permission: who holds the data, who holds the keys, who owns the chips, who owns the substation. It does attend to the tool itself in places: chips designed for other kinds of computation than today’s language models, sensors on grids and water systems that collect only what each job needs, and Olivetti’s joining of advanced technology to a human scale at Ivrea.¹² But it gives each warrant a boundary without giving a boundary to the system as a whole. In that respect it sits closer to the tradition Counterfoil has described as public interest research than to counterfoil research.

Counterfoil research, as Illich described it, measures the dimensions of tools against the people who use them: size, speed, energy, and how much of a task a tool takes over. It looks for the thresholds past which a tool stops serving people and starts managing them: overshoot, excess, and above all, monopoly. Its first task, Illich wrote, is “to provide guidelines for detecting the incipient stages of murderous logic in a tool.”¹³ Clark’s plan doesn’t identify any such thresholds. It rejects moratoria outright and takes for granted that AI will keep growing. It asks who should run the machine. It never asks how big the machine can get before it no longer matters who runs it.

A market is not a commons

Clark takes enclosure from the history of English land, and his remedy from property law: the easement. But a warrant market does not undo an enclosure. It surveys the fields, prices each crossing, and issues receipts. The plan calls the result a mutual-credit economy of computation. That may well be a fairer economy. It is still an economy, and a commons is more than a resource for equitable exploitation.

Illich drew this line in 1982, at a Tokyo symposium on the computer-managed society. A commons, he said, is used by custom, outside the market. Enclosure turns it into a resource, something to be allocated and priced. His example was the loudspeaker that came to his childhood home on the island of Brač in the early twentieth century, putting an end to silence as a common good. When the talk was printed in CoEvolution Quarterly, it ran under the line “Computers are doing to communication what fences did to pastures and cars did to streets.”¹⁴ (I took up that warning in “What would Ivan Illich make of large language models?”)

Clark’s best example shows how a warrant is meant to work. A health authority wants to know, in real time, where respiratory infections are rising and which hospitals have the staff and beds to respond, without pooling patient registries. The custodians authorize one specified computation, an agreed result is released, and a cryptographic receipt records who authorized it, what was released, and what was paid.¹⁵ That protection is real. But a receipt proves only that a query ran as authorized. If a patient’s record fed a result that cost her a bed, or her town its clinic, whom can she appeal to, by what process, and for what remedy? Clark raises the claims of the people represented in records himself, and he hands standard-setting to domain bodies; he doesn’t say who can challenge those bodies or the custodians. A patient can have a real right of appeal without being able to read a zero-knowledge proof. A public architecture for private information is still an architecture, and the question is who can contest what passes through its gates.

The university as gatekeeper

The university is where Clark’s plan and Illich’s perspective part ways most sharply. Clark wants to rebuild the public university and make it the anchor of the warrant system. Universities and libraries would set its standards and issue warrants over their own holdings; hospitals and other custodians would decide who may run what operations over theirs. Schoolchildren’s data would live in cryptographic sanctuaries managed for them.¹⁶ Who can challenge or say no to a custodian’s yes, especially when the person in the record has less power than the institution holding it?

Illich wrote Deschooling Society against exactly this kind of institution, which he came to see as patterned on the medieval church: one that claims a monopoly on the means of learning and then certifies only those who have passed its exams. Yet his alternative was also infrastructure, and parts of it come startlingly close to Clark’s plan. Illich proposed “learning webs” — open reference services for things to learn from, exchanges where people offer their skills, a peer-matching service, run by computer, that would let anyone find others who wanted to talk about the same book, and a directory of educators.¹⁷ Illich’s webs were publicly supported too, but the learner started each exchange, set its purpose, and could walk away. In Clark’s plan the custodian sets the terms.

Clark’s own historical notes undercut his arguments. His history of Stanford licensing shows the same university that seeded biotech firms with non-exclusive licences also chose an exclusive deal with Google for PageRank. Clark’s account of student debt shows public colleges shifting their costs onto students. Clark reads these historical patterns as deviations from an enduring public mission. Illich would read them as what schooling institutions always do as they grow. Clark even calls Wendell Berry as a witness, to argue that rural communities depend on the land-grant universities. Berry’s indictment went further: the colleges had helped undermine the communities they were founded to serve, turning publicly funded research and teaching toward industrial agriculture and abandoning the Jeffersonian purpose that runs from the Morrill Act through the Hatch Act, a republic in which, as Jefferson put it, “worth and genius” would be “sought out from every condition of life.”¹⁸ A plan that hands state AgTech universities a new monopoly over massive data access and permission to compute with it as context should say what would keep it from happening again.

