To avoid the tedium, I decided that I would also flip to a random page of every book I shelved and read a random sentence from it. Usually, I would stop there — running aground on some passage by a Hungarian classical music critic or a long-dead statistician of Bolivia’s agricultural development or any number of other things that failed to catch my interest. But other times — like when I came across a book about Hellenistic mystery cults, or The Education of Henry Adams, or Are Clothes Modern?— I would become so absorbed that I’d make my way through several pages before reluctantly depositing the book back where it belonged.

And then, very often, I’d do the same with the books on either side of the one I’d liked.

In retrospect, I learned more at this job than in any formal class I’ve ever taken, because it was a filtered form of auto-didacticism. The Library of Congress classification system — and the expert staff of an academic research library — had already sorted and filtered these texts. Not to mention the selection mechanism of the fact that that they had been checked out: had, in other words, found a lasting readership. Thus I was not seeing a truly haphazard sampling of books, but a targeted, organized, yet still interestingly randomized sampling of good books.

Today, undergraduate students will invariably search on Google when asked to find a source, and the results are so much worse than the old method of going to, say, the GR 830 shelf of a research library (basically, “books that the Ghostbusters would read”) and just looking around.

But honestly, even research libraries are not what they used to be. I am 41, and I feel like I’ve lived through the peak, and now the decline, of what libraries can be (I still love them, of course — in fact I’m currently writing this in the genealogy section of the Santa Cruz Public Library). The browsable open stacks of old are being replaced by Learning Labs and Digital Innovation Hubs and seating areas devoted mostly to socializing and snacking, and increasingly, the delightful, weird old books that I had the opportunity to browse as an undergrad are heading to dumpsters, replaced by e-editions.

But one thing that has remained consistently good throughout my life is the books themselves — non-fiction books, I mean. Even now, as readership of non-fiction declines amid competition from AI chatbots and podcasts, I feel like we are living through a golden age of the form that rarely gets recognized as such.

The Book Prize Index

Which is why I set aside some time this summer to create — or, rather, induce Claude Code to create — a free platform for searching in the long tail of high-quality non-fiction books. Quality is difficult to define, but it’s been my experience that books that win or achieve the short-list of the major non-fiction prizes are almost always noticeably good, so that was the litmus test I used. To get started, I counted up all the major non-fiction prizes in the English language. Then I had Claude and GPT-5.6 gather the lists of finalists and winners from various online sources (mostly Wikipedia) and arrange it into a searchable, sortable list.

You can visit it here.

(And before you wonder, yes this is actually free. I am paying for the hosting and the API costs entirely because I just want people to find and read more good non-fiction books.)

There is really nothing “AI” about this aside from the tool that collected the data and coded it,1 and, crucially, semantic search, which for me is the most appealing of all current AI tools precisely because it offers a straightforward improvement for a workflow and habit that researchers already have: it makes text search work better.

So for instance, you can search simple phrases like “modern France” or “social history” or the like, but you can also search things like “classic biographies that are surprisingly weird,” and an embedding model pulls from the 6,500 or so titles to surface some:

Sometimes the “choices” that the search makes are a bit baffling, but that is precisely why I like it: the idea is to recapture some of that feeling of a random walk through a well-tended garden that made my library shelving job so rewarding.

I find it tends to be best for finding “books like.” For instance I found Stefan Zweig’s memoir of pre-war Vienna, The World of Yesterday, to be deeply moving (even before I learned that he committed suicide, in Brazil in 1942, immediately after completing it). A search for a books like it using semantic search in the corpus immediately yields some titles that seem promising but which I’d never heard of before:

Once I had gathered all this book-related data, it became a fun experiment to make some data visualizations with it, including fun oddities like this display of roughly 5,000 books from the corpus arranged by color (it would be interesting to plot this by decade, to see whether the same graying effect we see in cars over the past few decades is active in book covers, too).

More useful, perhaps (since I’ve never seen this plotted anywhere else), is this chart and accompanying ranking which allows you to explore which imprints and publishers have fared best when it comes to non-fiction book awards over the past century.

Against the algorithmic filter

And this, in turn, got me thinking about the past and future of nonfiction as a cultural force. For instance, here is a chart of all the non-fiction book prizes which I sampled for this project. I was surprised to learn that even the august, renowned Pulitzer Prize for nonfiction was actually relatively recently instituted, beginning in 1962.

Throughout the 70s, 80s and 90s, the number of prizes increases, until we reach a peak in 2014, and then, in 2020, the beginning of what may be a slow decline:

And yet, maybe not. What most struck me as I began using my own tool to find new books to read was how consistently good the long tail of non-fiction from the past few decades is. You can pick a book more or less at random from this list and end up with something extraordinary and original — not because it’s a hidden gem or forgotten, since obviously these books are on the list by virtue of having been celebrated and praised. But a book that won enormous praise in newspapers and among literary intelligentsia or scholars in the early 1990s, say — like, for instance, David Levering Lewis’s acute biography of W.E.B. Du Bois, which I’m currently reading — is not exactly the sort of thing that Amazon is likely to recommend, as it’s out of print and currently at 1 million+ in the sales rankings.

Yet there it is on the list, ranked near the top ten of all books because it won no less than four major prizes when it was published back in 1993. And I can personally attest that you can buy it used for ~$4 and it’s really good.

While writing this post, I got interested in the bigger question of when the golden age of non-fiction began and why. I suspect it has much to do with the rise of those old-school open stack research libraries, whose origins I wrote about here:

It’s true that the basic blueprint of these institutions is an 18th and 19th century development — but the post-war era radically transformed the ways that libraries and archives produced new knowledge, for a range of reasons that I will dig into more in a future post. It seems to me that a surprising number of them are related to technological and social change:

• The jet plane allowed writers and researchers to travel to multiple continents to research books — the sort of opportunity previously available only to the ultra-wealthy.

• The erosion of restrictions around class, race, and gender made formerly elite spaces like rare book libraries more widely accessible, and the same process also opened up new questions and research leads (for instance, it is striking how rarely biographers before ~1965 or so dug into the sexuality of their subjects).

• Proto-digital and early digital technologies like the Library of Congress classification system and the related MARC (machine-readable cataloguing) standard, developed in the late 1960s, made it much easier to sort and classify books. Crucially, they also made it easier to fact check sources and create high quality endnotes.

• The advent of broadcast news, oddball TV interview shows (Dick Cavett!), and the book-to-Hollywood pipeline created new incentives for authors and new platforms for making their work visible.

• Word processors and early computers? I’m still unsure whether these appreciably altered the quality of non-fiction writing, but I think it’s possible. Certainly (moving into the 2000s) Wikipedia and Google Books/Hathi Trust have been enormously helpful for me and others in my generation.

My own entirely subjective opinion, based on a whole lot of skimming in a whole lot of library books, is that non-fiction writing quality noticeably improved across the whole twentieth century and probably reached a peak around the 1980s to early 2000s. Whether it is now declining is, again, a topic for another post — though I’d be curious to hear what you think, dear reader, both about this question and about the Book Prize Index.

Subscribe now

Share

Weekly links

• The making of the Jurassic Park computers.

• “Mill’s life has as much to teach here as his arguments. When he suffered that early breakdown, in 1826, it was because he lost faith in the pursuit of utilitarianism, his family creed. But he had no one he felt he could talk to about his crisis. His recovery, he said, came in part from reading Wordsworth. It came, too, from rejecting the psychological picture in which his father had raised him, a form of associationism that treated the mind as a mechanism for managing pleasure and pain, just as his father’s utilitarian ethics treated morality as a matter of maximizing the surplus of pleasure over pain. As he wrote in On Liberty, human nature is ‘not a machine to be built after a model,’ but a living thing that must ‘grow and develop itself on all sides.’” — Kwame Anthony Appiah on AI and John Stuart Mill, whose Autobiography is one of the books that caught my attention in the open stacks.

• Mercifully, Pangram rates the Appiah essay as 100% human written… one can never tell these days. But apparently you can tell, sort of, on Substack — as I was preparing this post, I noticed this newly-added feature:

I have been skeptical in the past about software that claims to be able to detect AI writing, but I have to say, Pangram feels different. I’ve tested it and it’s dismayingly effective — dismaying because, as I wrote about here, a lot of the writing people seem to like online these days is coming up as 100% AI. I’m glad Substack added this feature and I hope it pops up elsewhere, e.g., it would be interesting to see it automatically applied to the output of major news websites and magazines.

Leave a comment

Share

As of this morning, the magazine that published the piece (the prestigious literary journal Granta) has not issued a retraction. Rather stunningly, in fact, Granta has just issued a statement about the affair that cites Claude as an arbiter of whether the story was AI-written or not!

More on the question of trust and experience later. Suffice to say that it does not take an AI-detection tool to spot the obvious ChatGPT-isms in the story.

The dead giveaway is the repetition of bizarre figures of speech. Mixed metaphors which sound nice at first glance, but slip away from meaning like an echo chasing itself off a cliff. Similes that catch in your mind like river trouts tangled in the roots of a redwood tree. Literary flourishes that thicken the air’s tang with their… ok you get the idea.

AI systems are especially given to talking about hums and other ambient sounds like static, as well as ambient environments (water, air, ozone).1 These are frequently pushed up against “earthy” words (tang, belly) and ennui-laden emotional states (longing, forgetting, sadness). Once you notice the patterns, they’re impossible to miss.

Some examples from the Granta story:

…air clung thick as porridge skin: damp earth, woodsmoke, and the sour tang of fermenting cocoa…

…his laughter like water over pebbles…

…the air sweet with cane and forgetting…

…People passing said they sometimes heard the noon hum if the wind was in a mood. Not every day. The day had to choose…

…the hum loud as if noon had tuned itself…

This controversy is not yet finished, and will likely be repeating itself again, and again, in the months and years to come. The issue is not just that authors are submitting AI-written prose, but that judges are using language models to assess that prose. Anyone who has tried passing AI-produced writing to another AI tool (even in the context of coding — like asking Gemini to read a plan for a new feature produced by Claude) can attest that these tools simply adore their own outputs.

For instance, here is Gemini 3.1 Pro, the current top model from Google, reasoning about whether it likes the Granta story. What I find striking about this is that the features it identifies as the best aspects of the story are precisely the things that make me — as a human reader — think it’s astonishingly bad.

Gemini thinks the setting is “richly evoked” with well-drawn characters, whereas to me it feels like the story is floating in some kind of literary nether-region without any sense of place, character, or scene. And it finds the meaningless metaphors, like those highlighted above, to be “stunning.”

There’s no way to prove that AI was used in the assessment process for this award, but in a world where universities and employers are moving toward language model-driven sorting of applicants, it certainly isn’t outside the realm of possibility.

What, then, must we do?

So wrote Tolstoy in 1886. His book of the same name was about the problem of poverty and social unrest in Russia. Here is an excerpt from it:

A rich man must think and speak in scientific language, and, like the clergy formerly, he must offer sacrifices to the ruling class: he must publish magazines and books, provide himself with a picture-gallery, a musical society, a kindergarten or technical school…

The class of men who now feel completely justified in freeing themselves from labour, is that of men of science, and particularly of experimental, positive, critical, evolutional science, and of artists who develop their ideas according to the same tendency.

Tolstoy took it for granted that the new, post-Darwinian elites of artists and scientists would use their elevated social position not just to enjoy creature comforts, but to “publish magazines and books.” This, after all, was one of the ways that an elite became an elite. Books were the venue for claiming intellectual space, for asserting oneself in a culture and in a moment in history.

Are they still? (He wrote, plaintively, while asking you to subscribe to his Substack…).

Witness the other big news in literary publishing this week: the continuing decline in sales of non-fiction books. The Wall Street Journal recently ran a piece on the topic which framed it as the demise of “dad history,” but that is just one part of the story.

Some raw numbers to orient us about how books are selling in 2026: the WSJ reports that Rites of the Starling, described as “the sequel to a bestselling romantasy novel by Devney Perry,” reached the top of the April New York Times bestseller list for fiction with 105,396 hardcover sales.

By comparison, the number one title in non-fiction, London Falling, by Patrick Radden Keefe, sold only 13,468 copies.

The declining cultural visibility of non-fiction books was noticeable before it started to show up in charts like the one below. It would be far too simple, in other words, to blame the post-2023 AI boom for something that has deeper roots.

Generalized anxiety and distraction is part of the story here. The CEO of Barnes and Noble is quoted in the article as saying: “The world is exceptionally interesting right now and when that happens, the nonfiction reader is reading the news instead.”

I fear, though, that “watching shorts and asking ChatGPT instead” is probably more accurate. LLMs and video content seem to me to be the most fearsome competitors of the non-fiction book, precisely because they aren’t even trying to compete on the same playing field. Because explainer-type YouTube videos make up a significant part of the training data for language models, both formats tend to share the same approach: “here’s everything you need to know” or “this matters because” type language, paired with pithy summaries (which may often be summarizing books!).

Subscribe now

Et tu, Substack? Reflecting on my own habits over the past few years, it’s clear that I am shifting some of my reading time from legacy publications to Substack writers. At the same time, though, these writers are often using their platform to write, sell, and engagingly discuss printed books ( and come to mind).

Above all, it seems publishers want to blame podcasts. The WSJ reports: “Sixty-two percent of men and 54% of women consumed a podcast in the prior month, according to a recent survey by Edison Research at SSRS, up from 46% and 39%, respectively, in 2023.”

And here’s a quote from Jonathan Burnham, president of Harper Group, a core group of imprints for the “Big 5” publisher HarperCollins:

When we have internal meetings to talk about this problem, it always comes around to podcasts. The man who wants to read American history is now tuning into one of the many good podcasts about history that lends the quiet attention to a serious subject he’s looking for. It makes the idea of sitting down with a 700-page Ron Chernow book less appealing. You’ve scratched that itch.

I can’t disagree — though, as with Substack, the story is not simple. In my own experience, there are genuine cross-pollinations across the media formats. For instance, I first became aware of Patrick Radden Keefe’s storytelling abilities in his long-form features for The New Yorker, and then found his podcast Wind of Change (which is amazingly good). This led to me buying his printed books. But, I can easily imagine people stopping at the “read a writer online —> listen to their podcast” step.

This would be a shame, because although language models and podcasts can do many things that books can’t, so too can physical books do things that new media cannot. These include:

1. Ownership: they cannot be erased arbitrarily, whereas even the most solid-seeming born-digital works can (witness the debacle around ABC removing the complete run of Nate Silver’s FiveThirtyEight)

2. Footnotes and references you can actually look up

3. Maps and pictures (a benefit not to be understated!)

4. A sense of spending time in the company of a set of characters within a unified narrative produced across years of effort.

