The Chronicler’s Problem.

It began with Rhaenyra Targaryen.

A friend and I were pulling apart the Dance of the Dragons, the civil war at the heart of House of the Dragon, and chasing the history under it. George R. R. Martin has said the Dance drew in part on the Anarchy, the two decades England spent tearing itself apart in the twelfth century.¹ The parallel runs beat for beat.

Henry I of England lost his only legitimate son to a shipwreck in 1120. The throne was to pass to his daughter Matilda, widow of the Holy Roman Emperor, Empress ever after. The barons swore twice over to uphold her claim. Then Henry died in 1135, and her cousin Stephen of Blois raced for London and took the crown while she was across the Channel. Nineteen years of civil war followed. England remembers them as the Anarchy. It ended in exhaustion and a deal: Stephen kept the throne for life, and Matilda’s son succeeded him as Henry II. She never wore the crown at all.

As we tried to map what aspects of Martin’s tale were mirrors of the Anarchy, we slowly realized the events weren’t the only mirrors; how he told it was one as well. Everything about the Anarchy comes down through partisan chronicles with very different frames: the Gesta Stephani for Stephen, William of Malmesbury for Matilda,² the Peterborough monk who wrote that it seemed as though “Christ and his saints slept”.³ That one sentence fixed how England remembers an era historians now think was far more uneven.⁴ Fire & Blood runs the same way: the Dance comes down through a bawdy fool, a pious septon, and a courtly maester, so even in the fiction you cannot quite know what happened.¹ We were comparing two wars, and we could not see either one: only accounts we could not step outside of and could not check. All we could do was read the witnesses against each other.

The conversation stopped being about medieval succession. AI has made the chronicler’s problem vast and cheap: any account can now be manufactured. The answer we backed into was about education. What would a society have to teach its people to keep hold of the truth?

Mechanism, Not Resemblance

The parallel was worth trusting because it was structural: named heir, sworn oaths, usurping rival, ruinous war, a son who inherits what the mother could not keep. It was also rigged. Rhaenyra matches Matilda because Martin built her from Matilda, and a copy agreeing with its original proves nothing. The test earns its keep where nothing was copied. Valyria feels like Rome, all marble and roads, but nothing Roman runs underneath it. Stop asking what a thing resembles; ask what mechanism drives it.

Each time one of us reached for a resemblance that didn’t hold, the other caught it. I’ll come back to what that checking actually was.

The discipline sounds like fantasy trivia. It’s the working end of knowing anything. And it raised a question I couldn’t answer at the time: who checks my reading of the mechanism, when that reading is just another account I happen to prefer?

Too Big to Read

The chronicler’s problem is not behind us. Time severed the historian from the Anarchy: the war over, the witnesses dead, nothing left to check the accounts against. A thousand-page bill is public the instant it’s posted and no one governed by it will ever read it. What reaches you is testimony about it: the sponsor’s summary, the opponent’s, partisan chronicles of a document nobody visits. Complexity does to the living what time did to the dead.

I watched a version of it up close. I was a Catholic school kid in the years the Church’s abuse scandal broke, and what broke was not one crime but a bureaucracy. Priests were moved quietly from parish to parish, families were settled and sealed, and every parish saw an arrival, never a pattern. The truth was never hidden exactly; it was distributed, one fragment per parish, and the only desk that could read the whole pattern was the one moving the priests. When accountability finally came, it landed on individual priests, and only the ones the statute of limitations could still reach; the cardinal who arranged the transfers ended his career with one of Rome’s great basilicas.³⁸ That system had a man who decided, and it could not hold him. Most of the systems that run our lives don’t even have the man. No one decided this has become no one can undo this. Access, legibility, and power are three different things, and the distance between them is widening. Onora O’Neill saw the first of these gaps a generation ago: transparency was sold to us as a substitute for accountability and is nothing of the kind.⁵

When Forgery Is Free

Our oldest instrument for sorting true from false is provenance: where did this come from, who says it, who vouches. The chronicler’s problem was that provenance was thin: few could write, fewer could copy, and lineage decided what survived.

