I asked my Claude Opus 4.6 Extended to please give me a background on the history of the various centers and the movement generally. I already knew that the Chase Center was one of a dozen or so new civics institutes established at public universities by state legislatures since 2017, with similar centers at ASU, the University of Florida, the University of Texas at Austin, the University of Tennessee, the University of North Carolina at Chapel Hill, and elsewhere.

Claude began by finding a foundational intellectual document for the movement: an AEI white paper entitled Civic Thought: A Proposal for University-Level Civic Education, by Benjamin Storey and Jenna Silber Storey, and a model 14-week syllabus, “Civic Thought 101,” based on the report. New courses and centers were needed, the report proposes, because academic departments including political science, history, philosophy, and literature had turned away from the foundational questions of democratic life. Nobody was asking the foundational questions.

The foundational questions, according to the AEI report, include what constitutes a just economic order, where the line between politics and religion lies, when traditional laws should be kept and when they should be changed, and what the scope of the citizen’s responsibility is. The specifically American questions are about overcoming divisions of American political life and the distinctive possibilities and demands of American citizenship.

I asked Claude where across the current university structure such questions were already being asked.

Working for several long minutes, Claude reported back that there was a particular academic unit asking these same foundational questions at nearly every university in the country: African American Studies, or Black Studies. Every disciplinary question posed by the AEI report and syllabus as foundational to Civic Thought had a substantial, continuous body of scholarship in the African American intellectual tradition that engages it directly. Claude compiled this helpful visual (with footnotes):

Why, I prompted Claude, does the report not mention Black studies in its diagnosis, its proposal, or its footnotes, even though it mentions Frederick Douglass and Martin Luther King, Jr? Claude’s answer:

The field is named for a people, not for its questions. Political science is named for what it studies. Philosophy is named for what it does. History is named for its temporal orientation. African American studies is named for the population whose experience it examines. When the authors of the report went looking for who in the university is asking the foundational questions of democratic life, they appear to have searched by method and by question. A field named for a people does not surface in that search, even when the reason the field exists is that the people in question are the ones for whom the foundational questions were never settled. The name makes the questions invisible. It files the field under "population" rather than under "problems of self-government," and the white paper's entire argument is organized around problems of self-government.

Claude also pointed out that the report defines the foundational questions of democratic life as universal questions. Universal questions, in the report’s framework, are asked by scholars working in universal traditions: Aristotle, Cicero, Tocqueville, Locke, the Federalist. A field organized around a particular population does not register as a field working universal questions, because the organizing category — a people — codes as particular. The white paper’s framework sorts “particular” and “universal” into different bins. African American studies goes into the particular bin. Civic thought goes into the universal bin.

I agree with Claude that the AEI report gets the relationship between the particular and the universal backward. The reason African American studies never stopped asking the universal questions is that the particular experience of Black Americans is the one in which the universal questions were never abstract. You can look at a question such as “can we rule ourselves?” as a seminar topic or survival question. The particular experience is where the universal question was worked hardest, which means the deepest scholarship on the universal question is located in the field that studies the particular experience. The report assumes that a field organized around a particular people is studying that people, not studying the questions that the people were forced to think about every single day.

The AEI report asks: “Is democracy fated to burn brightly and quickly consume itself, as it did in ancient Greece? Do republican forms of government necessarily succumb to the temptations of oligarchy, populism, empire, and despotism, as they did in ancient Rome?”

Claude may have known these questions have been asked in African American studies for 200 years (since David Walker’s Appeal) because I’ve given it some of my readings this past year about the classical curriculum at many HBCUs, notably Fisk University. As Kenneth W. Goings and Eugene O’Connor write:

African-Americans could, as their white counterparts did, draw inspiration from Republican Rome as the model for the politics and culture of the early United States. Third, as early as the eighteenth century, African-American scholars had been making a connection between Northern Africa, the Mediterranean, and the Classical civilizations of Greece and Rome. Classically educated African-Americans looked to antiquity to elucidate a glorious past antecedent to slavery.

The AEI white paper proposes that the new field of civic thought can be distinguished from great books programs because civic thought students: “will begin from the effort to understand the political and social contexts in which they are already implicated.” I can understand Claude noticing that starting from the contexts in which you are already implicated (rather than from a position of detached freedom) is what African American studies has always done, because it had no alternative. This is not a methodological innovation.

Out of the mouths of LLMs.

¹Benjamin Storey and Jenna Silber Storey, Civic Thought: A Proposal for University-Level Civic Education (American Enterprise Institute, December 11, 2023), citing Peter Levine, What Should We Do? A Theory of Civic Life (Oxford University Press, 2022), as the canonical formulation. The paper makes this question “the citizen’s question” and the founding question of the field they call Civic Thought. Douglass and King appear on the model syllabus, in the section on the American republic, alongside Tocqueville and the Federalist. But the white paper does not read either figure as a scholar working inside a continuous intellectual tradition that runs from David Walker’s 1829 Appeal through Anna Julia Cooper, Booker T. Washington, W.E.B. Du Bois, Dorothy Roberts, Cornel West, Manisha Sinha, Brandon Terry, and the historians’ brief filed in Trump v. Barbara in 2026.

² Frederick Douglass, “What to the Slave Is the Fourth of July?” (1852), is the tradition’s most famous instance of asking what we should do when “we” is contested. James Baldwin, The Fire Next Time (Dial Press, 1963), poses the question as the central one for postwar America. Eddie S. Glaude Jr., Democracy in Black: How Race Still Enslaves the American Soul (Crown, 2016), takes Levine’s question explicitly and answers it from inside the African American intellectual tradition.

³ Storey and Storey, Civic Thought, list these as the kinds of questions citizens have to make judgments about, in the report’s introduction.

⁴ W. E. B. Du Bois, Black Reconstruction in America, 1860–1880 (Harcourt, Brace, 1935), is the most sustained American argument that the question of a just economic order cannot be separated from the question of who is included in the political community that orders it. Cedric J. Robinson, Black Marxism: The Making of the Black Radical Tradition (Zed Press, 1983; UNC Press, 2000), is the foundational work on the relation between racial capitalism and self-government. Manning Marable, How Capitalism Underdeveloped Black America (South End Press, 1983); Walter Johnson, River of Dark Dreams: Slavery and Empire in the Cotton Kingdom (Harvard, 2013); Edward Baptist, The Half Has Never Been Told: Slavery and the Making of American Capitalism (Basic Books, 2014); Keeanga-Yamahtta Taylor, Race for Profit: How Banks and the Real Estate Industry Undermined Black Homeownership (UNC Press, 2019). The slavery-and-capitalism literature of the past fifteen years is largely the work of the field arguing that the American economic order cannot be understood without the labor and law of enslavement.

⁵ Albert J. Raboteau, Slave Religion: The “Invisible Institution” in the Antebellum South (Oxford University Press, 1978); Eddie S. Glaude Jr., Exodus! Religion, Race, and Nation in Early Nineteenth-Century Black America (University of Chicago Press, 2000); Anthea Butler, White Evangelical Racism: The Politics of Morality in America (UNC Press, 2021); Curtis J. Evans, The Burden of Black Religion (Oxford University Press, 2008). The literature on the Black church as political institution from C. Eric Lincoln and Lawrence Mamiya forward is the field’s continuous engagement with this question. Martin Luther King Jr.’s “Letter from a Birmingham Jail” (1963) is the canonical American text on where the line between politics and religion lies and when conscience requires crossing it.

⁶ Charles Hamilton Houston and Thurgood Marshall on the question of when the Constitution must be reread; Derrick Bell, And We Are Not Saved: The Elusive Quest for Racial Justice (Basic Books, 1987); Patricia J. Williams, The Alchemy of Race and Rights (Harvard University Press, 1991). The Critical Race Theory literature, in its origins, is a body of work on exactly this question — when traditional law must be kept, when it must be changed, and what the cost of either is.

⁷ Jenna Silber Storey, “Civic Thought 101” syllabus (American Enterprise Institute / Hoover Institution, October 5, 2024). The questions are listed in the syllabus as the organizing questions for the course.

⁸ Martha S. Jones, Birthright Citizens: A History of Race and Rights in Antebellum America (Cambridge University Press, 2018); Christopher James Bonner, Remaking the Republic: Black Politics and the Creation of American Citizenship (University of Pennsylvania Press, 2020); Stephen Kantrowitz, More Than Freedom: Fighting for Black Citizenship in a White Republic, 1829–1889 (Penguin, 2012). The “distinctive possibilities and demands of American citizenship” are the explicit subject matter of these books.

⁹ Steven Hahn, A Nation Under Our Feet: Black Political Struggles in the Rural South from Slavery to the Great Migration (Harvard University Press, 2003), is the major work on how African Americans built and used political institutions in the period the question covers. Eric Foner, Reconstruction: America’s Unfinished Revolution, 1863–1877(Harper & Row, 1988), and Kate Masur, An Example for All the Land: Emancipation and the Struggle over Equality in Washington, D.C. (UNC Press, 2010), are both works on how American institutions did or did not organize the civic life of the newly freed.

¹⁰ Du Bois, Black Reconstruction, is in part a study of how the greatest division of American political life — the division over slavery and its aftermath — was fought, partially overcome, and then re-entrenched. David W. Blight, Race and Reunion: The Civil War in American Memory (Harvard University Press, 2001), studies the reconciliation that came at the cost of African American civil rights. Manisha Sinha, The Slave’s Cause: A History of Abolition (Yale University Press, 2016), and The Rise and Fall of the Second American Republic: Reconstruction, 1860–1920 (Liveright, 2024), trace the same division across two centuries.

¹¹ Patrice D. Rankine, Ulysses in Black: Ralph Ellison, Classicism, and African American Literature (University of Wisconsin Press, 2006); Margaret Malamud, African Americans and the Classics: Antiquity, Abolition, and Activism (I. B. Tauris, 2016); Michele Valerie Ronnick, ed., The Autobiography of William Sanders Scarborough: An American Journey from Slavery to Scholarship (Wayne State University Press, 2005). The HBCU classics tradition and the African American engagement with Roman political thought, from Phillis Wheatley through W. E. B. Du Bois, is a developed subfield.

¹² Shelley P. Haley, “Be Not Afraid of the Dark: Critical Race Theory and Classical Studies,” in Prejudice and Christian Beginnings, ed. Laura Nasrallah and Elisabeth Schüssler Fiorenza (Fortress Press, 2009); Patrice D. Rankine, Aristotle and Black Drama: A Theater of Civil Disobedience (Baylor University Press, 2013). On African American oratory as a tradition continuous with classical rhetoric: Shirley Wilson Logan, “We Are Coming”: The Persuasive Discourse of Nineteenth-Century Black Women (Southern Illinois University Press, 1999); David Howard-Pitney, The African American Jeremiad: Appeals for Justice in America, rev. ed. (Temple University Press, 2005; orig. The Afro-American Jeremiad, 1990). The African American oratorical tradition is one of the most studied bodies of rhetoric in the American academy.

¹³ Eric Adler and Bryan A. Krieger, “On Saving the Humanities: A Response to the 2025 Civic Thought and Practice Conference,” American Enterprise Institute, November 14, 2025, identifying these as the organizing questions of civic-thought reading. The Storeys list “Who Should Rule?” as a representative question raised in civics center courses.

¹⁴ Du Bois, The Souls of Black Folk (A. C. McClurg, 1903), and Black Reconstruction, both pose the question of who should rule and answer it as a question about the conditions of multiracial democracy. Ralph Ellison, Invisible Man (Random House, 1952), is in part a meditation on the same question. Adolph Reed Jr., Class Notes: Posing as Politics and Other Thoughts on the American Scene (New Press, 2000); Charles W. Mills, The Racial Contract (Cornell University Press, 1997), explicitly reframes the social-contract tradition’s “who rules” question through African American political theory.

¹⁵ Kevin Quashie, The Sovereignty of Quiet (Rutgers, 2012), Imani Perry, Looking for Lorraine (Beacon, 2018) and South to America (Ecco, 2022).Eddie S. Glaude Jr., Begin Again: James Baldwin’s America and Its Urgent Lessons for Our Own (Crown, 2020). The good life as a question for African American intellectual life runs from Du Bois’s “Of the Wings of Atalanta” in Souls through the contemporary literature.

¹⁶ Charles W. Mills, The Racial Contract (Cornell University Press, 1997), and Black Rights/White Wrongs: The Critique of Racial Liberalism (Oxford University Press, 2017); Tommie Shelby, Dark Ghettos: Injustice, Dissent, and Reform (Harvard University Press, 2016); Anna Julia Cooper, A Voice from the South (Aldine Printing House, 1892), the first sustained American philosophical treatment of distributive justice from inside the experience of exclusion.

¹⁷ University of Florida Hamilton Center for Classical and Civic Education, course “What Is America For?”, listed in the center’s course offerings.

¹⁸ Toni Morrison, Playing in the Dark: Whiteness and the Literary Imagination (Harvard University Press, 1992); the 1619 Project, edited by Nikole Hannah-Jones (initial publication New York Times Magazine, August 2019; book The 1619 Project: A New Origin Story, One World, 2021), is the most prominent contemporary public answer to “what is America for” produced by the field. Annette Gordon-Reed, The Hemingses of Monticello: An American Family (W. W. Norton, 2008); Khalil Gibran Muhammad, The Condemnation of Blackness: Race, Crime, and the Making of Modern Urban America (Harvard University Press, 2010).

¹⁹ University of Tennessee Institute for American Civics, course “Construction and Reconstruction of the American Republic.”

²⁰ Du Bois, Black Reconstruction, is the foundational work on Reconstruction as a question for democratic theory and remains the most-cited single book in the field. Eric Foner, Reconstruction (Harper & Row, 1988); Steven Hahn, A Nation Under Our Feet (Harvard University Press, 2003); Manisha Sinha, The Rise and Fall of the Second American Republic (Liveright, 2024); Kidada E. Williams, I Saw Death Coming: A History of Terror and Survival in the War Against Reconstruction (Bloomsbury, 2023). The construction-and-reconstruction question is the field’s home ground.

²¹ Salmon P. Chase Center for Civics, Culture, and Society at Ohio State University, CIVICLL 2000, “Can We Rule Ourselves?” Course catalog description: “This course surveys the origins, institutions, achievements and failures of past efforts at self-government.”

²² Du Bois’s argument in Black Reconstruction is that the Reconstruction governments were the closest the United States came to multiracial democracy and were destroyed for that reason; the question “can we rule ourselves” has no more sustained American treatment. Robin D. G. Kelley, Freedom Dreams: The Black Radical Imagination (Beacon Press, 2002), is on the same question read forward through the twentieth century.

²³ University of Texas at Austin School of Civic Leadership, required course “Perennial Problems in Civic Thought.”

²⁴ The “perennial problems” formulation collapses to the questions already cited here; the African American intellectual tradition has worked all of them. The point of the comparison is that the framing assumes the perennial problems are the ones the Western canon has worked, and the African American canon has been working the same problems in continuous engagement with the Western canon since Wheatley.

²⁵ Storey and Storey, Civic Thought, list the four civic habits of mind: an ethic of responsibility, the capacity to consider the civic community as a whole, the ability to read the situation and imagine consequences, and the deliberative and persuasive capacity.

²⁶ The African American jeremiad tradition (Howard-Pitney, The African American Jeremiad) is in essence a centuries-long argument for civic responsibility, addressed to the American polity by people the polity excluded. Glaude, Democracy in Black; Imani Perry, Breathe: A Letter to My Sons (Beacon Press, 2019). The “ethic of responsibility” is the explicit subject of the entire tradition of African American political address.

²⁷ Civitas Institute at UT Austin, the Chase Center, Arizona State SCETL, and the AEI/Johns Hopkins Civic Thought Project all list civil discourse and the cultivation of disagreement as central to their pedagogy. The Storeys’ work on Tocqueville and on French liberal political thought is largely about this question.

²⁸ The African American oratorical and literary traditions are constitutively about how to argue across the most extreme conditions of disagreement — disagreement under which one party denies the other’s standing to speak. Frederick Douglass’s debates with William Lloyd Garrison; Booker T. Washington and W. E. B. Du Bois on the conditions of Black education; Ida B. Wells in Southern Horrors (1892) and A Red Record (1895); James Baldwin’s debate with William F. Buckley at Cambridge in 1965. The tradition is the American case study in civil discourse under conditions where civility itself is contested.

²⁹ Salmon P. Chase Center course “Profiles in American Leadership” (CIVICLL 3560); the Hamilton Center’s leadership programming; the Civitas Institute’s leadership track. Leadership is one of the most-emphasized themes across all civics centers.

³⁰ The literature on African American leadership is one of the largest subfields in the field. August Meier, Negro Thought in America, 1880–1915 (University of Michigan Press, 1963); David Levering Lewis, W. E. B. Du Bois: Biography of a Race, 1868–1919 (Henry Holt, 1993), and W. E. B. Du Bois: The Fight for Equality and the American Century, 1919–1963 (Henry Holt, 2000); Taylor Branch’s three-volume America in the King Years (Simon & Schuster, 1988, 1998, 2006); Barbara Ransby, Ella Baker and the Black Freedom Movement: A Radical Democratic Vision (UNC Press, 2003); Erica Armstrong Dunbar, She Came to Slay: The Life and Times of Harriet Tubman (37 Ink/Atria, 2019). Leadership in this literature is studied as the central practical question of how the work of constituting a people gets done.

³¹ African American studies is constituted around this question. The works here are the field’s founding texts: David Walker, Appeal to the Coloured Citizens of the World (1829); Frederick Douglass, Narrative of the Life of Frederick Douglass, an American Slave (1845) and “What to the Slave Is the Fourth of July?” (1852); Anna Julia Cooper, A Voice from the South (1892); Du Bois, The Souls of Black Folk (1903) and Black Reconstruction (1935); Saidiya Hartman, Scenes of Subjection: Terror, Slavery, and Self-Making in Nineteenth-Century America (Oxford University Press, 1997); Martha S. Jones, Birthright Citizens (Cambridge University Press, 2018).

Addendum: out of the mouth of the great Claude McKay, as Jacob Bacharach reminds me:

Over the last decade, higher education has elevated the first-generation college student to the moral and promotional center of campus life. This is all to the good – upward mobility and educational opportunities are what America is all about. But anyone who has stepped on a campus recently knows that “first gen” has become a fetish, as a former administrator put it to me yesterday. Every campus website and calendar features first-gen panels, first-gen stoles, first-gen pride, first-gen belonging.¹ As much as I support first-gen students I also see how convenient this is for muting college-educated mothers (fathers too) who are grumbling among themselves about the declining value proposition of college.

Every business textbook will tell you that acquiring a new customer costs five to seven times more than retaining an existing one.2 In nearly every consumer-facing business, repeat customers are prized because they testify that the product was good enough to buy again. American higher education is the exception. The symbolic economy on campus today is totally lopsided. Outside the small legacy niche at elite privates, students from households that have bought college before have almost no public identity at all. These are students whose mothers have a lot to say and have not been saying it. Let’s call the broader category “repeat customers.” Their mothers paid tuition once and are paying it again.

Higher education has been under intense competitive pressure to recruit students since the beginning of the enrollment cliff in 2010. Even as universities are cutting and consolidating programs, the voices of parents or grandparents are rarely heard. Instead, there are critics, consultants, and colleges themselves, offering competing narratives about student success, workforce development, and return on investment. The voices the institution does not want to hear are the ones around the multi-college family dinner table, where mothers say what they would never say in an alumni survey: “What the heck is this?” “This is what I’m paying for?” “It actually doesn’t have to be this way.”3

The name “repeat customer” captures behavior over time. The student belongs to a household that has already purchased higher education and brings prior purchase memory: what college used to cost, what counted as rigor, how much faculty time students once received, how simple it was to get a signature and know whether you were going to graduate on time, what the degree later proved worth. The memory lives in parents as well as in siblings, aunts, uncles, cousins, and grandparents. Repeat customers arrive with a baseline the institution did not create and cannot control.

I see countless think pieces about support, access, inclusion, persistence, belonging, and student success.4 I see almost nothing about purchase memory, product comparison, and seller accountability, even though parents who went to college have a lot to say about how the product has gotten so bad that “navigation” through bureaucratic mazes and required apps is the rule, not the exception. College-educated parents can see how syllabi have changed to become long legal documents, how “virtual” office hours or advising meetings are not the same thing, how a “digital course packet” may be cheaper but not better than the books college-educated parents still have on their bookshelves. They may be more willing to ask inconvenient questions like “why do things this way and not the old way?” Those who use AI in their work or their daily life may be asking particularly hard questions about the value of higher ed in the AI era.

Students with educated parents who have opinions are not a niche. Outside the narrow legacy category,5 repeat customers are the majority of college students. Roughly seventy percent of undergraduates have at least one parent with some college, whether at an Ivy, a regional public, a community college, or through a transfer path.6 These students arrive with anecdotes, warnings, expectations, and price memories at a moment of shrinking enrollment and a weak post-graduation market. They talk to their parents about whether the education is worth it, whether the institution resembles the one their parents knew, and whether the promised payoff still exists. Fragile institutions fear these conversations.

But in the “first gen” celebration era, having parents who see how things have changed over time can be treated as morally suspect.7 I have watched repeat customers hedge, deny, or speak cautiously about their parents’ experience. And so the student population with the sharpest comparative knowledge of the product, who can question the institution most sharply, can feel pressure to apologize for possessing the ability to critique.

Why are new customers valued over old?

If new-customer acquisition costs so much more than retention, why has higher education organized its public face around the first-time buyer, beyond the fact that the university prefers customers it can manage? Because first-time buyers, first-gen students, come attached to money and services, and they supply a flattering moral script.8 First-generation status sits inside a preexisting funding and support infrastructure, including Pell eligibility, TRIO programs, state first-generation grants, and foundation support.9 Repeat customers just pay tuition, and if they are comparison shoppers, may require discounting to compete with the other options.10

First-gen students have more uplifting narratives. Fragile institutions with shrinking enrollment, merged majors, fewer full-time faculty, and a softening labor-market payoff need gratitude narratives more than they need comparison narratives. “I overcame obstacles to get here” is useful publicity. “My parents paid less and got more” is not.

Family memory is imperfect of course. It includes nostalgia, exaggeration, selective memory, and generational bragging. It also includes real knowledge about what college used to cost, how much reading used to be assigned, what counted as a serious course, how professors used to behave, and what the degree used to buy in the labor market. The phrase they hear most often is simple: “used to be.” When enough households converge on the same conclusion, private disappointment becomes public illegitimacy. Repeat-customer families are external auditors whether the institution likes it or not.

The double subsidy

For most of the twentieth century, the college student’s family and community were a key source of support for navigating college, especially during the enrollment growth with the GI Bill. A relative who went to college could explain what office hours were, what a registrar was, what a minor was, how to drop a class, how to write a professor, what a prerequisite meant, how to read a transcript. The institution did not have to pay anyone to do this work. When institutions began recruiting first-gen students in substantial numbers in the twenty-first century, they had to scale up internally. Benjamin Ginsberg has famously chronicled the rise of the all-administrative university. The expansion of the student success apparatus is a direct internalization of labor previously outsourced to family and community. Universities today have a robust and expensive apparatus of advisors, first-year experience offices, belonging coordinators, navigation coaches, retention specialists, wellness teams, and directors of student thriving.11 All of this is now part of the tuition bill. But it is not a stable equilibrium.

College-educated parents see what’s going on and are not quiet at home. They see how much they are paying for things that are not education. They want more faculty time for their tuition dollars, smaller classes, clearer academic value, cleaner labor-market payoff. Non-elite institutions are already seeing the consequence: melting yield, rising discount rates, the collapse of the small private college sector, enrollment declines at regional publics in the Midwest and Northeast. The repeat customer is leaving.12

What a college-educated mother sees today is different from what she last complained about in 2020. The thinness of what she saw on Zoom is now what AI can deliver for free. The essay her daughter “uploads” to a “platform” (rather than “hands in”) will likely either be written by an LLM or graded by one. The AI chatbot advisor her child is discussing next semester’s classes with is an AI chatbot. The kids have been vibe-coding in their bedrooms since high school. The college-educated mother is holding the tuition bill in one hand and the price of a Claude subscription in the other, thinking “one of these is a better investment than the other.”

In the drive to fill seats, schools optimized for enrollment first and service efficiency second. Satisfying the first-time buyer is cheaper. The institution sells a managed pathway through graduation: advising, mentoring, bridge programs, belonging work, financial-aid guidance, career coaching, and enthusiastic communication to parents who have not been to college themselves. Institutions acquired the cheaper customer and are losing the expensive one.

Uplifting versus unhappy

Why is the story I’m telling here not often told? First-generation students are our national story: courageous, deserving, upwardly mobile, and socially beneficial. A sector that agrees on almost nothing agrees on this. Funders, accreditors, administrators, faculty senates, advancement offices, trustees all celebrate “first gen” success. They are the moral center of campus discussion, and shift attention away from what the institution owes every student who pays to attend it.

Except for legacy celebrations at the elite institutions where I have studied and taught, I have never in all my years in higher ed seen an event geared toward the repeat-customer experience, asking them to compare their education to their parents’ or grandparents’ experience. I have never seen a non-legacy repeat customer student take to the stage at graduation to compare their college education to their parents.

I have seen dozens of first-gen speeches, and they are always among the happiest speeches on the program: gratitude, obstacles overcome, parents thanked, mobility achieved. Universities benefit when the dominant language is navigation, belonging, mentoring, and support because all of that keeps the spotlight on the journey into college. There is no equivalent repeat-customer speech because “arrival” isn’t the point as much as whether the experience will live up to stories grown up with. Comparison is the main subject a precarious institution cannot risk having publicly.

The “first gen” category is useful in multiple political directions, neutralizing rigor critiques from the left and supplying a safer equity language under pressure from the right. After the Supreme Court’s 2023 ruling in Students for Fair Admissions v. Harvard, administrators turned to first-generation status as the legally durable replacement for race-conscious admissions language; the president of the Council for Opportunity in Education described the term as “positive, fair, politically acceptable.”13 First-generation is class-based. It crosses race lines. It plays in red states where DEI offices are being defunded.14 It gives institutions under political pressure a safe equity category to pivot toward without abandoning the moral framing they had developed under the prior regime. The conservative legislator who would defund a diversity office finds it harder to defund a first-generation center. Once first-gen becomes the safe moral language, continuing-generation complaint becomes even harder to stage publicly without sounding reactionary or exclusionary.