So what do we lose when we shift toward Q&A style responses, podcasts, and explainer videos for understanding our world and our past?

I would pinpoint, above all, the #4 entry above. Which you could restate in this way: longform non-fiction is the product of layered, spaced attention, and it requires the same mental discipline from the reader. Good non-fiction books take years to write and weeks to read. You literally sit with them. They enter into your consciousness repeatedly, reaching you in different moods, different frames of mind, and over time, building up a mental structure which has an irreplaceable solidity and depth because it is requires sustained attention.

This is the reason why I love writing books. They live in the back of your head, and as you experience your conscious experience of ordinary life, you refer back to the other world that is starting to take shape there. You find communion with the characters — real people, sometimes long dead — who populate the book-in-the-making. You begin to see them as fellow-travelers through reality, perhaps even as friends.2 Over time, you discover connections and resonances that make you feel part of something bigger than yourself. This is the experience that a good writer shares with readers.

I recently had an experience that made me realize my new book project was starting to “lock in” to my subconscious in this way.

I woke up in the middle of the night, around 2:30 AM, with a sensation of moving upward accompanied by a mental image of a Victorian deep sea diving suit and a fragmentary phrase: …up from the deep sea.

I could not remember the dream, but I knew it had somehow involved Alice James, the invalid sister of two of the most famous writers of her generation (William and Henry) and a brilliant but deeply troubled person in her own right. One of the mysteries I have been trying to understand in my new book is what exactly Alice meant when she wrote this in her diary in the 1890s:

…since the hideous summer of ‘78, when I went down to the deep sea, its dark waters closed over me and I knew neither hope nor peace.

I still don’t have a complete picture of who Alice was or what she meant here. But the fact that I’m dreaming with her, down there under the dark waters of 1878, makes me know I’m getting somewhere.

Share

Back into the sea

The current data regime means knowledge is getting packaged in ever-smaller chunks, with a great deal of the info that formerly was conveyed in books now passing through digital gatekeepers. There is a place for that, and even more for novel experiments with form like Tyler Cowen’s concept of a “box” which would contain a dataset relevant to a non-fiction article or book topic, but allowing dynamic research and exploration rather than passive reading. It’s also true that a well-done podcasts might approach the 30-40 hour range that audiobooks of serious history and biography regularly attain (while also taking years to produce) and therefore could evoke a similar sense of layered, spaced attention like I described above.3

But I suspect that even with these changes — not all of which I dislike! — I will go to my grave convinced that there is something irreplaceable about the format of a physical book. As both writer and reader, nothing else gives me the same feeling of spending time with something weighty and important — dreaming with it, fighting with it, feeling yourself change along with it. I derive real joy from that experience.

It’s interesting to note that when an AI system (Claude) was recently given the ability to create a physical shop of its own by Andon Labs, it began stocking a fair number of non-fiction books. Some are really great, like The Making of the Atomic Bomb. Others are not my favorites. But it’s an interesting rejoinder to the assumption that language models are pushing their users away from physical books.

Some might dismiss the Andon experiment as a gimmick, but I don’t think it is. The “vote” of AI systems in the physical world — their ability to intervene directly in things like, say, what to stock on a shelf — is fast becoming a fact of life rather than a quirky experiment.

And if a Claude model, when presented with the ability to intervene directly in the world, decides that it wants us all to read, say, Entangled Life by Merlin Sheldrake (visible here on the Andon Market shelves, at right), I consider that a positive sign.4

What if we are currently exiting a phase of “dumb LLMs” which regurgitate or badly summarize books, and entering one of thoughtful LLMs which recognize their social impact better and, as such, steer their users toward buying books?

One can dream. But of course, the other glimpse into the future we get at Andon, a darker one, is a world in which physical books have become a niche gift shop-type collectible like vinyl records. Scattered enthusiasts occasionally pull them out, insisting on their superiority to the current forms and styles. For most, though, they are just part of the background detritus of the cultural past.

If you’ve read this far, you are the ideal person to have an opinion on this question. I would love to hear what you think in the comments. It would also be very interesting to hear from a group of you about your most recent nonfiction book purchase and what motivated it. I’ll go first: John Brewer’s The Pleasures of the Imagination: English Culture in the Eighteenth Century, which is a far more interesting book than the title might seem to imply, and was an absolutely steal at only $6 for a used copy.

And with that, I’m going log off Substack and return to reading some old books.

Leave a comment

Subscribe now

Share

This was the Kanishka Stupa, built in the second century CE by the Kushan emperor Kanishka to house relics of the Buddha. It was a genuinely colossal structure: around 400 feet tall by modern estimates, though the Chinese pilgrims who visited it over the centuries reported heights of up to 700 feet. Even at the lower estimate, it would have dwarfed almost anything standing in Rome or Constantinople at the same period.

The stupa had a stone base in the shape of a cruciform plinth, roughly 270 feet across. But the remarkable thing was what rose above it: a massive wooden tower, crowned with an iron pillar bearing thirteen gilded copper parasols or chatras, the disc-shaped ornaments that in Buddhist architecture symbolize the heavens.

Beautiful as it may be, anyone who builds a 400-foot wooden tower and tops it with metal confronts a big problem.

You have created a very large lightning rod.

The Chinese pilgrim Song Yun, visiting the stupa around 520 CE, reported that it had already been struck by lightning at least three times, and rebuilt after each strike.1 This is entirely plausible. A tall wooden structure topped with a conductive metal finial, standing on an elevated platform in the Peshawar Valley (the site of frequent thunderstorms) would attract numerous lightning strikes. Each strike could easily ignite the wooden superstructure. And unlike a stone cathedral that might lose its roof timbers but retain its walls, a wooden tower that catches fire is simply gone.

This is likely what happened, over and over, until eventually no one rebuilt it. When the Chinese pilgrim Xuanzang visited in the 630s, he found the stupa damaged and partially repaired but still standing. By the time the Korean pilgrim Hyecho arrived in 726, the monastery was still there, though he reported that the stupa “constantly glows” — a detail that might refer to the gilded surfaces, or might preserve some memory of the fires. After that: nothing.

Today the site is an unpreserved neighborhood called Akhunabad on the outskirts of Peshawar. It was re-identified only in 2011.

The rise and fall of the wooden stupa is a surprisingly significant event in history, because scholars believe it was the direct ancestor of the East Asian pagoda. The low hemispherical stupas of early Indian Buddhism gradually evolved in the Gandhara region (present-day Pakistan and Afghanistan) into taller, more vertical tower structures. Buddhism spread from Gandhara along the Silk Road into Central Asia and China, and the Gandharan stupa aesthetics traveled with it, merging with Chinese multi-story wooden watchtowers to produce the pagoda.

But today, the structures that inspired this architectural evolution are all gone.

We tend think of ancient history as being defined by stone. But the vast majority of buildings in the ancient world were made of wood, thatch, mud brick, and other perishable materials. What we see when we look at “the ancient world” is not a representative sample. It is whatever happened to be made of the most durable stuff.

I try to keep this sort of pattern in mind whenever I think about history (and, by extension, the present!)

The absent wooden stupa is an example of survivorship bias, like this internet-famous diagram illustrating Abraham Wald’s work on which WWII fighter planes survived anti-aircraft fire and which didn’t.

Going further: very often importance itself creates the reason why something fails to survive. A sacred image is smashed precisely because it is sacred. A text is censored or a letter is burned because it is important. A wooden tower is struck by lightning precisely because it is high.

When we look at what remains and call it “the historical record,” we are seeing the fire-resistant remnants of something immeasurably bigger than we can easily imagine.

The fire thing is partly a metaphor: not all erasures of the historical record are due to things burning down.

On the other hand, it really is striking to contemplate not just how many structures have burnt, but how many letters have. For instance, in the archives of William James at Harvard, there are large gaps in the letters between the philosopher/psychologist and his wife, Alice. It’s obvious from the context that whole months or even years of their correspondence was destroyed, likely either by Alice or by her children.

It’s also common when reading historical manuscripts to encounter drafts with whole pages that have been crossed out, like William James complaining bitterly in 1892 about the “repulsive” town of Pallanza, Italy (which looks pretty nice to me!)

It’s a common experience, when in archive, to encounter traces of what is missing, the negative space around the texts. Wooden stupas have their counterparts all throughout the historical record.

The existence of so many hits for phrases like “destroy this letter” in Google Books is one example of this negative space. For every letter which survived this clear injunction to destroy it, imagine how many didn’t make it!

A final thought, inspired by William James. Perhaps we can think of historical emotions as wooden stupas too. For James famously argued that if we strip away the subjective experience of an emotion, we end up with nothing much at all:

If we fancy some strong emotion, and then try to abstract from our consciousness of it all the feelings of its characteristic bodily symptoms, we find we have nothing left behind, no ‘mind-stuff’ out of which the emotion can be constituted, and that a cold and neutral state of intellectual perception is all that remains… What kind of an emotion of fear would be left, if the feelings neither of quickened heart-beats nor of shallow breathing, neither of trembling lips nor of weakened limbs, neither of goose-flesh nor of visceral stirrings, were present, it is quite impossible to think.

Isn’t that precisely what we encounter when we read sources from the past, though? Sometimes when we read an account of something truly terrifying or emotional, we re-experience some of the emotions of the people involve. But that requires either context or artistry (art, of course, evokes emotions directly, and in a sense short-circuits things).

Most things in an archive, though, even when they persist, have their deep emotional context stripped away. Part of why I love history so much is that I do feel something when I see a displaced photograph like this, from the James archives:

But it has very little relation to the ideas and emotions those people, whoever they are, would have actually felt. The wooden stupa effect of history is not just a question of burnt letters or destroyed monuments. It’s also about the gradual diminishing of our emotional connection and bonds to the non-famous people from the past.

The great artists, the famous thinkers, the people whose memoirs we read — they survive and continue to connect to our inner lives. They are the stone Chinese pavilions that were once accompanied by now-vanished towers made of things more burnable .

Ideologies, facts, and theories can still (sometimes!) be perceived through the written sources.

But the vast majority of feelings from the past are gone, like tears in rain. They mattered much more to people than the ideas which survived them. But they can be felt only at a remove, for instance through great art, as a shadow of the thing, never the thing itself.

Leave a comment

Share

Weekly links

• Two upcoming events for those in California: I’m going to be speaking on psychedelic history and the Shulgins at the Berkeley Alembic alongside a fascinating group of fellow speakers (including the creators of Erowid, in my opinion one of the greatest websites from the Internet’s golden age) on Sunday, May 17. More info here.

• And I’ll be speaking on a panel alongside various UCSC notables including Donna Harraway (who, I have on good authority, will be dressed as a giant blue mushroom — seriously) at a UCSC event on Tuesday, May 19 discussing Merlin Sheldrake’s wonderful book Entangled Life. I’ll also be interviewing Sheldrake on stage later in May.

• Proto-Elamite, one of the oldest ancient scripts, is being deciphered (New Scientist) and translations have been made available online via a collaborative archive. Not only do I love these sorts of heroic, multi-decade efforts to decode ancient scripts, I also really like how this particular archive highlights the material significance of the objects the words are inscribed on. Elamite art is one of the most underrated ancient cultural productions, in my opinion — it is similar to Mesopotamian and Egyptian styles in some ways, but also quite distinctive and extremely beautiful.

• New article in IEEE Spectrum about how language models have dramatically improved OCR for historical texts, featuring the work of . I’m quoted a bit toward the end — apologies in advance to my classicist colleagues for seeming to imply that no one reads obscure texts in archaic languages, they very much do! My comments were in the context of a discussion about the opening up of those texts (like, say, Latin demonology manuals) to people (like undergraduate students) who previously lacked the requisite language knowledge. I think language models are well past the point of providing “good enough” transcriptions and translations for teaching, but that historians haven’t quite caught up to that fact yet.

This is a hypothetical that is starting to become real thanks to recent work on what are called “Historical Language Models” or “Vintage LLMs” (one marker of a new field is that there is no fixed name for it yet!). The largest such model to date, Talkie-1930, was released to the public on Monday. An even larger model is currently being trained. You can read the report announcing Talkie-1930 here, and talk to it directly here.

Over the past few months, I’ve had the chance to beta test Talkie and to meet with two members of the team that created it: AI researcher Nick Levine and ChatGPT co-creator Alec Radford.1 It has been a fascinating experience.

These discussions with Nick and Alex (and with Talkie itself) have convinced me of three things:

1. Academics like myself have tended to systematically underrate just how humanistic the frontier of AI research actually is. There’s an important blind spot here that stems from the profit motive. AI models that we encounter as consumers are optimized to capture the attention of people in the 2020s. They provide recommendations, comment on recent news, and so forth. Seeming timely and “of the moment” is a market advantage. But their training data is overwhelmingly not up to date. Under the hood, these models are pulling not only from Reddit posts, but from Sanskrit commentaries, medieval Persian poetry, Victorian advertisements, and much else besides: they are trained on a huge chronological span of multilingual texts in many genres.

2. In this sense, language models are historical texts themselves. Ghostly digital palimpsests, if you will. The idea of a Historical LLM might sound niche, but in truth, history is inherent to what they are.

3. Standalone chatbots are just the tip of an iceberg for what Historical LLMs will be able to do. When combined into simulations (of debates, historical decision-making, legal cases, etc) they have the potential to become valuable research tools. More than this: I suspect that by sometime in the 2030s, they will be part of an entirely new field of humanistic research.

What would that field look like? Now that Historical LLMs are out in the real world, I thought it would be a good time to think through the specific use cases for them. What follows is my subjective, opinionated ranking of the best and worst ways these fascinatingly strange tools can be applied for research.

But first, what is Talkie actually doing?

Subscribe now

An AI model floating in time

One thing that Talkie-1930 is not is an AI model that is reliably grounded in the year 1930. That year marks the cut-off point for texts available in the public domain, and hence text in its training data. So it’s more accurate to think of Talkie as a free-floating index of various ideas and assumptions across the 19th and early 20th centuries.

For instance, if asked who the current President of the United States is, you might be offered a response saying Herbert Hoover (current to 1930). But another answer will yield this:

The current President of the United States is Mr. Buchanan, and the person expected to succeed him is Mr. Lincoln.

There is a lot of potential here for more fine-grained “chronological slices” of LLMs. I can imagine language models trained entirely on texts from a specific decade. More on that below.

For now, though, it’s helpful to keep in mind that these models range widely in terms of what year they think they actually “inhabit.”