Cheaper compute is more compute

The co-operative section is the most attractive part of Clark’s plan, and it’s the one where Illich’s warning is sharpest. Clark estimates that future public intelligence co-op members could pay about $1.90 per GPU-hour if the machines run 85 percent of the time, or $3.20 at 50 percent, against about $7.45 on the open market, with the buildings funded by low-interest federal loans and secured by advance purchasing commitments from public institutions.¹⁹ Rural electrification is a good precedent for who should own a grid. It says nothing about how much power a society should draw. In Energy and Equity, written during the 1973 oil crisis, Illich argued that past a threshold of energy per person, energy grows at the expense of equity, and that “participatory democracy postulates low-energy technology.”²⁰

Clark doesn’t leave growth to chance. His prices depend on keeping the machines busy, and he calls the arrangement a self-reinforcing loop in which cheaper co-op capacity “finances the next round.”²¹ Once universities, hospitals, and agencies have committed to buying that capacity, who can turn down the next round? Cheaper compute may also simply mean more of it — the old Jevons paradox (“Why more makes less”). The plan names nothing that would cap the load, the water usage, the next expansion, or anyone with the power to enforce a cap. The plan also holds up the Boeing 747 against the Concorde as its model of a winning technology made ordinary and affordable. Illich would have looked at the same jet and seen something else. In Energy and Equity, high-speed transport is his central case of radical monopoly: once a society is built around vehicles, people can no longer get where they need to go under their own power, and those too poor to afford airfare or a car are ghettoized — a rather deliberate outcome of modern American transport and urban housing policy. Affordability is Clark’s test. Illich’s test was whether a tool leaves people more able to act for themselves, or more dependent on others.

The word from the State College: In “Calvinist Farming,” Sietze Buning (the Calvin College poet Stanley Wiersma, who grew up in Sioux County) remembers his elders planting corn in straight, tidy rows as a matter of faith. Their grandsons plant on the contour, as the state agricultural college taught them. The soil holds better. The poem is not sure the community does. Throughout Wiersma’s work, “Ames” and “The Extension” are regarded with suspicion by the farming traditionalists of the years before Earl Butz. Versions of this narrative, where the Morrill Act figures as a betrayal, exist in other, northern and southern agricultural regions of the United States. It’s a repeated anecdote in paleoconservative, traditionalist, agrarian, and populist narratives of loss and decline. Farming there has since been thoroughly financialized, with grain-marketing desks at the elevators and co-ops and local brokers trading futures for farmers. The area has its own reasons to be wary of big, unchecked money and power, private or co-operative.²²

Clark leans hard on rural America, so it’s worth looking at a place with a long history of the co-op model he wants to revive.²³ Through the years of the Obama administration, I lived in Sioux County, in the northwest corner of Iowa. It gets the “conservatives in the mist” treatment from national reporters nearly every caucus season: the most ethnically Dutch (and Frisian) county in America, one of the most churchgoing in Iowa, and deeply Republican, yet cool to Donald Trump in the 2016 caucuses, where only 11 percent of its Republicans backed him as the party nominee. But that November, about 80 percent of its voters backed him as their presidential candidate.²⁴ In 2024 more than 84 percent of Sioux County’s voters chose Trump again.²⁵

Sioux County’s rural electric co-operative charges single-phase members 15 cents for each of the first 400 kilowatt-hours a month and 9.2 cents for every one after that. The largest town’s water utility gives its biggest industrial users the lowest rates.²⁶ The county has more large animal feeding operations than any other in Iowa and some of the most expensive farmland in the country.²⁷ Its fields and feedlots drain to the Missouri. Iowa sends on average more than half of that basin’s nitrate load and close to a third of what the Mississippi carries to the dead zone in the Gulf.²⁸ Local water supplies hover near constant crisis; Sioux Center now blends its well water with treated water piped more than sixty miles from an aquifer beside the Missouri River in South Dakota. Shallow municipal wells along the creeks and rivers carry nitrates and other farm runoff. The deep aquifers provide much less and much harder water, carrying naturally occurring radioactivity as well.²⁹ In 2022 more than half the county’s cropland was in corn; about a fifth was mapped as corn in both 2021 and 2022, and only about 6 percent was planted to a cover crop. One study estimates that erosion has stripped the topsoil from about a third of the cultivated land in its Corn Belt study area.³⁰ Farm debt is at a record high.³¹

This is what a co-operative grid looks like when affordability is the test: cheaper units, more units, and the costs carried downstream. Nor is it a place clearly waiting for a public intelligence movement.

Palantir now builds the software the USDA uses to enrol farmers in its programs and pay them, under a $300 million deal,³² and Anduril’s co-founder Trae Stephens preaches that building new military technology serves God’s plan for America.³³ Clark sees rural America as a site for public infrastructure. The people who live there built their modern public infrastructure in the 1950s and 60s; today they lean toward privatization, but key cooperative institutions remain. In 2026, the county board stopped new data centres until it had written its own rules for them.³⁴ It’s not inconceivable regions like this might buy into a public interest AI movement, but it would have to arrive the way the county’s moratorium did: slowly, locally, and with the right to say no. Clark’s plan asks them to host the machine, not to decide how big it gets.