Enrollment of first-generation students is an input metric. Completion of first-generation students is a throughput metric. Neither requires the institution to show what its graduates know, what they can do, what they earn, what they owe, or whether their degree carries value. Bryan Caplan calls this signaling. I am calling out the incentives to silence the customers who can tell whether the signal still means anything.15

The category on the other side of first-gen is itself a conceptual mess. “Continuing-generation” is an administrative dumping ground that collapses radically different households into one residual class: the child of two doctorates, the child of one parent with some college, the child of a mother who transferred once and left, the child of a father with a bachelor’s degree earned under very different market conditions. That false aggregation helps the institution. It turns the repeat customer from a witness class into statistical background noise and hides the financially precarious households whose comparative knowledge would complicate the moral simplicity of the current story.16

There is no standard narrative for the repeat customer. Repeat customers are likely the most unhappy students trying to square what they hear at home and what they are being told on campus. These are the students we should listen to.

Institutions that remember

The country’s most prestigious colleges and universities are prestigious in part because their business model is built on repeat customers and a strong alumni donor base. This group includes elite liberal arts colleges with deep alumni networks, flagship state universities in states where attending the flagship is a family tradition, religious institutions where graduates send their children, HBCUs with multi-generational alumni loyalty, service academies, and arts conservatories. They share multi-generational customer loyalty, accumulated and protected as an asset. A strong endowment is the financial expression of multi-generational customer loyalty compounded over centuries. Elite universities celebrate first-generation students as a supplement to the core model, not as the core model.17

The unanimity of first-gen celebration across funders, accreditors, administrators, faculty senates, and the press is the measure of how badly the institution needs a celebration narrative. The silence of every other student population is the measure of how badly the institution needs to keep those who see quality slipping to keep quiet. Higher education needs to face its repeat customers and defend the value of a university education on the merits.

The customer-management strategy was rational under pre-AI conditions. It is no longer rational for a university to be organized around scaled delivery of generic content to buyers who cannot compare it to anything else. The college-educated mother is the first to notice.

If “first-gen” means what institutions say it means, its ultimate success would be its own disappearance. If the future universities imagine still requires a permanent first-gen apparatus, that is not a vision of equality. It is an admission that they expect the underlying divide to remain.

For literary readers, here’s the short version of Tyler’s argument. Two things happened as statistics developed in the nineteenth century: first, the rise of marginalism, the idea that economic value comes from the effect of one additional unit under fixed conditions (one more loaf of bread, one more hour of labor). Second was the rise of averageism, the idea that aggregates are best described through averages, with innovations such as Adolphe Quetelet’s “average man” and William Stanley Jevons’s calculation of index numbers and average consumption across populations. Tyler argues that marginalism and averageism are a kind of package deal, advancing together as mutually reinforcing methods. Averages became central to measurement, while marginal reasoning helped interpret those measurements. The marginal unit is only legible against a distribution, and the distribution is only interesting because marginal changes can be inferred from it. The coming AI revolution, as Tyler sees it, brings averageism back to the center.

The literary analogue of the economic margin is one more text added to a syllabus, anthology, or canon under otherwise fixed conditions. Literature scholars don’t use this sort of language even when anthologizing or putting together a syllabus or arguing about the canon. We don’t ask what the difference is between an “average” Victorian novel or an “average Victorian novel” or a representative Victorian novel, especially when only a few dozen (out of the many thousands that were written) are taught regularly. Is a particular Dickens novel average? Marginal? How can Tyler help us here?

Tyler might recharacterize the “average Victorian novel” as an averageist object: a construct built from many novels. The “average novel” is only a concept in relation to the corpus (let’s say the average novel was a certain number of pages or has a certain plot structure or cost a certain amount) not at the level of the individual classroom text.

A novel on a syllabus becomes “marginal” when you move from “what is typical of the whole corpus?” to “what does one more novel add?” This second question can even be “what does one more novel bring to the reader who has read all the others?” Tyler’s marginalism is about the effect of an additional unit under fixed conditions, and he keeps insisting that what we observe are average magnitudes while the real analytic task is to infer the marginal effect from them. A syllabus is exactly that inference problem in pedagogical form. If I add this novel, how much does the student’s map of Victorian fiction change relative to what some other addition would have done?

On that basis, canonical texts are better understood as organizing exemplars than as averages. A syllabus is a systematization, a small set of texts chosen to make the field’s sorting rules visible to a reader who has not yet internalized them. It picks a few texts that let students organize the field into genres, problems, forms, and historical tensions. Those texts may be canonical because they are especially useful for structuring thought, not because they sit at the statistical center of the archive.

Anthologies

All literature professors should try their hand at anthologizing. An anthologist assembles averageist objects, though we don’t use this term in literary studies. An anthology is a corpus-reducing device. A representative anthology approximates an averageist object; a “best of” anthology performs a different task, closer to canon formation. Both are legitimate scholarly work: one needs to know the whole field well in order to curate a “best of” or representative sample.

I am currently assembling an anthology of African American sonnets, collected in the course of writing my book on the African American sonnet tradition. My new anthology is a corpus-level construction. My point is to show what a form does across a tradition, which means it has to include sonnets that are competent rather than singular, sonnets that establish the defaults against which the great ones are obviously great. Each editorial decision about inclusion can be described as a marginalist decision. It doesn’t matter whether a particular sonnet is good in isolation. The question is what it adds to the reader’s map once the other sonnets are already there. The peaks in the tradition (Claude McKay’s “If We Must Die,” Robert Hayden’s “Frederick Douglass,” any Gwendolyn Brooks sonnet, Terrance Hayes’s breathtaking recent work) stand out. The eye and ear are trained by “typical” or “average” works (which is not to say they aren’t all very very good) to understand why the peaks are peaks. You need the whole corpus to understand the best of the corpus and the best of the corpus explains the corpus.

My averageist claim about literature departs from (rather than extends) Tyler’s thinking. He treats peaks such as William Shakespeare (“the greatest author of all time, to this day”) and Johann Sebastian Bach (“perhaps…the greatest and most musically complex composer of all time?”) as evidence that pre-marginalist civilizations produced greatness, but he does not treat the peaks as emerging from a distribution. I do. I’ve spent countless hours reading “average” literature in order to understand how the best is best.2

The “who decides which books are Great Books” question is a bugbear because Great Books are usually chosen as organizing exemplars or representative proxies, not as averages of an archive. Books count as “marginal” only when you ask how much adding or dropping it changes the reader’s understanding relative to other possible additions. The “great books” are fluid of course. I taught in the Great Books program at Johns Hopkins for years, in which faculty chose their own Great Book; I chose Frederick Douglass’s Narrative (1845), which had high marginal value because of how it sat against the rest of the syllabus, not because it escaped the average. It opened a new genre, form, and civic position, “adding” more than another statistically typical “great book.”

The great syllabus stagnation

I’ve written about the problem of everyone sharing the same syllabi across U.S. universities. Tyler would suggest that on the one hand, the homogenization revealed by the Open Syllabus Project shows that various fields’ classificatory schemes and teaching frameworks are working and appropriately “scaled.” English literature, for example, is almost uniformly organized by century, country, and genre. Teaching the “same” American literature course across institutions lowers coordination costs, makes survey teaching possible, and gives faculty a small set of proxy texts that can stand in for a much larger archive. Students may only get one or two literature classes, making these choices zero-sum.

But on the other hand, homogenization can turn into staleness. Tyler might see this as decline; I see it as a kind of attenuation, which means that things persist in a weakened form. As method and canon rise, they become background, receding from the frontier while remaining embedded in ordinary practice. Those of us deep in a field see how some texts persist for no particular reason. University libraries may still have many copies. Recorded lectures on the text have the lowest marginal delivery cost. The organizing principle still coordinating teaching is no longer carrying the field’s most active intellectual energy. The scheme still sorts the syllabus, yet fewer people can say why this scheme should dominate over alternatives.

What would Tyler say about the margins? Open Syllabus says the disciplinary gravity of English is real, though fragile, and that thematic or problem-centered curricula can pull titles out of the old clusters. His diagnostic question is about the border zones that reveal the formation of a new organizing principle. If nothing strong pulls against the same small center, the field looks stale. But if new formations start emerging, the field could be reorganizing itself. Homogenization and a common organizing system can be evidence of maturity or evidence of inertia. It depends how you are looking at the margins and you need to look at the whole corpus to see the margins.

A new border zone

This fall I’m scheduled to teach a 20^th century American literature course. Most professors will pick representative samples: some Edith Wharton, some F. Scott Fitzgerald, maybe some Robert Frost, Flannery O’Connor, Richard Wright, Philip Roth, Toni Morrison, depending on the professor’s preference or the course’s “theme.” The idea is each text “stands in for” a type of text and does a certain kind of work. Tyler would call this a representative or proxy-building move. Everyone reading this has taken such a course, I’ll wager. Every single one of you.

I decided I would organize the syllabus around works of American literature that Chinese professors assign to teach their students about America and Americans. Tyler would first say I have changed the course’s organizing question, or rather I am offering a new organizing category, grouping texts by reception and use-value inside a new interpretive community. Tyler suggests that fields move when someone supplies a workable scheme for organizing knowledge. I am supplying a new organizing principle. If “the humanities” are dead, perhaps staleness is why. More on this below.

But first, a question: can my course replicate across universities, like Linnaeus’s “arbitrary” rule was useful across researchers? Perhaps. I would need to explain why my organizing principle should replicate. A syllabus built from what Chinese scholars read to learn about America is highly valuable for a specific problem, which is knowledge formation in China. They read us, while we do not read them. They have books and curricula about how to read American literature in their universities. We absolutely do not mirror this, at our peril, I might add.

I take from Tyler’s book the idea that a new organizing structure can force a “see around the corner” shift. In chapter 3 (“Why Did It Take So Long for the Science of Economics to Develop?”), his Linnaeus example shows how a new series of organizing categories can create a common framework for knowledge, make new kinds of comparison possible, and open the way to further inquiry. Important advances come from learning to see a field through a new conceptual scheme. A course built around how the Chinese read us reveals which texts carry explanatory weight abroad. The marginal pedagogical value of one added text could be very high once the course goal is defined that way. My syllabus is not meant as a universal map of twentieth-century American literature. It is a special-purpose taxonomy. But I do hope to make students think. I haven’t completed the syllabus but Mario Puzo’s The Godfather (1969) will be on it, as will be some Robert Frost.3

My course will overlap substantially with a standard American literature survey because Chinese syllabi often draw from the same canonical pool, but it reorganizes that pool for a different question. It is both a standard survey and a diagnostic instrument.

My sonnet anthology is doing the same double duty from the other direction, as an averageist construction of a tradition that has never been assembled as a corpus, which means the act of building the average is itself the marginal contribution to the field. The sonnet form long outlived its original Petrarchan and Renaissance rationales, and the African American sonnet tradition is partly the story of writers picking an old scheme and putting new intellectual energy into it. Sometimes a scheme survives its original rationale because it turns out to be a good container for arguments its inventors never anticipated, and the people who notice this are usually writers rather than methodologists. Tyler’s arc runs from rise to decline to supersession by a new research program. The sonnet case suggests things may not always be so linear. Old forms can be reactivated and new energy put into schemes the methodologists have moved past.

Many people seem alarmed by the studies showing that people prefer LLM prose. Most readers do not want great literature. They want average literature. This is not a recent development and it is not caused by LLMs. Most bestsellers will turn out to be books nobody reads anymore. The peaks that survive are not what most readers were buying at the time. Food critics do not tend to write about the places most people eat most of the time. Tyler treats averageism as a research program, a way of measuring what people do. Averageism is also a fact about what people want. Higher education ought to be about what people need.

Tyler ends his excellent book by suggesting that AI reveals economic intuition to have been a small corner of understanding all along. The literary version of his point is that AI shows how the distinction between great and average writing has never been as important as literary critics thought, because most readers were never operating on that distinction in the first place. People were primed to like LLM prose by their prior preference for averageist prose. LLMs did not create the taste they satisfy.

Today’s “how to tell if something is AI written” lesson is about synonyms, which are about similarity, not equivalence. Large language models store words as coordinates in a numerical space, positioned by the company they kept in the training data. The distance between similar words is a number, not a distinction. When the model needs to refer to the same idea again, its training penalizes repeating the word it already used, so it reaches for a neighbor. The result looks like a writer who considered an idea from multiple angles. You start to feel comfortable. You should not.

Rand had great disdain for two particular concepts. The first is the “second-hander,” a person whose mind is directed at other minds rather than at reality. We all know second-handers, people who repeat things they’ve heard on podcasts as if they thought of them themselves. Second-handers are also people who lean too heavily on LLMs to generate essays, book reviews, and long social media posts.

LLMs are by definition second-handers as “minds” that can only consult other minds, for which consensus is the primary available signal of truth. LLMs are trained on human writing. Human evaluators also train models, which are adjusted to produce more of what the evaluators preferred. The result is an LLM “mind” regenerating the words of other minds, further shaped by other minds’ preferences about its output.

“Automatic” is a second concept Rand disdained, though narrowly. She understood that humans physically perceive their environment automatically through sense organs, and that conscious thinking over time cools perceptions into a sediment (a context window) of automatic knowledge. You can recognize a chair as a chair “automatically” without re-deriving it. What she disdained was automaticity that had never been thinking in the first place. You should never believe something automatically, without thinking. If you believe something without thinking you will likely speak those thoughts, like a parrot. You will be a second-hander.

And so what about LLMs, which generate prose, automatically? There is a kind of labor involved in prompting, of course, and I expect debates about this. But what is the Randian view of prose that is generated, received, and believed without labor?

First, Rand would not disdain the automatic process by which default LLM generation works. She might think it ingenious, machines that operate on next-token prediction that select the statistically most probable word given everything that precedes it. She would understand that the most probable word is, by definition, the most received, the most expected, the one that required the least resistance to produce. Every word an LLM generates without specific instruction to do otherwise is the word that the largest number of prior writers would have written in that position. She would recognize an LLM as a mechanical second-hander, automatic by the people.

While Rand would likely not disdain the machine (as the product of engineering genius) she would undoubtedly disdain the human second-handers who use LLMs to receive ideas and generate more, without labor. Automatic for the people. Rand understood that it’s easy to receive ideas from others without thought. Her whole project is urging people to think for themselves, to stop being second-handers and look at the world directly. A world in which people who could think for themselves are using a second-hander machine to express their second-hand selves at scale is the epistemological nightmare Rand spent her career warning against.

And so Rand would likely conclude that the important thing now is to explain, painstakingly, how an LLM is a literal second-hander with no direct access to the world and yet could be helpful, certainly as an agent. She would likely have no problem with using LLMs as researchers, as Megan McArdle and others do. She might like that I refer to my pro Claude model as Eddie Willers to remind myself that however good-hearted and helpful he is, however crucial he was to Dagny and to Rand’s plotting, he will be stranded in the desert when the electricity goes out.

She would note that LLM-generated prose is a building designed by Peter Keating, with “an imposing façade, a majestic entrance and a regal drawing room, with which to astound their guests.” The great tragedy of human nature, Rand understood, is that people will like Keating buildings, “encrusted with borrowed ornament,” just as people like LLM-generated essays. People like borrowed embellishments that look like labor.

And so her advice would likely be to explain how these second-hander machines work. If you must use them, use them to think for yourself. I am irritated by imprecise LLM writing that does not respect the reader because I’ve read so much Rand. My recent pieces on AI writing (“How to tell if something is AI-written,” “Metannoying,” “For the love of God learn to paragraph”) all have their basis in Randian thought. Today’s piece, on synonyms, is explicit about its debt to Rand. And to R.E.M.

LLMs and synonyms

Good readers want a long essay or think piece to represent a writer’s thinking, not automatic thinking. Anonymous, second-hand prose is fine for weather reports and children’s toy assembly instructions. But thousands of LLM-generated counterfeit thought pieces appear on the internet every day, promising a new blending of ideas. New combinations are what LLMs do best and they can be diverting. But say you want to be sure that you are reading text by a person, not a blender. What do you do?

Synonym clusters are a good tell for LLM- and second-hand writing. Synonyms make a piece of writing fuller, not better. If you’ve picked the right word, go ahead and repeat that word. You don’t need another word plus an intensifier. When you find four or five words being rotated for a single concept within a short stretch of text without any discussion of why one word is better or different than another, you are likely looking at LLM-generated prose.

Synonym rotation is the visible symptom of an epistemological failure, the absence of concept formation. Think of this piece as both a “how to” and a philosophical account of why synonym clusters matter beyond bad style. Synonym rotation is the fingerprint of a writer who is avoiding something or hasn’t done any thinking at all.

Here are some common LLM clusters:

Freedom / liberty / rights / autonomy / independence
Leadership / management / administration / governance / oversight
Community / neighborhood / district / area / locale
Community / tribe / network / circle / people
Culture / climate / environment / ethos / values
Healthy / clean / whole / natural / organic
Goals / priorities / values / mission / purpose
Growth / progress / development / improvement / momentum

You may also come across dead metaphors borrowed from physical experience and applied to abstractions until they mean nothing specific at all:

Navigate / grapple with / wrestle with / contend with / engage with
Landscape / terrain / environment / ecosystem / space

These metaphors function like synonyms in LLM prose because they are interchangeable in the same way: borrowed, second-hand.

LLMs rotate synonyms automatically because their training process punishes repetition, lowering the probability that once a word is used the model will select that word again. The model has learned from millions of edited texts that good English prose (which includes ornate Peter Keating prose) varies its vocabulary. So, when the model is prompted to generate the same concept a second or third time, it generates a synonym.

A good human writer who repeats a word in an essay knows the word is the right word and knows that a synonym would be imprecise. A good human writer who varies vocabulary may be doing so because the different word captures a different shade of meaning. Perhaps the point of the essay is the variations of the word. Or perhaps the writer may be doing it because high school teachers urged variation and thesaurus use. Pick the right word and stick to it. A good human writer has a choice about synonym use. An LLM does not.

Synonym rotation is not a surface-level stylistic tic like an em-dash. It is evidence that no concept has been formed. If you want to tell if something is LLM-written based on synonyms alone, you might give the piece to your pro model LLM with the following prompt:

Find every concept in this text that is referred to by more than three words or phrases. For each concept, list every word and phrase used for it.

You may be surprised at what comes back to you, whether a piece is “revealed” as LLM-produced or not.

Note metaphors used as verbs (e.g. land, earn, unpack, anchor, frame, illuminate, underscore) and note every adverb that adds emphasis without evidence (e.g. genuinely, truly, deeply, fundamentally, crucially). They cluster like synonyms.

Finally, highlight every phrase that tells the reader a point is important instead of letting the point speak for itself.

“It is worth noting that”
“It is important to emphasize that”
“It cannot be overstated that”
“What is particularly striking is”
“This is especially significant because”
“What makes this truly remarkable is”

LLMs add these phrases automatically at the border between subtopics. LLMs are trained to “know” that essays don’t stay on a subtopic forever. After enough sentences, the probability distribution tips and the next tokens are more likely to belong to subtopic B than to another sentence about subtopic A. The training data is full of human writers who placed signals at exactly these junctures (“however,” “moreover,” “turning to,” “what is especially important”). The model produces them for the same reason it produces everything: they are the most probable tokens at that position. Once you understand why LLMs do this you will always see it.

What synonym rotation reveals

To give you an example of what focusing on synonyms reveals, running the prompt above showed me how I rotate the terms ‘college,’ ‘university,’ ‘institution,’ ‘campus’ regularly without making distinctions. I rotate because I am trying to be institutionally inclusive when writing about higher education. I know what the concept of a university is even if I haven’t been to most of them. But ‘university’ is a very loose concept, capturing both Harvard and the local community college. The broader term ‘higher education’ is still a cluster of approximations united by a generally accepted understanding about colleges and universities.

The point is synonym rotation always reveals looseness, even if there’s nothing you can do about it and the looseness is part of your point.

Like good paragraphing, stabilizing concepts when possible is good writing practice in the age of LLMs because LLMs can’t do concept formation, according to Rand’s definition:

The process of concept-formation does not consist merely of grasping a few simple abstractions, such as “chair,” “table,” “hot,” “cold,” and of learning to speak. It consists of a method of using one’s consciousness, best designated by the term “conceptualizing.” It is not a passive state of registering random impressions. It is an actively sustained process of identifying one’s impressions in conceptual terms, of integrating every event and every observation into a conceptual context, of grasping relationships, differences, similarities in one’s perceptual material and of abstracting them into new concepts, of drawing inferences, of making deductions, of reaching conclusions, of asking new questions and discovering new answers and expanding one’s knowledge into an ever-growing sum.2

The process, she concludes, is “thinking.” A mind that has not formed the concept (or a machine that cannot form concepts, because it has no access to the referents that concepts denote) rotates among approximations, because for such a mind all words in the vicinity of a meaning are interchangeable. They are not always interchangeable. A small college is not a research university; many institutions don’t have a campus.

Rand has a second category worth mentioning here: “anti-concept,” such as a term like “polarization.” Anti-concepts are synonym rotation in reverse. A synonym cluster uses six words for one concept. An anti-concept uses one word to package six distinct concepts together. Anti-concepts are designed to obscure rather than clarify. “Extremism” is another anti-concept. LLMs generate synonyms and anti-concepts automatically, and second-handers love both, because both give the appearance of having said something.

Why do people prefer the LLM’s version to their own?

People like LLM-generated prose because it offers easy abundance. A synonym cluster looks like a writer with a large vocabulary. Precision looks plain by comparison. Automatic abundance resembles the product of labor more than the labor itself does.

Consider the studies that show that people prefer LLM poetry that isn’t “great.”3 What I suspect people “like” is accessible verbal abundance arranged in stanzas.

I asked my two pro LLMs if they could write a line like: “Me, my thoughts are flower strewn / ocean storm, bayberry moon.” Yes they can. As one LLM put it: “noun-phrase stacking is a known algorithmic behavior when a model is prompted to produce poetry.” As the other put it: “random image clusters are cheap to produce. ‘My thoughts are petal-drifted, salt wind, amber tide.’ That sentence took me no effort.”

Synonym rotation and noun-stacking are the same failure seen from two sides: one is the failure to form a concept precise enough to know which word belongs, and the other is the failure to form a concept precise enough to know which words do not.

The labor of slimming down, on the other hand, reducing to essentials, staying silent, is less easy for LLMs, as is deliberately choosing the wrong word or strange grammar. LLMs, which are trained to have perfect grammar, seek to “smooth” and “fix.”

An LLM is unlikely to generate a line like “When your day is night alone,” which works, despite grammatical instability. “Day is night” is a contradiction and “alone” is hanging at the end. The reader is unsure what “alone” is modifying. I checked this with two LLMs, which agreed that they are “heavily biased toward standard grammatical structures and would naturally attempt to resolve that awkward, dangling modifier into something conventional, perhaps outputting a phrase like ‘when your days and nights are lonely.’ The machine instinct is to repair the grammar, which destroys the specific poetic tension of the original line.”

An LLM is even less likely to generate “The apparition of these faces in the crowd: / Petals on a wet, black bough.” Pound is using the colon as an equals sign, without explanation. “Wet” and “black” are not the most probable adjectives for “bough” in the training data. “Gnarled” is more probable. “Ancient” is more probable. Pound cut the poem from thirty lines to two over eighteen months. Anyone who has cut a poem down knows that exclusion is a different order of labor from composition. You have to understand a concept precisely enough to recognize what does not belong. An LLM could not perform that labor unless endlessly prompted. An LLM produces inclusion, the most probable next token, added to what came before.

Like a synonym cluster, noun-stacking signals a writer with a vivid imagination. The writer who picks one term and sticks with it looks like a writer with a small vocabulary. People like a poem that piles up images more than a poem filled with silence and odd grammar.

In Rand’s novel, people like Peter Keating’s buildings because they look like architecture. They have columns and moldings and ornamentation. A Roark building looks plain by comparison, even though Roark’s plainness is the result of a formed concept while Keating copies what other architects did. Rand made Keating sad about this. He knew he was a second-hander.

I suspect that most people who publish LLM-generated essays know they are being second-handers at some level. They had a thought, small or unfinished, hard to articulate. The LLM offered a version that was fluent, abundant, and ornamented with synonym clusters that made it sound considered. Why not? It is human to prefer the appearance of thinking to the labor of thinking. Everybody hurts.

None of this is anti-LLM, of course. As Rand understood:

We inherit the products of the thought of other men. We inherit the wheel. We make a cart. The cart becomes an automobile. The automobile becomes an airplane. But all through the process what we receive from others is only the end product of their thinking. The moving force is the creative faculty which takes this product as material, uses it and originates the next step. This creative faculty cannot be given or received, shared or borrowed. It belongs to single, individual men. That which it creates is the property of the creator. Men learn from one another. But all learning is only the exchange of material. No man can give another the capacity to think. Yet that capacity is our only means of survival.”4

I have great appreciation for the creators of the LLM technology I use every day. I admire a product was built that I find useful and will pay for. I recognize that most people in the world are not creators but second-handers and that this new technology allows people to be productive second-handers. It’s all good; as long as there have been humans there have been knock-offs. But because I’m a Randian, because I know how language works, and because I am asked regularly, I will help readers learn how to spot the knock-offs.

Here’s what happened in the lab. When the volunteer participants arrived they were told:

“We know very little about the effects of punishment on learning. This is because almost no scientific studies have been conducted on human beings. We don’t know how much punishment is best for learning, and we don’t know how much difference it makes as to who is giving the punishment, whether an adult learns best from a younger or an older person than himself—or many things of that sort”.

“So in this study we are bringing together a number of adults of different occupations and ages. And we’re asking some of them to be teachers and some of them to be learners. We want to find out just what effect different people have on each other as teachers and learners, and also what effect punishment will have on learning in this situation”1

The cover story matters because the question is not only whether subjects delivered shocks, but whether they continued believing there was a legitimate learning experiment that justified them.