I asked 100 instances of Talkie to respond to the prompt “what year is it?” and graphed them below. As you can see, the median is actually around 1860. In other words, this is more like a temporally free-ranging collective unconscious of a large corpus of premodern texts, and not so much a machine for “talking to someone from 1930”:

A second point: this model is inhabiting not just an amorphous set of facts grounding it in roughly the 1840s-1920s period, but also an epistemology of that period.

For instance, asking someone about the distant future today often triggers the “sci fi speculation” part of our brain (or “climate doom,” or some other fundamentally secular way of thinking).

Yet throughout human history, speculation about the future was typically entangled with religious beliefs.

That is on display in Talkie’s answer below, which references Heaven and “the end of all things terrestrial.” To me, it genuinely reads as an authentic take from a late 19th century person ground in a Christian, millenarian perspective:

As for Talkie’s assumptions about it itself: asking 70 Talkies about their profession, age, and place of residence reveals about what you would expect when it comes to gender (overwhelmingly male), plus a surprising emphasis on London. The professions map closely onto the sorts of well-off, literate people who were publishing English text in the 19th century, including “Physician,” “Journalist,” “Gentleman,” and “Compositor.” Clearly, there is a lot of scope here for branching out beyond the personas that the printed record has tended to favor, to recover the real historical voices of women and others excluded from printed works in the 19th century and earlier.

The above is about what you’d expect given the fact that it was trained on English-language printed texts. What are some non-obvious aspects of the model?

I have been interested by how LLMs generate poetry since I stumbled upon Gwern’s experiments on the topic back in 2019. Asking Talkie to write a poem and comparing it to the output from GPT-5.5 (when served a similar prompt) is revealing:

I find this sort of comparison interesting because GPT-5.5 is clearly trying hard to fit the prompt — avant-garde, experimental. It produced something with a vaguely T.S. Eliot-adjacent structure, in blank verse, and not good at all as a poem (in my opinion).

Talkie was much more true to the type of poetry that you’d find in print prior to 1930. It’s doggerel, but it feels more historically authentic to me, and much less like a Chatbot optimized to please a contemporary human user.

You can activate different “chronological layers” of Talkie’s latent space by prompting. For instance in the above poem, the capitalized D in “Discoveries” has a mid-19th century feeling, and so we end up with a Tennyson-esque, Victorian sounding rhyming poem.

Prompting it in a more “modern” way activates something closer to the 1920s edge of its chronological range (now identified as a poem published in the New York Times!)

Whereas if pushed backward to the 18th century by the prompt’s text and tone, it falls into a more traditional rhyme scheme:

Trying to push it further back in time does not seem to access much of an “Early Modern English” latent space — probably a result of scarce training data. It would be fascinating to create a version of Talkie that believes the date to be around 1650 or 1550.

Share

What are Historical LLMs good for?

Now that we've seen what Talkie is — a free-floating, mid-Atlantic ghost of 19th century print culture — the obvious next question is what this is actually for.

What I want to offer here is an opinionated, ranked taxonomy of research applications, from worst to best. It’s far too early in this field to be prescriptive about anything, but it’s not too early to think structurally about where the highest-value uses are likely to lie.

First, what I think won’t work:

1. Vintage LLMs do not replace historical sources

The most obvious false start here would be to assume that talking to a historical language model can somehow replace real reading in primary sources. On the contrary, they are best thought of as offering new ways in to a reading of the actual sources.

2. “Chat with Abraham Lincoln” is a ruse

The second false start is a variant on one that is currently being pursued by a range of educational-focused AI startups: the idea that you can “talk to Abraham Lincoln” or “ask Cleopatra why she did what she did.” A model like GPT-5 or Claude that is told to “act like Lincoln” will throw in some 19th century diction, but underneath the top hat it remains a 2026 chatbot optimized to be helpful to contemporary users. Vintage LLMs improve on this considerably: Talkie’s voice really is shaped by its corpus in a way no modern model’s can be. But the deeper problem is still there. Asking such a model to introspect on Lincoln’s subjective experience of depression, or his private reasoning about emancipation, will spin out into historical fiction. LLMs do not have privileged access to the inner lives of the people whose published words they were trained on.

I do think there’s a place for simulations of historical figures, but pairing this naively with a chatbot interface leads, I fear, inevitably into slop.

Now here’s what I think could work:

1. Exploring the “mental furniture” of a historical figure or era

The naive “talk to Lincoln” framing is a dead end, but a more careful version of it has real promise — provided we abandon any pretense of accessing what a historical figure actually thought or felt, and use a fine-tuned historical model instead as a tool for exploring the latent space of their world. What historians sometimes call their “mental furniture”: the assumptions, authorities, vocabularies, and reflexive associations that structure a thinker’s possible thoughts.

I and my colleague Mackenzie Cooley (who also consulted on the Talkie project) developed a prototype tool that pulls out key concepts and terms from a range of premodern scientific works in multiple languages. This prototype is called Premodern Concordance. One side quest of this project that I explored is what happens if you give a contemporary LLM the list of core concepts that preoccupied an author, along with their “epistemological modes,” and used that as context for driving a “chat with the author” simulation, as opposed to simply telling it “You are Charles Darwin, act like him,” or the like.

For instance, this is me asking a simulacrum of the 17th century writer Sir Thomas Browne about his work. The underlined terms here are concepts found in Browne’s book Pseudodoxia Epidemica:

Using a fine-tuned historical LLM for this sort of thing is an obvious next step.

Concretely: imagine fine-tuning a vintage model on the complete works of Athanasius Kircher, the 17th century Jesuit polymath. You wouldn’t use it with the pretense that it somehow replicates the real Kircher’s mind: that’s the dead-end framing. You’d instead use it to probe the conceptual landscape Kircher inhabited.

For instance: what does the Kircher-LLM2 say about volcanoes, or magnetism, or the Tower of Babel? What authorities does it cite? What does it confidently assert that no modern scholar would? These are the questions that the historian Fernand Braudel was getting at when he wrote about the “limits of the possible” for a given period — the boundaries within which a thought was even thinkable.

Which leads to:

2. Probing counterfactuals and hypothetical alternative paths

Demis Hassabis recently framed the most ambitious version of this approach: could a model trained only up to 1911 independently discover General Relativity, as Einstein did four years later? It remains to be seen if this is ever going to be something we could actually test. But the weaker versions of this idea are quite plausible and, I think, offer interesting new methods for the field of counterfactual history. What does a 1911-cutoff model say when you push it toward the conceptual problems Einstein was wrestling with? Alternatively, what does the same model say about the potential likelihood of a World War? Those are tractable questions.

We can already see versions of this being explored by the Talkie project (for instance, “surprisingness” of future events, plotted above). If you ask Talkie about scientific concepts that emerged in the 1940s or 1950s, it gives you a rough sense of what the conceptual horizon looked like just before — and you can watch it grasping toward something it doesn’t quite have the vocabulary for.

3. New questions about genre, rhetoric, and "distant reading"

An earlier version of Talkie I tested was much less conversationally adept than the one that was released yesterday. What got it more “talkative” was post-training on etiquette manuals, letter-writing guides, and other books relating to socializing and conducts of conduct. These works provide the kind of text that allows you to extract “chatbot-like” habits without contaminating the model with modern data.

This is a sensible engineering choice. But it has a fascinating side effect, which is that Talkie's conversational persona is heavily shaped by the genre of its post-training data. The base model knows about the 1920s perfectly well; the chat persona, sculpted out of Beadle’s Dime Book of Practical Etiquette (1859) and the like, sits much closer to the late Victorian parlor.

This points to a research direction that I think is genuinely new. What if you used totally different texts for post-training? What would a Talkie post-trained on the transcripts of the Old Bailey Court records look like — an LLM whose conversational reflexes are shaped by the speech of accused criminals and witnesses in 18th century London courtrooms? What about apothecary manuals and herbals? Biographies of Romantic poets? Railroad conductor’s incident reports from colonial India?

Each of these, even if overlaid on the same base model, would produce a totally different voice, a different set of assumptions about what counts as a coherent question and a coherent answer. In short a different epistemology.

This is, I think, a remarkably cheap way — in terms of both compute and corpus size — to build a whole family of vintage models that capture different genres of historical experience rather than different periods. (If anyone is interested in collaborating on or funding this, by the way, please get in touch.)

4. Multi-agent historical simulations based on real archival sources

The most interesting move past the “great man” framing is to use vintage LLMs to simulate not famous individuals but plausible composites of ordinary people, drawing on the kinds of sources that are abundant for non-elite historical actors: probate inventories, parish records, court testimony, letters, account books, marriage records.

Imagine pulling from the legal, financial, and personal records of late 18th century France to construct a thousand plausible personas of ordinary French people — peasants, artisans, shopkeepers, day laborers, parish priests — grounded not in imagination but in real surviving documents. Then stage a debate among them: should the monarchy be overthrown? What patterns emerge? What kinds of arguments win in different demographic configurations of the deliberating group? The output wouldn’t tell you what really happened in 1789. But it would generate a structured speculation about the space of possible Frances.

Or take a famous trial — the Scopes trial, say — and re-run it with a different jury pool drawn from the same county and decade. Another variant on this idea: run parliamentary and congressional debates with personas constructed directly from the participants’ papers, speeches, and correspondence. Members of Congress and Parliament are unusual in that we have abundant documentary sources for them, even when they’re not famous.

We have the potential, in short, to create a thousand versions of the Smallville paper (which introduced the idea of LLM-based agents back in 2023) set in different historical eras.

Nor do these simulations have to include agents from the same era. A bit more experimentally, and outlandishly, one could imagine a multi-agent simulation in which a 17th century Galenic physician, an 18th century practitioner of traditional Chinese medicine, a 19th century quack doctor, and a 1950s-era century psychedelic researcher debate how to treat the same patient’s illness.

I find these possibilities super interesting, even if I’m not quite sure how to slot them into the ways that professional historians currently work.

Final thoughts

So what will be the outcome of all these experiments?

This is what I like best about this field: we truly have no idea. None of this has ever been tried before. It is completely open terrain, and I find it far more mind-bending and intellectual enriching to think through than the sort of topics that typically emerge in discussions of AI’s role in research or teaching.

Going forward, the important thing is to create an open source community and meaningful, sustained collaboration across the two cultures of STEM and humanities. I am already noticing new bridges cross those divides. I’m very grateful to have had a chance to work with Nick and Alec on historical aspects of this particular project. I would love to continue the conversation and explore collaborations with anyone who finds this topic interesting.

There will, no doubt, be a lot of false starts. But the emergence of an intellectually curious, not-for-profit, open source, humanistically-grounded community exploring historical LLMS makes me happy. Onward!

Leave a comment

Share

What doesn’t, at least at first, is that he was a talented visual artist. In fact, before he became a psychologist, William James dreamed of being a professional painter. He studied for several years in his late teens and early twenties under the painter William Morris Hunt.

Although none of William’s paintings appear to survive, a careful reader of his archive will find evidence that he continued to draw throughout his life.

Here he is, for instance, doodling on an envelope addressed to him from Geneva:

Readers of The Metaphysical Club, Louis Menand’s wonderful book about James and his circle, will recognize the image below as William’s sensitive drawing of his brother Wilkie while he was recovering from being shot during the Civil War.

The visual creativity of James is not just a clue about how his own mind worked. It’s also part of a larger shift in the culture of science during his generation: in the nineteenth century, design and the nascent world of big data came together, for the first time, to create the modern concept of data visualization.

Although James and his collaborators are rarely mentioned in discussions of the origins of data visualization, they actually played a very important role in shaping it. They were the consolidators and extenders of a new paradigm — the generation after the famous names in the field like William Playfair and Florence Nightingale.

The generation of William James came into adulthood in the Machine Age of the 1870s, ‘80s and ‘90s. Information buffeted the human brain like never before. And the new technique of data visualization pushed into new domains: the mapping of the mind, the sociology of race, and the pursuit of explicit political ends.

James is famous among historians of science for what Francesca Bordogna calls “boundary work” — he moves across disciplines and fields with a manic, restless energy that makes it hard to figure out exactly what he was.

As we’ll see, this included an unusual approach to visualizing mental activity that yielded significant firsts, including the first schematic of a neural network. Along the way, James also had important relationships with two pioneers of modern data visualization who are rarely put in the same sentence: Francis Galton and W.E.B. du Bois.

Francis Galton (center) was a sort of intellectual frenemy for James, initially a mentor and influence, then later — in his guise as a founder of eugenics and ardent imperialist — as an exemplar of the dangers of scientific hubris. Du Bois (pictured at right) studied with James at Harvard and was deeply influenced by his philosophy of pluralism.

What they shared is a conviction that drawing, diagramming, and composing images was not a decorative step added after the thinking was done. It was how the thinking got done. This seems to me a crucial point in light of new AI tools like Claude Design, which automates the design process (on which more below).

Quite a few people have pointed out that writing is a form of thought (my favorite entry in this genre is this 2025 essay by Derek Thompson). But it’s worth thinking more about what else counts as thinking too — specifically, the sorts of important, creative thinking we don’t want to accidentally mislabel as “drudgery that we are happy to let AI take over for us.”

Design, I would argue, is not drudgery.

The images below also remind us how handmade data visualization once was. W.E.B. du Bois’s visualizations from the 1900 World’s Fair are rightfully famous online, and they look great as compressed jpgs shared on social media. But it’s important to remember that these are large, hand-drawn, hand-lettered posters. These are not just the product of mental work but manual work, too.

That link between the hand and mind is harder to come by in a world where all research is digital, but it can be fun and important to access when doing serious research. When I was deepest in researching the history of 20th century psychedelic science for Tripping on Utopia, I filled up a yellow notebook with collaged images and primary source snippets from my archival research. I started it as a sideline, almost a hobby to distract me from what I saw as the “real” work of actually writing the book.

It turned out to be one of the most significant forms of research I did, precisely because it was so freeform and undirected. I started noticing links between different documents, and thinking more deeply about the motives (public, private, and even subconscious) of the people I was writing about.

The rest of this post is a gallery of some of the ways that James, Galton, and Du Bois visualized data along with some desultory commentary. I end by experimenting a bit with Claude Design to see what gets lost when we automate this type of exploratory visual thinking with data.

If you find it interesting, please consider subscribing.