Knots only experts can untie, cooperatives that behave as cartels

Programmable cryptography is clever, and Clark is right to invoke Auguste Kerckhoffs, the nineteenth-century cryptographer who held that a cipher should stay secure even if the enemy knows everything about it but the key. Claude Shannon later put it more bluntly: “the enemy knows the system.” For Clark, that means open designs need not be weaker ones.³⁵ But open is not the same as legible. A convivial tool, in Illich’s sense, is one anybody can use, as often or as seldom as they like, for purposes of their own choosing.³⁶ Very few people can read a zero-knowledge proof. Clark’s plan calls the warrant user-governed, but the people whose records are knotted together will have to trust cryptographers, standards bodies, and settlement houses to tell them the knots came undone.

Clark sees part of this. He warns that whoever controls the market’s record of which computations proved useful will gain market power, and proposes shared clearing-house infrastructure, on the model of SWIFT, with a publicly owned common index to hold it.³⁷ That would create a new profession with a new centre. Illich called this kind of dependence overprogramming: more and more of what each person must know is set by someone else’s design.³⁸

It would be naive to believe that “shared utilities that no single bank owns outright” necessarily constitute a co-operative of equals that won’t be dominated by its most powerful members. Clark doesn’t say who would govern the index, or how. Legally, SWIFT is independent of the United States, though in practice it cannot ignore American financial hegemony and has increasingly become a tool of “weaponized interdependence.”³⁹ In this way it serves as a very good example of the problematic public interest role Clark wants to assign to a similarly structured organization.

Whose public?

Joyce calls Dublin: James Joyce left Dublin in 1904 and set Dubliners, A Portrait of the Artist as a Young Man, and Ulysses there while writing from Trieste, Zurich, and Paris. He told his friend Frank Budgen that he wanted “to give a picture of Dublin so complete that if the city one day suddenly disappeared from the earth it could be reconstructed out of my book.” Readers have walked its routes through the city ever since, on Bloomsday.

Machu Picchu, abandoned in the sixteenth century, was “discovered” by Hiram Bingham’s Yale expedition in 1911. Visitors will soon be able to fly directly to Chinchero International Airport, a controversial development in Peru’s Sacred Valley that’s due to open in late 2027.

The abbeys of Lucerne: After the French invasion of 1798, the new Helvetic Republic declared all monastic property national property. In 1848, after the Sonderbund War, the canton of Lucerne dissolved the Cistercian abbey of St. Urban. I imagine the empty buildings, the high-alpine forests, and the Nebelmeer (sea of fog) as connected to the abandoned Inca city and the Andean forests, which are often shrouded in clouds.

Martinique: Glissant was born there in 1928 and lived between it, Paris, and New York. His word for the world’s cultures in relation was the Tout-monde. After he co-founded the Front Antillo-Guyanais, which campaigned for the autonomy of France’s Caribbean departments, the French government confined him to metropolitan France from 1961 to 1965. From the Quiet Revolution on, Glissant came often to Quebec; the poet Gaston Miron, whom he met in 1959, called him “le plus québécois des Antillais.” In the winter of 1995, he gave four lectures at the Université de Montréal. They formed the basis of Introduction à une poétique du divers, published by the university’s press later that year.

La promenade de Champlain runs along the St. Lawrence in Quebec City. It opened in 2008 for the city’s four-hundredth anniversary, late in Glissant’s life, and I doubt he ever walked it, but the poem imagines him, or someone taking his place, there, where the river begins to open into its estuary.

Archipelagoes

Must a city withstand time? 
Is it enough for a writer to evoke a city?
May the city evoke the writer? 
Joyce calls Dublin calls Joyce into temporary being,
revealing the invisible passage 
that is their indigenous relation.

It is not a monument. 

The monumentalized city is a place you can fly to.
Machu Picchu with an airport, 
exhumed and put on display
because something is missing, elsewhere. 

Something is missing most 
where there hasn't been long, continuous use. 
The abbeys of Lucerne
had an invisible passage to the Andes 
only when they were suppressed.

An alpine archipelago touching the cloud forest,
when each place had no people. 
Or, when they had people, no travel but passage
from one to another in the imagination. 

To know what the snow is like in Quebec
in Martinique only by attention to a traveler’s story. 
It would be sacred.
It would connect the concert of the world in an island taxi 
to the fantasy of the world of white snow on black skin 
on la promenade de Champlain.