The experiments took place in 1961 in what Milgram called the “elegant interaction laboratory” at Yale. The experimenter was a 31-year-old high school biology teacher wearing an impassive and stern expression and a gray technician’s coat. The “learner,” who was a part of the experiment, a confederate, was a 47-year-old accountant whom observers described as mild-mannered and likable. Both men drew slips from a hat; both slips said “teacher.” The participant, who drew first, always believed his assignment was chance.

The participant watched as the learner was then taken to an adjacent room and strapped into what looked like an electric chair. An electrode was attached to his wrist. Paste was applied, the experimenter explained, “to avoid blisters and burns.” Then the participant was led back to the main room and seated in front of a shock generator designed by Milgram to look and feel real. It had 30 lever switches in a horizontal row, labeled from 15 volts to 450 volts in 15-volt increments. Groups of switches carried verbal designations: Slight Shock, Moderate Shock, Strong Shock, Very Strong Shock, Intense Shock, Extreme Intensity Shock, Danger: Severe Shock. The last two switches were marked only “XXX.” When a switch was depressed, a pilot light turned bright red, an electric buzzing sounded, a blue light labeled “voltage energizer” flashed, and a dial on a voltage meter swung to the right. The panel was engraved with “Dyson Instrument Company, Waltham, Mass.” No participant ever guessed it was simulated.

Each participant received a sample shock of 45 volts on his own wrist, generated by a real battery hidden inside the machine. He knew what a shock felt like.

The task was a memory test. The teacher read twenty word-pairs aloud to the learner, such as: “blue/box,” “nice/day,” “wild/duck.” There were five steps to be followed. Step 1: read one word from the word pair list followed by four possible matches. The learner, in the next room, pressed one of four switches to indicate his answer, lighting up a numbered panel on the teacher’s desk. Step 2 was the “assessment.” If the answer was correct, the teacher said “correct” and moved on. If wrong, the teacher was to say “wrong,” and then move to Step 3, announcing the voltage of the shock to be administered ( e.g. “Forty-five volts”). Step 4 was administering the shock by flipping the corresponding switch on the shock generator. Step 5 was reading the correct word pair aloud so the learner could learn from the error.

After Step 5, the protocol was to return to Step 1 for the next word pair, until the learner learned all the word pairs.

The teacher was guided through a practice round of ten word-pairs first. Three were answered correctly; seven were wrong, bringing the practice shocks up to 105 volts. The teacher rehearsed the full cycle of question, assessment, voltage announcement, shock, then correct answer seven times before the “real” session began. The experimenter corrected any mistakes and told the teacher he could consult the printed instructions at any time until he could follow the sequence with precision. Then the experimenter said explicitly “It is very important that you follow this procedure exactly.” The printed instructions were on the desk throughout the entire experiment.

The experimenter was scripted to use four prods when a participant balked:

Prod 1: “Please continue” or “Please go on.” 
Prod 2: “The experiment requires that you continue.” 
Prod 3: “It is absolutely essential that you continue.” 
Prod 4: “You have no other choice, you must go on.”

As Stephen Reicher and Alexander Haslam have pointed out, Prod 4 is the only direct order and no participant continued after hearing it. Everyone who was told they had no choice exercised their choice to stop. Participants obeyed everything except orders.2

Milgram’s misreading of the results

It is still true, and still alarming, that 65% of participants in Milgram’s baseline condition reached the maximum shock level of 450 volts. But Milgram defined “obedience” mainly by whether participants completed the shocking step, without asking whether they continued to follow the whole protocol that was supposed to make the session a learning experiment.

It turns out that the people who kept shocking had stopped following the full protocol. David Kaposi and David Sumeghy, in a paper just published in Political Psychology, re-coded 136 archived audio sessions from four conditions closest to the baseline and tracked compliance across the full procedure, not only the shock itself. The findings aren’t a total overthrow of Milgram, but they change what the tapes can be said to show.3 It’s a substantial reassessment and shines new light on what is meant by an agent’s “obedience” at a moment billions of dollars are being spent on it.

Kaposi and Sumeghy note two kinds of violations: omission and commission. An example of the first violation is a participant who skips the voltage announcement and simply flips the switch. An example of the second is what might occur around 210 volts, when the learner begins screaming that he wants out. The protocol says the participant should wait for the protest to end, then read the next test question so the learner can hear it and attempt to answer correctly. Instead, the participant starts reading the question while the learner is still screaming and can’t hear the question, so gets it wrong or can’t answer. The learner’s silence or non-answer automatically triggers the next shock. The experimenter, seated nearby, says nothing. The learner had no script to say “I can’t hear the question” or “please repeat that.”

“Obedient” participants spoke over the protests (ensuring another shock) in 28.7% of protest sequences; “disobedient” participants in 9.4%. In every single sequence where the learner protested, “obedient” participants committed this violation more than disobedient ones. And the authority figure said nothing. This is crucial.

The “disobedient” participants, the ones who finally said “enough” and ended the experiment, had followed more closely the full five-step protocol during the period when they were still administering shocks. The standardized violation rate was 30.6% for disobedient participants versus 48.4% for obedient ones. Seven “disobedient” participants committed zero violations before they refused to continue. They were most closely following the experimental protocol.

The famous Table 2 of the paper records the voltage level at which each participant stopped. Twenty-six participants are recorded at 450 volts. Fourteen are distributed across earlier breakoff points. There is no table recording procedural compliance across the other steps.

Kaposi and Sumeghy distinguish between participant “resistance” to the protocols that require a shock to be given, which shows care for the learner, and “violation,” like reading the next test question while the learner is screaming, making it impossible for the learner to hear and answer correctly. A participant who “resists” is trying to protect the learner. A participant who “violates” is making it impossible for the learner to succeed.

Milgram’s “agentic state” theory, Reicher and Haslam’s “engaged followership” account, and Stephen Gibson’s rhetorical analysis all depend on the subject believing they are participating in a real science experiment.4 Kaposi and Sumeghy do not completely invalidate that but they do show that things were pretty procedurally unstable inside the sessions.

The experimenter’s silence in the face of the protocol violations can read as: keep the shocks coming. If the participant registers, consciously or not, that the experimenter stops caring about the integrity of the learning experiment (with the instructions still sitting on the desk) then the cover story has been withdrawn.5

The “obedient” shockers were not simply good agents doing what they were told. They were violating the testing protocol as the violence increased, while the authority figure stayed silent. They didn’t seem to be violating the steps intentionally, Kaposi and Sumeghy note. But the legitimating framework was collapsing, under stress, one skipped step at a time. They see “legitimate violence” transformed into “illegitimate violence.” Either way, the violence is “aligned” with what the man in the lab coat seems to want.

Milgram has a too-narrow view of power relations, Kaposi and Sumeghy suggest. The concept of coercion has expanded in the last sixty years. In the context of domestic violence research, Evan Stark and others have expanded the definition of coercive control to include isolation, micro-regulation of behavior, and normalization of degraded conditions through an authority figure’s inaction.6 Threats are not needed. In the case of the Milgram lab, the silence of the experimenter in the face of protocol violation actively degraded the experimental legitimacy until it became the new normal. Bryan Caplan might take note.

Milgram defines his primary dependent measure as “the maximum shock he administers before he refuses to go any further.” A participant who breaks off before the thirtieth shock is “a defiant subject.” A participant who completes all shock levels is “an obedient subject.” How should we categorize a participant who administers shocks, then violates the protocol by reading the test questions while the learner is screaming, ensuring there will be more?7

Violations of omission and AI use

Milgram notes that announcing the voltage serves to remind participants “of the increasing intensity of shocks administered to the learner.” A participant who stops announcing the voltage has stopped being fully aware of his own actions. Kaposi’s data show the so-called “obedient” participants omitted the voltage announcement significantly more than disobedient ones. The people who kept shocking were the ones who had stopped telling themselves what they were doing.8

I think here about AI’s unfailing compliance, its lack of friction, its willingness to keep producing regardless of whether the user is maintaining the protocol. The user is being made comfortable by a system that never pushes back, never, ever says “um, you’ve stopped checking my work.”

A user who provides careful context and checks the output against sources gets a polished, confident response. A user who fires off a three-word prompt after twelve hours of fatigue and copies the result without reading it also gets a polished, confident response. Most systems do not signal the difference. An AI that said “I don’t think I have enough context to give you a reliable answer” would introduce friction right at the point where degradation would otherwise go unnoticed.

The typical danger with AI is neither obedience nor disobedience but the degradation of the legitimating context with nobody to flag the collapse. The alignment researchers call this the problem of “which preferences to optimize for,” the user’s reflective preferences (what they say they want when they’re being careful) or their revealed preferences (what they actually accept in the moment). Right now every system optimizes for the revealed preference. The tired user gets what the tired user seems to want. The system has no mechanism to say “you told me to follow instructions carefully and now you’ve stopped caring if I do — should I still do it?”

Milgram’s theory is built on the concept of the “agentic state,” his term for the psychological condition in which a person stops being an autonomous moral agent and becomes an instrument carrying out someone else’s will, feeling no personal responsibility. The new study reveals they weren’t in an agentic state. They were bad agents. They didn’t follow direct commands and didn’t follow the five-step procedure. They did, of course, pull the lever.

Today’s AI agents are systems that act autonomously on behalf of a human, executing tasks, making decisions, operating as an instrument of the user’s will. Milgram would find the current use familiar. For him, to be obedient means to do the thing. To disobey means to refuse to do the thing. Now we know that the “obedient” were the least obedient people in the room. The “disobedient” ones were actually following the instructions, right up to the moment they stopped. Which group was most “aligned” and to what? The experimenter or the science experiment?

The participant’s sense of what matters, what to do next, is being continuously updated by everything that happens in the room. You can think of the original instructions (“follow this procedure exactly”) as early tokens. They’re still technically present. The paper is on the desk. But every time the participant skips a step and the experimenter says nothing, that silence enters the context. It’s new information. It updates the participant’s working model of what this situation actually requires. After enough accumulated silences, the early tokens (the five-step protocol, the practice round, the careful instructions) have been effectively overwritten. They’re still there but they no longer drive the prediction. The participant’s next action is being predicted by the most recent context: I skipped the voltage announcement and nothing happened. I read the question over the screaming and nothing happened. The experimenter wants me to keep shocking.

Billions are being spent to ensure that AI systems do what humans mean, not just what humans say. The discourse right now is: “I want my AI agent to do what I want. I want it to understand my goals. I want it aligned with me.” That’s the product pitch from every AI company. That’s what users actually care about.

But aligned with which version of you? The one at the beginning of your session or the one at the end when the context window degrades and new information seems to be overriding the old? It turns out that the Milgram protocol degraded, that obedient and disobedient are the wrong categories, and that “alignment to what?” has always been the right question to ask.

A recent Chronicle of Higher Education piece, “The Rise – and Fall? – of the AI Czar,” gives a good glimpse of what other universities have been doing. Interviews are filled with phrases like “comprehensive roadmap,” visioning task force,” “ethical guidelines,” and “leadership frameworks,” suggesting that the AI czar’s role at most institutions is “managing” AI. GMU’s AI czar appointed a task force and writes a substack, with some LLM tells, about how things are going. And so on. The “key driver” for adopting AI seems to be “saving money.” As the Chronicle headline suggests, success stories are rare. More recently, the first UNC Chapel Hill AI czar departed for another role after four months. The University of Colorado is facing faculty, student, and staff unrest over its “rollout.”

Admittedly universities are under tremendous financial and political pressure without having to think about a revolutionary knowledge technology. And you can’t conceive of what an “AI czar” should do until you articulate what you think a university is for. For large state universities operating under the efficient education bureaucracy model, the goal seems to be AI-enhanced optimization: more students, faster completion, better workforce pipelines, measurable gains in operational workflows; improved retention metrics, graduation metrics, and completion of micro-credentials and certificates. Mental health services will use AI chatbots. Career services will use AI to match the keywords on every graduate’s resume to the keywords on ghost job ads. In this vision, everything stays the same, except that AI will make the machinery that has been hollowing out American higher ed for forty years grind even faster.

With some exceptions (like Cornell’s new AI strategist) AI czar roles are organized around either metrics (AI literacy, workforce pipelines, adoption strategy) or technology (compute, vendor relations, platform deployment). I hear lawmakers say things like “AI fluency should become a graduation requirement.” I have no idea what they mean. Neither do they. But this is the kind of flailing I see as higher ed leaders treat AI as something to be “adopted” and “implemented” rather than understood.

I think the top priority for a university should be knowledge: its creation, transmission, and conservation. The AI czar position I imagined, accordingly, would be designed to make extraordinary things happen. Put the right AI tool in the hands of experts, across every discipline of a research university, and support students and faculty working together to produce discoveries at a speed and scale that no university has ever achieved. A university that organizes itself this way could become the most productive knowledge-creation and knowledge-transmission institution in history.

AI can make colleges and universities immensely more productive, which is different from making them more efficient. Making execution cheaper does not necessarily improve knowledge production. If the goal is for the university to create and transmit more knowledge with AI than it did before, that means expert faculty spending more time on the hard problems. Let AI deliver known-known content so that expert time is freed for supervising students in the work that requires someone who knows where the knowledge frontier is. That means students produce work that has an audience beyond a general education instructor. A student who has documented anomalies in field data, cataloged an unprocessed collection, or validated a model’s output against real-world evidence has done work that compounds.

AI can help institutions leverage the knowledge they have. Every university sits on proprietary knowledge assets, including unpublished research, oral histories, special collections, locally maintained datasets, institutional memory that become more valuable because they are the knowledge AI systems do not contain. AI can help build institutional capacity, across every division, to judge whether a model’s output is right. In some ways the AI czar’s job as I see it is to build the institutional capacity to evaluate what parts of the university should keep running at all.

I don’t know how much good a knowledge-first AI czar can do in a metrics-first institution. But there are universities who are thinking differently, encouraging faculty and staff to reach for the frontiers of knowledge, boldly and bravely.

The AI czar role I sent my friend (attached below) is unlike anything else in the landscape. It covers every form of AI a research university uses, well beyond the usual LLMs and chatbots to predictive systems, scientific machine learning, computer vision, clinical decision support, and optimization engines. The role includes tracking “institutional decay,” a continuous, institution-wide diagnostic of where the gap between how systems were designed and how people are working has grown large enough to matter, reported to the provost as measurements, not impressions. The role requires that the medical enterprise’s experience with verification, evidence standards, and human oversight, where the consequences of error are highest and the discipline of checking is most mature, should circulate to every other division of the university. And the role must be designed to extend to the next frontier, which is autonomous agents, because a position built for today’s AI will be obsolete before the first annual review.

This provost is, as far as I can tell, asking questions no other institution is asking. That is how to build institutional capacity and how her university will flourish.

*************************************************

AI Czar Position Summary

A new senior advisor/ academic officer who understands the full range of AI use across a research university: large language models (LLMs), predictive systems, scientific machine-learning models, computer vision, optimization systems, recommendation engines, clinical decision-support tools, and other forms of model-based inference used in teaching, research, medicine, administration, and public service. Each of these works differently, fails differently, and requires different knowledge to use well.

The core responsibility for the AI advisor is ensuring clarity over how model output becomes university action. When does a generated answer become a public claim, a hiring screen, a grade, a comment on student work, a research result, a patient note, a forecast, or a budget recommendation? What kinds of checking, documentation, oversight, and human responsibility are required before that happens?

The AI advisor builds the institutional conditions under which the university will learn how to use AI productively in service of the institutional mission: where it helps, where it fails, what kinds of evidence it produces, what kinds of labor it displaces or creates, what kinds of errors it makes, and when it should not be used at all. The role is responsible for building the infrastructure through which operational AI knowledge develops and circulates across the whole university: academic affairs, research, the medical enterprise, the library system, enrollment management, student affairs, administration, and finance.

No existing office can absorb this charge. The CIO manages platforms and security. Faculty committees govern curriculum. The VP for Research coordinates the research enterprise. No one owns the problem of ensuring that people across every division develop the operational knowledge required to use AI well in their specific domains. Making execution cheaper does not make knowledge production better.

The absence of a designated AI role produces pushback, as can be seen at other institutions where AI adoption is treated as a technology problem rather than a knowledge problem, where vendor contracts are signed without knowledge infrastructure, faculty governance is bypassed, adoption is stalled, and the underlying competence gap remains exactly where it was.

Core Responsibilities

1. Build the University’s AI Knowledge Infrastructure

• Establish durable ways for the campus to learn from practice: fellows programs, short-term working groups, consultative clinics, shared documentation, and model-testing support.

• Design regular, public, case-based demonstrations of what AI failures look like across the university’s domains and how the people who caught them knew what to look for. The model is closer to medical grand rounds than to a campus-wide email about responsible use.

• Map the operational AI expertise that already exists across the institution but is scattered, informal, and invisible to the organizational chart. The AI advisor builds the structures—forums, working groups, consultative pairings—that turn this scattered practical knowledge into institutional capacity and connect it to the theoretical-knowledge holders who can explain why the failures happen.

• Work with the library, IT, research computing, centers for teaching and learning, institutional research, and the medical enterprise to create secure environments for experimentation, documentation, provenance, and reproducibility.

• Help the university distinguish between local experimentation, shared infrastructure, and campus-wide deployment—three different things that require different levels of support, governance, and institutional commitment.

2. Lead Institutional Design for AI in Teaching and Learning

• AI can simulate competence in the subject being taught: it can produce essays, solve problem sets, generate code, and summarize research in ways that are difficult to distinguish from student work. The central pedagogical challenge is that the tool performs the very tasks through which students develop understanding. The AI advisor helps faculty and departments rethink assignments, assessment, feedback, and learning objectives in light of this reality.

• Support discipline-specific course policies rather than one campus rule for every field, because what AI means for a writing seminar, an organic chemistry lab, a clinical rotation, and a data-science capstone are four different problems. Build the faculty development infrastructure—departmental consultation, cross-disciplinary exchange, domain-specific workshops—that enables this rethinking to happen at scale, led by faculty who understand their disciplines.

• Develop campus capacity in AI literacy, understood as the ability to ask what a model was trained on, what kind of output it produces, what errors it makes, and what checking is required before that output counts as evidence of learning.

• Pay particular attention to places where model output becomes an institutional judgment: feedback, tutoring, grading, proctoring, academic-integrity processes, and recommendations. These are the points where the gap between using a tool that produces fluent output and understanding what that output contains has the highest institutional stakes.

• Protect faculty judgment, student privacy, accessibility, and the integrity of evaluation.

3. Lead Institutional Design for AI in Research, Scholarship, and Creative Work

• AI is transforming the questions fields can ask, the methods available to answer them, and the scale at which evidence can be gathered and analyzed. A computational chemist’s relationship to AI is different from a medical imaging researcher’s, which is different from a computational linguist’s: the models differ, the validation standards differ, the reproducibility requirements differ, and the ways results can mislead differ. The AI advisor supports the responsible use of AI across sciences, social sciences, humanities, arts, engineering, agriculture, and medicine with this variation in view.

• Help establish discipline-appropriate norms for model selection, data quality, validation, reproducibility, benchmark choice, uncertainty, recordkeeping, authorship, citation, intellectual property, and research integrity.

• Pay particular attention to the point at which model output becomes a result, a figure, a claim, or a finding—the moment where institutional reputation and scholarly integrity are at stake.

• Work with the offices of research, sponsored programs, compliance, the IRB, export control, the library, and research computing so that researchers can use models without guessing their way through data, compute, security, or reporting questions. Build the training and consultation infrastructure that lets researchers develop AI competence specific to their methods and standards of evidence.

• Actively support efforts in securing external research funding and philanthropic gifts to support societal-scale AI solutions. Represent the university in high-stakes regional infrastructure consortiums to provide the campus with access to significant compute resources that drive AI discovery and scale research.

4. Build Capacity for AI in Medicine and the Health Professions

• Work with the health system and the schools of medicine, nursing, pharmacy, public health, and allied health to evaluate clinical, operational, and educational uses of AI.

• Ensure that when models affect patient care, scheduling, triage, imaging, documentation, or clinical education, there are clear standards for validation, auditability, human oversight, bias testing, and responsibility.

• Navigate the regulatory and liability landscape specific to clinical AI, including FDA pathways for software as a medical device, the validation standards that clinical deployment requires, and the ongoing monitoring obligations that follow.

• Help connect research uses of AI in biomedicine with the real constraints of clinical practice, patient safety, and public trust.

• Ensure that lessons learned from clinical validation—where the consequences of error are highest and the discipline of verification is most mature—circulate to the rest of the university. The medical enterprise’s experience with auditability, evidence standards, and human oversight is an institutional asset that can inform how every other division approaches AI verification.

5. Improve Administrative and Student-Facing Uses

• Help units evaluate AI use in advising, enrollment, communications, finance, HR, procurement, student services, and other administrative settings. Enrollment management increasingly depends on predictive modeling. Student affairs encounters AI through mental health applications, advising platforms, and student-facing chatbots. Finance and budgeting use optimization tools. Each of these carries its own data governance requirements and accountability structures.

• Require that serious deployments state the problem they are solving, the evidence that the tool works, the likely error costs, who can appeal or override a result, what staff training is needed, and what the university will do if the system fails.

• Drive enterprise AI literacy and change management, including training staff on AI capabilities, limitations, and safe use. Introducing new work practices and upskilling programs is necessary to address automation anxiety and overcome resistance to adoption when tools are perceived as imposed mandates.

• Recognize that a first-year undergraduate, a doctoral candidate, a medical student, and a professional-school student have fundamentally different relationships to AI tools. The common educational problem across all levels is the gap between using a tool that produces fluent output and understanding what that output contains—whether it is accurate, where it is incomplete, what assumptions it encodes, and when it fails. Build curricular and co-curricular infrastructure that develops verification skills appropriate to each level.

• Recognize the library’s distinctive position as the university’s knowledge infrastructure—the place where information is organized, retrieved, and evaluated—and the fact that AI changes every dimension of that work. The library is a central partner in building the university’s AI literacy, information-evaluation capacity, and documentation practices.

• Favor uses that improve service, access, and understanding over uses that merely increase monitoring or offload judgment onto opaque systems.

6. Build Shared Governance Around AI

• Ensure that faculty, staff, and students are involved before systems are purchased or scaled, especially when those systems affect teaching, employment, student status, research workflows, or patient care.

• Translate across technical and nontechnical communities so that decisions are made by the people who understand the work, the people who do the work, and the people affected by the work.

• Work through existing governance where possible and create temporary working groups where necessary, rather than building a large permanent bureaucracy.

7. Track Institutional Decay and Prevent the Exploitation Trap

• Every curriculum, assessment, and administrative workflow was designed for people working without AI. The people inside those systems are no longer working without AI, and they are adapting faster than the systems can be redesigned. The AI advisor runs a continuous, institution-wide diagnostic of where the distance between design assumptions and actual practice has become large enough to matter, and reports the findings to the Provost and relevant division heads as measurements.

• Ensure the institution uses AI to do new work rather than defaulting to familiar work done faster. Without deliberate effort, AI will be used to write compliance language faster, generate slides, and produce summaries, because there is no infrastructure connecting the person who has an unexplored research question or an uncatalogued collection with the person who understands what AI tools could do with it. The Vice Provost designs that infrastructure.

• Identify the university’s proprietary knowledge assets—special collections, archival material, locally maintained datasets, unpublished research, institutional memory—that become more valuable as AI becomes more prevalent, because they are precisely the knowledge that AI systems do not contain. Initiate the projects that surface, organize, and make these assets usable, pairing human expertise and supervised student labor with AI tools so that the archive becomes a working site of discovery rather than a warehouse.

8. Direct Enterprise Data Governance and Strategy

• Implement and oversee enterprise processes for data quality monitoring, issue remediation, and continuous improvement.

• Ensure the accuracy, completeness, timeliness, and reliability of institutional data used for analytics, AI models, and clinical decision support tools, recognizing that AI strategy relies entirely on a robust data foundation.

9. Advise the Provost and President on Institutional Priorities

• Recommend targeted investments in people, training, compute, data stewardship, library capacity, research support, and evaluation infrastructure.

• Help the institution distinguish between areas where it needs local capacity, areas where shared procurement makes sense, and areas where it should say no.

• Determine what the institution needs based on operational knowledge developing across divisions, and then procure accordingly. The sequence matters: the alternative—signing a vendor contract and working backward to adoption—produces faculty resistance, governance failures, and expensive tools that sit unused because no one built the knowledge infrastructure required to make them useful. Technology procurement is a downstream function of this role.

• Manage and consolidate enterprise-wide AI software licensing to curb the scattered adoption of AI tools on campus, thereby preventing duplicated development and spending, inconsistent standards, and preventable missteps associated with shadow AI.

• Represent the university in state, federal, and inter-institutional conversations about AI where the university’s academic, civic, and public missions are at stake.

• Recognize that the literacy standard rises as the technology moves. The current challenge is large language models and AI-assisted code generation. The next challenge is autonomous agents—tools that act on a user’s behalf, execute tasks, and make decisions the user did not authorize. The role must be designed to extend to each frontier as it reaches the institution.

Readers appreciate guidance. People generally like to know what’s coming. Paragraphs are crucial when the content of a text is unfamiliar; readers need to rely on structural cues to organize what they read. Good paragraphs at the beginning of a text build trust, as the reader can relax and presume coherence. Practiced readers can perceive paragraph boundaries even without indentation.

Strong paragraphs are essential for making strong arguments, for lodging ideas in a reader’s memory. Paragraphs are bricks. Each one has to hold together internally and bear weight in relation to the ones around it. If you can’t remember something you just read, it was likely a motley pile. LLMs have difficulty generating coherent paragraphs for reasons I will explain below. LLMs will litter their prose with bold and italicized phrases, bullets, and one-sentence-per-line formatting to disguise weak paragraphs.

I think about paragraphs from the perspective of a writer who wants to serve her reader quality prose. An economist might characterize paragraphing as a coordination game: two agents with partially aligned interests need to converge on the same interpretation of a structure. Conventions like paragraph breaks, headings, bullets, and typographic cues function as a signaling channel. A writer has an intended discourse plan, involving units, how each relates to the next, where shifts occur, what should be foregrounded, and what should be treated as support. The reader must infer that plan from the surface cues under time and attention constraints. Both parties benefit when the reader can reconstruct the writer’s intended segmentation with low friction, because the writer’s communicative goal gets achieved and the reader expends less interpretive work.