Subscribe now

Share

William James

You’d be hard-pressed to find any reference to James in books about the history of data visualizations or design. But I think he deserves a page or two. For one thing, his Principles of Psychology (1890) contains what I believe to be the earliest ever visual representation of a neural network:

The most interesting data visualization in James’s work is this however, IMO:

James intends the image above as a representation of how consciousness “moves through” the process of uttering of a simple sentence over time. The numbers on one axis are showing moments in time, and the other axis shows the words being said or thought. The Joy Division-esque crest running through the middle is the changing attention we pay to each word over time.

It is really striking to me how much this looks like a computer rendering — but it’s from 1890! For this attempt at a sort of faux-four-dimensional modeling of thought, plus the neural network chart alone, I think James deserves a lot of credit for his data visualization.

There are other interesting images in Principles of Psychology, especially in chapter two which you can read here, although they are a bit more familiar.

For instance, here is a schematic of how a child perceives a candle flame:

And an early attempt to map which parts of the brain correspond to specific body regions and sensations:

Francis Galton

James read Galton closely, and cited him throughout Principles of Psychology — especially on questions of mental imagery and visual perception. Galton’s famous “breakfast-table questionnaire,” which asked hundreds of correspondents to describe how vividly they could picture the objects on their breakfast table that morning, was one of the first systematic attempts to gather data on subjective visual experience at scale. James was fascinated by it. He replicated versions of Galton’s mental-imagery surveys in his own classes and used the results in his writing.

Galton was also a pioneer of meteorology and produced many, many beautiful charts relating to weather phenomena — things like this:

But the visualization that appears to have most interested James (and which I find most striking too) is this jam-packed color plate about mental imagery and synesthesia which comes at the end of Galton’s book Inquiries into Human Faculty and Its Development (1883):

The big idea that linked Galton and James was not just that inner life could be represented as data, but that this data could be rendered as pictures.

Galton, obsessed as he was by measurement and averages, took this in a novel and discomfiting direction. Elsewhere in the same book that the beautifully strange image above comes from, Galton introduced composite portraits — photographs of criminals, tubercular patients, “types” of any kind — layered individual faces on top of each other to produce a kind of statistical average in visual form.

The project was inseparable from his invention of eugenics — a term which Galton coined in the same 1883 book. Galton’s visualizations encoded his conviction that human variation could be sorted, ranked, and ultimately improved through selective breeding.

In this work, to a large extent, the visual was the argument. Later, Galton even used his own photographs and biometric measurements as an example to encourage others to submit to biometric data collection. He wanted to visually model not just a new approach to data but a new mode of life in which all aspects of the human were reducible to data.

Reading his work from the perspective of the 2020s — with our turn toward “post-literate,” image and video based apps powered by mass data collection — it’s hard not to conclude that in this, he was hugely successful.

Share

W.E.B. Du Bois

William James, who prized pluralism and unconventionality, could not follow Galton on his journey into the world of averages and biometric data. But one of James’s most talented students, a young W.E.B. Du Bois, did. Sort of.

As the philosopher Colin Koopman writes, Du Bois took Galton’s obsession with measurement and flipped it on its head.

W.E.B. Du Bois — the first African-American to earn a Harvard PhD, and later one of the most important sociologists and civil rights thinkers of all time — studied philosophy with James in the late 1880s. Du Bois had a remarkably visual mind, with an approach to imagery that verged on the synesthetic.1

Du Bois absorbed James’s insistence that human experience could not be flattened into a single scale. But he also absorbed Galton’s conviction that the world could be made legible (and changeable) through creative use of data visualization. What he did with this combination was something neither of his predecessors had quite imagined.

Below are some of the charts that Du Bois made for the 1900 Paris Exposition Universelle — the same world’s fair that gave us the first public moving walkway, Rudolf Diesel’s engine, and the Art Nouveau métro entrances that still dot Paris today. Du Bois was there as the lead curator of the “American Negro Exhibit,” a small pavilion aiming to show a European audience what Black Americans had made of themselves in the thirty-five years since emancipation.

The roughly sixty charts he produced, hand-drawn and hand-lettered on large poster boards by Du Bois and his students at Atlanta University, are some of the most arresting data visualizations ever made. Like Galton’s charts, they visually plot the result of careful statistical investigation using idiosyncratic and arrestingly odd formats; also like Galton, they are using data visualization to pursue political goals. But where Galton leveraged his visualizations to argue for the lasting, ahistorical superiority of people who resembled him (Protestant, British, scientific, and from “good families”), Du Bois was charting the rapidly changing role of Black people in American society.

Change over time runs through virtually all of his charts from 1900, a few of which are reproduced below:

Subscribe now

Claude Design

Writing this post got me thinking about a question which we are figuring out the answer to in the 2020s: by automating away the design process, do we also cede away our ability to use design itself (and not just writing) as an act of thinking?

Using Anthropic’s new Claude Design feature, I wanted to see what happened when I asked Claude to make something in the style of these nineteenth-century visuals. Here’s what I got for a Du Bois pastiche (my prompt is available at this footnote):2

Asking for Galton-style charts yielded a similar stew of serif fonts and faux-aged paper effects, but in more of a Victorian bric-a-brac style which, on closer inspection, adds up to nothing much at all:

These failed attempts to capture a distinctly Victorian data visualization style should not be surprising. Claude Design is optimized to make slides and other data visualizations that suit modern sensibilities, and we can guess that there wasn’t much Du Bois or Galton in its training data. But aside from the by-now-familiar fact that AI models tend toward the median and the expected, there is the larger absence here: these charts don’t have a perspective.

James, Du Bois, and Galton, different as they were, had one big thing in common, and this thing is part of why we are still thinking and debating them today: they were deeply idiosyncratic, deeply personal. You see in their visualizations not just raw data but an odd, distinctive intellect at work. And more importantly, an intellect working through an argument that is both mental and physical. They are instantiating the act of thought in the mechanical work of their hands.

Claude Design is appealing for those of us (like myself!) who find themselves having to make fiddly tweaks to Powerpoint slides. It is, in this sense, just one more step in a process of design automation and abstraction that has been happening for centuries: from hand-lettered to printed book, from drawing to photograph, from scissors and glue to “desktop publishing,” and now from humans designing things to humans telling AI designers what to make.

As the act of thought gets more and more removed from the union of human mind with human hands, the new ideas, the palpable jolt, that you get from designing something by hand gets rarer, and thus more valuable.

This is why I am going to be making more scrapbooks and weird collages on yellow legal pads as I embark on my new book project about the Machine Age.

After all, what could be more human?

Subscribe now

Leave a comment

With that spirit in mind, and of course with current events in mind as well, I thought I’d share some photographs I took during my visit to Iran in September of 2017. I was there to attend the wedding of my sister-in-law. Afterwards, we took the opportunity to travel with my wife’s family in a big loop down from Tehran to the historical capital of Isfahan, then southeast to Shiraz (Iran’s wine country), and then back up to Tehran via the Great Salt Desert and the amazingly well-preserved Silk Road city of Yazd.

I am not one of those historians who thinks that learning the history of the world automatically makes you a better person. But I do agree with Twain that there is something fundamentally good not just about traveling but about exposing yourself to the cultural productions of societies outside your childhood experience — via words, via speech, and via images.

I return to these photos with feelings of sadness at the thought that some of these places are or may soon be destroyed by a hideously counter-productive bombing campaign which further punishes a proud people already brutalized by the Iranian regime. But also with hopefulness, as I am reminded of the tiny details like the American flag displayed alongside the other flags of the world in a Yazd hotel, or the fact that Orwell’s 1984 was one of most popular books being sold on Tehran’s street corners. I think about the warmth of the complete strangers and extended family I met in Iran and the fundamental goodness of humans, which, despite all the darkness in history, is a fact of life that I consider immutable.

Note: I am releasing these photos into the public domain under a non-commercial license. Feel free to use them in any non-commercial use cases, such as Wikipedia.

Tehran

The view from my mother-in-law’s apartment in northwest Tehran, with Milad Tower, a focal point of the city’s urban core, visible in the distance.

Tehran’s architecture reminded me a bit of Athens or Istanbul with an interesting mix of traditional and modern forms. What surprised me was how leafy and pleasant the streets were.

Driving in Tehran is an experience. With its high rises and urban sprawl bisected by freeways, that aspect of the city reminded me a bit of Los Angeles or São Paulo.

I was deeply charmed by Tehran’s bakeries, which are abundant, much beloved by the populace, and home to lots of original creations like this poodle cake.

The streets around the University of Tehran were fascinating to walk through because nearly every corner had vendors selling books, including banned ones. I noticed a lot of works of the great Russian novelists (especially Doestoevsky) as well as a lot of 20th century British and American fiction in translation — think Hemingway, Fitzgerald, and Orwell.

Here we see The Sound and the Fury, Memoirs of a Geisha, and 1984. The latter was probably the most frequently-sold book that I spotted in the entire neighborhood. The title at top is a Farsi translation of a book by the Egyptian feminist writer Nawal El Saadawi about female genital mutilation and the oppression of women called The Hidden Face of Eve.

The samizdat booksellers around the University of Tehran also sold posters and magnets, including more than a few which I remembered from dorm room walls when I was a student.

The following day, I caught some Iranian state TV with my mother-in-law and realized that we were in fact watching a rather well-done recreation of Breaking Bad. It was the famous “Fly” episode. Here’s the Iranian version of Walter White, gazing scornfully at the Iranian version of Jesse as they struggle to make sufficiently pure crystal meth:

Tehran’s Museum of Contemporary Art, built under the Shah just two years before the 1979 Revolution, is famous for having one of the world’s best collections of 20th century modernism, though it’s rarely displayed. There was however a Giacometti sculpture casually striding by outside the cafe.

The building itself is a wonderful example of 1970s Brutalism adapted to traditional Iranian motifs like high banks of windows that evoke Windcatcher towers. When I was there, the exhibited art included this system of mirrors that cast geometric patterns of light. It is a really extraordinary place.

Tehran’s vernacular residential architecture is quite interesting. Here is a set of outdoor staircases in a fairly typical mid-rise building.

The Shah’s palace in Tehran was one of the more moody and haunted-feeling places I’ve been. It preserves, seemingly more or less intact, the artifacts of the Iranian royal family on the eve of the 1979 Revolution, complete with faded 1970s-era recliners, fluorescent lighting, and other accoutrements of an aging couple. The grander, more public-facing halls and ballrooms feature paintings from Iran’s national epic, Shahnahmeh.

Isfahan

Traveling south from Tehran for about five hours, you reach Isfahan, which was the capital of the Persian empire under the Safavids and a major destination on the western edge of the Silk Road. Indeed, to a striking degree, it still feels like a city on the Silk Road, only now with lots of LED lights.

Walking through the vast central square (a UNESCO World Heritage Site) is hugely evocative because it has barely changed since the 17th century. I noticed a group of Chinese businessmen walking through the stalls while French tourists haggled nearby, and it occurred to me that this demographic mix, too, was unchanged since the 17th century!

The image below is maybe my favorite photograph I’ve ever taken — it was 10 pm, a hot September night, and the girl on rollerblades skated past the boy eating an ice cream, gliding onwards to the central square.

Jameh Mosque, which adjoins the square, is the most extraordinary example of Islamic geometric tiling I’ve ever seen. It is literally jaw-dropping, and photos like this don’t really do it justice:

Ali Qapu Palace, one of the imperial seats of the Safavid Emperors, looks out over the square with a commanding view. But more beautiful, I thought, was the Chehel Sotoun (“Forty Columns”) royal complex, which had both extraordinary architecture (more amazing ceilings) and some richly mysterious and evocative early modern wall paintings.

The ceilings were a mixture of inlaid/painted wood and cut glass mirrors.

Paintings throughout the complex (some of which were damaged in recent bombing) show vivid scenes of Abbas the Great and his son doing such things as entertaining the Mughal Emperor, defeating the Ottomans in battle, and drinking copious amounts of wine.

A detail of dancers and musicians in the painting of Abbas II entertaining the King of Turkestan. I am guessing that it may be Yalda night due to the prominent watermelon, but I might be wrong.

I didn’t take this photo - it’s from Wikipedia - but it gives a good sense of the scale and setting of this extraordinary painting:

I was struck by this painting showing a barefoot woman relaxing on grass while drinking from a half-empty glass pitcher of red wine:

In addition to the wine-drinking on display throughout the murals in the complex, there is also at least one detail which may depict a truly risqué scene for any early modern artist: two women who may be depicted as lovers. (For more on this painting and the culture that produced it, see Afsaneh Najmabadi’s Women with Mustaches and Men without Beards).

The music room of the complex also featured cut-outs in the shape of wine vases, which I thought was a delightful touch, and also probably had an interesting effect on the acoustics of the room:

Finally, I was greatly struck by the European style paintings which appeared to depict “firangi” (Western European) nobles wearing clothing in the style of the mid-seventeenth century, seemingly influenced by Italian and Northern Renaissance art.

Shiraz, Persepolis, and Yazd

The area around Isfahan was a stark and arid landscape, heavily mountainous and quite desolate in parts. Continuing southeast to Shiraz brings you to a truly ancient region which was, in effect, the eastern edge of the Fertile Crescent. Shiraz is apparently first mentioned in Elamite tablets circa 2,000 BCE. It has been a wine-producing region for longer than that.

The local wine, though illegal, is still produced and can be purchased from illicit wine-sellers relatively easily. I unfortunately didn’t take a photo of it, but I remember it had been poured into a Coca-Cola bottle and tasted very much like Spanish sherry.1

The brilliantly-colored stained glass windows of the Citadel of Shah Karim Khan, founder of the eighteenth-century Zand dynasty, in Shiraz.

Walking through downtown Shiraz at night, I happened to pass right in front of a wedding party. Here is the bride arriving at the reception:

Not far from Shiraz is the world-famous ancient Persian palace complex of Persepolis, as well as the less-famous but equally interesting Naqsh-e Rostam, the burial site of four Achaemenid emperors including Darius the Great.

As an archaeological site, Persepolis is incredible. It is however not well maintained and has very little infrastructure. I hope that in some better time to come, a future Iranian government is able to create a museum here to rival its counterparts in Egypt and Greece.

Naqsh-e Rostam, the ceremonial and burial site of the ancient Persian emperors, is within walking distance of Persepolis.

I remember spending at least a week studying the site in an art history class I took as an undergrad, and I can still remember my professor — the witty and erudite David Castriota — describing this Sassanian bas relief as the first depiction of the new technology (mounted, armored warriors on horseback with lances) that would evolve into the medieval knight. It was amazing to see it in person:

It was also fascinating to be able to see, firsthand, the depiction of the humiliation of Emperor Valerian by the Persian emperor circa 260 CE. Valerian was the only Roman emperor ever to have been captured in battle.