May 2022

I’ve received (and often redistributed) the gifts of Clark’s brilliant and beautiful Isolarii publishing project from its very beginning, and my favourite of its little books is The Archipelago Conversations, the late philosopher Édouard Glissant’s dialogue with Hans Ulrich Obrist.⁴⁰ Reading it in May 2022 was a near-ecstatic experience. I wrote the poem laid out here, “Archipelagoes,” as I read it, seemingly in sync with Glissant. Clark ends his plan by defending exactly that kind of experience, the transcendent and the ecstatic, against being reduced to an object of technological manipulation. Isolarii’s archipelago of small books, each its own island, is the best case I know for the world he wants. That is why his map of how to get there troubles me. My poet’s gut distrusts the place you can fly to; Clark’s plan is built on Boeing’s 747.

Clark grew up in Tasmania, founded Isolarii in New York, and now publishes from London to subscribers in some fifty countries.⁴¹ His plan draws its spirit from Glissant and its political hopes from students in Dhaka, Kathmandu, Nairobi, and Mexico City. He treats sovereignty with contempt, as a “fetish” that makes boundaries ends in themselves. I think he’s not wrong about that, but it doesn’t sit well in a proposal so focused on American institutional foundations, and so in need of the “legitimacy” the Pax Americana can claim in 2026 mostly for want of serious competition.

Clark’s own sources, as I read them, make the case against him. His author’s note credits a decade of conversations with Jo Guldi, Wendy Chun, “Vitalik Buterin and friends around the Ethereum Foundation and 0xPARC,” as well as David Graeber and Glissant, “whose legacies we have done what we can to help steward.”⁴² Graeber, an anarchist, wrote against the kind of bureaucratic rule-making a warrant market run by universities would need.⁴³ Chun showed how machine learning turns correlation into segregation, a problem in what the models do, whoever owns them.⁴⁴ Guldi’s history of land reform spans the globe; Clark’s plan narrows to American institutions.⁴⁵ Buterin argues for decentralized, defensive technology and is wary of concentrated power, yet Clark would hand his cryptographic keys to universities, co-ops, and the American state.⁴⁶ With a nod to Glissant, Clark borrows the archipelago but builds on the continent. He takes their diagnoses and their mechanisms, then builds on institutions most of them distrust.

Nearly every lever the plan pulls is American: land-grant universities, the Rural Electrification Act, Bayh–Dole march-in rights, an antitrust decree against AT&T, federal loans, Stanford. Its clearing house is modelled on SWIFT and the American settlement utilities. Its criteria for success are American too. Clark retells the American translation of the French revolutionary maxim — liberté becomes consumer choice, égalité is access to affordable goods, fraternité is membership in a shared material world. Then he says the American dream and the legitimacy of American dominance have long rested on technology becoming ordinary, and he makes affordabilité the test of national technology policy. It reads like a plan for flyover America written in the Mamdani moment at the close of 2026.

The essay is addressed to the world, but even when it looks outward, it is parochial. The world may have better historical analogues and present opportunities to put digital commons under public governance farther from the imperial centre, but Clark’s plan rarely leaves home. “America” turns up 22 times; Canada never appears. Glissant is cited three times, but Martinique, the island his whole thinking grew out of, never appears. The rest of the world shows up on Clark’s terms: as rival, labourer, or crowd. China is the rival to beat, and every mention of it is about AI. Kenya supplies data labellers doing what Clark calls clerical work “of an almost colonial kind.” Nairobi, Dhaka, and Mexico City lend him students who have shaken their governments. Clark rejects “beating China” as a reason to protect monopolies, then uses the same rivalry to sell his co-ops.⁴⁷

So the plan sets sovereignty aside as a fetish while leaning on one state’s institutions. What role would anyone outside the United States have in setting the standards or governing the shared infrastructure? The students Clark counts on are elsewhere; the institutions he wants them to force open are American, run by the administrators he admits are compromised. When North American students did occupy their campuses in 2024, many of those administrators called in the police, and more than 2,000 people were arrested in a few weeks.⁴⁸ The encampments echoed Occupy Wall Street, which filled parks and plazas in 2011 and stands in most memories as a strange failure snatched from the jaws of success. What Clark’s plan offers everyone outside the United States is opacity: the right not to be fully legible to systems whose keys they will never hold. That right matters, but how exactly it would coexist with the Intelligence Community — eighteen federal agencies — and the Five Eyes alliance, Clark does not say.

Illich knew this arrangement well. In April 1968 he told a room of American college students bound for summer projects in Mexico that they could not help being salesmen for the American way of life, and that good intentions had little to do with it. “To hell with good intentions,” he said, and the line became the speech’s title.⁴⁹ Clark’s intentions are good. But a public intelligence that has to be guaranteed by reformed American universities, American courts, and an American-style clearing house that names state surveillance but never says how the system would resist the state’s demands for access is still help administered from the centre. A plan that took usufruct seriously would start somewhere else: with tools small enough for people to hold, check, and set down wherever they live, without a warrant from anyone beyond their immediate community.