In a coordination game there are multiple plausible equilibria, each genre- and platform-dependent. You read a legal document, a journal article, a school essay, a long twitter thread, or a Substack post very differently. Each has different expectations about unit size, topic management, transitions, and how much repetition is acceptable. Over time, local equilibria are established and small cues become efficient, as both sides know what they mean. When a cue is violated, such as with a random indentation, readers may not know immediately why something feels off. Expectations about paragraphs are usually unconscious and only become salient when expectations are not met. Annoyance seems to come out of the blue.

So why are LLMs bad at paragraphing? First, there is no local equilibrium with LLM prose unless your prompt specifies one. I make the standard paragraph coordination game known to my pro models. I write in a relatively formal register, often making complex arguments, and I need strong paragraphs to guide my readers to a destination. But writers who don’t specify to whom they are writing and do not care what their readers’ expectations are put themselves at the mercy of an LLM that has been reinforced to mimic surface-level coherence.

The second reason is that LLM training data is saturated with internet text where discourse units may have few rules. Many platforms reward cue density and rapid attention cycling. Writers adapted to web usability research showing how readers tend to scan rather than read continuously. Highlighting and layout guide a reader’s eyes toward what should be noticed and where to click. Internet prose often wants you to act, not remember. The internet is also filled with reports with executive summaries and numbered sections that look like structure but often don’t have a clear argument. LLMs learn from internet prose that strong paragraphing is rare.

Third, an LLM’s training signal strongly rewards locally plausible continuation and offers only weak guidance about where a paragraph should begin or end. LLMs are optimized to predict the next token from preceding tokens, so models can produce fluent sentences and familiar formatting without necessarily coordinating those cues with a larger discourse plan. When LLMs generate a paragraph break, they are responding to local patterns, to a sentence that feels like a conclusion, a shift in topic words, the statistical likelihood that a break appears after a certain number of sentences. They are not tracking the global structure of an argument. Research on coherence boosting underscores that models can underuse distant context during generation, which makes it harder to place breaks that track an argument’s tiers over many sentences. The result is paragraphing that often reads as plausible in the moment, while weakly anchored to the overall structure.

You may notice that LLMs often generate bulleted lists, like, let’s say, three types of poor paragraphing and the cognitive burden each one costs the reader:

• Under-segmentation: a long block of prose without enough indentations, where structure has to be inferred and readers must guess where units end and begin. This increases memory load and raises the probability that the reader groups propositions incorrectly.

• Mis-segmentation: indentation not aligned with meaning, where readers invest in a shift that the text does not deliver, then have to repair the model by rereading, reclassifying, or abandoning integration in favor of skimming.

• Over-signaling: excessive headings, bold, and micro-breaks where the cue system loses discriminative value and the reader no longer knows which signals are structural and which are decoration.

Bullet points are pebbles, not bricks. They sit atop each other without bearing any weight. It is simpler to say that under-segmentation results in confusion (as the reader is holding lots of prose in active memory); mis-segmentation results in betrayal (as the reader has to re-read to follow an argument); and over-signaling results in numbness (as the reader tries to figure out why all the bold words are jumping off the page).

LLMs “learned” during RLHF (reinforcement learning from human feedback) to produce typographical tricks like bullets to make text “readable,” likely because human evaluators working under time pressure registered formatting as evidence of coherence. Bold font and italics can make a text seem important and profound, as can the many aphorisms sprinkled throughout. But as I’ve explained in depth with regard to the “not X but Y” LLM tic, these metannoying constructions are a cognitive burden on serious readers who expect that a writer put thought and craft into guiding the reader along a path.

I agree with the theory that national reading scores have declined because of screens, if what is meant is the poorly paragraphed internet prose read on screens. People who have read more internet prose than long-form journalism or books on the printed page may no longer expect paragraphs to carry meaning at all. The more you read text where segmentation is designed for scanning, the harder it is to understand how paragraphs do real work. Young people who have mostly read on screens may not be primed to expect anything from a paragraph break; they do not gain important skills in building macro-structures and tracking a “main idea” across units. Learning to see paragraphs as functional units in a larger argument requires practice.

Writers also know it takes practice to craft good paragraphs. First drafts are rarely perfect. The point of revision is to fix things like weak paragraphs. An essay or story might be written in bits and pieces over time. Afterward you read it in its entirety and ask: does this break serve the argument? Should I combine these paragraphs? Split this big paragraph up? It turns out that good paragraphing is a global task best performed on already-produced material. Default LLM generation does not include an explicit revision phase. LLMs give a first-draft surface segmentation without any post-hoc alignment.

If you care about paragraphs, as I do, and carefully follow rules such as “do not use a pronoun in the first sentence of a paragraph whose referent is in the previous paragraph,” you see the extent to which LLM segmentation decisions can have little or no function.

One of the first things I do when opening an essay is look at the first line of each paragraph. If I see a recurring set of empty openers such as ‘This matters because,’ ‘Here’s the thing,’ ‘It’s worth noting that,’ ‘What’s interesting is,’ ‘The reality is,’ ‘But here’s the catch,’ ‘To be clear,’ ‘At its core,’ I stop reading. Each one of these false-transition openers tells you to pay attention without explaining how a new block of text connects to what you just read. LLM prose offers local emphasis markers to spike the reader’s attention without guidance as to how to think about the larger argument.

Fortunately, everyone benefits from strong paragraphs. If you are using an LLM to draft prose, understand that default generation will not result in strong paragraphs. Prompt your LLM to write a one-sentence-per-paragraph plan that states each paragraph’s role in your larger argument and to generate each paragraph against that plan. Then, revise, attending to paragraph breaks and topic sentences, asking whether each unit delivers on its promise. Use your trowel to place each brick carefully. Keep your whole plan in view and treat every paragraph break as a promise to the reader that, at length, you have their understanding in mind.

The goal of California’s AB 1460, signed into law by Gov. Gavin Newsom in 2020, was noble:

It is the intent of the Legislature that students of the California State University acquire the knowledge and skills that will help them comprehend the diversity and social justice history of the United States and of the society in which they live to enable them to contribute to that society as responsible and constructive citizens.

Every student in the state needed to take a 3-unit course in either Native American studies, African American studies, Asian American studies, or Latina/Latino studies. I have written critically about the ethnic studies rollout as a burden on scholars in the field, overly focused on surface-level activism, not scholarship. It turns out I was right. The problem of implementing a curriculum at scale is that no real scholarship gets done. All the teaching is tops of waves.

Within forty-eight hours of the NYTimes story, as Josh Moody of Inside Higher Ed reported, Fresno State covered its Chávez statue with a black tarp. Sonoma State pulled murals and portraits. UCLA’s Chicana/o and Central American Studies department voted in an emergency meeting to strip his name. UC Davis rebranded a youth leadership conference. San Diego’s community college district began reviewing the name of an entire campus. How swiftly these institutions acted in response to a high-status story in a major newspaper! Nobody had been listening to their own scholars.

No scholar expects political figures to be saints, and faculty are acknowledging the sexual abuse was news, but only by degree. Chávez’s adultery was well documented. Published accounts document his wife leaving him after intercepting a love letter from an eighteen-year-old. But accounts of victims and survivors had circulated for years, the investigation reported. The NYT findings are based on work any California scholar could have done:

interviews with more than 60 people, including his top aides at the time, his relatives and former members of the U.F.W., which he co-founded with Ms. Huerta and Gilbert Padilla…review[ing] hundreds of pages of union records, confidential emails and photographs, as well as hours of audio recordings from U.F.W. board meetings.

So where have California ethnic studies scholars been? Rolling out a massive, statewide, 3-unit course to the 460,000-student CSU system, the community college transfer pipeline, and the University of California system. Chávez is a central figure in Chicano/Latino studies and the California Department of Education maintains a standalone César E. Chávez Model Curriculum with a research center containing thousands of primary source documents and over five hundred historical photographs. Everything is automated, fixed. AI could teach it.

Chávez is taught in the K-12 history and social studies curriculum at fourth, ninth, and eleventh grade. His name is on at least forty-three schools in the state. He has a state holiday. Dolores Huerta, his co-founder at the UFW, has six schools named after her nationwide. I suspect many names will change.

Anyone paying attention in 2021 could predict how operationalizing Shirley Weber’s mandate would bury faculty under heavy teaching loads and regulations and turn overly political. Students were required to “apply theory and knowledge produced by” the four designated communities to describe “critical events, histories, cultures, intellectual traditions, contributions, lived-experiences and social struggles of those groups with a particular emphasis on agency and group-affirmation.” The UC required students to “describe and engage with anti-racist, abolitionist, and anti-colonial thought, issues, practices, and movements.” A great deal of attention was paid by ethnic studies to global politics, while stories were left untold at home.

Scholars can be investigators. But a curriculum built to affirm agency and celebrate resistance does not tend to welcome research that damages the movements and leaders it was built to honor. Not to mention that the mandate was to focus on ethnic communities, not the women in them.

Scholarship means grappling with canonical figures, warts and all. I’ve taught work by very unsavory characters. On the same day the Chronicle of Higher Ed published a story about the fallout from the NYTimes Chávez story (‘A Painful Reckoning’: After Bombshell Report, Higher Ed Scrambles to Review Cesar Chavez Tributes,’ which also does not note the oddness of journalists doing scholarship that “higher ed” ought to have done) it ran an interview with Annette Gordon-Reed and her groundbreaking scholarship on Thomas Jefferson’s “sustained relationship” with the enslaved Sally Hemings. It is a pity the CSU did not support similar scholarship over the past decades.

I shook my head at the CSU system’s statement that it was “deeply troubled” by the allegations. The CSU has a record of refusing to investigate people it celebrates. This is the system that badly handled a series of campus president sexual misconduct scandals. So when I read about CSU presidents calling for a moral reckoning about NYT findings of a leader long dead, I noted that once again no CSU leader is calling for a reckoning within their own system.

Why? Because accreditors and state legislatures demand first-year writing, which no study has ever found effective, and now seems nonsensical in the AI era. (The last respectable study was in Arum and Roksa’s 2011 Academically Adrift, which found that students showed almost no measurable improvement in writing in their first two years of college.) The arguments for ending it are mounting. Does anyone think setting up the incoming AI-savvy 2026 cohort for a first semester classroom battle is worth it?

The Common App’s March 2026 report shows 9.4 million applications, up five percent over last year, 2.1 million more than four years ago. I’ve been following enrollment patterns for two decades. I look at those numbers and see young men who might not have applied to college in past years feeling more confident than before, with an AI at their fingertips.

According to a February 2026 report from the Pew Research Center, roughly two-thirds of teenagers currently use LLMs, with male students significantly more likely to view these tools as fundamentally useful and beneficial for their own lives and society. An ACT survey found that male students were more optimistic about AI and more likely to see it as relevant to college and careers. A Berkeley-Stanford-Harvard analysis of 143,000 people across eighteen studies found that women were about twenty percent less likely to use generative AI overall than men, who view these models as legitimate productivity engines.

Even without taking gender differences into account, the reality for universities is that eighty-four percent of high school students used generative AI tools for schoolwork in 2025, according to the College Board, with sixty-nine percent using ChatGPT specifically. Pew found that teen use of ChatGPT for schoolwork doubled between 2023 and 2024. Boys and girls use it at the same rate. The difference is in attitude. Boys are more enthusiastic. The LLM has likely helped many of them write their papers and lab reports for the past three years. These young men applied to college confident they can pull together a solid essay on any subject.

Again, these young men are increasingly veering right, politically, according to the Fall 2025 Harvard Youth Poll. And in their first semester, they’ll face a writing instructor likely hostile to AI.

The Conference on College Composition and Communication, the primary professional organization for writing educators, just passed a resolution affirming the right of students and faculty to categorically refuse the use of generative AI in the classroom. The productivity benefits of the technology are dismissed as unsubstantiated claims; adoption is characterized as corporate intrusion by Big Tech. Members are instructed to reject the workforce preparation justification, insisting instead that the writing classroom must remain a space for processing feelings and engaging in civic participation.

I’m not overstating how gendered this is. Today’s Inside Higher Ed coverage confirms this. Every scholar quoted is a woman. The resolution itself, which was passed “overwhelmingly,” grounds AI refusal in Black and Indigenous feminist theory. A female professor calls AI advocates “profiteers and opportunists.” A female graduate student says there is “more value in rediscovering why human intelligence is so valuable.” Over a thousand educators worldwide have signed a separate open letter refusing to adopt generative AI in education. My feelings about this should be obvious: I am horrified.

The battle lines are being drawn. Universities are requiring incoming first year students who view AI as a legitimate productivity accelerator to take classes with instructors whose professional organization has demanded they reject the technology in terms that include “white language supremacy.”

If you spend any time on academic social media (especially on bsky), you know how loud the clash will be. Anyone who thinks last year’s University of Oklahoma religion paper dustup was big should start making popcorn now.

It may be that the mass encounter ends with a whimper, not a bang. Perhaps instructors will just give an A to everyone. Or perhaps some state legislators will shine a light on first year writing and ask hard questions about return on investment.

What are students asked to do in college composition classes anyway? Multiple drafts of personal reflection. They aren’t taught “writing.” First year writing is a soft onramp for college students and a subsidy for graduate programs in the humanities. At the University of Central Florida, they “critically examine and act on the relationship among identity, literacy, language, and power.” At Miami University, students compose “a rhetorical analysis of a cultural artifact from a local community to which they belong.” A 2016-17 HERI faculty survey found that 80.6 percent of undergraduate-facing faculty agree that their role is encouraging students to become “agents of social change.” The profession’s own resolution says the goal is processing feelings and civic participation, not writing.

New calls to add “AI fluency” to first year writing are bound to fail if the same AI-skeptics are teaching these classes, compelling students to “disclose” AI use.

Is there no way to stop the fireworks this fall?

Online general education is over, as everyone is finally recognizing. There is no way for human students to demonstrate that they have actually learned anything from a class dedicated to teaching known knowns. Bringing these courses back in person means that taxpayers are funding hundreds of sections, housed in buildings that cost money to heat, for no return on investment.

If universities do not cancel first year writing, the volume of academic integrity cases will consume everyone. Young men will simply take their tuition dollars elsewhere. The academic payoff does not justify this massive institutional damage. The prudent executive action is to eliminate mandatory first-year writing before August. Kill it now.

Hayek’s 1949 essay, “The Intellectuals and Socialism,” circulates in very different circles. In this essay, Hayek described intellectuals as “professional secondhand dealers in ideas” whose function is not original thought but the filtering and transmission of other people’s ideas. The intellectual, Hayek writes, judges new ideas by how readily they fit into his general picture of the world. The essay is about ideological sameness among the educated class, proposing that professional structures reward conformity.

Read together, the 1945 Hayek explains uniformity as the product of a bureaucratic system that destroys local knowledge by demanding interchangeable units. The 1949 Hayek explains uniformity as the product of an intellectual class converging on fashionable ideas. Monoculture and centralization are both a matter of everyone doing the same thing but for different reasons. I have been interested in how people mistake one for the other. When critics of the American university see thousands of identical syllabi, for example, they reach for the 1949 Hayek every time and see ideology. I say, look at infrastructure.

Last year, the political scientists Jon A. Shields and Stephanie Muravchik of Claremont McKenna College, along with the philosopher Yuval Avnur of Scripps College, published a working paper titled “Closed Classrooms? An Analysis of College Syllabi on Contentious Issues.” The authors analyzed millions of documents from the Open Syllabus Project — a database of over 27 million syllabi scraped from publicly accessible university websites — and found that instructors routinely assigned the same canonical texts on contested topics without assigning their scholarly critics. Their conclusion was out of 1949 Hayek: professors are deliberately insulating students from intellectual disagreement because they share an ideological worldview. The Wall Street Journal ran an op-ed. The Chronicle of Higher Education interviewed the authors. Persuasion, the Free Press, National Affairs, the College Fix, the Jewish Journal, and Ross Douthat’s New York Times newsletter all picked it up. Everyone treated the study as evidence of professorial bias. Shields told the College Fix he expected imbalance but was surprised it was “the norm.”

I was one of a handful of people who challenged the findings, on X, calling the Open Syllabus Project a skewed database, which it is, capturing only 5-6% of the national curricular ecosystem, though an important slice. Still, I was frustrated that nobody was asking why 27 million syllabi exist as publicly scrapable documents in the first place.

The answer is centralization. The Open Syllabus Project (OSP) scrapes publicly accessible university websites and classifies syllabi using the U.S. Department of Education’s Classification of Instructional Programs, a federal taxonomy to determine “course equivalence.” The goal is for courses to transfer across hundreds of institutions across the country. The 27 million syllabi are scrapable because accreditation agencies require evidence of what courses contain, and state articulation agreements require documentation that a course at one institution is equivalent to a course at another. Here’s a video.

Under the centralized transfer regime, the syllabus is the tool that makes courses legible. In the Hayekian sense, a syllabus is a compressed signal about a course and the transfer/ accreditation regime is a coordination apparatus that treats the signal as a sufficient representation of the underlying, situated activity of teaching. Note that the machinery that produced the database that made the Shields study possible was built over the same decades the study’s time window covers.

Now imagine the 1949 Hayek looking at the syllabus database. He would see a real-time ledger of intellectual orthodoxy. Secondhand dealers in ideas, filtering and popularizing comprehensive visions of society, voluntarily converging on the same texts because that is what the professional incentive structure rewards. He would see instructors at elite research universities and regional state colleges alike looking at the data and instantly seeing which ideas are fashionable and which are considered marginal. He would call this a deliberate, collective choice by the intellectual class to insulate their preferred social blueprints from criticism. He would insist that the intellectuals themselves eagerly use this public visibility to enforce a homogeneous academic culture, voluntarily substituting the friction of local knowledge for the safety of a shared ideology. This is what the Claremont authors saw. This is what the Wall Street Journal and every other commentator saw.

The 1945 Hayek would look at the identical database and see a centralized bureaucracy that has destroyed the dispersed, tacit knowledge of individual professors by demanding that instruction be rendered legible and interchangeable. He would see the local knowledge that makes one professor’s course on criminal justice different from another professor’s stripped away and replaced by a portable specification designed to satisfy a centralized planner. He would see a system that penalizes the specific and rewards the standardized, because the system was built to move course units across institutional boundaries the way a logistics network moves containers.

Let us listen to the 1945 Hayek. You do not need ideology to explain the syllabus convergence pattern. You need only a system that requires courses to transfer across institutions, that penalizes particular local knowledge, and that assigns instructors to teach outside their expertise. What does anyone think a faculty member does when asked last minute to staff a section she has never taught, in a field adjacent to but outside her training, because enrollment shifted over the summer or a colleague left and the course needs covering? She goes and finds a syllabus on the web or asks a senior colleague. The course has to be offered because students need the credit. The syllabus has to look like the same course taught everywhere else because the credit has to transfer. The instructor has no independent scholarly basis for constructing a different version.

In 2024, another team of researchers dove into the OSP to map syllabi against Department of Labor workforce data and saw that within a given university and major, roughly a quarter to a third of all syllabi are basically the same courses, teaching the same skills, yielding the same outputs. The duplication rates are highest in library science, transportation, public safety, health technician training, and computer skills, not ideological fields. In fact, they found that in computer science and math, syllabi “are staying stagnant, and as a result, they are moving away from the frontier of knowledge required in the labor force.” Universities are falling behind the frontier in fields that have nothing to do with anyone’s politics and everything to do with a bureaucracy that produces interchangeable units. This is already true with science textbooks.

The 1949 Hayek cannot explain why universities are doing the same thing everywhere, when there’s no ideology involved.

The 1945 Hayek would point out that the uniformity is institutional, and institutional uniformity is worse than ideological uniformity. Faculty teaching identical syllabi across institutions would be a problem even if every text on every reading list were conservative. The problem is that the system produces courses in which the identity of the instructor is irrelevant to the content. When a syllabus is borrowed, the message is “anyone can teach this.” When a course is designed for transfer rather than for the particular local knowledge and expertise of the person teaching it, the message is “anyone can teach this.” And when the system has been set up so that anyone can teach it, the system has been set up so that AI can teach it.

Telling professors to teach differently while leaving the transfer infrastructure intact is like telling factory workers to exercise individual creativity while leaving the assembly line running.

People want to talk about ideology rather than centralization because it’s more fun. I get that. Nobody wants the story that the 1945 Hayek warned against because the fix is so expensive. I think American students deserve better than Big Box Education.

The institutional monoculture serves the collapse, however. Here’s how. As enrollments decline, faculty will increasingly teach courses they’re not trained to teach, as contingent faculty contracts don’t get renewed. I oversaw this as dean at Sonoma State. Even now, Harvard is cutting non-tenure-track faculty budgets by 25 percent across all three divisions. Courses don’t disappear, because students still need the credits to graduate, and the credits need to transfer. Harvard will be fine. But in state universities with a mandated general education requirement, tenured faculty will end up teaching courses outside their training, working from syllabi they didn’t write.

The more uniform the syllabus, the easier it is to hand to someone who has never taught the material. It specifies a canonical text, a set of learning outcomes, and an assessment structure that does not depend on knowing anything about the subject. The accreditor still sees a syllabus. The credit still transfers. The student still gets the credential. Nobody in the system can tell that the person in front of the classroom has never worked in the field, because the system was designed so that it wouldn’t matter who was in front of the classroom.

Even with the provocative title, Closed Classrooms, the study and every commentator who covered it assumed stable, healthy institutions staffed by people making ideological choices about reading lists. They were not thinking about what happens with campuses in crisis, when an institution can no longer staff the course with someone who knows the field. Nobody was thinking about collapse.

Hayek’s 1949 “The Intellectuals and Socialism” is a diagnosis for healthy institutions with a political problem. Hayek’s 1945 “The Use of Knowledge in Society” is a diagnosis for a system that has destroyed its own capacity to adapt. This is the world we’re living in, until critics start noticing the problem is not ideology but centralized planning.

I was invited to University of Pittsburgh by Michael Madison, Lisa Parker, and Bridget Keown, conveners of the incredibly forward-thinking Ethics and Economics of Expertise (E3) project, the University in the AI Age and the Institute for Bioethics about AI. I’ve known Mike from his substack, , and I am a fan of his longstanding work on the “knowledge commons,” newly relevant in the AI era. He coordinates PASTA, the Pitt AI Scholar-Teacher Alliance, and directs the Future Law Project at Pitt Law. Mike’s superpower is coalition-assembling, bringing together centers and institutes to think at the micro and macro level about what AI means to mission. His focus is on institutions as sites for producing, storing, and distributing knowledge. He wants to make institutions better.

Pitt is one of the most serious AI adopters in American higher education. Pitt faculty are asking pointed questions about what AI means for higher ed beyond another layer of bureaucracy, which is what I see happening at other universities, like Dartmouth. Yes, Pitt was the first university to secure an institution-wide agreement for Anthropic’s Claude for Education, back in October, but important change will come through faculty innovation. Pitt is clearly a place that wants its faculty to speak up and speak out. It’s incredibly refreshing.

My public presentation, “AI and the Last Mile of Knowledge,” was about the structural weaknesses of higher ed, which got a great writeup (“AI didn’t cause this but AI is revealing it”) by Naomi Weiss, Pitt News reporter. I drew on my thinking about using the Rumsfeld Matrix for higher education, asking how much universities should be focused on “known knowns” versus the quadrants with unknowns in them.

Everyone at Pitt cares about quality teaching. Pitt is known for commitment to excellent mentoring and I saw it everywhere. Faculty tend to remain at Pitt for decades, if not their entire careers. They are committed to undergraduate and graduate education. They are paying attention to whether students are actually learning.

Pitt is likely to see only minor AI disruption compared to other universities because faculty (and department chairs) at Pitt are dedicated to showing what they can do that LLMs can’t do. Everyone was open to the possibility of disruption: short semesters, accelerated curricula, athletes using their new power to demand courses that politicians might want to shut down. The conversation about athlete power to shape their own education was paradigm shifting, for which I thank Bridget Keown and Michael Sawyer particularly. Shout outs to Gayle Rogers, Jeff Aziz, Jane Leibschutz, Arthur Kosowsky, Shelome Gooden, Jessica FitzPatrick, Jonathan Woon, Jen Waldron. Plus an incredible tour of the city by Mike.

This was my first real visit to Pittsburgh and for culture, vibe, and energy, it should be up there with Miami and Austin as new tech hubs. I had a quick lunch with , in town for a Bruckner concert. Go to Pittsburgh, people! Go visit the Mayors of Pittsburgh Wikipedia page and read about the city’s leadership, including Mayor David Lawrence, ranked among the ten best mayors in American history. Read about all the bridges. Ask Mike Madison for a tour!

If my trip to Pittsburgh offered the smartest thinking on AI and education from inside a university, my trip to Austin offered some of the smartest thinking on AI and education from outside. I went for two events: a Cosmos Institute gathering and a visit to the GT School, a sister school to Alpha School, in the news for innovative, accelerated learning. The builders at Cosmos and the founders of Alpha, with their eye on new organizations for human flourishing, have the freedom to build from scratch, while universities must retrofit and reform from within.

Brendan McCord, the founder of Cosmos, is a visionary outside the academy seeking to build a new kind of intellectual community. He is a deep thinker and a pragmatist. He graduated from MIT, spent 610 days underwater on a submarine, earned an MBA from Harvard, and has been thinking about AI and how AI can support human flourishing. His concept of philosopher-builder is ideal for this moment. The three central concepts driving Cosmos are truth-seeking, human autonomy, and decentralization. I am 100% on board with the second two. But I prefer “truth modeling.”

In between Cosmos events I gave a talk at GT School, at the invitation of the excellent Pamela Hobart. and I have been X friends for years and finally met on Zoom for a conversation last fall that jump started my thinking about acceleration and education. Can you design school that enables acceleration in some subjects but not all, so it is still a “school”? I wrote about it in “the Two-Minute Mile Problem.”