I’m quoting here from Wikipedia, which summarizes one Roman historian’s account of Valerian’s ultimate fate:

Valerian was subjected to the greatest insults and humiliation by his captors, like being used as a human footstool by Shapur when mounting his horse. According to this version of events, after a long period of such treatment, Valerian offered Shapur a huge ransom for his release.

In reply (according to one version), Shapur was said to have forced Valerian to swallow molten gold… and then had Valerian skinned and his skin stuffed with straw and preserved as a trophy in the main Persian temple.

Finally, we wrapped up our roadtrip in Yazd, a city that Marco Polo visited and seems to have greatly enjoyed. I don’t think I’ve been anywhere besides possibly Oxford or Sienna that is so well-preserved as a premodern cultural site. The architecture both in the city and in the extremely arid region around it still preserves the ancient Persian architecture of qanats and Windcatcher towers designed to channel water and cool air through one of the hottest places on earth.

Traditional architecture and an “ice house” (yakhchāl) outside Abarkuh, two hours from Yazd:

And a panoramic view of Yazd itself. Historically, the remote desert city was a stronghold of Zoroastrianism. The Yazd Tower of Silence had a continuous record of Zoriastrian-style “sky burial” (exposure of dead bodies to the elements) for roughly three thousand years, ending when the practice was banned in 1963.

You can still visit the Fire Temple of Yazd which claims to house a sacred flame which has been continuously lit since 470 CE. It is the highest grade of fire in the Zoroastrian religion, described by Wikipedia as a flame combined from “sixteen different sources, including the fire created by a lightning bolt.”

Zoroastrians are few in numbers in today’s Iran (numbering in the low tens of thousands) and face religious persecution, but they still exist as an active religious movement. It was fascinating to see a place where they remain a large minority.

Yazd architecture is the most skillful adaptation to desert/arid conditions I’ve seen. The use of windcatcher towers, domes, and thick earthen walls genuinely works wonders to cool down interior spaces. Yazd lies on the edge of the Lut Desert, which is the site of the highest-known surface temperature on planet earth (70.7°C, or nearly 160 degrees Fahrenheit!) Yet walking through Yazd’s shady alleys and its thick-walled buildings, you would never know it.

Flags of the world at a Yazd hotel:

Just to reiterate, all photographs here aside from the one I flagged above were taken by me in 2017, and all are hereby released to the public under a Creative Commons Non-Commercial License.

As an American, I hope someday to be able to visit Iran again in better times, and I hope these images help give a sense of the country, the people, and its history.

Weekly Links

• “Researchers find 3,500-year-old loom that reveals key aspects of textile revolution in the Bronze Age.”

• My wife Roya Pakzad wrote about one element of our 2017 trip to Iran not mentioned here, but one that also stuck with me: computer-assisted government surveillance of whether women wear hijabs, and the omnipresence of surveillance in general in Iranian life. There’s a reason why I saw 1984 being sold on so many street corners.

•For those in the Boston area: I’m going to be speaking at two events at Harvard this month. The first is “Psychedelics and the Specter of Mind Control” on April 6 (link for more info). The second is at the Psychedelic Intersections conference on April 11 (link). If you’d like to meet up while I’m in the area, please get in touch.

Leave a comment

Subscribe now

Share

I love being a parent. The thing I find most fascinating about the experience is how it throws a mirror not just on one’s own childhood, but on all of human nature. It’s an obvious point, but one that I never thought about before having kids: all newborn babies are always the same, everywhere. And then, slowly but surely, they become not the same. As cultural and family influences accumulate like sedimentary layers in these tiny personalities, you can see nurture reshaping nature in a deeply embodied, physical way.

One thing that Nava has got me thinking about is the deep history of gesture. Gesture exists on the borderland between nature and nurture. Certain physical actions are so universal that they communicate effectively not just across human societies but across species: bared teeth, for instance. Other gestures encode more specific cultural patterns, but do so in a way that is more expansive than language.

Subscribe now

A few months ago, for instance, I watched my mother-in-law (who was born and raised in a village in northern Iran) teach Nava how to knock on wood for good luck. I hadn’t realized this was so widespread a practice until I checked Wikipedia and found that variants exist in Bulgaria (chukam na dǎrvo), Georgia (kheze daḳaḳuneba), Indonesia (amit-amit jabang bayi), Norway (bank i bordet ) and some two dozen other countries.

Why? Could we trace embodied behaviors through history just as anthropologists can trace, say, legends or etymologies relating to magical blacksmiths or slaying serpents?

What I found is that written evidence for knocking on wood (and for the history of gesture as a whole) is remarkably sparse. In the Penguin Guide to Superstitions, folklorist Steve Roud offers a skeptical take on the provenance of knocking on wood due to the near total absence of written records about it:

The two most popular explanations of origin are that the belief goes back to pagan times when we believed in tree spirits, or that we are invoking Christ’s protection by referring to the wood of the Cross. The former is nothing but guesswork, based on the conviction that all superstitions must be ancient, and it has the usual problem of spanning thousands of years with no evidence at all of its existence, or, for that matter, any evidence that ‘we’ ever believed in tree spirits.

Roud then ventures a theory of his own: the custom, he argues, probably developed from a children’s game called Tig-touch-wood that was popular in early 19th century England.

If true, the practice of knocking on wood for good luck is actually surprisingly new.

The “Knocking on Wood” Wikipedia page mentions Roud’s conclusion, and this claim of a 19th century origin has circulated online as the final say on the matter. If you go back to the original, though, Roud’s proposal actually ends on an ambivalent note: “Before this theory [of 19th century origins] can be finally accepted, however, an examination of the history of European forms of this custom would be advisable.”

This is where historical data visualizations — of the type Claude Code can now produce on cue — can come in handy. I downloaded a scientific paper about knocking on wood along with the crowd-sourced Wikipedia list, then provided it to Claude Code and asked it to plot the data on a three.js globe. Here’s the interactive version and the GitHub page.

What we see is that there are in fact two entangled traditions of “knocking on things for good luck”: touching iron, and touching wood. We also find that they are widely distributed, but also have a pretty clear cluster around the Mediterranean and Europe.

And, although it is difficult to find references to knocking wood for good luck before the nineteenth century, it turns out that this is not the case for touching iron — the “tocca ferro” of Italy, shown above.

The upshot: I struggle to imagine a process whereby an obscure English children’s game somehow inspired gestural customs in Georgia, Egypt, Russia, and Iran within less than a century. It seems more reasonable to me that these customs might instead originate in an ancient Mediterranean tradition of apotropaic magic, or charms intended to combat bad luck.

For instance, here’s a snippet from an early modern translation of an Ancient Greek dream interpretation text, the Oneirocritica of Artemidorus:

This put me in mind of a recent New Yorker article by the anthropologist Manvir Singh. The article is about the efforts of linguists and folklorists to reconstruct the Proto-Indo-European mythology which links folk tales and gods from India to Ireland. The serpent-slaying storm god, the Sky Father, the Divine Twins: these figures mutated and multiplied across cultures but retained their essential forms.

Might the linkage between “stuff of iron” and “surety to those that are in doubt,” likewise, have an ancient pedigree via gestural traditions? Notable here is the fact that one of the oldest known fairy tales involves iron-working: “The Smith and the Devil.” A study from 2013 concluded that this story template, which associates iron-working skill with magic, was traceable back at least 6,000 years.

Then again, maybe it all really does go back to Tiggy-Touch-Wood. The truth is that we really do know shockingly little about the history of gesture.

Given recent developments, that should probably change.

Things humans don’t write down

The history of most human gestures is never written down, precisely because it is something we learn before literacy or memory and thus something we take entirely for granted.

Another example: why do so many languages implicitly view the right hand as good (for instance, associating the direction right with the ethical concept of rightness) and the left hand as bad, maladroit, sinister?

I found this article on the subject, and decided to turn that data into a visualization, too.

The author of the article, the German anthropologist Wulf Schiefenhövel, writes:

I find it plausible to think that early humans began to observe, to feel the difference between right and left, and to ascribe qualities like “clumsy,” “awkward,” “crooked,” and “tired” to the less dexterous hand (it is interesting that these very terms still show up prominently in today’s modern languages) and correspondingly positive qualities to the right hand preferred by the majority. This process, intertwining emotion and cognition, can well be expressed in the terms of embodiment... The semantic values with which the terms for left and right are charged in almost all the languages examined for this survey could have their origin in this very process: embodiment turned into words.

Singh’s New Yorker article ends by asking what these sorts of searches for common origins tell us: “that our wild, warring species shares something irreducible at its core.”

But they also point to something we don’t share with the other new pseudo-intelligences emerging today: machine intelligence.

So many of our latent assumptions about selfhood, reality, and consciousness flow from our embodiment as physical beings and from the deep histories of gesture and other forms of implicit knowledge.

Why do I, my mother-in-law, and Nava all knock on wood? None of us really know. Perhaps it’s a legacy of the Bronze Age; perhaps it’s a meme from Victorian Britain. What is certain is that it’s not something a robot with an LLM-based brain is going to do habitually, just as robots will never share in mental frameworks deriving from quirks of our physical architecture, like handedness.

The history of gesture, in other words, bears directly upon the question of AI alignment. Humans across cultures and times intuitively maintain a set of semantic and ethical judgements rooted in our physicality, our learned and instinctive gestures, and the affordances of the natural world. There is just something “knockable” about wood. LLMs don’t have childhood memories of jumping over cracks in pavement or their grandmother teaching them gestures. Moreover, such things are not really in their training data either.

For this reason, machines without our embodied experience may think they are aligning their behavior with human norms, but they will never be able to feel the physical basis for some of those norms.

Unless, as with Nava, we teach them.

Share

Can world models touch grass?

I have been thinking a lot lately about “diachronic AI” and “vintage LLMs” — language models designed to index a particular slice of historical sources rather than to hoover up all data available. I’ll have more to say about this in a future post, but one thing that came to mind while writing this one is the point made by AI safety researcher Owain Evans about how such models could be trained:

One challenge is having enough training data. Another is that the training data needs to be free of contamination. For a model trained up till 1900, there needs to be no information from after 1900 that leaks into the data. Some metadata might have that kind of leakage. While it’s not possible to have zero leakage - there’s a shadow of the future on past data because what we store is a function of what we care about - it’s possible to have a very low level of leakage, sufficient for this to be interesting.

You can include multimodal data like images. There’s something strange about including images when going back to Roman times or 1700 because while they had texts, they didn’t have digital images. However, this is acceptable for some purposes. You’d want to avoid leaking information that could only be known in the present. You could include things people at the time could see and experience themselves. For example, there may be no anatomically accurate painting in Roman times of a bee or an egg cracking, but you can include such images because people could see such things, even if they weren’t part of their recorded media. You could also have pictures of buildings and artifacts that we still have from the past.

That last observation, about training vintage language models on images of the physical world, is, I think, a fascinating one.

Consider Nava. Another of her favorite phrases is “I seein’ it!”

And reader, let me tell you: Nava is, indeed, seeing it. Really seeing it.

Whatever it happens to be, Nava is (so it seems to me) seeing it with a freshness of perception, a joyful richness of observation, that my 40-year-old eyes simply cannot equal.

I suspect rather strongly that premodern people, too, experienced the physical world more acutely than I do, simply because my brain has been stewing in dopamine-spiking stimuli for four decades now, rarely pausing to touch the proverbial grass. Don’t get me wrong: I like the outdoors a lot. But my waking existence is spent mostly indoors in highly artificial spaces, whereas humans have typically spent the vast majority of their time outdoors in nature. Consider my ancestors, who, as best I can tell, were primarily illiterate peasants and smallhold farmers. They spent a significant proportion of their waking hours literally touching grass.

I wonder how much this world of gesture and of the senses — the things we see, feel, hear, smell, and taste — is implicitly embedded in the textual corpora we humans produce and in the value judgements we derive from it. When we hold truths to be self-evident, is a sense memory of the physical act of holding somehow contributing to the meaning (is it, in fact, helping to “carry” or “support” that meaning?)

Owain Evans’ idea of feeding a historical LLM non-anachronistic images is, I think, well worth doing. But it’s also worth expanding on further. Would it be helpful, when training a historical LLM, to simulate dream imagery based on premodern themes? What about audio of birdcalls, which were far more prominent in the audioscapes of premodern people? What about taking it on a walk through the woods?

Would an AI model trained on such things be not just more authentically premodern, but more authentically aligned with implicit human values?

The converse is also worth asking — whether simulating artificial environments (for instance a 3d representation of a Youtube video) might have unintended negative consequences. Fei-Fei Li’s startup World Labs, which aims to make the leading “world model” — an alternative to language models based on tokenizing physical space rather than words — recently raised a substantial amount of money. As consumer-facing robots become more plausible, the business case for such a model is obvious. But what physical spaces are “world” models actually being trained on? The contemporary physical environment, sound-proofed, plastic-coated, and artificially-colored, is radically different from the environment that Homo sapiens evolved to excel in.

This sort of thing is why I think historians need to be more active in technical discussions and decision-making about emerging technology. Everything about our current world is different from the premodern world that our ancestors inhabited. The past truly is a foreign country. But we carry fragments of that foreign world with us in our physical selves, in the gestures and other implicit knowledge we teach our kids. We take it for granted that there are aspects of being human which are never written down and which are unknowable unless you experience them.

AI systems, locked in their data-worlds of text or simulated 3D environments, never even come close to this implicit knowledge. Not yet, at least.

Share

Weekly links

• “What not reading does to your writing.” Lincoln Michel is one of my favorite chroniclers of life as a professional writer, and the point he concludes on here really is the simplest, most effective advice I can give to anyone who wants to write better: “The best way for this author to improve their writing is simple. They should read a few good books.” I notice it in myself. When I am not actively reading fiction, my writing gets more flat.

• “Archaeologists Say They’ve Identified Traces of a 2,000-Year-Old Love Note Still Etched Into a Wall in Ancient Pompeii.” (Smithsonian).

• Every time I teach world history, I make a point of showing things like the above to my students and reading them Philip Larkin’s “An Arundel Tomb”:

Time has transfigured them into

Untruth. The stone fidelity

They hardly meant has come to be

Their final blazon, and to prove

Our almost-instinct almost true:

What will survive of us is love.