Open Questions

• At what size, speed, or energy draw does “public intelligence” become counterproductive, whoever owns it?
• What knowledge stays outside the warrant market, used by custom and never priced?
• Who can check a warrant without hiring an expert?
• If universities hold the keys, what stops them from becoming the next monopolistic platform?
• What would it mean, in Clark’s plan, to build less?

Notes

1. Rural power co-ops brought the infrastructure for affordable electricity to rural regions that private utility companies ignored as unprofitable. In the mid-1930s, nine out of ten rural homes had no electric service; by 1953, more than 90 percent of US farms had electricity. National Rural Electric Cooperative Association, “History”. ↩︎
2. Sebastian Clark, Plan for Public Intelligence (Isolarii, 2026), author’s note and introduction. See the Counterfoil catalogue entries on Local and Cloud LLMs. ↩︎
3. Illich held many university posts but only briefly, first as vice-rector of the Catholic University of Puerto Rico in the 1950s and later as a visiting professor at Fordham University, UC Berkeley, the United Nations University, the University of Marburg, the University of Hagen, Pennsylvania State University, and the University of Bremen. He was never in a tenured or tenurable role, and he stayed suspicious of universities all his life. See Todd Hartch, The Prophet of Cuernavaca: Ivan Illich and the Crisis of the West (New York: Oxford University Press, 2015). ↩︎
4. Clark, Plan for Public Intelligence, secs. I(a) and I(b). In England and Wales, section 60 of the Youth Justice and Criminal Evidence Act 1999 repealed the rule that computer evidence had to be shown to be reliable, and courts came to presume that computers worked properly unless shown otherwise. ↩︎
5. The term “algorithmic governance” (and its French counterpart “gouvernance algorithmique”) dates back to early theoretical work on software codes and technical architectures regulating social behaviour, with Antoinette Rouvroy and Thomas Berns using the latter term in 2009, and Inga Müller-Birn, Jakob R. Dobusch, and James D. Herbsleb first conceptualizing the English term “algorithmic governance” in 2013 as a coordination mechanism opposed to social governance. Christian Katzenbach and Lena Ulbricht, “Algorithmic Governance,” Internet Policy Review 8, no. 4 (2019). ↩︎
6. Ivan Illich, Tools for Conviviality (London: Calder & Boyars, 1973), chap. 3, “The Multiple Balance,” sec. 6, “Frustration.” ↩︎
7. Édouard Glissant, Poetics of Relation, trans. Betsy Wing (Ann Arbor: University of Michigan Press, 1997), “For Opacity.” ↩︎
8. Clark, Plan for Public Intelligence, secs. II–IV and conclusion. ↩︎
9. Illich, Tools for Conviviality, chap. 3, “The Multiple Balance.” ↩︎
10. The Cooperative Extension System is a publicly funded educational network in the United States that connects land-grant universities with local communities. ↩︎
11. Illich, Tools for Conviviality, chap. 2, “Convivial Reconstruction.” ↩︎
12. Clark, Plan for Public Intelligence, paras. 118–25 (chips), 153 (purpose-limited sensing), and 145–46 (Ivrea). He also locates the value of public institutions in “their situatedness in people’s lives, rather than any operational scale” (para. 9). ↩︎
13. Illich, Tools for Conviviality, chap. 3, sec. 6, “Frustration.” In the introduction, Illich argues that when an enterprise grows past a certain point it “first frustrates the end” it was designed for and then threatens society itself; chapter 3 names six such threats: biological degradation, radical monopoly, overprogramming, polarization, obsolescence, and frustration. Illich’s own word for the good life a convivial society protects was not eudaimonia but Thomas Aquinas’s eutrapelia, graceful playfulness (introduction). ↩︎
14. Ivan Illich, “Silence Is a Commons,” CoEvolution Quarterly, Winter 1983. The line about computers and fences is the magazine’s display text over the talk, which was first given at the Asahi Symposium on Science and Man, “The Computer-Managed Society,” Tokyo, March 21, 1982. ↩︎
15. Clark, Plan for Public Intelligence, paras. 66, 74, and 81. On the claims of those represented in records, see para. 23; on domain bodies setting standards, paras. 76–77. ↩︎
16. Clark, Plan for Public Intelligence, sec. III. On refusal as the first act, see para. 117; on frontier models training under warrants, para. 104. ↩︎
17. Ivan Illich, Deschooling Society (London: Calder & Boyars, 1971), chap. 6, “Learning Webs.” ↩︎