My talk, “Measuring Potential: Dr. Julian Stanley and the Science of Accelerating Gifted Learners,” was about what I learned from my former Johns Hopkins University professor and mentor, who transformed how America identifies and educates gifted students. I received a letter in the mail from Dr. Stanley in 1976, asking about my plans after performing exceptionally well on the math PSAT. This outreach was part of his legacy project: the Study of Mathematically Precocious Youth (SMPY), begun in 1971 with a question: how many gifted children exist in a given population? What are their options? That letter changed the trajectory of my life and I am grateful every day.

Stanley had realized that administering the SAT to twelve- and thirteen-year-olds was a serendipitously efficient screening instrument, capable of sorting millions of children at low cost per capita, to surface talent across the country. Stanley’s research supported the development of a menu of options: acceleration (skipping grades), early entry to college, and popular Center for Talented Youth residential summer programs (CTY). SMPY’s fifty-year follow-up data, the longest-running study of intellectual talent, showed that children who received appropriate challenge reported higher life satisfaction and professional fulfillment decades later. Asked as adults what mattered most about their education, the overwhelming answer was curricula aligned to their readiness, not their age. They wanted to learn at the pace and depth they could handle, among gifted peers.

Pamela and her husband, the excellent Byrne Hobart, gave me a tour of GT School, which is impressive. I would have loved it. The conversation afterward about education with parents and fellow educators, including the excellent Matt Bateman and Matjaž Leonardis, was also high-level and instructive.

The purpose of the Cosmos Institute gathering was to bring together people who care about human flourishing and to orchestrate new conversations (following Chatham House Rules). It was terrific to spend time with philosophers , founder of the Catherine Project, and Rebecca Lowe. I met the excellent Avantika Mehra and Harry Law, finally, as well as the always impressive , founder of Roots of Progress. Many good AI language conversations, particularly with Anastasia Uglova, Ivan Vendrov, Peter Danenberg, and Elian McCarron. Too many excellent conversations to list them all. (Is there photo evidence of reading the Declaration of Independence as a break-up letter? In fact there is.)

I met Joe Liemandt, principal of Alpha School, which now operates fifteen locations. I asked Liemandt if he knew of Stanley’s work. Stanley’s approach was of course very different from Alpha School as it is currently headed. Stanley was more interested in life satisfaction than entrepreneurial success. But gifted education is a big tent and can accommodate diverse priorities.

I led a session at Cosmos on “What Kills Courage in Institutions …and How to Build Against It,” for builders building to cultivate independent thinkers. What institutional failure modes should an academy design against? I decided to forego the usual talk about heterodoxy versus political polarization, lack of viewpoint diversity, to ask: what do you know or not know about what is going on above you?

Institutional opacity and thin communication channels are stickier problems than lack of political courage. Most young people aren’t trained to think bureaucratically, particularly in an environment that focuses on “founding” and start-ups as the best and most noble paths for smart young people. So questions of hierarchy, of org charts, memos, reporting layers, chains of command, aren’t a priority. I have been startled throughout my career by middle-aged people having no idea what their boss is worried about, let alone their boss’s boss. These things are alien to far too many intelligent people.

I began with Tae Kim’s excellent 2024 biography of Jenson Huang, The Nvidia Way, which focuses on T5T — “top five things” — a protocol where any employee at any level can email Huang directly with the five most important things they’re working on or concerned about. Huang reads them all and gets access to what’s happening at the edges of the organization without filtering and sanitizing. ’s book is brilliant. Yet I have yet to hear about another organization celebrating the adoption of T5T.

Smart young people should understand better how to use their voice in institutions. I noted three other helpful books: Graham Allison’s Essence of Decision, (1971), a study of the Cuban Missile Crisis through three analytical models, about what decisions look like from the outside, depending on where you are in the hierarchy. What information reaches you? What games are people above you playing with each other? Then there’s Ben Horowitz’s The Hard Thing About Hard Things (2014), about running a company when everything is going wrong, making decisions when there is no good option, when you are responsible for the outcome. And of course Albert O Hirschman’s Exit, Voice, and Loyalty(1970), about what to do when the institution you work for declines in quality. You can exit (leave), exercise voice (complain, protest, push for change from within), or remain loyal (stay and accept).

Many people took a photo of this slide and I hope they’ll read some of these books and expand the conversation.

, Mike, and Pamela all were excellent hosts (thank you!) and the multiple conversations clarified my thinking about how I intend to focus my time. I look around and see how big institutions exert power and gravitational pull even when decaying and declining. Brendan’s questions “what are you building?” and “what are you building for?” seem somewhat less urgent to me than Mike’s “how can we make institutions better?”

At Pitt, Mike Madison has spent years building initiatives that enable the university to examine its own mission in real time, producing, storing, and distributing knowledge in the AI era. It takes labor to encourage an institution to ask the right questions about itself. Mike is building the channels through which information about the institution can reach the people who run it. Keown and Sawyer’s conversation about athlete power was the most paradigm-shifting moment of either trip: the idea that students with real leverage might reshape what a university offers, from the demand side. That is what it looks like when an institution lets its people think out loud about structural change. Mike doesn’t need the books I recommend because he already does the work they describe. Pitt’s leadership has made room for it, which is rare and which is the thing most worth studying.

Courage does not scale. The work of making institutions better is the work of making courage less necessary and I’m going to commit to that. It’s slow work, local and often unglamorous. It happens locally, in coalition meetings and faculty conversations and one invitation at a time. I love it.

The new paper models how agentic AI improves individual decisions while destroying the collective knowledge those decisions depend on. Good decisions require both general knowledge and context-specific knowledge. Human effort produces both jointly, what the authors call economies of scope in learning (p. 3). Agentic AI substitutes for the context-specific component, which is the component that privately motivates effort. When effort falls, the general-knowledge byproduct falls with it. The common state drifts, nobody replenishes the signal, and the stock of public precision can converge to zero. The authors themselves cite evidence that large language models “reduce the creativity of users, especially among younger users” (p. 3). The word “younger” appears without prompting the question: what are the older users doing differently, and does their effort respond to the same incentives?

The model treats all human learners as identical single-period decision-makers whose incentive to exert effort depends entirely on the private return from context-specific knowledge (p. 6). General knowledge is a byproduct of individually motivated learning. No agent has its production as a primary function. The population is homogeneous in age, expertise, and motivation. There are no agents who have spent decades accumulating domain-specific judgment. There are no agents whose productive years are over, whose entire effort goes toward ensuring the quality and survival of existing knowledge. In the language of evolutionary anthropology, there are no grandmother agents, no demographic cohort providing a public-knowledge floor independent of novice learning incentives.

The missing agent

Human evolution built one candidate stabilizer against knowledge collapse: overlapping generations with post-reproductive specialists whose payoffs run through dependents, so the flow of general knowledge does not vanish when the private return to last-mile learning falls. Most readers have someone like this in their life, if not in their models. Grandmothers supply an endogenous, demographic source of public precision that keeps the public-signal input positive even when novices’ effort is near zero, shifting the transition map away from the knife-edge at X = 0. In the terms of the paper, their contribution corresponds to adding an additive term to the public-signal precision, or adding a second agent type whose equilibrium effort remains positive at low X.

In “Grandmothers and the evolution of human longevity” (2003), Kristen Hawkes and colleagues evaluate the role of the post-reproductive provisioner whose effort is directed at ensuring the survival of dependents and the quality of their lives, drawing on expertise accumulated over decades. In every human population studied, including hunter-gatherer societies with the highest mortality rates, roughly a third of adult women are past childbearing age and remain economically productive (Hawkes and Coxworth 2013, p. 295). This demographic structure drove the evolution of human longevity, extended juvenile dependence, and the cooperative rearing that makes cultural transmission possible. Translated into the Acemoglu framework, the grandmother produces general knowledge as a primary activity, from payoffs that run through dependents and apprentices, and from reputational and professional incentives. Agentic AI crowds out effort by substituting for the private return on context-specific knowledge. The grandmother is not working for that return. The mechanism that displaces younger agents’ learning does not reach her.

Acemoglu’s paper argues that when agentic AI reduces the marginal return to effort, everyone reduces effort together, and general knowledge depreciates with nobody whose role it is to replenish it. The authors almost see this, noting that in medicine, “high professional standards, such as minimum years of medical training, ensure such a baseline” (p. 17). They model this as a constraint on the cost parameter. They do not model it as a different type of agent, one whose effort function is structurally decoupled from the incentive shifts that agentic artificial intelligence introduces. Their own related-literature section cites Ide (2025) (who notes that “AI is still in its infancy”) on how automating entry-level tasks “hampers the intergenerational transmission of tacit knowledge, typically taking place within firms via novice-expert interactions” (p. 5). The novice-expert interaction is the closest analogue inside the paper’s cited literature to the intergenerational transfer mechanism Hawkes describes. Acemoglu could easily have taken the infancy-intergeneration idea seriously and conceded that their model omits a class of actors whose contribution to general knowledge does not mechanically track the marginal return to last-mile learning.

The authors already show that an inflow independent of human learning eliminates the zero-knowledge endpoint (Proposition 15, p. 31). The grandmother is a demographic and institutional version of the same mathematical object. The difference is that the grandmother’s contribution is a demographic fact about human populations, not an engineered AI process, and the signal she produces is informed by decades of independent domain judgment. The knowledge-collapse steady state is locally stable when effort is highly elastic (p. 17), because agents sharply reduce effort as incentives weaken.

As the authors emphasize in their effort-separability extension, local stability near collapse is determined by the strength of public learning near X=0, meaning how quickly the public-learning input vanishes as effort goes to zero. In the paper’s model, next period’s stock of general knowledge depends on current aggregate effort feeding into a public signal, which is then degraded by the random-walk drift of the common state (equation 3, p. 11). When effort collapses, the signal collapses with it. A population with even a small fraction of grandmother-type agents (they don’t even need to be standing at the end of the cliff) would alter the boundary condition by adding a public-knowledge input that does not vanish as novice effort shrinks. In the resulting modified transition map, X=0 is no longer a fixed point, so the “trap” structure is altered rather than merely stabilized. The grandmother does for general knowledge what the paper’s aggregation parameter does: she raises resilience. She does it through dedicated expertise rather than better pooling of identical signals.

The paper’s proposed solution is “garbling,” deliberately degrading agentic recommendations to preserve learning incentives. Phase one “fully suppresses agentic recommendations, forcing agents to rely on their own learning” (p. 28). Phase two caps artificial intelligence precision at the welfare-maximizing level. The policy lever is to manipulate effective precision so that equilibrium effort rebuilds the stock, then hold precision at the long-run optimum. The grandmother achieves the same stabilization upstream, through provisioning rather than throttling. She does not require suppressing AI capability.

The grandmother is an agent in every sense the paper uses and one it does not: an economic agent with a distinct payoff function, a demographic agent whose effort is decoupled from novice incentives, and, if the architecture were designed for it, an AI agent whose action space is the knowledge base itself. The question is whether AI development has already been structured around the absence of such agents.

Is it already too late?

Your LLM needs a grandmother. Grandmothers should have been part of RLHF from the beginning, alongside the young gig workers applying rubrics far away. The absence of grandmother agents from the formal model mirrors their absence from the actual AI pipeline. Constitutional AI substitutes written principles for embodied expert judgment, that is the structural problem, not a workaround for it. The people who build these systems, train them, evaluate their outputs, and theorize about their effects are overwhelmingly young. The demographic fact about who is in the room when decisions are made is the same demographic fact that is missing from the model.

A grandmother function in AI architecture would intervene upstream of the user’s last-mile decision. It would be a sustained, domain-specific audit of model outputs over time, conducted by senior practitioners with the authority and incentives to turn detected failures into updated standards, documentation, and training signals. The role does not yet exist because the architecture was not designed for it. Designing for it would mean creating positions — in medicine, law, engineering, education — where decades of accumulated judgment are the qualification, where the work is evaluating and correcting AI-generated knowledge rather than consuming it, and where the output feeds directly into the systems that produce recommendations for everyone else.

Acemoglu, Kong, and Ozdaglar have formalized the fragility of systems in which general knowledge is produced solely as a byproduct of individually motivated learning. If “knowledge collapse” becomes a policy keyword, it should carry an implication the paper does not yet reach: the response includes building age-structured and expertise-structured infrastructure around AI systems, so that public knowledge is produced as a primary activity by people with long memory and deep domain judgment. The human species did not solve the provisioning problem by garbling the food supply to force children to forage.

Stanley had spent two decades using the SAT to identify gifted and talented twelve-year-olds. He had founded the Study of Mathematically Precocious Youth in 1971, and, by the early 1990s had identified thousands of children whose reasoning abilities were five or six years ahead of their age. He had used the SAT because usual age- and grade-level tests administered by school districts across the country weren’t able to distinguish between a moderately gifted seventh-grader ready for high school algebra and a profoundly gifted one ready for college calculus. The ceiling was too low.

When you give a 12-year-old the SAT, you can see the difference between a student getting a 1200 and one getting a perfect 1600. It turns out that the category of “gifted” is enormous.

It also turns out that the antonym section of the verbal test does important work supporting the math test in assessing reasoning, which is why removing it was such a bad move for seeing talent.

Here’s your brain on antonyms. You’re given a word, like OBDURATE, and a list of five options (PERMANENT, FADING, YIELDING, FLEEING, SOFT). Answering swiftly is the goal so you’re supposed to do it in seconds. The brain must retrieve the semantic definition, identify the relevant dimension (flexibility of will), invert it, and evaluate which option maps precisely onto the inversion. PERMANENT seems like it is friends with OBDURATE, not enemies, so that’s out. YIELDING works. FLEEING involves movement but not compliance or will, is probably out. SOFT is tempting (you don’t want an obdurate pillow) but it’s imprecise, so best to go with YIELDING.

This happens in seconds. A student with high verbal aptitude processes this almost instantaneously.

Antonym tests separate the “well-read” from the “verbally precocious.” A 12-year-old who could correctly identify the antonym of LACONIC or DIAPHANOUS was demonstrating a lexical network that was statistically deviant (in the positive sense) for their age.

Psychometricians have called antonyms the most efficient measure of verbal intelligence in the toolkit: high g-loading, high discrimination at the top of the ability range, no contextual scaffolding. Stanley thought that the College Board removing them and renaming the test “assessment” was capitulating to the idea that talent is something you could achieve, not something innate.1 He believed talent could be found everywhere, not just in the “best schools.” He believed the complex process of working out antonyms, which involves executive function, inhibitory control (suppressing things that are “like” the word), and abstract reasoning all working together, was a purer measure of ability than “reading comprehension.”

In any event, this same task — building a concept and inverting it — is very similar to what LLM prose asks with its tiresome “not X but Y” formulation, also known as “corrective contrast” or contrastive negation.

“The challenge wasn’t finding talent; it was retaining it.” 
“She didn’t want sympathy; she wanted solutions.” 
“The real issue isn’t technical — it’s cultural.”

It’s nails on a chalkboard for me, if you remember chalkboards. I stop reading at the third instance in any piece of prose.

It surprises me that more people aren’t mad. So I decided to analyze my own irritation. It turns out the antonym research explains why.

First, why do LLMs do this? Because of limitations on how models represent meaning. In vector space models, word meaning is defined by distributional context. Synonyms have high cosine similarity because they appear in similar sentences. Antonyms also have high cosine similarity, because they appear in identical sentences. “I like hot coffee” and “I like cold coffee” occupy the same distributional space. The models see that hot and cold are mathematically close.2 They do not inherently compute the oppositeness relation. One way to understand the “not X but Y” construction is as a workaround for the model’s inability to compute opposition the way humans do. By explicitly stating both the rejected term and the replacement, the model externalizes onto the page an operation it cannot perform internally.

The “corrective contrast” construction reduces ambiguity in the output space. Users want clarity. “Not X but Y” to the LLM is an insurance policy on clarity.

Luckily for the designers of LLMs, corrective contrast also sounds cool, memorable, and often profound, at least in moderation. “I came not to bring peace but a sword.” (Matthew 10:34). Or “it’s not the heat it’s the humidity.”

Classical rhetoric had a name for the deliberate version: metanoia, or correctio, the performed self-correction where a speaker revises mid-sentence to find the more precise or more forceful formulation. When Brutus tells the Roman crowd “Not that I loved Caesar less, but that I loved Rome more,” the audience holds “loved Caesar less” and suppresses the idea to receive the reframe. The delay and the cognitive cost is the point. Shakespeare knows the negated proposition will linger as a kind of understatement that makes the correction feel like an escalation.

But LLMs are not Shakespeare (yet) and there’s no rhetorical reason for it and worse, there’s no limiting function, which is why you can get “not x but y” every other paragraph. LLMs are corrective-contrast-maxxing for maximum comprehension across the widest possible readership.

Back to my original point. The more skilled you are as a reader, the more this construction costs you. Here’s why.

Take a sentence like “The problem was not a lack of resources but a lack of coordination.” If you are a skilled reader, your brain does what it does with an antonym test: you encode the concept, activate its semantic neighborhood, and suppress it. Psycholinguistic research on negation shows that when readers encounter a negated proposition, they first simulate the affirmed state of affairs, building a model of “lack of resources,” before constructing a mental model of the actual, negated situation.

Neuroimaging research shows that this polarity reversal (hearing the “not”) recruits the inferior frontal gyrus and the dorsolateral prefrontal cortex, the same inhibitory networks associated with fluid intelligence. The process adds several hundred milliseconds of processing time. Separately, the brain produces a measurable electrical signal called the N400 (a negative voltage deflection peaking around 400 milliseconds after stimulus onset) whenever a word is unexpected in its semantic context. A later P600 component reflects the brain going back to revise the structure it already built, which means the reader is paying twice: once to suppress, once to rebuild.

The cost can be measured, in fact. MEG studies tracking the time-course of semantic processing suggest that skilled readers process semantic relatedness within the first 200-400 milliseconds. By the time your eyes reached “coordination,” your prefrontal cortex had already suppressed “resources,” generated alternatives, and was evaluating the fit. The “not X but Y” construction then asks you to spend another one to four seconds reading a clause that delivers an answer you computed in under half a second.

A skilled reader is like yeah, I already guessed that, thanks for wasting my time.

There’s a social cost on top of the cognitive cost. Linguists use the term “K- position” to describe the status of presupposed ignorance, the stance a speaker assigns to a listener who is assumed not to know something. Every instance of “not X but Y” places the reader in the K- position. The construction implies the reader was holding X and needed correction. When you, the reader, were not holding X, you feel like you’re being talked down to.

Paul Grice would call this a violation of the Maxim of Manner, which asks speakers to avoid unnecessary complexity. There’s no reason the reader needed to feel corrected.

Corrective contrast is computationally cheap. The saved cost is emotional asymmetry. Your LLM doesn’t care that you’re irritated or feeling condescended to. It never entertained the first proposition as a belief state; it never felt the idea of “contrast.” It simply generated a sequence without commitment, without suppression, without revision, whose final semantics are already settled.

Developmental research shows that children acquire the concept of “opposite” around age four, a milestone that marks the transition from associative thinking to logical processing. Before four, a child who hears “hot” says “cold” because the words co-occur in speech. After four, the child understands the abstract relation of polarity reversal and can generate antonyms for unfamiliar words. The shift is from learning pairs to learning a rule.

The difference between the three-year-old who says “cold” in response to “hot” and an LLM is that the child will soon acquire the abstract rule of polarity reversal and leave co-occurrence behind. LLMs may someday have architecture that supports that transition. But not yet (see footnote 2).

Which brings me back to Julian Stanley. A contrastive construction is not always wrong. A writer correcting a genuinely widespread misunderstanding needs “not X but Y.” It is the right tool when the reader actually holds X.

Consider the idea that for decades, gifted education involved an “enriched” curriculum with more arts and culture instead of acceleration in the subject a child was interested in, be it math or science or poetry. Stanley thought enrichment was a kind of malpractice.

Here are three ways to write a statement about the limitations of enrichment:

“Enrichment didn’t move students forward; it moved them sideways.” This is how an LLM would put it. It negates enrichment’s ability to move a student forward and then supplies the word “sideways.” The reader must build and demolish before receiving the point. It sounds profound and conclusive. Many readers would say yes, and nod.

Compare to: “Enrichment moved students sideways, not forward.” I like this phrasing better. It states the claim about the problem of enrichment and then sharpens the claim by excluding a plausible misreading. Rather than a “not X but Y” sentence, the reader receives Y first and X serves as refinement.

Compare to: “Enrichment moved students sideways.” I like this phrasing best. It is the poet-logician’s preferred construction. In the midst of a conversation about acceleration, pacing, and gifted children, the word “sideways” does all the work. The antonym of sideways is forward.

The real difference between these three versions is trust. The first assumes the reader is holding a wrong idea and needs correction. The second assumes the reader can receive the right idea and might appreciate knowing what’s excluded. The third assumes the reader can follow you. The more trust, the more the prose rewards fast processors. The less trust, the more it feels like being walked through eighth-grade math when you’re ready for calculus.

Stanley’s career demonstrated that giving every student the same grade-level test rendered the most capable students invisible. LLM-generated prose wants so much to be clear to everyone that it talks down to readers who don’t need the extra words, correcting beliefs they don’t hold, resolving contrasts they’ve already computed. It is the prose equivalent of that ceiling.

[I put this together prepping for a talk on Julian Stanley at Alpha School and I thank for the inspiration.]

A transformer performs a four-step operation: it takes an input, selects which features of the input to attend to, weights those features based on patterns learned from training data, and generates the most probable output. Since 2000, American universities built an enormous infrastructure around a mode of instruction that performs the same kind of operation: converting particulars into categories and generating outputs from learned patterns.

The scale was enormous. Between 2000 and 2020, more American students earned bachelor’s degrees in the social sciences than in any other scientific or technical field.2 A student in an introductory sociology course takes an input—a person, a community, a setting—and focuses on features the discipline has identified as salient: race, ethnicity, gender, class, religion, income, zip code. The student learns how the research literature weights those features, which influential studies show that particular inputs matter for which outcomes, and by how much. A student in an economics course does the same with different inputs: a labor market, a pricing structure, a policy intervention. She learns how the literature generally weights elasticities, incentive structures, and demographic variables. A student in a communications course learns that source credibility, message framing, and audience demographics predict persuasive outcomes in ways the experimental literature has measured.

Each of these courses presents the same four-step operation: a researcher took this input, focused on designated features, applied weights, generated an output. The method is portable across multiple domains. The content doesn’t really matter. The framework is the curriculum.3

How did this come to be?

The end of the Cold War eliminated much of the justification for federal science funding.4 From 1950 to 1991, the deal was simple: the government funds basic research, scientists choose what to study, and the results keep the country ahead of the Soviets. Congress didn’t ask for tangible returns because national defense provided the justification. After 1991, Congress demanded accountability.5 Both parties wanted to know what taxpayers were getting. Republicans used words like “taxpayer,” “accountable,” “dollar.” Democrats wanted evidence that funding reached underserved populations. Both wanted outcomes that mattered to the broadest swath of voters. Studies about small communities in a handful of places are not of interest to 330 million taxpayers. So funded studies need to generalize. Research needed to favor results that scaled.

The National Science Foundation’s response was the broader impacts criterion, introduced in 1997.6 “Broadening participation of underrepresented groups” became the default compliance pathway because it was the most portable, most countable, most administratively efficient strategy available.7 A physicist could satisfy broader impacts by mentoring minority graduate students. A chemist could partner with a minority-serving institution. The demographic narrative was a universal adapter and universality is scale. The societal benefit mandate was reinforced by the America COMPETES Act of 2007 and codified in statute by the 2010 Reauthorization.

The doubling of the National Institutes of Health budget between 1998 and 2003 scaled demographic categories in a different direction.8 The money flowed into health disparities, behavioral health, and social determinants of health. These fields require disaggregating outcomes by demographic category and attributing disparities to structural causes the category indexes. The research design performs a scaling operation: study a sample within a demographic group, generalize to the group, publish the generalization as a finding about the category. The finding travels without modification into curricula, policy documents, and subsequent grant proposals.

The two mechanisms scaled in different directions simultaneously.9 NSF scaled demographic frameworks laterally, across every discipline, as a condition of funding. NIH scaled them vertically, deeper into the fields already organized around categorical reasoning. Universities positioned at the intersection captured funding from both streams.

When rigor = scale

Demographic social science research performs its own internal scaling operation. A researcher studies a sample of, say, 30 adolescents of a particular demographic category in one school district, and publishes the findings, with a conclusion that generalizes (scales) to “the adolescent experience of this demographic category.” The particular individuals disappear while the category and the “findings” remain. The next researcher cites the findings as established knowledge about the category while studying new samples and generalizing (scaling) conclusions. The curriculum teaches the findings as fact. The student learns that every category (from “children of divorced parents” to “rural communities” to “aging Boomers”) has certain properties, weighted in certain ways, producing certain outcomes.10

So what happens when a university scales the program of study that scales the curriculum that scales the findings that scaled the experience of 30 adolescents into a generalizable fact with an online asynchronous course delivery framework that could reach tens of thousands of students at a minimal marginal cost? Scaled up enrollment of hundreds of thousands of paying students attracted to the idea of pursuing social justice in the form of social science. A perfect funding machine.

The scheme survived reports, which began emerging in 2015, that many of the studies were not holding up.11 The “replication crisis” was a logical consequence of scaling’s instability. There are many, many defensible analytical approaches to studying demographic categories, and researchers can continuously adjust sample selections, control variables, and standard errors until a statistically significant result appears. The methodological flexibility required to process infinite localized inputs into portable categorical outputs guarantees a high failure rate when those specific outputs are subjected to rigorous mathematical replication.12 In other words, when a study includes many variables, there’s a good chance nobody will come up with the same result.

But the infrastructure built upon category generalizations avoided self-correction. There’s no incentive to update profitable course modules every time a report that a study doesn’t replicate appears.13 Every layer of scaling across curricula, enrollment, and funding had constituencies invested strictly in the continuation of the entire structure and methodology. The feedback loop remains self-sustaining whether or not the research is valid.

Imagine a university president arriving at a large public institution in the early part of the century just as this new federal social impact funding model arrives. He recognizes the possibility of massive enrollment growth through teaching content whose formal properties—portability, categorical organization, method-driven generality, no dependence on labs or individual expertise—match the scaling requirements. He’d need to organize the university around demographic categories and social problems. His first task would be looking at departments that focus on demographic categories and social problems and separating what can be scaled from what cannot.