Ours is that AI has made it cheap to fake. Provenance was never truth; it was a proxy that held while forgery stayed expensive. When any text, image, or recording can be manufactured at essentially no cost, forgery becomes indistinguishable from the genuine article, and the proxy stops correlating with truth. Nearly every method we had for telling them apart ran through provenance. Engineers are building a cryptographic replacement, signed capture and content credentials,⁶ and it may work. But it relocates trust into whoever controls the signing rather than restoring it. And underneath it all sits the harder asymmetry: producing false claims now scales; verifying them does not.

The Church had closed questions before. In 325 the council at Nicaea decided the nature of Christ and named the losing view heresy. The creed was later remembered not as a decision backed by power but as the discovery of something always true. AI does this at industrial scale, in three quiet steps: what it learns from is chosen, its answers are trained toward a house style, and a handful of systems serve the result to billions. Ask one a question the culture still contests and it answers in the voice of settled fact, or declines, which is the same ruling delivered as etiquette. But Nicaea’s closing stayed openly contested for fifty-six years, because it was visible: a room, a date, signatories, a text to petition against.⁸ The AI’s creed has none of those,⁷ and that, not the closing itself, is the new thing.

A wrong answer can be corrected. The danger is a picture of the world so seamless and authoritative that you stop asking whether the world is shaped that way.

No View from Nowhere

Two things are true at once. There is a floor: the world is some way regardless of belief; the primes don’t care what we think. And there is a frame: everything we say about that world arrives already selected, described, and weighted by a mind that decided what mattered. There is no view from nowhere. And nobody grades their own homework.

Which is why the answer is not to find the one wise mind and hand it the frame. Civilizations have chased that since Plato’s Republic idealized the philosopher-king,¹⁰ and none has solved the step where someone must crown the philosophers. The case for many hands was never that the many are wiser than the one; it is that a structure with many hands can be wrong in one place and corrected from another. Popper made the argument in 1945: democracy was never a claim that crowds find truth but a mechanism for removing rulers without killing them.¹¹ What matters is whether an error can be undone before it is fatal.

Education’s New Job

If nothing can vouch for itself, the checking has to come from outside. And at the scale of a civilization, outside cannot be one more authority; that only moves the question of who checks the checker one step down. The checking has to be distributed.

This is where the conversation was headed, because it reframes what education is for. Education’s central task can no longer be handing down settled conclusions, because conclusions can now be forged at machine speed. The task becomes producing people who can evaluate: probe a claim, know what would count against it, weigh confidence against evidence, and hand their judgment into a process someone else can contest.

Conclusions don’t leave the curriculum. You cannot probe a claim about vaccines or interest rates without knowing something about vaccines or interest rates.¹² So conclusions change jobs: no longer the thing you memorize, now the material you practice judging on.

This already runs in classrooms. One school district put it in its high-school government courses: students taught to read laterally, leaving a page to check who stands behind it before trusting it, the way professional fact-checkers work. Against classmates taught the standard unit, they came out measurably harder to fool.²⁷

Evaluation is also not one skill. Some claims you check by measuring, some by whether they survive attack, some by whether they work under load, some the way a historian checks a chronicle, and some by nothing but whether the judgment can be defended out loud. The master skill is knowing which check the question calls for.

The old worry, running since Lippmann and Dewey,¹³ is that a half-trained crowd, each member newly certain, is more dangerous than one that defers. Fair. The ask is narrower: that people learn the edge of their own competence, so they know when to judge and when to defer.¹⁴ Though deferring is not free either; choosing whom to trust is itself a judgment.¹⁵

Beneath that worry sits a harder one, and the founding case states it. The author of the Gesta Stephani was not incompetent. He was loyal. The chroniclers could evaluate; they evaluated for their side. Modern studies find the same thing: give partisans more numeracy and they read the same table into a sharper version of their tribe’s conclusion.³⁹ On contested ground the problem is not what people can do but what they want to do, and a curriculum can sharpen the weapon it means to retire. I have no clean answer. The loyal do miss together, though, and missing together can be seen from outside.

For the checking to be worth anything, it has to be owned broadly, or it is the closed council again. It has to be collective, because reasoning runs sharp in company and lazy alone,¹⁶ and mixed groups often beat abler ones that all think alike.¹⁷ And it has to be practiced, or it hardens into dogma.