18. Clark, Plan for Public Intelligence, para. 157; Wendell Berry, The Unsettling of America: Culture and Agriculture (San Francisco: Sierra Club Books, 1977), chap. 8, “Jefferson, Morrill, and the Upper Crust,” especially on the Hatch Act. Reviewing the book in 1978, Russell Kirk agreed that the land-grant colleges had betrayed Jefferson’s America: Russell Kirk, review of The Unsettling of America, by Wendell Berry, Birmingham News, June 4, 1978, reprinted in “Classic Kirk,” Russell Kirk Center for Cultural Renewal. Jefferson’s words are from his letter to John Adams, October 28, 1813, describing his plan for public schools in Virginia. The line often credited to him, “an aristocracy of achievement arising out of a democracy of opportunity,” I can’t find anywhere in his writings; the best-known modern restatement of his idea is James B. Conant, “Education for a Classless Society: The Jeffersonian Tradition,” Atlantic Monthly, May 1940. ↩︎
19. Clark, Plan for Public Intelligence, sec. V, paras. 149–51. The prices are the essay’s own figures for October 2026: about $1.90 at 85 percent utilization and $3.20 at 50 percent. ↩︎
20. Ivan Illich, Energy and Equity (London: Calder & Boyars, 1974), chap. 1. ↩︎
21. Clark, Plan for Public Intelligence, paras. 151 (utilization and price) and 154 (the self-reinforcing loop). ↩︎
22. In 2018 the Commodity Futures Trading Commission found that Kooima & Kaemingk Commodities, an introducing broker in Sioux Center, had made unauthorized trades in its customers’ accounts that caused net customer losses of about $11.9 million between 2012 and 2016: U.S. Commodity Futures Trading Commission, In the Matter of Kooima & Kaemingk Commodities, Inc., Lauren Kaemingk, and Bradley Kooima, September 26, 2018; “Sioux Center Commodities Brokers Fined,” N’West Iowa Review, 2018. In 2021 two former officers of Sioux Center’s Farmers Cooperative Society went to federal prison for directing workers, from 2011 to 2017, to blend cheaper oats into soybeans and hide it from customers and USDA inspectors: Associated Press, “Ex-Official Gets Prison Time for Iowa Grain Blending Scheme,” February 2021; “Ex-Sioux Center Co-op Officer Gets 3 Months Prison for Scheme,” Sioux City Journal, 2021. ↩︎
23. Sioux Center has run its own municipal electric utility since 1949, first on diesel and, from the 1950s, on hydro power from the Missouri River dams; in 1965 its mayor and city manager helped start what is now Missouri River Energy Services, through which the town still buys its wholesale power: City of Sioux Center, “Department History” (Electric); Dave Den Herder, “City’s History of Leadership,” N’West Iowa Review, 2021. The town’s telephones came from the shareholder-owned Mutual Telephone Company, founded in 1906, which as Premier Communications began laying fibre across northwest Iowa in the mid-1990s: profile of Premier Communications, Sioux City Journal. Ross Vernon managed Mutual from 1937 to 2002 and sat on the city council from 1946 to 1998, through the hydro connection, the water towers of the 1960s and 1986, and the early planning for the Lewis & Clark regional water system: Ross K. Vernon obituary, December 2011. ↩︎
24. Elizabeth Dias, “‘Christianity Will Have Power,’” New York Times, August 9, 2020, which reports that only 11 percent of Sioux County Republicans caucused for Trump in 2016 and 81 percent of its voters chose him that November; Timothy P. Carney, Alienated America: Why Some Places Thrive While Others Collapse (New York: Harper, 2019), and “The Most Interesting County in Iowa: Trump’s Worst County and His Best County, in One Place,” Washington Examiner, 2019, which puts the county at 46 percent Dutch by ancestry, the highest share in the country, and its churches at the centre of community life. ↩︎
25. Iowa Secretary of State, 2024 General Election Official Canvass by County: in Sioux County, Donald Trump received 16,053 votes (84.49 percent) and Kamala Harris 2,626 (13.82 percent). ↩︎
26. North West Rural Electric Cooperative, “Electric Rates”, single-phase rates effective June 1, 2026; City of Sioux Center, “Rates”. In Sioux Center, commercial water over 40,000 gallons a month costs $5.60 per thousand gallons rather than $6.10, and industrial water over 100,000 gallons costs $4.90 rather than $5.60. ↩︎
27. Iowa Department of Natural Resources data reported in The Gazette (Cedar Rapids) count 472 medium and large animal feeding operations in Sioux County, the most of any Iowa county. Sioux County farmland averaged $15,983 an acre in Iowa State University’s 2025 Land Value Survey, in the district with the state’s highest values, and one parcel there sold for a record $32,000 an acre: KIWA Radio, “Ag Land Record Set in Sioux County.” ↩︎