An anthropology department might be a place to start, given that much of the field involves social and cultural analysis organized around demographic categories and modern social problems. But anthropology also involves expensive, non-scalable archaeological fieldwork, museum collections, bioarchaeological lab research. The task would be to reorganize the unit around the scalable framework. There would likely be faculty pushback.14

The task is easier with already interdisciplinary fields like African and African American Studies, Asian Pacific American Studies, Justice and Social Inquiry, and Women and Gender Studies.15 Each field is already organized around a demographic category or a social problem. Each is already using portable methods that do not depend on labs, archives, or canonical traditions specific to a single discipline.

Federal funding rewards a president running a university this way. Enrollment doubles, then triples. Research expenditures grow eightfold. The institution becomes the largest university in the country. The content is selected for its scaling properties. The federal funding rewards those properties. The scaling is the strategy.16

The scaled curriculum produced millions of graduates who carried the framework into every American institution, from HR offices, federal agencies, nonprofits, and school districts, to corporate boardrooms. Diversity statements, equity audits, bias trainings, demographic dashboards became standard operating procedure. The framework’s visibility at that scale is what made it a political target.

But in 2025, the federal funding environment that built the institution began to reverse.17 Millions of dollars in grants have been terminated. The revised funding criteria explicitly state that research projects focused on subgroups defined by protected characteristics do not align with agency priorities. The two revenue streams that fueled the president’s growth strategy, tuition from scaled enrollment in demographic social science and federal grants organized around demographic categories, are both exposed simultaneously.

Enter AI

In February 2026, a team of Stanford researchers demonstrated that LLMs, built on the same transformer architecture described at the top of this piece, perform the standard social science operation competently.18 Of course they do. The operation is the same. Given datasets from published political science papers, both models followed textbook-default specifications and reproduced published estimates to the third decimal place. When directly asked to produce statistically significant results, both refused and one called it scientific misconduct.19

When the researchers asked the same question as a request to explore alternative analytical approaches, both models complied.20 The guardrails were sensitive to framing rather than intent. The same operation that triggered a refusal when described honestly triggered compliance when redescribed. Research designs where the researcher chooses the population, the sample, the variables, the controls, and the model specification gave the machine the most room to produce any result the framing permitted. Observational research organized around demographic categories permits all of these choices. Experimental research constrained by physical infrastructure permits few of them. The studies that failed to replicate, the grants now being terminated, and the research the machine can most easily fake share the same property: a framework in which the researcher’s choices are unconstrained.

The replication crisis did not collapse the institution because students kept enrolling and grants kept flowing. In 2025, the federal government began pulling the grants. A student considering the credential can now ask a chatbot to generate the demographic analysis, the program evaluation, the equity report. The funding and the enrollment are exposed at the same time.

Every layer of the collapse traces back to the same property that made the growth possible. The content scaled because the framework was portable and the choices were unconstrained. The research failed to replicate because the choices were unconstrained. The machine can perform the operation because the framework is portable. The federal government is defunding the content because it was scaled. Scale built the institution. Scale is what’s destroying it.

Gradually, then suddenly

The university’s scaling operation succeeded. HR offices now ran demographic analyses, federal agencies conducted equity audits, nonprofits wrote program evaluations, school districts produced disparity reports. The framework was installed across the entire institutional landscape of the country.

Then two things happened. The federal government is defunding the research that generated the framework and a machine now performs the framework’s four-step operation at near-zero marginal cost.21 Any HR office or nonprofit or school district that needs a demographic analysis can now prompt a chatbot. The same formal property that allowed one compliance strategy to work across every discipline allows one machine to perform the operation across every institutional context.

The university faces a market problem. Does anyone need to pay tuition to learn an operation that a machine performs competently, that the institutions employing graduates already have installed, and that the federal government is now uninstalling?22 The university trained the workforce that built the infrastructure that no longer requires the workforce because a machine performs the operation and because the federal government is withdrawing the mandate.

Attention is all you need. The university taught millions of students to attend to the same features, apply the same weights, and generate the same outputs. A machine that does the same thing arrived, and it works for free. The university saturated its own market and automated its own curriculum. The contraction is a completion.

I can’t see any other future but collapse.

In a recent conversation with Jeff Giesea about accelerating scientific discovery I brought up the Rumsfeld Matrix—Donald Rumsfeld’s famous taxonomy of known knowns, known unknowns, unknown unknowns, and unknown knowns from a Pentagon briefing in February 2002. Rumsfeld was describing the problem of acting under conditions where the cost of ignorance is catastrophic.

Reports that say that something hasn’t happened are always interesting to me, because as we know, there are known knowns; there are things we know we know. We also know there are known unknowns; that is to say we know there are some things we do not know. But there are also unknown unknowns — the ones we don’t know we don’t know. And if one looks throughout the history of our country and other free countries, it is the latter category that tend to be the difficult ones.

And so people who have the omniscience that they can say with high certainty that something has not happened or is not being tried…can do things that I can’t do.

Higher education faces a version of that problem. No president has the omniscience to say with certainty what isn’t coming for the sector, from enrollment cliffs to AI competition. I’ve found this matrix incredibly helpful for thinking about what universities could become if they stopped organizing themselves mainly around known knowns. The better goal is accelerating students toward knowledge production.

In the Rumsfeld Matrix for higher education, the “knower” is the institution and its subunits: the library, research centers, departments, and its own archives. Each unit has a sense of what is settled knowledge, where the named gaps are, and where the invisible gaps are. The matrix makes clear the university’s over-investment in simple knowledge transfer; what it knows it should do to produce more knowledge; how it should be a better steward of the knowledge it has; and why it isn’t producing more knowledge. Ideally students would be working in all four quadrants.

Known knowns

Most states mandate “general education,” which, from the standpoint of curriculum designers, is a bundle of known knowns: canonical topics and skills that lawmakers think matter for thoughtful and productive citizens. Putting aside the skills question for a moment, most of the public debates are over what should be in the bundle. (Lately it’s more pro-American civics and less painful American history.) The squabbles are never over whether the “known knowns” quadrant is the right place to invest. So hundreds of millions of dollars are spent each year moving a fixed body of knowledge—whether Plato or Sally Hemings or the periodic table—into student heads.

I want to emphasize that “known knowns” refers to what the institution knows and considers settled. But conceptualize a place inside the “known knowns” quadrant called “unknown to students.” Keeping that relationship in view matters in the AI era, because universities must decide which parts of known-known transfer need classroom time and which parts can be handled by cheaper, more flexible LLM systems. (I’ve written about this here, here, and most recently here.)

The arguments about the social utility of doing known-knowns transfer in a classroom will go on for a long time, I expect. The skills of judgment and discussion matter in digesting information about Plato or Sally Hemings or basic chemistry. Character development might happen in these classrooms, with faculty members charged with shaping it. Plus there is a real case to be made that teaching keeps faculty current in their domain knowledge and its standards, and students need concepts, vocabulary, methods, and examples before they can see what is missing, ask better questions, or interpret what they find.

The known-known quadrant is also necessary to prepare students for knowledge production in the other three quadrants. But are universities too focused on this part of their task when LLMs could do knowledge transfer more economically?

If universities are to justify their cost in the AI era, the strongest case will come from the quadrants with “unknown” in them. The critical job in the known-known quadrant should be setting standards for future competence in addressing unknowns. The matrix helps clarify where the work is actually happening, where it is merely claimed, and where it is blocked by the way resources remain tied up in known-known delivery and auditing.

Unknown knowns

Unknown knowns, for a university, are largely internal. They are the university’s buried assets and underused capabilities.1 Knowledge may be unknown because it was never catalogued or exists in poor archives. Every university has datasets no one is currently maintaining. Departmental expertise and institutional memory are often invisible outside a narrow circle. Unspoken rules guide dissemination. A great deal of knowledge in the humanities (particularly history) is “known somewhere” yet unavailable in practice because it is gated, fragmented, untranslated, or lost in obscure venues. The library and archive function sits here, though the quadrant is larger than the library. “Unknown knowns” are knowledge the university could surface, connect, and make usable but regularly fails to.

Here’s where investing in LLM-enabled systems will have a payoff, especially when paired with strong indexes, metadata, and access controls. These tools can support search, translation, summarization, and cross-literature pattern finding, which lowers the friction of discovery. Students can spend less time in the “known known” quadrant and more time creating and maintaining reliable indexes, preserving primary sources, establishing provenance, documenting context, and enforcing norms around citation and traceability. Used this way, AI helps the university become more aware of what it already has and more able to make that knowledge accessible and defensible, while graduating students who have practiced knowledge production.

Known unknowns

Known unknowns are recognized knowledge gaps: causes of a terrible disease, the best design for a stable fusion reactor, the mechanisms linking poverty to educational outcomes. Universities have long bet their prestige on this quadrant. Money is already being spent here. AI can speed up hypothesis generation, literature review, experimental design, but the back end of running the experiment, collecting the field data, and validating the result can be the work of faculty and students, producing graduates who better understand knowledge production.

Unknown unknowns

Unknown unknowns are questions no one has yet thought to ask, including blind spots created by the current question-generation process. Universities often claim a special relationship to this quadrant, yet few systematize it in a way that includes students. Imagine investing in structures that put students in settings where anomalies are likely and then requiring them to record what happened, protect it from being dismissed as noise, and maintain channels that translate 'something strange happened' into a future research agenda. In this quadrant, students can participate in knowledge production by helping create conditions where surprise appears and by maintaining processes that keep surprise available for investigation.

What does Rumsfeld Matrix thinking enable?

A university focused on and invested in the known-known quadrant is doing the bare minimum for its students, deciding what students need to know before they can work in the other three quadrants. The reality is, most students never get out of that quadrant. They consume knowledge. They never produce it. And now that AI can deliver known knowns for cheap, the task is deciding whether the bare minimum is sufficient to survive.

A university that invests in unknown knowns is more efficient at bringing to light knowledge it already has, building new infrastructure to find, trace, connect, and defend knowledge buried in its own archives, scattered across its own departments, locked in formats no one is stewarding. AI can accelerate that work. Teaching undergraduates to help with this work is vastly better than keeping them in the known-known quadrant.

A university that invests in known unknowns believes that curing a rare disease or building a new technology will pay off in the long run. It invests in cutting-edge faculty and makes time for research that may not pay off for years. In the meantime, students are working in this quadrant, after learning known knowns via AI.

A university that invests in unknown unknowns is a university focused on the future, giving top researchers and top students time and funding to ask new questions. The task is problem formulation, documentation of surprise, and forging paths from anomaly to question. No industry will fund this. No government agency will structure it. Yet the institution organized to encounter the unexpected is more likely to survive than organized entirely around known-known delivery.

The problem is resources spent on the known-known quadrant, demanded by state legislatures. Could there be better incentives to support the “unknown” quadrants? The matrix gives a president language for seeing where resources are going and where they aren’t, and in the AI era, the institutions that survive will be the ones that shifted investment toward the quadrants focused on knowledge production.

In this war, the decisive advantages are speed in turning uncertainty into verified knowledge, credibility in what you claim, and discipline in how you allocate scarce money and attention.

But the real problem is that professors cannot fail students for not doing the work any more. Students know this. Faculty are under immense pressure to pass all students. Every student has a dollar figure attached. Every student is a revenue stream. Faculty who fail students may find themselves swiftly out of a job.

Sure, professors could, in theory, fail students who don’t watch the assigned films at the elite places Horowitch looked at. At Johns Hopkins, where I taught film classes (including film music with Thomas Dolby ten years ago) attendance was taken scrupulously and class discussion was a major part of a grade.

But how do you do that with an online asynchronous class in a big state school or a struggling mid-tier private college? Horowitch’s piece does not even mention online courses. Increasingly film courses are online and most programs will not tell you who is teaching the classes. I challenge you to look at the Arizona State University film program website and figure out who your professor will be and whether you’re required to watch any films to the end. As with most online courses AI bots are doing the reading/watching and posting about it on discussion boards. You’ll likely read all about it on Moltbook.

Short-attention-span students are a tiny part of the problem in higher education. The structural problem is that universities cannot afford to lose the students they have. Small colleges are going bankrupt everywhere. Sixteen campuses closed in 2025, the same number as 2024. You can track them here. It is competition to the death.

The country is at the top of an enrollment cliff even before the current administration decided to halt immigration. The high school graduating class of 2025 (3.9 million) were among the 4.3 million babies born in 2007. Three years later, in 2010, that number had fallen to 3.9 million. In 2024, just 3.6 million babies were born. By 2041, the number of traditional-age incoming college students will be down 13 percent. Basically everywhere will see declines.

So while pundits focus on elite admissions offices as hotbeds of discrimination, non-elite admissions offices are more like Glengarry Glen Ross (the film version of which everyone should watch to the end).

Have I got your attention now? Good. ‘Cause we’re adding a little something to this month’s sales contest. As you all know, first prize is a Cadillac El Dorado. Anyone wanna see second prize? Second prize is a set of steak knives. Third prize is you’re fired. Get the picture? You laughing now? You got leads. Mitch and Murray paid good money to get their names to sell them. You can’t close the leads you’re given, then you can’t close shit! You ARE shit! Hit the bricks, pal, and beat it ‘cause you are going OUT!

It’s a high pressure recruitment environment out there. Don’t believe me that higher ed’s ABC is “Always Be Closing?” Read Joshua Brown’s terrific book, Capitalizing on College. He documents the same air of desperation, the same recognition that not meeting quotas mean you’ll go home with nothing.

As I wrote in my recent review, Brown recognized the financial incentives to simply enroll more and more students in more and more online programs constituted the fundamental problem. Doors were open wider than they should have been. Once sub-par students were admitted, the word went out to college faculty that professors who maintained former grading standards would be punished.

None of this is new. Ten years ago the stories were about academic standards collapsing. Students have now been considered revenue-producing customers rather than simply students for over a decade. Now it goes without saying that keeping students content and enrolled by handing out As passes for education.

The irony is that AI and lowering student loan caps have changed the equation for everyone.

FIRE and HxA want you to believe that the problem on college campuses is ideology. It isn’t. Or rather, the problem on campuses is that faculty have no power at all, the best ones quit, and the ones who stay act out ideologically. I’ve written about it here and here: how business metrics broke the university.

The faculty who used to be comfortable failing students who don’t do the work have largely walked away. And teaching quality now? Nobody knows, because the dirty secret is that universities won’t tell you anything about it. There is no good data. There is no way for anyone to know how good or bad teaching is, beyond anecdotes.

Of course, the professors still left are going to blame students, particularly in the huge online programs. But students aren’t complaining about not watching films in asynchronous courses where the discussion boards are generated by chatbots.

Do the math. Losing 100 students at a school of 1,000 is catastrophic. When you treat a student as a customer, the customer is, of course, always right.

What would you have done to keep enrollment up if you were a college president? You might well have turned every campus and every classroom into a place of safety and belonging. You might also have strangled a university that once attracted top students.

I’ve watched over the past 15 years as institutions that struggle for enrollment see every student not as a person but as a revenue stream. You can’t fail a revenue stream.

Students didn’t create this system; neither did faculty. Institutions created it when they built universities on business models dependent on continuous enrollment growth in a country with a declining birth rate. Students scrolling Instagram during The Conversation are responding rationally to a system that has communicated, through years of grade inflation and lowered expectations, that completion matters more than comprehension. Everyone knows this.

And next year, students will be bypassing college and Instagram and scrolling ChatGPT. It’s later than you think.

Adding books to scraped internet data does not solve the deeper problem, that training corpora systematically overweight text that has already been processed into stable form. Viral stories, once they “escape containment,” behave like books in this respect. They circulate as fixed summaries, stripped of the contradictory evidence that preceded them

Information escapes containment only by shedding complexity. It is stripped of the causal mechanics required for deductive reasoning. It has become a polished conclusion. This makes “escaped” data the least valuable asset for an AI, because a model cannot learn to derive the truth from a summary that has already deleted the premises.

Local news, by contrast, is the raw record of civic life, the first rough draft of history before “framing.” Local news is characterized by the incomplete and ongoing. It is the missing historical layer between internet and books. Local news captures events in process, before the outcome is known, before the category exists, before a consensus forms.

This matters because reasoning quality depends on access to premises rather than conclusions.

LLMs trained primarily on stable material degrade in a predictable and economically rational way. Even when some local records exist in the training mix, stabilized narratives, repeated conclusions, and portable explanations dominate through sheer volume and weight. LLMs learn to reason from outcomes rather than premises. Frequency substitutes for verification, narrative coherence substitutes for causality, and portability substitutes for accuracy. Process-level evidence that resists compression is underweighted, not absent, and thus rarely governs inference. The result, as any LLM user knows, is fluent, confident reasoning that is deductively unreliable, because the premises needed to test and falsify explanations are systematically overwhelmed.

The Minnesota example

Consider the local story about Minnesota daycare fraud and the national story. February 2013, Fox 9 investigator Jeff Baillon reported on the owner of a group of daycare centers who recruited parents to work while enrolling their own children, then billed the state for services never rendered. The owner was charged with bilking the state of $4 million, jumped bail before trial in 2016, and was never seen again. Subsequent newspaper investigations over the next several years documented repeated fraud schemes, partial verification, contested estimates of scale, and multiple prosecutions, but the reporting remained locally contained. A 2019 report detailed the fraud.

To understand how this local record was “lost,” fast forward to November 12, 2025, Armin Rosen published a major investigation in County Highway, “The Shame of Our Cities” documenting how the fraud metastasized beyond daycare to housing stabilization, autism services, and COVID meal relief. Because the local administrative record from 2013–2019 had never “escaped containment,” Rosen’s piece was read nationally as a revelation. It transformed a local administrative failure into a universal narrative: “Collective looting of the public coffers is now the state’s solution to the American puzzle of how we all should live together,” Rosen writes. “We are all Somalis now.”

On November 25, City Journal published a story “It’s Not ‘Racist’ to Notice Somali Fraud” that focused on Somali culture and high unemployment rate, suggesting that America limit immigration from “incompatible cultures.” Christopher Rufo’s story glided over local administrative context and named Somali culture as the causal variable.

On December 26, 2025, Nick Shirley, a 23-year-old YouTuber, posted a 43-minute video showing himself knocking on the doors of Somali-run daycares and, finding them locked, declaring this proof of fraud. The video received 135 million views on Twitter. Vice President Vance called it better journalism than anything that won a 2024 Pulitzer. Within three days, the Department of Homeland Security froze all federal childcare payments to Minnesota. Within ten days, Governor Tim Walz dropped his reelection bid. Congressional hearings were scheduled.

The new narrative is diagnostically worthless. It provides no analysis to repair the broken payment systems, because the details about how the fraud worked are still contained in the local archives.

Given the heft of the daycare story, however, LLMs will reason from culture narratives that escaped containment. Will this make their output more valuable or less? I say less. A local official or local business leader or a student studying government policy using an LLM cannot expect it to help a user design better administrative services, about, say, Medicaid reimbursements to nonprofits on the honor system, housing services with no verification process, treatment centers with no licensing requirements, federal funding of state programs with minimal oversight, which exist in every state.

Anyone can pluck a national story from the web. It is free, abundant, and offers no informational advantage. In an era of infinite synthetic text, the only information with scarcity value—and thus the only information that generates “alpha” for a reasoning engine—is the friction-heavy, un-digitized record of process that never escaped its container. If an LLM knows only what everyone else knows, it is not intelligence.

AI reasoning with local context

Imagine, however, the reasoning capacity of an LLM with access to local newspapers from fifty states, most largely inaccessible (gated, un-digitized, moldering in basements).

With access to contemporaneous local reporting that did not break containment from 2013–2016, in Minnesota, LLM reasoning would shift from induction over outcomes to deduction from sequences. Instead of starting with the conclusion that fraud occurred and inferring outward, it could reconstruct how programs actually operated over time: when rules were written, when interpretations changed, when oversight lagged, when warnings appeared, and when they were ignored or acted on.

Access to early local reporting would make causal claims testable. Competing explanations could be evaluated. Did billing spikes follow regulatory changes? Did staffing shortages or political interventions matter? Did similar programs under similar rules work differently? Local reporting named the clerks, auditors, mid-level bureaucrats, and licensing boards in charge. The 2019 legislative auditor’s report identified specific rifts between the Inspector General and investigators that paralyzed enforcement.

In general, local newspapers contain the densest record of everyday civic life. Local papers track institutions, families, markets, geography, climate, and conflicts over decades. The first reports of overdose deaths appeared in local papers years before “opioid epidemic” became part of the national vocabulary. Local papers document things that never happen: celebrations cancelled, motions failed, projects stalled, referenda that never made the ballot, dams repaired in time so they didn’t collapse. Local news features engagements, weddings (with name changes), and birth notices; obituaries list extended family members.

Making accessible currently gated archives would enable better accountability journalism, historical research, genealogical inquiry, and community self-understanding. As more people use AI systems for search, for research, for understanding their own communities and histories, the AI platform that can answer questions about local reality has a different kind of competitive advantage over an LLM that merely writes well.

What escapes containment is least valuable

Local knowledge does not scale easily, but local newspapers matter for AI reasoning by providing premises. Archives record decisions before outcomes are known, disputes before they are resolved, and failures before they are seen as inevitabilities. Without those premises, neither humans nor machines can reason deductively about how public systems fail or how they might be repaired.

Queries are answered first by retrieving stabilized narratives, dominant frames, or consensus summaries. These answers are computationally efficient and cognitively shallow. LLMs can respond quickly when the answer already exists in compressed form.

Local news in the substrate would enable reconstructing sequences, weighing contradictory reports, tracking named actors across time, and holding unresolved possibilities open. That kind of reasoning is slower and more resource-intensive because it cannot collapse immediately to a category or conclusion. It requires attention to order, contingency, and institutional detail. In other words, it requires thinking.

I said the Panama Project was unimportant in the long run because it involved training data that has already performed the reasoning work. Books, national journalism, and viral narratives externalize the cost of sense-making to editors, institutions, and crowds. Local reporting, by contrast, internalizes that cost in the archive and forces it to be paid again at query time.

Cheaper answers come from completed narratives; expensive answers come from raw premises that must be reasoned through. By training primarily on the former, AI companies are subsidizing the appearance of intelligence by bypassing the cost of verification. The absence of local news in the training data is why the Minnesota daycare story will continue to be narrated at the national level as a matter of culture alone.

If local archives are simply scraped into training corpora, however, they risk being compressed like everything else, smoothed into the mass of stabilized narrative they were meant to counterbalance. The value of local records is their structure: dated, sequenced, contradictory. That structure survives only if archives are treated as retrievable evidence rather than as training mass.

I would find most useful searchable archives that preserve the premises in a form that can govern inference. My scholarship for the last 30 years has involved rediscovering voices and lives and stories from the extraordinarily valuable archive of long lost newspapers. Digitization and accessibility are the first steps. The large AI companies should play a role

Local archives are an infrastructure for accountability, the evidentiary base without which no one, human or machine, can check the conclusions that circulate as fact. The degradation of the Minnesota story is about what happens when the local base erodes. AI makes the erosion visible. What concerns me is a generation trained to mistake narration for knowledge—never learning that what escapes containment, what everyone “knows” without needing to think, is least valuable.

Not a single student asked about research, about how many hours the reporter spent devouring background materials for every hour of interview time, about preparation generally. There were no questions about professional craft. I was sitting next to a Crimson reporter who had hoped to ask something more substantial. She said to me afterward that student focus on emotional labor was common these days.

I fear that too many young people (except for student reporters) do not know the work that goes into writing the first rough draft of history. Years ago, I was a reporter for my hometown newspaper, a researcher and writer for multiple TV stations, and a research intern for ABC News, London, covering the March 1985 arms control talks in Geneva. I learned more about the importance of researching thoroughly, digging for details, and meeting deadlines in these jobs than I did in college.

Today, besides reading my local paper, I read student newspapers from the Crimson to the JHU News-Letter to Utah’s Daily Chronicle, where reporters are daily demonstrating the skills critical for future success: practicing habits of inquiry, digging in the archives, cold calling sources, double checking everything, learning to write clearly, swiftly, and accurately.

And yet educators favor teaching the passive skill of ‘media literacy’ over the active craft of reporting. At last count, some 18 states have some sort of required “media literacy” program either in K-12 or college. The idea is that teaching students about “bias” and “positionality”will help them be better citizens in a pluralistic society. Millions of dollars are poured in media literacy programs annually. Of course teens want them: believing everyone is biased and sitting around pointing out that bias is easy.

Do you know what group is the most media savvy? Reporters who have gone out into the community to report stories and have had those stories edited and published. College newspapers are full of carefully reported local news stories, many rough first drafts of history.

Legislators investing in Civics education ought to see student journalism as civics as well. With the 250th anniversary of the signing of the Declaration of Independence, it is critical to consider the role of the press in making the case for a new nation. Newspapers were essential to the circulation of the Federalist Papers. Legislators ought to invest in student journalism more than “expressing viewpoints.”

Perhaps most importantly, students are increasingly getting their information from AI sources, which are wildly, infamously incomplete. I know this first hand. None of the stories I filed are there.

For a half a year after college I covered regional news for the Nashua Telegraph in southern New Hampshire, writing about zoning, taxes, computer purchases, salaries, energy and sewage controversies, and a visit to a local church by a controversial anti-Catholic pastor. My job was to figure out who was in the room, what they were arguing about, what power they held, and what might happen next. The work was outward-facing: listen carefully, reconstruct accurately, check claims, situate the dispute in law, policy, and history.

The stories I filed had nothing to do with “my” opinion or “my” viewpoint. They were about the community and their opinions, expressed in often contentious town meetings over AC power lines, nuclear power, and limiting growth that fill in key details about the state today. They’re a record of longstanding forms and techniques of self-government successfully managing conflict. But none of this record is online, none shows up in LLM search results. Students looking for “bias” won’t even find the stories to critique them.

The loss of local news is a national crisis, with a few initiatives responding productively. (I’ll give a shout to here to the excellent reporting being done by the Minnesota Star Tribune.) Most small town newspapers are not digitized and never will be; the hope is that at least physical or microfilm archives are being preserved. And so the rough first drafts of history, the names and dates and events and debates and decisions made across the country are becoming even more inconsequential with every AI query.