The strongest rival view locates the fix in institutions: courts, journals, newsrooms.^19,32 I am betting one level lower. Institutions hold only if the people staffing and checking them can evaluate; a body checked by people who cannot is captured from inside. And no one can evaluate everything, so institutions carry what one mind cannot, keeping track records visible enough that trusting well is a lookup, not a second career.³⁷

Westeros never changes, eight thousand years of frozen feudalism, because Martin withholds the printing press; the ability to evaluate stays bottled in one guild of maesters. Our press broke that monopoly, and the old order came apart within a few centuries.²⁰ A civilization’s ability to correct itself tracks how widely that ability is spread, and education is how a society sets the spread on purpose.

The catch: a crowd can agree with itself and still be wrong. People drawing from the same source don’t check each other; they echo, and the agreement feels like verification. It is one error copied a thousand times. Judges are only worth having in numbers if they miss in different ways,²¹ and the reverse is true too: many decision-makers running the same excellent model do worse than a motley of weaker, independent ones.²² If we all see the world through the same few systems, we are one evaluator wearing a thousand faces. So watch the independence itself: when everyone starts agreeing for the same reason, that is the alarm. And a few frontier models push the wrong way by default.

Which brings a correction the essay’s own terms demand: the friend in the second sentence was an AI. The conversation that produced everything above, the Matilda match, the checking discipline, the second pair of eyes I praised, was with a language model. I let “a friend” stand for five sections because the experience of not being able to tell is the argument. I won’t oversell the device; the Guardian ran an op-ed under a robot byline in 2020,³³ essayists have composited friends for as long as there have been essays, and I controlled both the deception and its lifting. What AI broke is only the default: “a friend” no longer even implies a human. The deeper cut is that the check was never independent. The model trained on the corpus; I was raised in the culture the corpus was drawn from. Where we agreed, we may have been not two witnesses but one. Correlated misses are invisible from inside the agreement. That is why the dialogue can never be the test, and why the checking has to go to minds that were not in the room.

I should have seen the strongest evidence sooner. How do historians know the Anarchy was more uneven than the monk’s sentence suggests? Not by finding a neutral witness; there is none. They triangulate.

The chronicler’s problem was never solved by a better chronicler. It was solved, partially and over centuries, by exactly this kind of shared checking.

Two things I don’t have. The tools that could make the world legible are owned by the few who profit from its illegibility, and I have no story for how broadly owned checking gets built against that interest; the nearest playbook is how commons get governed against the people who would strip them.²⁵ And I have said judgment must be spread wide while saying nothing about how a million judgments become one verdict: who counts, what weighs, what rule combines them. Pieces exist. Forecasting tournaments score judges against outcomes,²⁹ and Community Notes surfaces the notes that win agreement across raters who normally disagree.³⁶ Nothing joins them yet.

So the AI’s right place is inside the process: used, doubted, one input among many, never the authority that ends it.

You Can Only Trust What Can Be False

What I’ve built is a single frame. It’s coherent, it fits a suspicious range of things, and reaching it feels like arrival. The pleasure of coherence is not evidence of truth. Nicaea and the Anarchy each support other readings; I chose the ones that fit. An essay is the wrong place to test a frame, because the writer controls which objections appear. So this ends in a commitment rather than a conclusion: a companion piece states the claim plainly enough to be tested, and invites you to break it.

Christ and his saints slept, wrote a man who could not see the whole of what he described. Neither can I. So take up the companion piece, and come find the crack.