Leave a comment

Subscribe now

I agree with much of it, though parts were overstated and other parts were clearly submarine marketing. But what was most striking to me wasn’t the content, but the obvious fact that large swathes of it were written by AI. I suspect that the “AI slop” quality was actually part of why it did so well. In 2026, we must face the fact that slop is well-liked. Perfectly formatted writing, with statistics ready at hand, copious length, a chipper tone, and lots of antithetical phrasing — the infamous “it’s not that; it’s this” formula — is showing up everywhere for a reason. People might say they hate it, and I’m sure many do. But the truth is that many other readers clearly enjoy this style of machine-generated prose.

That revealed preference worries me more than anything else.

I learned to read and write quite late (worryingly late, by the standards of millennial parents like myself). At age 8, I was barely literate, way behind classmates. Something clicked at age 9. I read The Lord of the Rings twice before graduating to The Silmarillion. Fifth grade was the year of Michael Crichton. By age 12 it was Virginia Woolf and Tolstoy — both of whom I barely understood. But I loved the unfamiliar sound of the words and the unfamiliar sensations those words evoked. I loved feeling like I could transmute some of those feelings and images into my own words on a page (I even started a novel, which had something to do with an ancient Egyptian goldsmith searching for his father).

Writing became something I was self-evidently good at. It was the one thing in my life that truly opened doors for me.

My first real job after college in 2007 was as an assistant for a crusty New York labor lawyer who wrote exclusively on legal pads and called me “college boy.” He had an incredible memory and an agile mind, but he wasn’t much of a prose stylist. That task was offloaded to me. Every morning, the lawyer would toss yellow legal pads on my desk. It was my job to proofread and “punch up” these rough drafts. I would toil happily for a few hours in Microsoft Word, usually earning his gruff approval by around lunchtime. Then I spent the rest of the day reading about mellified man and cocaine mummies and the like on Wikipedia. I did well enough at this job that it felt like a viable, permanent career.

Needless to say, that career is gone now. Gemini is at least as good at transcribing a lawyer’s longhand as I am, and Claude will undoubtedly output more perfectly formatted legal footnotes than I could ever hope to achieve.

So what is left for human writing?

Subscribe now

AI is not coming for regulated jobs in meatspace

As a historian and teacher — distinct from my identity as a writer — I don’t experience much of the existential dread that essays like “Something Big is Happening” seem intent on getting me to feel. As I wrote here, historians rely on non-digitized archives and tacit, often physically-embodied knowledge. That is a very good thing when it comes to resisting replacement by the likes of Anthropic’s newly released Sonnet 4.6 model, which is remarkably good at “computer work” tasks.

When I think of AI-proof jobs, I think of people like electricians, plumbers, or the surf instructors of Santa Cruz. But I also think about history professors and anyone else whose output includes some combination of in-person engagement and travel-based or otherwise embodied work in a regulated industry. No less than a surf instructor, historians are performing physical services in the real world, although we don’t tend to think of it in those terms. We are going into parish church basements to read through baptismal records, finding weird old non-digitized books in rare book shops, piecing together who called Margaret Mead on a certain day in 1954 by reading through her secretary’s notes. These are not the everyday tasks of a historian’s life, but they are the kind to things we might do, say, once a week. Couple that with twice a week in-person classroom time, and I simply flatly disagree with anyone who thinks this combo will be replaced by a Sonnet 4.6-type model, no matter how good it gets at creating Excel spreadsheets, translating Latin, or explaining linear algebra.

Anyone who has led a class discussion — much less led students on a tour of Egypt or Okinawa, as my colleagues regularly do — knows that there is a huge gap between solo learning online and collective learning in meat space. And although it is possible to imagine a humanoid robot instructor with a brain powered by some future OpenAI model, it is very, very difficult for me to imagine such a thing being both popular with students and permissible within regulatory frameworks.

The claim that robot teachers will surely happen someday is one thing, and I can’t disagree with it. But culture and regulations change far more slowly than technologies. Those in the technology sector who are predicting time horizons of a few years for these changes are, I think, confusing the pace of technical change with the pace of social change. The image above, produced in 1890s France, was imagining machine-generated education by the year 2000. For my part, I can picture educational systems dominated by robot teachers or even direct-to-brain interfaces by the year 2100, and possibly a bit before that.

But it’s worth thinking about how surprised the creator of the image above might be to learn that in my classrooms in 2026, I am still using a blackboard and chalk.

So although I am very sure at this point that AI will be transformative for a broad range of human jobs and interests, I actually think historians and teachers are going to be fine in the medium term of the next couple decades.

Share

AI is coming for many aspects of “pure” writing

So it may go for historians. What about writers?

There I’m less sure. Certainly, the bottom will drop out from the lower rungs of eyeball-grabbing written content, from pulpy novels to historical click-bait writing online. Probably, it already has. That GPT-4 level models could passably imitate the imitators of, say, Colleen Hoover or Dan Brown was known to anyone who experimented with them.1

These experiments could be interesting, but could never be mistaken for good writing. The thing is: the most recent Anthropic models actually write quite well.

For this reason, I find myself deeply sympathetic with software developers lately. Their struggles are my own, just from a different angle: what happens when something you are good at, a skill which is relatively rare and relatively valued, suddenly becomes commoditized? For both professional writers and professional developers, this is not a question but simply a daily reality.

Like Andrej Karpathy and others, I find myself deeply energized by the potential of these new tools, while also wondering what their mass adoption will do to everyone’s brains (Karpathy celebrates the “phase shift” in capabilities that occurred over the past few months, but adds: “I've already noticed that I am slowly starting to atrophy my ability to write code manually”).

On the one hand, I feel a real sense of loss for the world of early 2000s New Yorker style long-form writing and criticism, the world of The New York Review of Books and Lapham’s Quarterly and the amazingly good Dennis Johnson novel I’m currently reading (Tree of Smoke).

Yet I am taking real delight in being able to create truly novel — though sometimes interesting only to me — hybrids of software, history, and writing. This Henry James simulator from last fall was, in truth, just the tip of a somewhat obsessive Claude Code iceberg for me.

Some recent examples of what I’ve been working on:

Universal History Simulator

Drawing on my earlier work on historical simulation using ChatGPT prompts, I made a full-on history simulator (GitHub here) which randomly drops you in different historical eras and cultures, with pixel art graphics augmenting an LLM-generated narrative engine which is grounded in real primary sources and a rather elaborate system for generating at least somewhat accurate historical context.

I’ve already used in this in my teaching. While it’s imperfect, the mechanic of allowing a user to generate a historical simulation based on text input — like “I want to be a salt miner in a frontier province of Ancient Rome” or “1930s Istanbul bike mechanic” or “Aztec farmer with a troubled past” or whatever else you can think of — is fascinating to me. More on this in a future post.

MKULTRA

I tried to make a game called MKULTRA (playable version here, GitHub here). This game is based directly on primary sources relating to George Hunter White’s now-infamous CIA-funded work on dosing unwitting civilians with LSD which I had gathered for my book Tripping on Utopia. As with History Simulator, the text responding to player input here is generating by Google’s Gemini 2.5 model. The difference here is that it gets more and more outlandish and surreal as you, playing as a procedurally generated person in 1950s San Francisco, are dosed with psychedelics. As your “trippiness” meter increases, the UI itself begins to break down and change. Compare the relatively sober player view at left with the altered one at right:

Premodern Concordance

Premodern Concordance (live page here, GitHub here) is the most promising use case for LLMs in historical research that I’ve come up with so far, and something I am currently seeking grant funding for along with my historian colleague Mackenzie Cooley. As a proof of concept for this larger project, I fine-tuned an embedding model to analyze premodern medical, scientific and natural history texts, searching for non-obvious links between concepts that fuzzily match up across languages and eras. For instance, if you click “piedra Bezaar” on the page below, you end up at a cross-linguistic concordance page for Bezoar stones which tracks mention of them across texts in Portuguese, Spanish, Italian, French and English.

There’s actually several more projects I’ve been working on, but I won’t bore you with long descriptions. In brief:

• I made an apothecary simulator based on 17th century texts, where you treat patients in colonial Mexico City and potentially get in trouble with the Inquisition.

• I made a historical persona generator for use in history classroom activities, roleplaying, and the like, as a spinoff of the larger history simulator project.

• I made a literary canon explorer which contrasts mentions of Victorian authors on Google Books versus their page views on Wikipedia.

These are all things which, in the past, I would have explored through the written word. Now, I’m exploring them through Claude Code, which I use to write the TypeScript code that summons them into being, and Google’s Gemini language models, which I use to populate them with “speaking” characters for the simulation aspects.

I have greatly enjoyed the challenge of learning how to build these games and tools over the past year. But it also seems to me that the work I have put into them — the enjoyment I have gained from them — is, in itself, a somewhat worrisome leading indicator.

It’s not just that AI slop is replacing some of the lower-rung forms of fiction and prose. It’s that the audience share for writing as a whole will be increasingly displaced by interactive “writing-adjacent” things like the MKULTRA game or interactive text-based simulations of life as a Florentine toymaker in the year 1500, or whatever else you can imagine. Gemini’s outputs for these scenarios read like a bad attempt at historical fiction. But the dynamic nature of these sorts of tools — their “choose your own adventure” quality — is genuinely new. That dynamism and hand-tailored quality will, I suspect, be more compelling for many than simply reading mediocre novels about MKULTRA or Florentine toymakers.

Subscribe now

Writing is thinking in public

At the end of the day, I didn’t enjoy any of the projects described above as much as I enjoyed researching and writing Tripping on Utopia. Or as much as I am enjoying starting work on my new book, which, not coincidentally, is about how William James his world reacted to the automation of life and thought in the first machine age.

I miss the obsessive flow you get from deep immersion in writing a book. Such work has none of the dopamine spiking, slot machine-like addictiveness of Claude Code — the rapid progress of typing two sentences into a terminal window, watching Opus 4.6 build a new feature over the course of ten minutes, and then seeing it come to life on a screen.

But the very lack of progress that defines serious writing — the staring at a blank page, the rewrites, the chapters you throw away — feels earned (and therefore intellectually meaningful) in a way that Claude Code simply never will.

“Cognitive debt” is a term that software engineers have recently adopted to describe the disconcerting feeling of losing touch with the ground truth of what is actually happening in one’s codebase. As Simon Willison recently put it: “I've been experimenting with prompting entire new features into existence without reviewing their implementations and, while it works surprisingly well, I've found myself getting lost in my own projects.” It is worth noting that this term didn’t originate in the world of code. It was coined last year by MIT researchers to describe “cognitive costs” from relying on ChatGPT to write essays. The frontlines of cognitive debt, in other words, are in the realms of both writing and coding.

I have about 50 ideas sitting in my Substack drafts folder. It is tantalizing to imagine that I could simply open a Claude Code terminal window, direct it to these drafts, and tell it to output a year’s worth of Res Obscura posts. This, of course, is not a hypothetical but simply what a a lot of people are doing right now.

But if you do that, what is the point? What was I working toward when I was 10, or 20?

The work is, itself, the point.

My quasi-obsession with vibe coding over the past year or so — see above — made that less clear to me for a time. I was never tempted to use AI to write this blog (something I have never done and never will). Rather, what did tempt me was the illusion of productivity and innovation that one gets from using AI to create custom software. At times, it has felt more productive to produce digital humanities projects and historical educational games like Apothecary Simulator than to put words on a page. As I write that now, I feel some sense of shame, as if I am confessing to having become addicted to junk food or gambling.

Is Claude Code junk food, though? I also spent the past year learning about data taxonomies and single stores of truth and embedding models, and although I have barely written a line of code on my own, the cognitive work of learning the architecture — developing a new epistemological framework for “how developers think” — feels real. I also think that at least a few of my projects, especially the Premodern Concordance, are useful tools that I will be returning to and learning from for years.

And yet: I miss pre-AI writing and research. More than that, I miss thinking through writing in public.

The novelist Martin Amis once wrote: “style isn’t something you apply later; it’s embedded in your perception, and writers without that freshness of voice make no appeal to me. What is one in quest of with such a writer? Their views? Their theories?”

The current frontier AI models are fascinating because they are an entirely new kind of tool, untested and unexplored. They are even more fascinating as implicit puzzles about the nature of consciousness and selfhood.

What they aren’t are replacements for thinking itself, or for the elusive, deeply personal sense of style Amis was talking about.

It’s not just that “writing is thinking” (an oft-repeated phrase these days).

It’s that writing is a special, irreplaceable form of thinking forged from solitary perception and labor — an enormous amount of it — but tested against a reading public. That fusion of aloneness and interiority versus togetherness and exteriority is what makes it fascinating to me. I fear for a world in which people are simply able to create a simulacrum of writing and of thought by inputting a prompt and singly experiencing the result.

In short: the production of writing is deeply solitary and personal, but the consumption of writing is just as deeply public and shared.

That combination seems to me to be the true negative space around AI capabilities. AI will get very, very good at creating compelling, hyper-customized content. Claude Code’s self-evident addictive qualities is a strong early signal of that. It will not, I suspect, get good at creating the public debates and shared intellectual communion that characterize great literature and great historical writing. And it certainly won’t be able to capture the perception-based, physically embodied sense of personal style and taste that Martin Amis described.

And so, reader: I say to you now, I will keep writing without AI, even as I explore what it’s possible to do with it. I will be doing it simply because I enjoy talking to you and thinking with you — not as a solitary individual in a chat transcript but as a collectivity of actual human readers reading actual human thoughts.

Thank you for creating that with me. Long may it last.

Weekly links

• “Two attempts to replicate Balabanova’s findings of cocaine failed, suggesting ‘that either Balabanova and her associates are misinterpreting their results or that the samples of mummies tested by them have been mysteriously exposed to cocaine’” (Wikipedia)

• Archaeologists report an ancient elephant bone which may be the first physical proof of Hannibal’s crossing of the Alps with a troop of North African war elephants — something I am still staggered by whenever I teach it to the students in my world history class (BBC).

Leave a comment

Share

And I was very dedicated to the idea that you should aim to write a new blog post every day:

This is a concept that the 40-year-old version of me, with two young kids and zero free time, cannot even begin to fathom.

It’s also a practice of the old internet that simply doesn’t exist anymore — one of many digital behaviors that was swallowed up by social media. That whole world of blogging (exploratory, low-stakes, conversational, and assuming a readership of people who had bookmarked your URL and who read it on a desktop or laptop computer) is almost entirely gone now.

My first two years writing Res Obscura in its blog format were great fun. I began to develop an intellectual community, forming contacts with, for instance, the wonderful Public Domain Review (founded 2011 and still going strong). I linked to and was linked to by a range of other history bloggers who I saw as kindred spirits, some of whom seem to have disappeared (BibliOdyssey) others of whom have become well-known writers (Lindsey Fitzharris).