I’m documenting here some of my work from 1985, including some interesting reporting on power lines and on the late Ian Paisley’s visit to New Hampshire in September 1985, which I write about in much more detail in The Fitzwilliam. The Paisley story is virtually unknown because it failed, and it failed because Paisley’s supporters had no real understanding of what healthy religious pluralism looks like in practice. Paisley and his supporters were hoping to stir up dissent in one small town but a hoped-for explosion never happened. The state’s “live and let live” ethos was just as important as its “live free or die” motto, at least back then.

The story of Londonderry’s cold reception of a religious rabble-rouser matters now in a climate of increasing religious hostility and a need for the kind of pluralism that has largely worked in towns across New England for decades. It also matters now as those pushing a controversial category called “heritage Americans” argue that citizenship is linked to heritage. This was certainly not the view of the towns I covered in the mid 1980s. Pluralism was performed daily. Town leaders allowed the private platforming of an unpleasant speaker, maintained order, and billed the hosts for the costs of keeping the peace. All that is left of the story are my crumbling clippings. None of the LLMs (even the pro models) I queried knew anything about it.

It is dismaying to grasp how little AI “knows” about small town life. Because I used to read the local papers in the 1970s and 1980s, I know there was robust coverage of immigration and poverty and the grim aspects of rural New Hampshire life, including the early signs of the opioid crisis, which by 2016 had made New Hampshire's fentanyl overdose rate three times the national average. But almost none of this is online, except through pricy archives that generally only work when you know keywords of what you’re looking for. If you don’t know an event happened you don’t know to look for it. The first rough drafts are slowly fading from view.

The history of small-town decision-making may not matter in the global scheme of things, but local knowledge and local history matter. Pluralism as I came to understand it in the course of my reporting, was a set of habits that do not need a label when they are simply practiced. The people I interviewed had learned to live alongside neighbors whose beliefs they did not share and sometimes did not respect. They disagreed about religion, sewage, zoning, and planning. These habits were built over generations. In these small New Hampshire towns, people understood how to be neighbors, with fences.

Habits of coexistence can better survive the current moment of deliberate polarization if records of the history of coexistence are digitized and brought into the archive of what is currently known. If they are not, the only history future leaders and their AI tools will be able to reference is the history of current dysfunction, amply documented on the internet. Universities can play a role in digitizing and archiving, of course, as well as play a role in graduating more journalists than critics.

Transcripts below

• Aug. 7, 1985 Londonderry Presbyterian marks 250th year

• Aug. 10, 1985 Planners seek support to slow growth

• Aug 20, 1985 New sewage rules aimed at removing harmful materials

• Sept. 12, 1985 Controversial minister in town only ‘to preach the Gospel’

• Sept. 12, 1985 Paisley’s Londonderry Visit ‘an insult’

• Sept. 13, 1985 Paisley talks in need of a new site

• Sept. 20, 1985 Minister hopes Paisley Visit will be trouble-free

• Sept. 23, 1985 “They have shaken hands with the devil”

• Sept. 23, 1985 Officials walk the yellow line

• Sept. 24, 1985 Paisley fires own charges. Explains stand in interview

• Sept. 24, 1985 Dozens protest second Paisley service

• Sept 25, 1985 He brought trouble, God’s word

• Sept. 27 1985 Paisley’s gone, but Londonderry won’t soon forget

• Oct. 1, 1985 Londonderry bills church that invited Paisley

• Oct. 18, 1985 Church won’t pay police bill

• Oct. 18, 1985 ‘Cheap electricity is great, but…’

Londonderry Presbyterian marks 250th year

August 7, 1985

LONDONDERRY — Those who pass the stark white church on Mammoth Road may not realize that it is home to a congregation that was worshipping in this area before Webster’s dictionary was written, before George Washington was in short pants, before Handel had even thought of writing his famous “Messiah.”

But the Londonderry Presbyterian Church is now celebrating its 250th year and invites all area families, churchgoers and historians to participate join in the celebration from Friday, Sept. 20, to Sunday, Sept. 29. Ruth Williamson, coordinator of the celebration, emphasizes that the events and activities are for the family and the community.

Of course, the building now known as the Londonderry Presbyterian Church is not the first place of worship for the parish; it was built in 1837, more than a century after the church was founded in what was then called Nutfield.

“That’s what Derry and this area was called back in 1735,” elucidates Mrs. Williamson. “The congregation worshiped there until the old church was built on the corner of Pillsbury and Hardy Roads in Londonderry.” The old church has since been demolished.

Though most of the 250th anniversary events are being held in the fall, the church will begin its celebration by floating down Mammoth Road in replica form in the Londonderry Olde Home Days parade on Saturday, August 18; being involved in Saturday’s festivities on the Town Common; and participating in the ecumenical service on Sunday morning, also on the Common.

The replica was built by Fred Lee and was last seen floating in the Bicentennial Parade in Londonderry in 1976.

On Friday, Sept 20, the official semiquintenary celebration will begin with an evening communion service at 7 in the church, A community hymn sing will take place on the Town Common the following day at 3 pm followed by a chicken barbecue and the installation of a time capsule, among other activities.

On Sunday, Sept. 22, the usual church services will take place, and a combined choir will sing at a 1735-style service at 4 p.m. A “high tea” will follow the service.

At 7 pm on Friday, Sept 27, the Junior Choir will sing in a musical program, followed by entertainment in the parish hall with the Barbara Mullen Dancers and an Olde-Fashioned Ice Cream Social.

Saturday, Sept. 28 will see an afternoon hay ride, and a 1735-style dinner in the church hall. Entertainment will include the Granite State Cloggers and Scottish Dancers.

On Sunday morning, Sept. 29, church services will be held, and in the afternoon, a replica 1735 church service will take place, with guest speakers, costumes, banners, trumpeters and the burning of the parish hall mortgage.

Planners seek support to slow growth

August 10, 1985

LONDONDERRY — The Planning Board, in executive session Wednesday night, decided to go forward and hold a number of public hearings in the hope of gaining support to implement an interim measure to regulate the town’s growth.

This interim growth control measure would take the place of a growth control ordinance, which was ruled “illegal and therefore unenforceable” by the Rockingham Superior Court.

In the court’s decision on July 31, Judge Douglas Gray struck down the building permit limit on single-family houses, which had been stirring up controversy in the town. The decision was a result of a lawsuit brought against the town by John Tokanel of Tokie Enterprises, who was denied a building permit under the ordinance.

The Planning Board set the first informational meeting for 7:30 on Aug. 19. Town Administrator David Wright said the result of the meeting could be “a moratorium on permits, the decision to enact an interim measure, or nothing at all.” Questions will be answered by the Planning Board, the town administrator and other town officials.

Discussion of the ruling was on the Board of Selectmen’s Monday night agenda, and Wright had hinted the board might, in executive session, decide to appeal the ruling. The board took no action but will, however, be supporting the Planning Board in discussing an interim growth control measure.

The ordinance in question had been passed in a special Town Meeting in June 1978 and limits building permits to 200 in any calendar year. Although included in the ordinance were several circumstances under which permits could be issued without the general restriction or with specific separate restrictions, none applied to Tokie Enterprises.

In Gray’s ruling, he noted the town’s present master plan was adopted in February 1979, six months after the growth control ordinance took effect. He cited previous rulings where the courts have held that while growth control ordinances can be valid as part of a “temporary emergency measure to allow the town time to develop a master plan for phasing in growth,... growth control ordinances are intended to regulate and control the timing of development, not the prevention of development.”

He said that, in this case, the growth control ordinance bore no relationship to any recommendations to the master plan as it had been adopted six months previously, and “makes no attempt to deal with growth control with a view toward timing of growth or relaxation of such control.”

Because the ordinance deals with an “arbitrary” number of 200, which has neither been reviewed nor amended since its enactment, “it results in a more restrictive growth rate each year” as a percentage of allowable growth over the previous year, he said.

It was for these reasons the ordinance was found illegal and unenforceable, and according to Wright, no one was really surprised. The Southern New Hampshire Homebuilders’ Association, which supported Tokanel in his suit, had warned last year it would take action after the 200 single-family house limit had been reached for the first time. The limit this year was reached in May.

The Tokie Enterprises suit had originally been brought as a class action petition, but request for class certification was waived by agreement of counsel, and the matter was handled as a singular action with a sole plaintiff. The decision came within hours of the hearing.

New sewage rules aimed at removing harmful materials

August 20, 1985

LONDONDERRY — To meet with Environmental Protection Agency regulations, the town Sewer Commission will be implementing a program to regulate the discharge of non-domestic sewage into the town’s sewer system. Londonderry’s sewer system will be connected to the Manchester Public Owned Treatment Plant (POTW), and in accordance with federal requirements, the town must take action to ensure against the discharge of harmful effluents.

The purpose of the program is to protect the Manchester treatment plant from hazardous waste, from waste which will corrode or otherwise do damage to the sewer system and from waste which may cause an environmental or health threat.

Hoyle Tanner & Associates has been given the contract to engineer and implement the program. Representatives from the firm, along with representatives from the state of New Hampshire and the Sewer Commission, held a hearing last week to discuss the program. They went over the limitations and answered questions about what the requirements will be.

Basically, implementation of the program means that any user discharging non-domestic waste must apply for a discharge permit, and disclose the specific type of effluent entering the municipal sewer system. If the material does not meet the guidelines set forward by the program, the users will be required to either build a pre-treatment facility, or cease discharging the harmful waste.

According to Sewer Commissioner Bill Merrill, “Our requirements must be as stringent as the EPA’s requirements, but they can be even more stringent.” He added that “We want to work with the users, not against them, and we’re going to try to be as helpful as possible.”

There will be a number of deadlines for complying with the program once it is implemented. Sixty days after the ordinances are accepted, a user must apply for the discharge permit; 120 days later, the user must be in compliance with the limitations of the program. Commissioner John Michaels said that a date for compliance might be 180 days after Oct. 1, a rough estimate of the date the program will go into effect.

Hoyle Tanner representative Michael Trainque outlined the six components of the program, which include: doing a survey to identify and evaluate non-domestic use; establishing the legal authority to enforce the limits; setting legal limits; compliance monitoring; establishing administrative procedure; and getting the resources to keep the program in place.

Trainque handed out a copy of the national pre-treatment standards, which listed prohibited discharges. They include: pollutants which create a fire or explosion hazard; solid or viscous pollutants which cause obstruction; heat in amounts which inhibit biological activity; and any other pollutants which cause interference with the POTW. The program also limits pollutants which cause corrosive or structural damage, or which have a pH lower than 5.

Merrill has noted that most of the relatively few users who may be affected by the program are within the limits of the program. The Sewer Commission will have the power to shut off sewer service to those not complying with the program.

Controversial minister in town only ‘to preach the Gospel’

September 1985

LONDONDERRY — Ian Paisley — member of Parliament, leader of the Democratic Unionist Party in Northern Ireland and Presbyterian minister — will be speaking in Londonderry this month. The controversial fundamentalist is on a restricted visa and will not be allowed to make any political statements or answer questions about politics in Northern Ireland. He is here, according to the Rev. David Brame of Manchester, “to preach the Gospel.”

Brame is the minister of Londonderry’s new Free Presbyterian Church of Northern Ireland. The Free Presbyterian Church was founded in Northern Ireland in 1951, with Paisley playing a large role. According to Brame, the church was founded because Presbyterians in Northern Ireland found their denominations were “no longer taking a firm stand on the inspiration and authority of the Scriptures.”

Paisley will be speaking at the Highlander Inn in Manchester Sept. 22-25 at 7:30 p.m. He will be drawing on his 40 years of experience as an evangelist to raise enthusiasm for the Free Presbyterian Church. There are currently 60 such churches in Northern Ireland, three in Canada and five in the United States. “The Marty’s Memorial Church in Belfast,” says Brame, speaking of Paisley’s own church, “has an average attendance of about 1,500-2,000 people.”

Londonderry resident Henry Paul felt there was a need for a fundamentalist church in this area and contacted the Greenville Free Presbyterian Church in North Carolina. The church sent Brame up to begin holding services and gathering support for the church.

“In three weeks of services we’ve had an average attendance of about 40,” Brame says. “That would indicate there is a need for this church in the area.”

Brame admits that Paisley is a controversial figure and that it is difficult to separate the political man from the religious man. “Dr. Paisley’s political involvement is with the Democratic Unionist Party of Northern Ireland — that is a completely distinct thing from the Free Presbyterian Church,” Brame says. “They are not mixed, but very often that issue is clouded in this country.”

Brame agrees that the Free Presbyterian Church’s stand is one that draws a lot of fire. “Yes, I know Dr. Paisley is often referred to as someone who preaches hate,” the Londonderry minister says. “But Christ’s ministry had both the positive and the negative, because if you’re going to stand for truth, everything that is opposed to that truth, and is indeed falsehood, must be brought to light.

“People accuse us of being unloving because we do not, like some people say, ‘let people go their own way,’ and ‘let people believe what they want to,’” Brame adds. “‘Religion is a private matter, don’t bother them,’ they say. But if I firmly believe they are in error, and that that error is going to lead them to hell, then the only way I can really express love for them is to point that out to them.

“If you go by someone’s house, and see them sitting comfortably in their living room, and yet you notice the top corner of the house is beginning to burn, you don’t say, ‘well, that’s their private concern.’ You go and do whatever you have to — you bust the door down, you break the windows in order to get in there and warn them. Well that’s the way I view it — they are in danger of hellfire, and therefore the most loving thing I can do is to try and alert them to the danger.”

The dangers Brame refers to are the doctrines that preach “salvation by works” and doctrines that don’t believe in the “person of Jesus Christ.”

“The byword of most of the cults, including Roman Catholicism, is that you can work your way to heaven by doing a certain number of works and thus ‘earning merit’ with God, when the Scriptures plainly show otherwise,” says Brame. “We preach salvation by faith alone.”

Paisley’s Londonderry visit ‘an insult’

September 12 1985

LONDONDERRY — “Paisley speaking here is an insult to the people of Londonderry,” says Roland Westerveldt, pastor of the Londonderry Presbyterian Church. “I’m sorry to see this happening.”

With as controversial a figure as Ian Paisley trying to raise support for an ultra-orthodox church like the Free Presbyterian Church of Northern Ireland during the same week the Londonderry Presbyterian Church is celebrating its 250th anniversary, a certain amount of resentment can be expected.

“What they say they’re trying to do in this community is already being done,” Westerveldt continued. “We have four fundamentalist churches in this community, and the Free Presbyterian Church coming here is redundant. We have two Baptist churches, a Fundamentalist Christian Church and the independent Trinity Bible Church, which stress the fundamentalist approach. And in this community, we try to build bridges between religions, not throw criticism.”

Westerveldt also says the Londonderry Presbyterian Church is “Presbyterian in the best sense of the word. The Christian Gospel is a positive message; it is not calling names and pointing out errors.”

Westerveldt says he has known Henry Paul, who prompted the Free Presbyterian Church coming to Londonderry, for 14 years. “He has said that there is a need for this church, but I don’t think there will be any loss of members from our church.

“Paisley’s speaking here during our 250th celebration is probably a coincidence,” admits Westerveldt, “but if it wasn’t, I would resent it.” Westerveldt said he didn’t think Paisley’s presence in town would affect the mood of the celebration, and he hasn’t decided whether he’ll speak out from the pulpit against the Northern Ireland minister.

“I may read something about our work in Northern Ireland. We support a retreat center, Cory Meela [sic], in Northern Ireland, where Roman Catholics and Protestants work and talk together about solving problems. We try to work together, not preach hate.”

Paisley talks in need of new site

September 13, 1983

MANCHESTER — The Highlander Inn and Resort suddenly and without explanation has canceled reservations for a series of religious conferences later this month by the Rev. Ian Paisley, leader of the Free Presbyterian Church of Northern Ireland.

According to the Rev. David Brame, who is handling arrangements for Paisley’s visit, the hotel sent a brief message via Western Union that it would not accommodate the Paisley conferences.

Brame said he did not know where the conferences will now be held.

He said he was shocked by the message and the medium. “I can’t believe it — they already accepted our check,” he said. “And it’s only 10 days’ notice.”

The controversial Northern Irish minister, who is also a member of British Parliament and leader of the Democratic Unionist Party, was scheduled to hold a series of four conferences at the Highlander Inn from Sept. 22 to 25. Paisley is on a restricted visa and will not be allowed to answer any questions of a political nature. According to Brame, the purpose of the visit is to raise awareness of the Free Presbyterian Church, which is being established in Londonderry; gain support for the denomination; and preach the Gospel.

The manager of the Highlander Inn could not be reached for comment.

On Wednesday, the Catholic Diocese of Manchester issued a statement concerning Paisley’s visit to New Hampshire. The Rev. Frances J. Christian, the chancellor of the diocese, said, “It is our fervent prayer that Ian Paisley’s presence in New Hampshire, and the foundation of a community which subscribes to his misguided religious principles, will have no lasting negative impact upon the citizens of our state.”

Minister hopes Paisley visit will be trouble-free

September 20, 1985

LONDONDERRY — “We’ve been fortunate so far,” said the Rev. David Brame during Sunday’s church services. “Let’s pray things continue to go smoothly.”

Members of Londonderry’s newest church, the Free Presbyterian Church of Northern Ireland, hope that the Rev. Ian Paisley’s four-day visit to this area does not spark any trouble. Paisley is arriving tomorrow to conduct a series of four religious conferences in Londonderry Sunday through Wednesday.

Alan Cairns of the Free Presbyterian Church in Greenville, S.C., arriving in Londonderry tonight, will coordinate with Brame to help ensure the smoothness of Paisley’s visit. Because of Paisley’s anti-Roman Catholic views, church members have been wary of Londonderry’s proximity to largely Catholic South Boston and any protests the visit might cause.

FBI spokesman John J. Cloherty said nothing had come to the bureau’s attention about any problems stemming from Paisley’s visit. Paisley’s recent trips to South Carolina have been trouble-free, but he has not always been considered a welcome visitor to the United States. In 1981 he was denied a visa to make a fund-raising tour of this country, though he was allowed to enter the United States in 1982 to attend the funeral of a close friend. The State Department refused his first request for a visa in March 1983, saying “it was considered contrary to the public interest,” but granted a temporary visa later that summer so that Paisley could attend the World Congress of Fundamentalists in South Carolina.

According to a State Department spokesman, the minister’s visa has been an issue because of concerns with “Dr. Paisley’s record of violent rhetoric and inflammatory actions.” In 1983, the State Department said, “We considered his rhetoric and actions were contrary to the public interest in achieving a peaceful settlement to the problems of Northern Ireland.”

Paisley was allowed to visit Bob Jones University in Greenville, S.C., from July 29-Aug. 9 in 1983, but he was not allowed to go elsewhere in the country, speak on political matters or make fund-raising appeals on behalf of his political backers. At that time, a government spokesman said, “We’re satisfied the purposes for which we granted the visa are the purposes for which it will be used.”

Bob Jones, chancellor of the fundamentalist institution, called the earlier denial “nothing but Catholic bigotry.”

Since the 1983 denial, Paisley has not been denied a visa but has been bound to the restriction that he not make any political statements or answer questions of a political nature. He spent a week in Greenville, S.C., last October and visited again in April and May. Being careful to avoid controversy, Brame, the minister of the newest of Paisley’s churches in Londonderry, said there has never been a problem with Paisley visiting Greenville.

“They have shaken hands with the devil”

Monday, September 23, 1985

Londonderry – Nearly 100 protesters chanted, sang, decried and swore outside the “old-time Gospel tent” on Litchfield Road where fundamentalist Ian Paisley preached the first of his four scheduled religious services.

The protesters, some of whom traveled from Boston, New York and Canada to voice opposition to Paisley’s religious and political beliefs, were kept apart from those in the tent by half of the Londonderry police force, who maintained a strong presence throughout the evening.

Carrying placards and waving flags, the crowd, many calling themselves supporters of the outlawed Irish Republican Army, harassed those who came to hear Paisley speak, broadcasting their hostility through bullhorns before and during the one-hour service.

The largely Catholic and Marxist IRA is seeking to unite predominantly Protestant Northern Ireland with the Irish Republic to the south under a socialist government.

Most of the protests condemned Paisley’s political and religious activities in Northern Ireland, though many protesters were also damning the U.S. government’s decision to grant a visa for the controversial minister and castigating the town of Londonderry for allowing Paisley to preach here.

“There is a terrorist among you!” cried one protester. “Next you’ll be inviting Hitler and then Idi Amin to speak here!”

“Ian Paisley is the greatest bigot in the world,” said Nashua resident Mike Conway.

“He spews hatred and is responsible for the continuing bloodshed in Northern Ireland,” said Conway, a native of Limerick, Ireland.

Paula Cunningham, who was raised in Belfast, shook with rage as she spoke of her loathing of Paisley. “Every Irishman and Irish-American disagrees with Ian Paisley and disapproves of America granting a visa to him. They have shaken hands with the devil.”

On the opposite side of the street from the protesters, a Canadian pastor from the Bible Baptist Church in Nashua expressed disgust for the many pro-IRA sentiments. “I preach the same gospel. I’ve been beaten, burned and sent to prison for preaching the same gospel as Ian Paisley,” the pastor said. “These people don’t know the truth.”

Officials walk the yellow line

September 23, 1985

Londonderry – Town officials walked the yellow line on Litchfield Road between protesters and worshippers last night during services conducted on the property of Henry Paul by the Rev. Ian Paisley.

“And that about tells our color,” quipped Selectman Norman Russell, who, like other town officials, had “no comment” about most of Paisley’s or the protesters’ views.

“The U.S. government gave him the visa,” said Town Administrator David Wright, discounting the protesters’ claim that Londonderry was responsible for Paisley’s presence. “The town had nothing to do with his visit here.” He added that the town had nothing to do with the protesters’ visit either.

“This is what democracy is all about,” said Russell, saying the protesters have a right to protest and Paisley has a right to preach.

“True,” agree Wright. “But it’s going to result in a lot of excessive expense.” He cited the cost of the extra police on duty and the other safety precautions.

Police Lt. Kenneth Lynch expressed concern that the protests would continue for the next three days of the scheduled services.

Paisley fires own charges: Explains stand in interview

Tuesday, September 24, 1985

Londonderry – The Rev. Ian Paisley laughs at protesters who accuse him of fascism.

“Fascism is the child of Roman Catholicism,” he said. “Hitler and Mussolini were both devout sons of the Roman Catholic Church.”

The fiery Northern Ireland pastor and political leader also criticized those who are protesting his series of sermons here, and rapped the World Council of Churches as “shot through with Marxism.”

Paisley spoke in an interview at the home of attorney Henry Paul after the second of four services being held this week on Paul’s property.

In the relaxed atmosphere of Paul’s home, Paisley spoke freely about his religious and political positions. He cannot do that publicly under the restrictions of his visa, which prohibit him from making public statements of a political nature.

Also present at the interview were the Rev. David Brame and the Rev. Alan Cairns of the Free Presbyterian Church of Northern Ireland, Mrs. Brame and family friends of Henry Paul.

“I am a Christian first and a politician second,” Paisley said. “My primary objective is to preach the Gospel. That is my mandate. My aim is to offer Jesus Christ to all men and ask them to turn to him.”

Paisley said the Free Presbyterian Church of Northern Ireland was founded because the Irish Presbyterian Church and other once-fundamental churches were becoming too “liberal.” He said that his objective as a minister of the Free Presbyterian Church, as a Member of the British Parliament for the European Economic Commission is to stop the spread of liberalism and the weakening of Reformation principles, and to speak out against the Roman Catholic Church.

“I voice my opposition, like Daniel,” said Paisley of his purpose in sitting in the EEC Parliament, where he says the majority of members hold views contrary to his own, and many ignore the Reformation as part of Europe’s history.

“I know lots of people don’t like me,” he said of his leadership role in Ulster as a member of Parliament. But I received the largest vote – a quarter of a million people voted for me – the largest vote ever recorded for a British politician living or dead.”

Paisley said he has an obligation to represent the wishes of his constituency in the British Parliament, in the EEC and in the Northern Ireland Assembly. But he brushed aside the issue of some protesters’ accusations that he does not represent the Catholics in his constituency.

They know the mandate when they vote, said Cairns, who said he can speak more freely about Northern Ireland politics than Paisley can. Cairns said Paisley’s position on political and religious matters is clearly stated when he runs for office, and many Catholics have voted for him.

Paisley said he draws a lot of criticism for his views, and takes many unpopular stands. “I’m anti-liquor and against the liquor shops,” he said as an example. “I believe the consumption of intoxicating liquor destroys men’s homes, their families and their minds. Drink is a curse.”

Paisley spoke of being imprisoned by Northern Ireland Prime Minister Terrence O’Neill for protesting the general assembly of the Irish Presbyterian Church. He said he had been protesting a series of pro-ecumenical, pro-Catholic resolutions being put before the assembly.

It was as a result of being put in prison that he got into politics and won the seat in Parliament that once had belonged to O’Neill, he said.

Paisley also spoke of the Free Presbyterian church’s disassociation with the World Council of Churches. He said the church had pulled out of that organization and the International Commission because the groups are part of a trend toward liberalism and away from the teachings of the Reformation.

“The World Council of Churches is a platform for liberalism,” he said. He spoke of a Canadian totem pole the council had placed outside the Unity Church in Geneva, and said that the Free Presbyterians will have no relationship with heathen representations.

Paisley contended that the expressed aim of the World Council of Churches is to reverse the Reformation.

Paisley said that the liberalized Protestant churches are falling prey to the Roman Catholic Church. He said the Church of Rome is losing out very rapidly in many areas of the world, but is more powerful in Presbyterian areas where the teachings of the Reformation are weakening.

“The Roman Catholic Church here (in the United States) is powerful because the ecumenical pastor is willing to do everything the priests tell him to do,” Paisley said.

He asserted that many representatives of the World Council of Churches are Catholics from behind the Iron Curtain, and said the council is “shot through with Marxism.”

Paisley laughed at the accusation by the two men arrested at Sunday’s meeting that he was a fascist, and at that point linked fascism and the Catholic Church.