28. Christopher S. Jones et al., “Iowa Stream Nitrate and the Gulf of Mexico,” PLOS ONE 13, no. 4 (2018): e0195930. On average Iowa contributed 29 percent of the nitrate load of the Mississippi–Atchafalaya River Basin and 55 percent of the Missouri River Basin’s. ↩︎
29. On the shallow wells: Kate Payne, “Study: Nitrate Contamination in Water More Likely to Affect Lower Income Communities in Iowa,” Iowa Public Radio, June 28, 2021, on the Lyon and Sioux Rural Water System, whose nitrate has risen toward the federal limit of 10 mg/L and which blends its water because it cannot afford nitrate removal. I did the soil probes for the project described in Iowa Farm Bureau, “Farmer Works to Improve His Community’s Water Supply,” March 26, 2014, on nitrate in Sioux Center’s shallow city wells. On the deep aquifers: Robert Rowden, Groundwater Quality Evaluation of the Dakota Aquifer in Northwest Iowa, Water Resources Investigation Report 1C (Iowa City: Iowa Geological and Water Survey, 2008), which finds the water very hard across most of the region and detects gross alpha radiation in 65 of 73 samples; and the University of Iowa’s Center for Health Effects of Environmental Contamination, “CHEEC Supported Study Finds Radium in Private Well Water,” which found the highest radium levels mostly in northwest Iowa. Sioux Center says its own wells were “limited by the available water and mineral content” of the local aquifers until it began blending in piped water from the Lewis & Clark Regional Water System in 2023: City of Sioux Center, “Lewis & Clark Water Makes Difference in Sioux Center,” April 24, 2024. One field study near Sioux Center measured the soil nitrate left after harvest, which can leach into those wells, under five cropping systems from 2009 to 2013; I helped collect its data: Robert L. De Haan, Matthew A. Schuiteman, and Ronald J. Vos, “Residual Soil Nitrate Content and Profitability of Five Cropping Systems in Northwest Iowa,” PLOS ONE 12, no. 3 (2017): e0171994, https://doi.org/10.1371/journal.pone.0171994. ↩︎
30. USDA National Agricultural Statistics Service, 2022 Census of Agriculture, County Profile: Sioux County, Iowa: 221,894 acres of corn for grain and 40,668 acres for silage, about 58 percent of 454,943 acres of cropland. The census counts cover crops on 27,896 acres, about 6 percent of cropland: 2022 Census of Agriculture, Iowa county tables. Overlaying USDA’s Cropland Data Layer maps for 2021 and 2022 shows about 85,000 acres in corn both years, roughly 19 percent of the census cropland; the 2023–24 pair is similar. These figures do not show which acres had both consecutive corn and no cover crop. On topsoil: Evan A. Thaler, Isaac J. Larsen, and Qian Yu, “The Extent of Soil Loss across the US Corn Belt,” Proceedings of the National Academy of Sciences 118, no. 8 (2021): e1922375118, which estimates that erosion has removed the carbon-rich A horizon from 35 ± 11 percent of cultivated land in its study area. ↩︎
31. Reuters, “US Farm Debt Is at a Record. Official Data Might Be Undercounting It,” October 8, 2026, which puts it at $605.1 billion. ↩︎
32. Todd Neeley, “USDA Signs $300M Deal With Palantir,” DTN, April 23, 2026. The $300 million agreement supports the USDA’s National Farm Security Action Plan and its Farm Production and Conservation agencies: the Farm Service Agency, the Natural Resources Conservation Service, and the Risk Management Agency, which handle farm loans, conservation payments, crop insurance, and disaster aid. ↩︎
33. Kiera Butler, “Tech Billionaires Want Christians to Believe in AI,” Mother Jones, April 16, 2026. ↩︎
34. In June 2026 the Sioux County Board of Supervisors put a one-year moratorium on new data centres and battery-storage plants in the county’s unincorporated areas; Woodbury and Plymouth counties passed their own the same month, and Ida and Emmet counties followed. The proposal that set it off was PathOne Data Centers’ plan for a site near Hull, starting at 10 megawatts. On October 27 the supervisors voted 4–1 to amend County Ordinance No. 1 with rules for any future project, including a 25-megawatt limit per site, limits on noise and light, and a decommissioning plan, and voted 4–1 again to lift the moratorium. A week later the county gave Zach Lahn, the Republican candidate for governor, 79.9 percent of its vote. Both Lahn and the winner, Democrat Rob Sand, had campaigned against tax breaks for data centres; Lahn also backed a pause on new construction and a property tax on data centres five times the industrial rate. An Emerson College poll found 62 percent of Iowa voters opposed to a data centre in or near their