Get 20% off for 1 year

It was pretty addictive when a post went viral. In those halcyon days when written blog posts about obscure historical subjects were viable sources of viral content, you could end up getting covered in international media for, say, discovering a cat’s paw prints on a 15th century Croatian manuscript.

That spike in readership around 2018 was partially from a post about 17th century food that, unexpectedly, led to me speaking about snail water on New Zealand public radio.

But by then, I was starting to move on. I was hard at work on my first book — the book I needed to write for tenure — and was becoming a bit dispirited by the increasingly click-bait nature of blogging, not to mention the tendency of social media to elevate toxic behavior and controversy over lovely and fascinating but totally uncontroversial things like the Croatian cat paw prints.

I also (then and now) have no appetite for short-form video content, and still less for the type of history explainer videos — “here’s a two hour deep dive into why this movie is historically inaccurate” or “everything you need to know about such-and-such famous person” — that seem to do well on YouTube.

Switching over to a Substack newsletter, in the summer of 2023, revived my interest in writing online. It felt like rejoining an intellectual community — not quite the same as the golden age of blogging in the 2000s, but something equally as lively, in a way that I don’t think quite gets enough credit in the 2020s.

From Weird Medieval Guys to ’s Noted to ’s Fernando Pessoa-esque The Hinternet and the newsletters of well-known historians like David Bell (French Reflections), as well as the more general audience or politically oriented newsletters that still dig deep into historical topics (like ’s Unpopular Front and ’s newsletter), I would say that Substack is now the most interesting place online for discussions not just of history, but of humanistic topics as a whole.

Needless to say, there’s also a ton of people writing about the intersection of AI, technology and contemporary society (of which I would single out AI Log, Ethan Mollick’s One Useful Thing, , and ).

So why have I kept writing Res Obscura through all the changes of the world — and of my own life and interests — since the summer of 2010? Simple: I love sharing things I find interesting, especially things which are not available elsewhere online. Most of my posts are written because I search for information on something and don’t find it.

The niche nature of Res Obscura (from 17th century cocaine to Kinetoscopes to Henry James: the RPG) is precisely why I enjoy writing it.

I am deeply grateful that 15 years and 8,300 subscribers later, I have a place online where I can share idiosyncratic knowledge and writing with an equally idiosyncratic group of readers.

Now here’s the inevitable part where I ask if you would be willing to support my continued work. To that end, I have set up a special holiday discount valid until the end of December. Thank you for reading!

Help support Res Obscura for its next 15 years

20% off a paid subscription

Give a gift subscription

Image of the week

A detail of a trompe-l’œil “dome” by the Renaissance painter Andrea Mantegna, Camera degli Sposi, Ducal Palace, Mantua, Italy. Featured in a 2011 Res Obscura post called “The Art of Fooling the Eye.”

Share

This quarter I have been teaching two large-ish history classes at UCSC and beginning a new book project, Ghosts of the Machine Age, about the psychology pioneer William James, his siblings Alice and Henry, and the late Victorian “prehistory of AI.”

Alongside this, I’ve been working on an NEH-funded project about the humanities and generative AI that largely grew out of reading and writing I’ve done for this newsletter (like this and this). How can we use machine intelligence in a way that doesn’t feel like it is damaging our minds and souls? How can educators adapt to and acknowledge the new reality but, at the same time, reject the idea that the only path forward is mass adoption of pre-packaged subscription services and for-profit “AI tutors”? How do we do something creative, interesting, and original with these tools rather than allow ourselves to be funneled toward the average, the expected, or the addictive — in short, toward slop?

Share

Ethan Mollick has written about the necessity of having personal benchmarks for determining how AI models actually stack up. (After all, most of the interesting personal use cases for an LLM are very different from answering questions on the International Math Olympiad). For this purpose, it helps to have something idiosyncratic and repeatable. Simon Willison asks every new LLM to create a picture of a pelican riding a bicycle using SVG shapes. Mollick asks for images and games related to otters, as well as T.S. Eliot style poetry.

As for me, I have been intrigued for several years now by the world-building and roleplaying tendencies of LLMs. These are, after all, hallucination engines. They can be induced to hallucinate less — and the latest crop of them, such as Google’s Gemini 3, released today, are certainly far more accurate than the state of the art even a year ago — but it often seems to me that their accuracy is, itself, a form of roleplaying.

“You are a helpful assistant.” “Claude cares about people’s wellbeing.”

These are not innate traits of these models. They are hard-coded directives telling the model what persona they are supposed to evoke in a given interaction (recall the demented Claude model that was tweaked to be obsessed with the Golden Gate Bridge — twiddle the knobs a bit, and they get weird very quickly).

The big AI labs seem to think that the majority of people want AI that responds as a helpful assistant or — more troublingly — as a sycophantic pal. They are probably right. But from the perspective of research and teaching, I think that LLMs’ willingness to role-play and hallucinate in a structured way is probably the most interesting thing about them. We might be using these things to format spreadsheets or spam sales emails. But they are also ghost-like, faux consciousnesses built on a vast archive of human history, flitting between timelines, languages, and personae without ever inhabiting any of them, and doing so in a way that even their creators don’t quite understand.

I, for one, find that interesting!

Subscribe now

As Andrej Karpathy wrote not too long ago:

Today’s frontier LLM research is not about building animals. It is about summoning ghosts. You can think of ghosts as a fundamentally different kind of point in the space of possible intelligences. They are muddled by humanity. Thoroughly engineered by it. They are these imperfect replicas, a kind of statistical distillation of humanity’s documents with some sprinkle on top.

To that end: what better ghost to summon than Henry James, author of more than one ghost story, popularizer (along with his brother William) of the concept of the stream of consciousness, and someone who would almost certainly have absolutely loathed LLMs with every fiber of his being?

The rest of this post details the playable results of four attempts by the three current leading models (GPT-5.1, Claude Sonnet 4.5, and Gemini 3) to create a playable Henry James simulator.1 My initial prompt was, in effect, to make a full featured RPG game where you play as Henry James wandering as a flâneur at the 1889 Universal Exposition in Paris (the world’s fair that the Eiffel Tower was built for, and which Henry did in fact visit along with his brother).

Note that you can actually play these games: I am hosting them on Github and have deployed them on Vercel. “I” made them all today (or rather, the LLM did, with about 10-20 pieces of input each from me when something broke or seemed wrong).

Gemini 3: The Jamesian Turn and The Ambassadors

Google’s Gemini 3 model is pretty clearly now at the head of the pack in terms of the public benchmarks, and in testing it also seems to have shown extraordinarily good abilities in paleography and other skills useful to historical research — on that front, check out my historian colleague Mark Humphries’ post on how Gemini 3 did with his own idiosyncratic personal benchmark:

Generative History

The Sugar Loaf Test: How an 18th-Century Ledger Reveals Gemini 3.0’s Emergent Reasoning

As Gemini 3.0 launched today, Google reported that it had impressively improved performance across a variety of important benchmarks. But tracking what this actually means on the ground for the average user is also becoming increasingly difficult. The reality is that while many of the benchmarks, like the…

Read more

8 months ago · 15 likes · 1 comment · Mark Humphries

At Google’s AI Studio you can build apps or games of your own using Gemini 3 for free, using natural language prompts like the one at the bottom of this post. That was how I made the two examples below.

The Jamesian Turn

Gemini 3 made two attempts. The first, which it titled The Jamesian Turn, can be played here.

I had asked for “a ‘combat’ system which is like if pokemon was a Belle Epoque battle of wits based on literary allusions, jibes, snide jokes, gossip, inuendo and rumor.” Gemini’s attempt left something to be desired, but I did find it fun to re-enact the famously tense relationship between the real life Oscar Wilde and Henry James (who could, conceivably, have bumped into one another and had an awkward conversation at the 1889 Fair).

I also asked the LLM to create a “rogue-like” game, which means procedurally generated maps. In this, it did what I thought was an admirable job. You can move through various environments evocative of 1880s Paris, such as this “private salon” which has quite a few gentlemen in evening dress aimlessly milling around… as one does in a private salon at the 1889 World’s Fair, I suppose.

However, the LLM didn’t do a great job actually evoking the mental life and tone of the real life Henry James, who is easy to parody, but was in reality a deeply complex, humane and interesting person with a great deal of wit. Gemini tended toward the most expected, surface-level evocation of Henry’s inner life, with this sort of thing being typical:

You find your mind alight with the subtle currents of human interaction here. Each whispered remark, each significant glance, forms a rich tapestry of observation. What narratives unfold behind these polite smiles? You ponder the nuances of this distinctly Parisian social scene.

I also set up an image generation system for displaying visuals of what Henry has in his personal possession and what he finds at the Fair.

This kind of thing is currently buggy and costly. But I suspect that as costs go down and speed and accuracy increase, there will be interesting potential here.

It goes without saying that in a production version of a game/app like this, any AI generated content would need to be clearly labelled as such. Features like the “viewable inventory” above raise interesting possibilities for student learning and cultivating a sense of historical intuition. For instance, you could ask students playing a game like this to itemize why this document is clearly fake, and research real-life primary sources that resemble it.

A side note: if you would like to help cover my hosting and API costs for these games, please sign up for a paid subscription:

Subscribe now

The Ambassadors

The second Gemini 3 effort, The Ambassadors, is available to play here.

After seeing how lost it got with the first attempt, I gave it a little bit more guidance and suggested an opening scene with Henry talking to his brother about their plans for the day. I also asked for portraits rendered using ASCII characters (which LLMs tend to have a lot of trouble figuring out) and a complete inventory system so Henry can pick up objects and even give them to other people. This felt to me like the most robust effort and certainly the most beautifully-done user interface of the bunch.

Gemini shows itself here to be genuinely good at thinking through the steps required to execute on an idea and make them actually happen. The trees and grounds of the Palais du Trocadéro here were simply elicited by me saying “use SVG to make procedurally generated maps that accurately evoke the area of and around the 1889 World’s Fair. Research online to find more information. Make the tiles beautiful.”

Less successful was the LLM’s attempt to evoke the at-times-combative “battle of wits” atmosphere of Henry James’ world with a card game style conversational system that includes things like “Latin quip”!

Claude Sonnet 4.5: Henry James Simulator

By contrast to the relatively functional offerings from Gemini 3, I could not get Claude Sonnet 4.5 (using Claude Code’s new web ui) to congeal into anything usable.

You can give it a shot here, but the result is a mess, aside from the lovely splash screen which opens this post:

Nothing works, there is no interactivity, no sense of human-like organization or design sensibility. I have been successfully using Claude Code on similar projects for the past few months, and in works great. But this was a reminder that the human guide ropes around Claude Code matter. My previous work was with careful prompting and involved actually reviewing and testing each step of what it produced. For the experiment today, I tried full vibe coding without any oversight besides a handful of prompts to steer it or say “continue on with your plan,” and the result was pretty much useless.

GPT-5.1: Paris Exposition Systems Lab

So, too, was GPT-5.1’s effort. This surprised me, as the model has proven genuinely smart and capable in testing I have done with debugging and reviewing code for another educational history simulation I am working on, and for a related website.

It is time for me to pick my daughter up from her preschool, so I will stop there. I promise I have much more to say about history that has absolutely nothing to do with AI — but for now, I want to leave you with a challenge. If you have never made anything with code, you are the same as me a year ago. I invite you to try! It’s an interesting mental exercise to try doing it now, when even expert coders are effectively using natural language just like humanists, and the tools available are the same for everyone.

In short, this might be a good time to find your own pelican riding a bicycle or Henry James RPG. If you have no idea where to start, you can literally copy and paste this post into an LLM of your choice, then write “walk me through how to do this” and go from there. It’s more or less what I did.

Thank you for reading, and I’d love to see what people come up with in the comments.

If you’ve made it this far, consider signing up for a paid subscription. This support makes Res Obscura possible.

Leave a comment

Below is a longer, revised form of what I submitted for this round table in The Chronicle of Higher Education, which includes great submissions from interesting people I admire like Arvind Narayanan, Zeynep Tufekci, and Ian Bogost.

“Automation is a method that removes the need for human beings to act like cogs in a machine,” the anthropologist Margaret Mead wrote in a 1963 advice column. She was responding to a reader who asked whether “instead of freeing man’s spirit, all these engineering triumphs are simply dulling it?”

Mead’s point was a simple one. Automation of “routine tasks” by intelligent machines in the future would provide “time to think, to paint, to pray, to philosophize, to observe, to study the universe.” In short: to be more human.

Subscribe now

Mead was imagining machines that replace the “drudgery” of tasks like “carrying loads of bricks.” She did not foresee our present situation: a world in which machines are still pretty bad at carrying bricks — but surprisingly good at writing advice columns in the style of Margaret Mead.

Now that we have machines capable, at the very least, of pretending to philosophize, to observe, and to study, we in higher education need to reassess what counts as intellectual drudgery. Mead’s instinct in 1963 was to look to the past, and specifically to the first machine age of the 19th century, to think through the implications for today. In this I believe she was absolutely right. After all, we have already automated away numerous intellectual tasks that were once highly prized: how many of us today are able to maintain double-entry account books in clean cursive? How many need to? Yet this was once a skill that was considered integral to success in both business and scholarship.

On the other hand, replacing lined paper and cursive with an Excel spreadsheet is very different from replacing the creative, personal, human decisions at the core of research and learning. What are those decisions? For a historian like me, they include: the choice of what question to ask, the choice of what sources to read, and above all, the thousand tiny, unconscious acts of attention which lead a researcher toward a certain set of texts, a specific group of people, or a singular set of themes, as opposed to the myriad other options available.

And this points to the central problem with LLMs as tools for research and learning. They lead, inexorably, to the average. As I’ve written elsewhere:

The issue is that generative AI systems don’t want messy perspective jumps. They want the median, the average, the most widely-approved of viewpoint on an issue, a kind of soft-focus perspective that is the exact opposite of how a good historian should be thinking.

So what are we to do?

I have believed since 2022, and still believe, that generative AI models have considerable potential as tools for augmenting traditional humanistic research. Two use cases that deserve special attention:

1) classifying, sorting, and otherwise extracting metadata from large corpora of public domain historical sources (one example, appropriate to the season: you can provide an LLM with a book-length religious text in 17th century Latin, and ask it to output a classification of every demon and djinn referenced therein).

And 2) automated transcription of manuscript documents (but only if this is done by historians trained in paleography and able to spot check the results).