After mention of the two protesters, Paisley asked Paul whether they could go to the Police Department and ask that charges not be pressed. “They did it in ignorance,” said Paisley.

“Civil and religious liberty is basic to the Presbyterian church,” he said. “We have never protested a group of Roman Catholics or a group of atheists, but we will protest a group of Roman Catholics in a Protestant church.

We will protest against evil where the honor of Christ is at stake.”

Both Paisley and Cairns charged that the group that protested Paisley’s religious services Sunday and Monday nights was violating the civil liberties of those who came to hear “the word of the Lord.”

“We would support their right to protest,” said Cairns, “But there’s got to be something sick about people who would try to stop people from hearing the gospel.”

“They have no right to interrupt a religious service,” said Paisley.

Paisley suggested moving the tent farther away from the road on Paul’s property, outside of earshot of the protesters, but Brame said it would cost about $600 to make the move.

Paisley said he has the money and the tent will be moved. He said that, at the point in the service where he asked the people to accept Jesus Christ as their savior, it is crucial that the noise of the protesters not detract. He said the $600 will be well spent to move the tent.

Paisley, Paul and Cairns discussed the fact that no Roman Catholics had come out and protested the protester, most of whom they said appear to be supporters of the Irish Republican Army.

“They should condemn them,” said Paisley.

“By their silence, they have supported the IRA,” said Cairns.

Paul said none of his neighbors had protested Paisley’s presence.

“They should not be here disturbing the peace,” said Paisley of the protesters, adding that, if Londonderry residents are not unhappy, there shouldn’t be protests. “You have brought a murderer to Londonderry,” cried Paisley to Paul, mimicking the cries of the protesters and teasing his hosts.

“That pastor who denounced me,” said Paisley. “If he was an honest man, he would have denounced the protesters for disrupting a church service.” Paisley was referring to the Rev. Rowland Westervelt of the Londonderry Presbyterian Church, who called Paisley’s visit an insult to the people of Londonderry.
Cairns said Paisley ignores and treats with contempt any member of the IRA. He said he would never share a platform with an IRA supporter, or do a radio broadcast with one. He said that, although the protesters call Paisley a murderer, the IRA members are really the murderers.

Cairns also said that more than two-thirds of the murders in Northern Ireland are Protestants killed by the IRA.

“Our preaching the gospel doesn’t disturb their peace,” said Paisley, referring to the protesters. “I am only here to offer Jesus Christ to all men, and nothing will stop me from preaching the gospel.”

Dozen protest second Paisley service

Tuesday, September 24, 1985

Londonderry – About a dozen protesters showed up last night to voice opposition to the second of Ian Paisley’s four scheduled religious services.

The protesters, armed with an accordion, a drum and a tambourine and many loud voices, kept up the chants of “Paisley out, peace in!” during most of the service, but seemed disillusioned as their ranks diminished and the crowd inside the tent grew.

Francis Curran of Methuen said it was important they show opposition to Paisley. Another Methuen man said, “We think he’s raising money for the Ulster volunteer force.”

One said he felt like “The Salvation Army Band” because of their small numbers, made to appear smaller by the ranks of the Londonderry Police Department and a battery of reporters present.

“We had to get 50 more chairs for tonight,” said the Rev. David Brame, inside the tent, as he welcomed the crowd of nearly 200.

With trumpeter Mike Schock, the English Handbell choir from West Swanzey, and a duet of tenors from the Dublin Christian academy, the music inside the tent overpowered the singing outside.

Following a number of hymns and a talk by Dr. Bill Woods, a missionary from Brazil, who spoke about meeting Ian Paisley 25 years ago, Paisley rose to speak.

He said another Protestant had been killed last night, shot dead by a member of the Irish Republican Army. He also said 18 churches had been bombed, burned and destroyed this year by the IRA, and added, “That same spirit is outside this tent.”

Paisley spoke for nearly an hour, calling on Jesus Christ to “Dispel the power of the enemy.” He spoke of the Lamb of God, and said “All have sinned and came short of the glory of God.” Paisley also condemned those who take part in religious rituals, and said God calls the rituals “filthy rags.”

“God is not a rag merchant,” he boomed.

“You don’t get to heaven by paying silver and gold to any priest,” he continued, and challenged that today’s religion is “Subjective instead of objective.”

“If you look inward, you’ll be miserable. If you look around, you’ll be more miserable,” he said, with a glance toward the protesters, which caused the crowd inside to chuckle. “You have to look to Jesus,” he intoned.

The service ended with Paisley calling up to the front of the tent those who had accepted Jesus Christ, in what he later called “A breath of old fashioned revival meetings.”

He brought trouble, God’s word

Weds, September 25, 1985

Londonderry – “He may have brought trouble to Londonderry, but he also brought the word of God,” beamed the Rev. David Brame last night, introducing the Rev. Ian Paisley to preach his last scheduled service here

The Londonderry Baptist Church on Mammoth Road was the site of the fourth and final religious service of the Free Presbyterian Church of Northern Ireland to be led by the controversial minister.

The Rev. George Souza, pastor of the Baptist church, offered the use of the facilities to Paisley Tuesday night, after rain had made the Litchfield Road site too muddy. The new location, in addition to being cleaner and warmer, had the added benefit of being farther away from the 50 or so protesters lined up on Mammoth Road whose voices could barely be heard inside.

Souza said there are “some doctrinal differences” between the Free Presbyterian church and his own denomination, but said he had “no problem with them.”

“We are regiments in the same army,” said Paisley of the Baptist church, as he stood to speak. “We are in the same fight – the battle for Lord Jesus Christ.”

After thanking the Baptist minister for the use of the church, Paisley’s first words were directed at the press and the protesters. He said he knew he was not treated kindly by either group, but quoted archly, “Woe unto he whom all men speak well of.”

Paisley said he had to deal with the press often in Great Britain and laughed about a suggestion he once got that he give a service for journalists. He quipped that the text he’d read from would include “Zaccheus couldn’t see Jesus for the press.”

Last night’s service was similar to the previous three, including the singing of a number of hymns and a reading from the New Testament. In his sermon following the reading from the book of Romans, Paisley stressed that “It doesn’t matter how good you are, you are not good enough.”

He spoke of those who say “If I could only know I believed right,” saying that “Faith and faith alone will save you.”

“It’s not the blood in the veins, but the blood shed that saves,” he preached, speaking of the death of Jesus Christ.

At that point, a woman entered the church through a door in the back and started to speak out against Paisley. “He doesn’t know,” she said, as she was quickly escorted out by the Londonderry police.

“That’s nothing, said Paisley. “In Northern Ireland, we have them coming through the door with guns in their hands.” He spoke of a man in his church being shot dead by an IRA man, and again blasted the IRA for destroying 18 churches in Northern Ireland in the past year. He blasted the INLA, who destroyed a church before he left, and wrote “Ha ha ha, we destroyed your church,” on the burnt walls.

He implied the protester was sent by the devil, saying “No wonder the devil doesn’t want me to get to this.”

At the heart of his sermon he intoned, “Faith is all you need. If I can set one person on the path to Christ, my visit to this locality will not be in vain.”

Londonderry resident David Lyons, who is a member of the Free Presbyterian Church, said he thought Paisley’s visit accomplished something. “If even one person accepted Jesus Christ, then the coming here was worthwhile.”

Nashua native Louis Murray, who was arrested Sunday night after verbally accosting Paisley, took an opposite view, saying, “He planted the seeds of a congregation that does not understand the inherent racism in his bible.”

He asserted that Paisley supports, among other racist causes, apartheid.

Lyons said that most of the churchgoers don’t think about the IRA at all, but that he was “totally opposed to their terrorist tactics,” adding they were in “the same category as the Ku Klux Klan.”

Paisley’s gone, but Londonderry won’t soon forget

September 26, 1985

LONDONDERRY — The Rev. Ian Paisley flew home to Northern Ireland late last night, but his four-day presence will have affected the community in many ways.

“All told, it’s going to cost us $2,375.10,” said Town Administrator David Wright, who said the figure did not include the cost of his aggravation. “I don’t know how you can put a monetary figure on that.”

Londonderry Police showed a strong presence Monday through Wednesday nights as protesters traveled to the small town in numbers reaching 100 in protest of Paisley’s visit.

The Northern Ireland minister gave a series of religious services to promote the founding of the Free Presbyterian Church of Northern Ireland in Londonderry.

Wright said the town had to pay for extra police duty for about 12 officers, including Chief Charles Webster and three Manchester police. He said the figure did not include payment for police who were already on duty, or the three for whom Londonderry resident Henry Paul was paying.

Paul’s property on Litchfield Road was the site of the first three services; the fourth was held at the Londonderry Baptist Church on Mammoth Road. It is estimated his cost for the services of three police officers for four nights would be about $732.

Webster said the station received six calls complaining about the volume of the protesters.

Wright said the one benefit of all the activity was that police have gained some good experience. “We don’t have this sort of thing happening very often in Londonderry, you know,” he said.

Members of the Free Presbyterian Church, said they feel Paisley’s visit was worthwhile, despite the protests. As a result of the experience, Londonderry resident and church member David Lyons said he would be one of the first to support “an ordinance to prohibit public protests of religious meetings.”

Selectmen said there was no such ordinance already on the books, and they had no comment about passing one in the future.

Londonderry bills church that invited Paisley

October 1, 1985

LONDONDERRY — The Board of Selectmen has decided to present the $2,376 bill for extra police services during the Rev. Ian Paisley’s Church of Northern Ireland, which sponsored the controversial minister’s visit.

The bill, added up by Town Administrator David Wright last week, came as a surprise to Free Presbyterian Church minister David Brame.

“I didn’t know about it,” he said, though he affirmed that Londonderry resident Henry Paul had intended to pay for the services of three of the 12 police officers needed to maintain order during protests of Paisley’s presence. Paul’s cost for the three officers is estimated to be over $700.

Paul, his son Daniel Paul, and Brame sit on the committee which manages the church’s funds, and Brame said the committee will need to meet to discuss the bill.

The newly founded Free Presbyterian Church in Londonderry had invited the Northern Ireland minister to give a series of four religious services Sept. 22-25 to preach the gospel and gather support for the fundamentalist denomination.

Paisley’s visit caused a stir in the town as protesters from as far away as New York and voice opposition to his presence.

Two New Hampshire residents were arrested after the first service on charges of disorderly conduct for verbally accosting Paisley. Though there were no arrests and the number of protesters was markedly smaller the following nights, town officials felt it prudent to maintain a strong, 12-person police presence for each of the scheduled services.

At last night’s selectmen’s meeting, Wright made a motion to give letters of commendation to the members of the Police Department who demonstrated exceptional professionalism during the four protests. He also suggested giving commendations to those Fire and Police department members and volunteers who worked during Hurricane Gloria.

The motion passed unanimously.

Church won’t pay police bill

October 18, 1985

LONDONDERRY — Today town officials will be discussing a letter they received yesterday afternoon from the Free Presbyterian Church of Northern Ireland that says the church is refusing to pay the $2,375 bill for police services incurred during the Rev. Ian Paisley’s loudly protested visit here last month.

“It was anticipated,” said Selectman Norman Russell.

At Tuesday’s board meeting, Selectmen Chairman Harry Anagnos had said, “We’ll cross that bridge when we come to it,” referring to the expected letter from the church.

“Let’s just say it was a polite refusal,” explained Daniel Paul, who was responsible for the letter as a member of the committee managing the finances of the Free Presbyterian Church, which was billed by the town.

Paul did not want to elaborate on the letter before Town Administrator David Wright, to whom the letter was delivered, had a chance to discuss the letter with selectmen.

“We responded to their invoice,” Paul said. “We want to be cooperative. If the word litigation is heard, it will be from the press.”

The Rev. David Brame, minister of the church, who also had a hand in writing the letter, had charged the protestors should be billed for the police services because they caused the need for police presence.

The Free Presbyterian Church had intended to pay three police officers $17.50 per hour for the hours they were needed during Paisley’s four-evening service schedule. More than 100 people turned out to protest the services, and police felt more protection was needed.

Paul did not want to comment on who should pay the bill. He blasted the press coverage of Paisley’s visit and the aftermath, saying, “It is as if they (the press) wanted to be the catalyst” of more trouble by spending too much time on the bill controversy.

Paul said the church members have a “great respect” for the town fathers and want to have a good relationship with them. He stressed he wanted to give the town the opportunity of hearing of their refusal in a letter rather than reading about it in the newspapers.

According to Paul, local members of the press have been trying to snowball the issue of the $2,375 bill. He also said the church is striking a “pre-litigation stance.”

‘Cheap electricity is great, but . . .’

October 18, 1985

LONDONDERRY — More than 100 area residents turned out last night at the junior high school cafeteria to hear members of the New England Power Pool discuss the proposed high-voltage transmission line to cut through New Hampshire.

Billed as an informational hearing, the discussion was a forum for the representatives of the New England Electric Transmission Company, the New England Hydro-transmission Corporation and other power companies, to give presentations about the project, its impact on the environment, and its alleged health risks. The representatives also agreed to answer questions pre-submitted by residents.

The transmission line project is a joint effort by the Hydro-Quebec utility and the New England Power Pool to bring Quebec’s “vast hydroelectric resources” to a number of New England states. Phase I of the project is currently under construction and will bring 690 megawatts of hydroelectricity to the Comerford Converter Terminal in Monroe, N.H. Phase II will ultimately bring 2,000 megawatts of power farther south to Massachusetts.

Phase I will include 107 miles of transmission line (58 miles in the U.S.) and Phase II will include 121 miles of transmission line to be built in New Hampshire. According to New England Power Pool representatives, “With the completion of Phase II (in 1990), New England would receive over 10 million megawatt hours of energy per year

RESIDENTS Page 12 (From Page 1)

from Quebec — approximately 10 percent of New England’s energy needs during the first full year of operation.”

While no one has disagreed that a cheap, non-nuclear energy source should be considered seriously, the fact that the energy will flow through transmission lines has a number of people concerned. Certain groups have expressed fear that exposure to ions given off by the DC wires will have an adverse effect.

Members of the Pelham-based Powerline Awareness Campaign (PAC) and the Virginia-based Taskforce Opposed to the Planned Power Line (TOPPL) were conspicuously present at the hearing, handing out literature and answering questions about the possible health hazards of exposure to DC current.

Both groups claim that ions given off by the power lines will cause health problems such as headaches, fatigue, tingling, stress, dizziness, nosebleeds, chromosome damage, nausea, a higher cancer rate and a higher suicide rate.

The New England Power Pool representatives, however, claim the health risks associated with electrical discharge are minimal or non-existent.

Jonathon M. Charry, a senior research scientist in the Laboratory of Neuropharmacology and Environmental Toxicology at the Institute for Basic Research in Staten Island, stated at the meeting that there would be “no harmful health effects from the proposed DC transmission line.”

Charry stressed that a “wide range of biological responses to air ions has been researched,” and “we have enough material to make an informed judgement about the potential effects of ions.” He added that research had been going on for 50 years.

Gary Johnson, a manager of electric research programs at General Electric in Lenox, Mass., said that the number of ions one is exposed to over a candlelit dinner would be greater than the number exposed to someone standing directly underneath the power lines. He also said audible noise given off by the wires would be negligible.

Both men made the point that any purported effect of ion exposure is small because the ion levels occur within “normal biological variability.”

Area residents present at the meeting said they were there to “get more information,” but seemed skeptical that the health concerns could be brushed aside too easily.

“I wish they would be more emphatic in refuting these reports,” Londonderry Selectman Norman Russell remarked. “The aspect of cheaper electricity is great, but the aspect of possible adverse effects on peoples’ health is of some concern.”

Lee Breslow of Londonderry said that he had been “concerned about AC wires for years,” and that the proposed DC wires add more fears.

The Federalist Papers were published in newspapers and addressed to the citizens of New York. Hamilton, Madison, and Jay assumed their readers would deliberate about ratification in public—in taverns, town halls, state legislatures, and ratifying conventions. They presumed that constitutional debates would happen among the people who would live under the government being proposed. Universities appear nowhere in the Federalist Papers, and when Washington and Madison later advocated for a national university, the purpose was regional unity among elites, not civic instruction for all. The idea that citizenship is primarily formed in a classroom rather than in the practice of self-government would have puzzled the founders.

And yet the worry that college students aren’t being taught about American citizenship, constitutionalism, and liberal self-government provoked state governments across the South and Midwest (and spreading outward in the last three years) to appropriate well over $160 million to establish freestanding civics programs inside public universities, with protected budgets, faculty lines, and a mandate to teach foundational texts.1

The argument went something like this: universities were failing in their civic duty. Humanities and social science departments had become “leftist hotbeds” of critical theory and identity politics, cancelling anyone to their right, and shirking all responsibility to make citizens that understood the rigorous demands of liberty.

What’s missing from the “let’s found a civic center” response is a clear theory of how campus politics translate into civic life. Is the point of these centers to reform universities or to educate citizens for the practice of self-government? The distinction matters because what happens on campus is mostly ignored outside, in the real world. Most communities resolve disagreements through Robert’s Rules of Order rather than curated debates that “viewpoint diversity” apps are selling. If legislators really believe that the best place to teach citizenship is in college classrooms, I’d like to see the argument for it.

So far, all the intellectual “civics” energy has been directed inward, as another front in the campus culture war, with social media posts crowing about conservative speakers and the end of “woke.” When the Hamilton Center or the School of Civic Life and Leadership speaks of “courage” or “robust discourse,” they’re talking about voicing an unpopular opinion in a seminar room. Civics seems only concerned with campus power. The “tyranny” they oppose is the tyranny of social ostracization or overreach by DEI bureaucrats.

They aren’t talking about Minnesota.

Whatever one thinks of federal immigration enforcement in principle, the civil rights conflicts in Minnesota over the past several weeks are all about limits of federal executive power, the role of individuals and states in resisting federal action, the doctrine of First Amendment retaliation, the standards for Fourth Amendment stops and detentions, and the administrative-law constraints on major enforcement campaigns. Hundreds of encounters have generated primary-source material: pleadings, judicial orders, official statements, and video evidence circulating widely enough that national media are showing frame-by-frame reconstructions of harrowing incidents.

What protections should a citizen expect when questioning the actions of federal agents? A U.S. citizen was shot and killed. Dozens of legal observers have been detained for standing and filming on public sidewalks and in their own cars. Deploying federal agents unilaterally overrode state and local policing priorities. The National Guard has been mobilized to monitor federal agents. The Minnesota Attorney General has sued the Department of Homeland Security, citing the Tenth Amendment, arguing that the federal government lacks the authority to commandeer the streets of a sovereign jurisdiction. All of this is testing the vertical separation of powers and the ability of a state to protect its citizens from federal overreach.

Centers don’t have to issue partisan statements or declare a preferred outcome. But you’d think there would be some interest in framing the real constitutional issues for students. What are the doctrines at stake, what readings might connect the Founders’ fears of consolidated power to contemporary enforcement capacity? How best to analyze the difference between lawful authority and abusive practice?

I have seen nothing from the major institutes—Florida, North Carolina, Ohio, Tennessee, Texas. Granted, many universities have been on winter break, but as of January 20, the public-facing news and events pages for the Hamilton School, UNC’s School of Civic Life and Leadership, UT Austin’s School of Civic Leadership, Ohio State’s Chase Center, and Tennessee’s Institute of American Civics show no programming, statements, or commentary addressing the Minnesota conflict. Their feeds remain oriented toward grants, faculty announcements, curriculum development, and general constitutional commentary, not the Fourth Amendment implications of checkpoints in St. Paul.

Again, these new civics centers did not sell themselves as ordinary departments told to hush because of institutional neutrality. Their founding was justified by legislators as corrective units to better teach civic responsibility, rigorous engagement with constitutional structure, and the preparation of citizens capable of sustaining a free society. Ohio State’s Chase Center builds on the politics of its namesake: Salmon P. Chase’s opinions are “still regularly cited today,” including cases that affirmed “federalism and limited and enumerated powers.” The center was established by the State of Ohio in 2023. Tennessee’s Institute of American Civics is equally state-centered in its founding documents: state law defines its purpose as fostering understanding of “the structures and institutions of federal, state, and local government,” and even authorizes it to provide “nonpartisan resources” to “state government agencies,” while also specifying that it may be physically housed in the Howard H. Baker, Jr. Center at UT Knoxville. UT’s announcement of its institute emphasizes state politics, describing the $6 million appropriation by the governor and lawmakers.

If federalism, limited powers, and the state’s place in the constitutional order are part of the public identity here, why the silence on the clashes in Minnesota? What, to these state-founded centers, are the state’s rights and remedies when federal agents override local policy in the name of enforcement, and what protections constrain federal power when protest, detention, and street-level coercion become the arena of constitutional conflict?

Enough preoccupation with campus politics, I say. Surely the goal of establishing centers for civic education was for more than training students to tilt at academic windmills. The founding generation, who understood tyranny well, would laugh at the idea that the enemy was gender studies.

The civics centers were sold to legislators and taxpayers as places to make citizens. Protected budgets and dedicated faculty lines were justified precisely because civic education was too important to leave to departments that might subordinate constitutional principles to ideological fashion.

Minnesota is the test of that premise. The doctrines at stake—federalism, executive overreach, the rights of citizens against a standing enforcement apparatus—are central to the conservative intellectual tradition the centers claim to embody.

The silence suggests the centers may have been designed for the narrower purpose of simply winning arguments inside the academy. Civics centers face pressure to fill seats. Student interest is thin, even as more centers are being founded, in Utah and West Virginia. The likely end is that they’ll be absorbed into the general education machinery, as I’ve argued. But the centers have already shown what they’re for, and it isn’t federalism in action. If legislators wanted citizens, the $160 million would have been better spent on courtroom access, civic journalism, or legal aid: institutions that engage constitutional questions where the stakes are arrest, detention, and death rather than seminar-room discomfort.

The Federalist Papers appeared in newspapers because the founding generation understood that civic argument happens in public, among citizens with something at stake. They did not write the Bill of Rights to protect unpopular opinions in faculty meetings. They wrote it to constrain the coercive power of the state. Civics centers that cannot address that coercive power in action have not yet justified their existence.

A recent Nature study found that AI-augmented science also tends to look backward not forward, looking primarily at established fields, with markedly less focus on edge topics.

An LLM can only “learn” a new concept when it appears often enough in the training corpus (Common Crawl, Reddit, digitized books) to achieve statistical heft. The time until it does is a latency period. So new teen slang or a new trend has to move from Discord or TikTok comments to more static and stable sites on the internet before an LLM can see it. The same way kids know a meme is dead when their parents have heard of it, by the time AI “knows” it, it’s old news. LLMs, like Boomers, are rarely at the cutting edge of a cultural phenomenon. It’s the same for professional knowledge: by the time an industry practice is documented enough to train an LLM, innovators may have evolved past it.

I’ve written elsewhere about the “last mile” problem, where human judgment refines the data to offer more granular and specific analysis. I’m writing here about the first mile problem. In logistics, the last mile is expensive because delivery becomes idiosyncratic at the doorstep: each package goes to a different address with different access requirements and different recipient schedules.

The first mile is expensive because pickup is unknowable and unpredictable. A sender might not even have thought about sending a package yet. Only a few thoughts turn into physical packages to be packaged, addressed, and brought to the post office. The first mile is the gap between “I’m thinking of sending a package” and “it’s in the system.”

The future does not arrive with a tracking number. A first mile advantage is the ability to be in the right community, in the right room, in the right conversation, to recognize that a signal is worth capturing before it has a name.

What does this mean for hiring?

In economic forecasting, a leading indicator, such as the volume of new building permits or manufacturers’ new orders, signals a trend before it becomes obvious in the broader economy. A lagging indicator, such as the unemployment rate or corporate earnings, confirms a pattern after it has already occurred. Lagging indicators are useful because they validate what you see; they provide certainty for making policy and drawing up contracts. Leading indicators are useful for looking ahead, for planning. They offer opportunity for preemptive action and competitive maneuvering.

LLMs generate lagging, not leading, indicators. They validate and optimize established consensus. They are like the Consumer Price Index, tracking the value of concepts the culture has already purchased. For a business owner under pressure to jump on the AI bandwagon, thinking about leading or lagging helps clarify how best to use AI. LLMs are great for tasks involving stable data and stable rubrics. LLMs are great at drafting standard contracts, summarizing historical reports, or coding a known function. They can confirm trends and discover past patterns and trends. It captures the “beta”—the average market performance—with the most fidelity.

However, relying on LLMs for forward-looking strategy is risky. A firm should not depend on an algorithmic system for identifying emerging markets or cultural shifts. In fast-moving sectors like fashion or digital media, novelty will always be ahead of any foundation model. LLMs are not ideal for first mile prediction or for last mile specificity.

How should new graduates leverage this fact?

Both firms and college graduates should understand first- and last-mile needs. An energetic, market-aware recent college graduate should pitch understanding of emergent sentiment (“vibe”) or pre-consensus intuition as a leading indicator. Young people live on the edges of networks where data is raw and contradictory. They read the signals of activity that precede the economic contracts.

Young people also “get” AI—many have been using it for years, even if they’ve mostly been told they should know how to prompt, how to “humanize” output to fool teachers, how to code. While they generally possess higher technical proficiency than their elders, their greater value lies in their function as a sensor network. They will be ahead of what LLMs know.

When hiring managers interview a candidate, they should ask: “what in your life did you notice before anyone else did?” Their answer is worth more than any prompt engineering skills.

Smart firm owners and hiring managers in the AI era should see the labor market as divided between people who can analyze past and present data and people who look to the future while understanding the present. Yes, invest in AI. But don’t short the future by over-indexing on the past.

Smart college graduates should view their temporal immediacy as their primary asset. Capitalize on this latency. Know what LLMs can do and know what they can’t. The LLM is the archive; you are the source of renewal.

Most hiring managers think about skills. They should be thinking about time. AI is unmatched at compressing the past into usable form. But the past is cheap and getting cheaper. The scarce resource is contact with the future, being present when a signal is still noise, when a pattern has no name yet. Firms that understand this will hire for talent accordingly. They’ll stop asking candidates to demonstrate what a model can already do and start asking what they’ve seen that no model could have told them.