community. Siouxland Public Media, “More NW Iowa Counties Joining Woodbury in Placing Moratoriums on Data Centers”, KWIT, June 26, 2026; Siouxland Public Media, “Growing Trend: More Siouxland Counties Place Moratoriums on Data Centers”, KWIT, August 26, 2026; Sioux County Radio, “Residents Seek More Details on Potential Sioux County Data Center”, WOWT, August 12, 2026; KTIV, “Sioux County Approves Data Center Rules, Votes to Lift Moratorium”, October 28, 2026; KTIV, “Sioux County Approves Strict New Rules for Future Data Centers”, October 27, 2026; NewsNation, “Data Center Backlash: 5 Key Races to Watch in the 2026 Midterms”, October 2026; Decision Desk HQ, “2026 Iowa Governor”, county results, unofficial, updated November 9, 2026. ↩︎
35. Clark, Plan for Public Intelligence, para. 130. Auguste Kerckhoffs, “La cryptographie militaire,” Journal des sciences militaires 9 (January 1883): 5–38, and (February 1883): 161–91. The second of his six requirements for a military cipher is that it “must not require secrecy, and it must be able to fall into the hands of the enemy without inconvenience.” Claude E. Shannon, “Communication Theory of Secrecy Systems,” Bell System Technical Journal 28, no. 4 (1949): 656–715, at 662. The principle is why modern encryption standards are published and tested in the open. ↩︎
36. Illich, Tools for Conviviality, 32. ↩︎
37. Clark, Plan for Public Intelligence, sec. II. On the common index, para. 82. ↩︎
38. Illich, Tools for Conviviality, chap. 3. ↩︎
39. SWIFT is a member-owned co-operative incorporated under Belgian law, and it disconnects banks when European law requires it, as it did with Iranian banks in 2012 and major Russian banks in 2022. Henry Farrell and Abraham Newman show how the United States turns its jurisdiction over central nodes like SWIFT into coercive leverage: Farrell and Newman, “Weaponized Interdependence: How Global Economic Networks Shape State Coercion,” International Security 44, no. 1 (2019): 42–79. In 2022 the United States, the European Union, and their allies also froze about $300 billion of the Bank of Russia’s roughly $630 billion in reserves: Congressional Research Service, Russia’s War on Ukraine: Financial and Trade Sanctions, IF12062 (2023). Russia and China now say they settle more than 95 percent of their trade in rubles and yuan (Interfax, August 21, 2024). Yet after the United States authorized secondary sanctions on foreign banks in December 2023, banks in Turkey, China, the UAE, and Central Asia began delaying or refusing Russia-related payments: Cleary Gottlieb, “Impact of Recent U.S. Secondary Sanctions Authority Targeting Foreign Financial Institutions,” February 22, 2024. ↩︎
40. Édouard Glissant and Hans Ulrich Obrist, The Archipelago Conversations, trans. Emma Ramadan (Isolarii, 2021). ↩︎
41. Elena Clavarino, “India Ennenga and Sebastian Clark,” Airmail, April 2, 2022; “London Book Fair 2026: Isolarii’s Avant-Garde Ambitions Get Bigger,” Publishers Weekly, March 2026. ↩︎
42. Clark, Plan for Public Intelligence, author’s note. ↩︎
43. David Graeber, The Utopia of Rules: On Technology, Stupidity, and the Secret Joys of Bureaucracy (Brooklyn: Melville House, 2015). ↩︎
44. Wendy Hui Kyong Chun, Discriminating Data: Correlation, Neighborhoods, and the New Politics of Recognition (Cambridge, MA: MIT Press, 2021). ↩︎
45. Jo Guldi, The Long Land War: The Global Struggle for Occupancy Rights (New Haven: Yale University Press, 2022). ↩︎
46. Vitalik Buterin, “My Techno-Optimism,” November 27, 2023, published by Isolarii as My Techno-Optimism (2025). For discovery in the warrant market, Clark proposes 0xPARC’s PODs, a zero-knowledge proof format. ↩︎
47. Clark, Plan for Public Intelligence, secs. III–V and author’s note. In the essay, “America” and “American” appear twenty-two times and Canada not at all. Europe appears as France and as Ivrea, Italy (para. 145); Delhi and Belgrade appear only in the author’s note. “China” appears six times, all in passages on AI and technology; in section V Clark calls America’s energy deficit “its real competition with China.” Clark’s defence of “the transcendent and ecstatic” is in the conclusion. ↩︎
48. Ryan Pearson et al., “At Least 2,000 People Arrested in Pro-Palestinian Protests on US Campuses, AP Tally Shows,” Associated Press, May 2, 2024. ↩︎
49. Ivan Illich, “To Hell with Good Intentions,” address to the Conference on Inter-American Student Projects (CIASP), April 20, 1968. The usual headnote places the talk in Cuernavaca; Todd Hartch places it at the CIASP training session at Saint Mary of the Lake Seminary in Illinois. Hartch, The Prophet of Cuernavaca. ↩︎