1. George R. R. Martin, Fire & Blood (Bantam, 2018). Martin has acknowledged the Dance of the Dragons drew in part on the Anarchy; the book's frame (Archmaester Gyldayn reconciling the irreconcilable testimonies of Mushroom, Septon Eustace, and Grand Maester Munkun) reproduces the source condition of the real war.
2. Gesta Stephani, ed. K. R. Potter, rev. R. H. C. Davis (Oxford Medieval Texts, 1976); William of Malmesbury, Historia Novella, ed. Edmund King, trans. K. R. Potter (Oxford Medieval Texts, 1998). The pro-Stephen and Angevin-leaning witnesses, respectively.
3. The Peterborough Chronicle (Anglo-Saxon Chronicle, MS E), annal for 1137.
4. Edmund King, King Stephen (Yale University Press, 2010); David Crouch, The Reign of King Stephen, 1135–1154 (Longman, 2000). The revisionist case that the disorder was severe but regional, argued substantially from non-narrative evidence: charters, writs, and coinage struck through the war years.
5. Onora O'Neill, A Question of Trust (Cambridge University Press, 2002). The BBC Reith Lectures arguing that transparency has been mistaken for, and cannot substitute for, accountability.
6. Coalition for Content Provenance and Authenticity (C2PA), Content Credentials technical specification, c2pa.org. The principal industry effort to rebuild provenance cryptographically.
7. William MacAskill, What We Owe the Future (Basic Books, 2022), on value lock-in: the risk that systems built at scale entrench one era's judgments.
8. R. P. C. Hanson, The Search for the Christian Doctrine of God: The Arian Controversy, 318–381 (T&T Clark, 1988). The standard account of the fifty-plus years of contest between Nicaea (325) and Constantinople (381).
9. Kurt Gödel, "Über formal unentscheidbare Sätze der Principia Mathematica und verwandter Systeme I," Monatshefte für Mathematik und Physik 38 (1931): the incompleteness theorems. Used here as illustration of the shape of a limit, not as proof about societies.
10. Plato, Republic, Book VI (the ship of state, 488a–489d): the navigator who ought simply to be given the helm.
11. Karl Popper, The Open Society and Its Enemies (Routledge, 1945), esp. vol. 1, ch. 7: replacing "who should rule?" with the question of how bad rulers can be removed without bloodshed.
12. Daniel T. Willingham, "Critical Thinking: Why Is It So Hard to Teach?" American Educator 31, no. 2 (Summer 2007): the case that critical thinking is substantially domain-bound.
13. Walter Lippmann, Public Opinion (1922) and The Phantom Public (1925); John Dewey, The Public and Its Problems (1927). The original debate over whether a general public can be competent to judge.
14. Justin Kruger and David Dunning, "Unskilled and Unaware of It," Journal of Personality and Social Psychology 77, no. 6 (1999): miscalibration concentrated precisely where competence is lowest.
15. Elizabeth Anderson, "Democracy, Public Policy, and Lay Assessments of Scientific Testimony," Episteme 8, no. 2 (2011): second-order criteria (track record, conflicts, responsiveness to criticism) by which laypeople can judge experts; Harry Collins and Robert Evans, Rethinking Expertise (University of Chicago Press, 2007): meta-expertise.
16. Hugo Mercier and Dan Sperber, The Enigma of Reason (Harvard University Press, 2017): reasoning as an evolved social capacity, biased in solitary use, effective in argumentative exchange.
17. Lu Hong and Scott E. Page, "Groups of diverse problem solvers can outperform groups of high-ability problem solvers," PNAS 101, no. 46 (2004); for the contested generality of the result, Abigail Thompson, "Does Diversity Trump Ability?" Notices of the AMS 61, no. 9 (2014); Hélène Landemore, Democratic Reason (Princeton University Press, 2013).
18. Helen Longino, Science as Social Knowledge (Princeton University Press, 1990). Objectivity as a property of a social practice meeting four conditions: recognized avenues for criticism, uptake of criticism, public standards, and tempered equality of intellectual authority.
19. Jonathan Rauch, The Constitution of Knowledge: A Defense of Truth (Brookings Institution Press, 2021).
20. Elizabeth L. Eisenstein, The Printing Press as an Agent of Change (Cambridge University Press, 1979), including print's double edge: pamphlet wars and witch manuals alongside the republic of letters.
21. Leo Breiman, "Random Forests," Machine Learning 45, no. 1 (2001): ensemble generalization error bounded in terms of the strength of individual members and the correlation between them.
22. Jon Kleinberg and Manish Raghavan, "Algorithmic monoculture and social welfare," PNAS 118, no. 22 (2021): many decision-makers adopting the same superior algorithm can lower aggregate outcomes.
23. Marc Bloch, The Historian's Craft (posthumous, 1949; written before his execution by the Gestapo in 1944). The critical method: cross-examining witnesses who cannot be recalled.
24. Marquis de Condorcet, Essai sur l'application de l'analyse à la probabilité des décisions rendues à la pluralité des voix (1785). The jury theorem: with independent, better-than-chance judges, group accuracy rises with size; below chance, it falls.
25. Elinor Ostrom, Governing the Commons (Cambridge University Press, 1990): empirical design principles for governing shared resources against extractive incentive.
26. Neil Postman and Charles Weingartner, Teaching as a Subversive Activity (Delacorte, 1969), ch. 1, "Crap Detecting": education's job as building the detector, not delivering the conclusions.
27. Sam Wineburg, Joel Breakstone, Sarah McGrew, Mark D. Smith, and Teresa Ortega, "Lateral reading on the open Internet: A district-wide field study in high school government classes," Journal of Educational Psychology 114, no. 5 (2022): classroom-taught source evaluation measurably improved.
28. Jon Roozenbeek and Sander van der Linden, "Fake news game confers psychological resistance against online misinformation," Palgrave Communications 5 (2019); accessibly synthesized, with decay caveats, in van der Linden, Foolproof (W. W. Norton, 2023).
29. Barbara Mellers et al., "Psychological strategies for winning a geopolitical forecasting tournament," Psychological Science 25, no. 5 (2014); Philip E. Tetlock and Dan Gardner, Superforecasting (Crown, 2015): brief calibration training measurably improved forecast accuracy.
30. Peter Lipton, Inference to the Best Explanation (Routledge, 1991; 2nd ed. 2004).
31. Mark R. Leary et al., "Cognitive and Interpersonal Features of Intellectual Humility," Personality and Social Psychology Bulletin 43, no. 6 (2017): the young measurement literature on intellectual humility, the trait P2's kernel proposes to train.
32. John Wihbey, "AI and Epistemic Risk for Democracy: A Coming Crisis of Public Knowledge?" (SSRN working paper, 2024); and "In Post-Authenticity AI Age, Knowledge Institutions Matter More than Ever," Tech Policy Press (November 14, 2025): the nearest recent argument that provenance collapse makes knowledge institutions more, not less, decisive. Institution-centric where this piece is education-centric.
33. "A robot wrote this entire article. Are you scared yet, human?" The Guardian (September 8, 2020): an op-ed generated by GPT-3, prompted by the paper and assembled by its editors from eight outputs (run by Liam Porr); an early instance of AI authorship deployed as a rhetorical reveal, the device the essay's confession both uses and disowns.
34. Ethan Mollick, "Post-apocalyptic education," One Useful Thing (August 30, 2024): redesigning education around AI, away from transmitting conclusions and toward judgment.
35. Archon Fung, Mary Graham, and David Weil, Full Disclosure: The Perils and Promise of Transparency (Cambridge University Press, 2007); Mike Ananny and Kate Crawford, "Seeing without knowing: Limitations of the transparency ideal and its application to algorithmic accountability," New Media & Society 20, no. 3 (2018): disclosure scholarship distinguishing information disclosed from information usable and actionable.
36. Stefan Wojcik et al., "Birdwatch: Crowd Wisdom and Bridging Algorithms can Inform Understanding and Reduce the Spread of Misinformation," arXiv:2210.15723 (2022): the design of X's Community Notes (formerly Birdwatch), a bridging-based ranking that surfaces notes rated helpful by contributors who normally disagree; a deployed aggregation rule for contested claims.
37. Philip Kitcher, "The Division of Cognitive Labor," The Journal of Philosophy 87, no. 1 (1990): how a community should distribute investigative effort across rival approaches; what the community learns depends on the allocation, not on each member's rationality alone.
38. The Investigative Staff of The Boston Globe, Betrayal: The Crisis in the Catholic Church (Little, Brown, 2002). The Spotlight investigation, awarded the 2003 Pulitzer Prize for Public Service: known abusers reassigned between parishes, settlements sealed by confidentiality. Cardinal Bernard Law resigned that December; in 2004 John Paul II appointed him archpriest of the Basilica di Santa Maria Maggiore in Rome.
39. Dan M. Kahan, Ellen Peters, Erica Cantrell Dawson, and Paul Slovic, "Motivated Numeracy and Enlightened Self-Government," Behavioural Public Policy 1, no. 1 (2017). The most numerate subjects polarized most: identical data read accurately as skin-treatment results and tribally as gun-control results; evaluative skill deployed in service of identity.