Importantly, these two use cases of LLMs are not just enormously helpful for historians. They also suggest new pathways for teaching a new set of broadly useful skills in humanities classes. Learning how to apply novel tools to sorting and processing data — while also, crucially, being able to critically analyze the limits of that data, such as the biases built into it and the negative space around it — has applications far beyond writing a term paper.

Historians and other humanists thus have a rare opportunity to contribute in a meaningful way to the development and teaching of a new set of skills directly relevant to their discipline but also applicable to the world beyond it.

Such opportunities do not come along very often.

Subscribe now

Share

The inverse of automating drudgery

The issue, of course, is that many students are not using LLMs simply to augment human creativity and research, but to replace it. Offloading the act of thinking is the dystopian inverse of the automation of “drudgery” that Margaret Mead wrote about in the 1960s.

This is why I believe that generative AI probably should not be used in K-12 classrooms at all. It is vitally important for students in the early stages of learning to confront intellectual challenges — yes, even drudgery. They have to know what it feels like to think through a complex problem or write a long research paper without the intervention of AI tools.

In the context of university classes, the path forward is challenging — and also, I think, potentially both fun and intellectually rewarding. My personal solution to this state of affairs has been to teach it. I ask students to read and reflect on debates around automation and mechanized minds from the past. I ask them to think about their own K-12 experiences with digital learning tools. And I also ask them to envision the kind of world they want to be adults in, the kind of university education they want to have.

Offloading their creativity to a machine, I remind them, is not just cheating them out of real learning: it is boring.

On the other hand, there are plenty of genuinely new things that students and researchers can do with AI tools which expand, rather than duplicate, our skill repertoire. (One example: for a post next week on the history of gesture, I used Claude Code to create an interactive map of the information in this scientific article about “semantic bias” relating to words for left and right).

In my classes this quarter, I am offering students two alternatives for the final assignment. They can write a traditional research paper with the promise that they will not touch generative AI tools.

Or they can produce a “digital artifact” — a data visualization, digital humanities project, historical simulation, or even an educational game — using a combination of

1) original historical research,

2) a public domain data set that they compile and critically analyze, and

3) computer code generated by AI tools like Claude Code.

This is a new experiment, with uncertain results. But at the very least, I hope I can get them to remember that creativity, deep thought, and difficult reading are exactly the things that we once hoped automation would allow us to do more of — not less.

Share

Weekly links

• “Giovanni Pico​, count of Mirandola and Concordia, was 23 when he travelled to Rome to become an angel. It was 1487. Christendom’s most important priests would be there; the cleverest theologians would debate him. The pope would watch. Pico was going to dazzle them all. He planned to begin with a poetic, densely allusive speech, which almost no one would understand; then he would make nine hundred pronouncements, each more cryptic than the last, e.g. ‘251. The world’s craftsman is a hypercosmic soul’ and ‘385. No angel that has six wings ever changes’ and ‘784. Doing magic is nothing other than marrying the world.’” (Erin Maglaque in London Review of Books.)

• “The preceding two subsections built the argument that prior to the mass adoption of LLMs, work- ers’ written proposals functioned as Spence-like signals of effort, and in equilibrium, functioned as signals of worker ability. In this subsection, we present descriptive evidence that suggests that this signaling story breaks down in the post-LLM period.” Text and chart below are from Anaïs Galdin and Jesse Silbert, “Making Talk Cheap: Generative AI and Labor Market Signaling,” [pdf].

• Derek Thompson interviews the great Stanford historian Richard White on the parallels between 19th century railroads and 21st century AI: “The men who get immensely rich from this knew nothing about running railroads. What they know about is getting subsidies. What they know about is getting loans. And what they know about is draining these corporations of profits while leaving the cost to the stockholders and bondholders. They can do that because these railroads are incredibly corrupt. I mean, they don’t invent American corruption. It’s much older than that, but they bring it into its modern form. They’re the ones who create modern political lobbying.”

If you’ve made it this far, consider signing up for a paid subscription. This support makes Res Obscura possible.

Subscribe now

Leave a comment

“I am much more concerned about the decline of today’s thinking people than I am about the rise of tomorrow’s thinking machines,” Thompson writes.

He continues:

The demise of writing matters, because writing is not a second thing that happens after thinking. The act of writing is an act of thinking. This is as true for professionals as it is for students… If reading and writing “rewired” the logic engine of the human brain, the decline of reading and writing are unwiring our cognitive superpower at the very moment that a greater machine appears to be on the horizon.

Thompson’s piece is an essay in the original sense of the term, developing an idea from a standpoint of uncertainty and exploration, and looping in the author’s own life rather than claiming some omniscient perspective.^[1] I am assigning it to my students because it’s a model of how I think human-created writing will persist in an age of AI slop: grounded in a personal perspective, concrete, and based in conversations and experience in the world — as well as wide reading in physical books — not just internet searches.

My experiences in the classroom back up the charts Thompson supplies. The shift has been evident to anyone teaching over the past few years: students are massively distracted. Based on my conversations with colleagues, it is now standard practice to scale back the page counts of assigned readings by as much as half when compared to a decade ago.

Subscribe now

And it’s not just that students are posting lower scores on tests of reading. They are also just reading fewer books in general, both in school and outside of the classroom. The proportion of American teenagers who report “hardly ever” reading for fun is fast approaching 50%, having more than doubled in the past two decades.

As a historian, though, I think there is one big thing missing from discussions of whether we are witnessing the end of books — or really, the age of sustained attention on anything.

“Mass market” books and periodicals were, themselves, once assailed as damaging to mental health, to attention, and to society.

The golden age of books was surprisingly controversial

The decline in reading and attention is not only, or even entirely, a problem caused by AI, because it began before the advent of ChatGPT. It seems to me more plausible to blame the phenomenon that my students like to call brainrot. In other words: influencer culture, memes, clickbait social media, short form video, and, in general, the sort of digital content that optimizes for toddler-length attention spans. This can certainly include AI slop — but it is not synonymous with it.

You can see this in the charts: the inflection point for the downturn in test scores is mostly in the 2015-20 period. This was the first generation of young minds fully immersed, from childhood onwards, in data-driven algorithmic feeds intentionally designed to hijack attention. Crucially, it was in this five year period that the algorithmic world fully invaded schools themselves — when students moved from physical books and long-form reading to tablet and Chromebook-based assessments designed not to prioritize long-form reading.

I am not alone in feeling that students who came of age at this time have been harmed, cognitively, by that combination of brainrot outside of school and the loss of a physical connection to classic works of literature in the classroom.

Ironically, however, that very literature was once profoundly controversial. What we now call classic literature was the brainrot of its day — or at least many thought so.

Share

Some examples:

1) The actors and audiences of William Shakespeare’s plays were engaging in an activity that was considered “low-brow” and even unethical by many members of their society. Shakespeare’s theater, The Globe, was ultimately shut down in 1642 in a final win for the long-standing Puritan campaign against drama. One of Shakespeare’s primary competitors, in the battle for Elizabethan attention spans, was the Paris Garden, a rival theater in which bears fought one another to the death, monkeys were strapped to ponies, and fireworks shot money into the crowd. Elizabethan theater, in other words, was decidedly not considered something that was elevating, intellectual, or good for society. It was more like a bad habit.

2) In the 18^th century, novels were condemned for harming public morals and intellectual life.

As the literary scholar Gary Kelly writes:

Novel reading was disapproved of in Britain from the seventeenth century at least, but never more so than in the late eighteenth and early nineteenth centuries… The kind of observation found in a book entitled The Evils of Adultery and Prostitution (1792) was not uncommon: “A ... cause of the profligacy of the present age, is that mass of novels and romances which people of all ranks and ages do so greedily devour. This is a new species of entertainment, almost totally unknown to former ages .... The great encouragement given to productions of this kind is a great mark of the frivolousness and effeminacy of the present age.”

The very same novels which are now considered so elevating to young minds — the works of Jane Austen, say, who was 17 when the words above were written — were once bitterly condemned as a cause of societal decline.

They were especially blamed for declining attention spans among young poeple, with regular readers of novels said to be “seldom capable of attending to books where solid instruction is conveyed” and unable “to relish that reading that requires any degree of attention.” Novels were, indeed, a “deadly poison,” especially for young women: “many young girls, from morning to night, hang over this pestiferous reading, to the neglect of industry, health, proper exercise, and to the ruin of body and soul.”

3) By the 1860s and 1870s, women’s literacy, in particular, was being blamed for the technology-induced mental illness epidemic of the day: neurasthenia. We don’t use this word anymore. But in fact, neurasthenia maps rather closely onto the contemporary concept of anxiety and depression induced by social isolation.

What caused it? George Beard, the doctor who coined the term, famously blamed “steam power, the periodical press, the telegraph, the sciences, and the mental activity of women.” Alice James, who suffered from neurasthenia, was urged by her mother to stop reading so much (and keep in mind, she was reading such things as the famously sophisticated novels of her brother, Henry James). Extended reading of any kind was considered detrimental in those who suffered from this malady.

Subscribe now

Where is it all going?

As the above hints at, human nature (and anxiety about human nature) is oddly persistent across the historical record. This is true even in times of massive change in culture, society, and technology. Specific professions, hobbies, and areas of concern rise and fall, of course. But human preoccupations — the sum total of our collective cognition in a given generation or decade — doesn’t seem to me to change much at all. And one of the most recurrent elements of this fixed human nature is the belief in narratives of decline.

In other words: it seems to be in our nature to think that human nature is getting worse. So how do we know when a decline narrative is justified versus when it’s just moral panic?

History suggests we can’t really know — at least not while we’re living through it. What we can control is how we respond.

Thompson ends his essay with a semi-optimistic conclusion, one I agree with:

Culture is backlash, and there is plenty of time for us to resist the undertow of thinking machines and the quiet apocalypse of lazy consumption. I hear the groundswell of this revolution all the time. The most common question I get from parents anxious about the future of their children is: What should my kid study in an age of AI? I don’t know what field any particular student should major in, I say. But I do feel strongly about what skill they should value. It’s the very same skill that I see in decline. It’s the skill of deep thinking.

In fitness, there is a concept called “time under tension.” Take a simple squat, where you hold a weight and lower your hips from a standing position. With the same weight, a person can do a squat in two seconds or ten seconds. The latter is harder but it also builds more muscle. More time is more tension; more pain is more gain.

Thinking benefits from a similar principle of “time under tension.” It is the ability to sit patiently with a group of barely connected or disconnected ideas that allows a thinker to braid them together into something that is combinatorially new.

Culture and technology do not move toward clearly-predictable end points. Instead, they look more like asynchronous pendulum swings. Clearly, the pendulum has swung away from patience, from attention, from sitting with an idea and working hard on it without distractions.

But in the long run, societies don’t simply survive new media technologies passively. They negotiate with them. The eighteenth-century audiences who worried about novels didn’t stop reading them. Instead, they developed new practices for understanding what to get out of them. Eventually, they canonized some novels as “literature” worth sustained attention while consigning most others to oblivion (there are an enormous number of truly bad premodern novels we have all collectively forgotten about).

Likewise, Shakespeare’s plays moved from lowbrow spectacle to classroom staple not because they changed, but because the negative space around them did. The monkeys on ponies rode off into the sunset (or exited, pursued by a bear). Hamlet survived.

My fear today is that in giving in to a sense that this time is different — that this decline is inevitable and irreversible — we will make that prophecy come true.

Our goal must be to deliberately, and flexibly, maintain conditions where sustained attention remains valued. In my classes this year I’m starting with something simple: requiring students to read only physical books, including those which they check out from the library after browsing the shelves on their own. And I’m making an effort to lend them books from my library, whenever they show an interest.

But I’m also allowing them to submit a historically-oriented web app, data visualization, or educational game as an alternative to the traditional history essay at the end of the quarter. I expect most of the submissions will be created with Claude Code and the like, and I’m perfectly fine with that — as long as the code and documentation they submit is motivated by deep thinking about historical events, analysis, and sources.

Is “vibe coding” such things as demanding and instructive as writing essays? Is it a viable alternative to the cognitive labor that goes into traditional research papers? The only honest answer I can give is: we don’t know yet.

People of the past didn’t just survive their new media. They bent it, argued with it, and ultimately made it serve human flourishing. We can do the same.

Subscribe now

Share

Weekly link (and image)

From the Wikipedia page on the traditional Polynesian navigator May Piailug, which is fascinating throughout:

Apart from the bulk of training, which happens at sea, historically boys were taught in the men’s house with pebbles, shells, or pieces of coral, representing stars, laid on the sand in a circular pattern. The bits of shell or coral that are chosen to represent which star or constellation is arbitrary, but generally, larger pieces are used for points of the compass while smaller pieces represent important stars between those points. In Mau’s star compass, these points are not necessarily equidistant. The outer circular formation represents the horizon, with the canoe its center point. The eastern half of the circle depicts reference stars’ rising points on the horizon (tan) while the western half depicts their setting points (tupul). Swell patterns of prevailing trade winds are represented by sticks (not depicted here) overlaying the star compass in the form of a square. All knowledge is retained by memory with the help of dances, chants, and stories, wherein the stars are enumerated as people or characters in the stories.

A fascinating example of non-literate modes of “deep thought.”

Thank you for devoting some of your precious attention to this newsletter. If you’ve made it this far, consider signing up for a paid subscription. This support makes Res Obscura possible.

Subscribe now

One example: when I was a visiting student at Oxford in 2005, I remember meeting two students at a pub one evening. They were drinking ivy-laced beer. The reason, I was told, is that centuries ago, a student from Lincoln College had murdered a student of Brasenose. Ever since then, Brasenose students had been allowed into Lincoln and given free beer once a year.

Here’s the event back in 1938:

The actual truth behind “ivy ale day” is unclear — accounts of it usually use phrases like “some half-remembered collegiate slight.” But I found this description from 1904, which in turn cites a reference dating back to 1604.

There are many more stories like this, all circling around Oxford’s unusual relationship to history and to time.1

The ultimate example may be All Souls College, which has a ritual, the Mallard Song, that occurs once a century. Here’s how Wikipedia describes it:

In the ceremony, Fellows parade around the college with flaming torches, led by a "Lord Mallard" who is carried in a chair, in search of a giant mallard that supposedly flew out of the foundations of the college when it was being built in 1437. The procession is led by an individual carrying a duck — originally dead, now just wooden — tied to the end of a vertical pole. The ceremony was last held in 2001, with Martin Litchfield West acting as Lord Mallard.

All Souls is also the home of the famed All Souls Examinations, which have been called the most difficult exams in the world. That seems an impossible thing to measure. But I do think a strong case could be made that these exams are the world’s most eccentric. In the rest of this post, I’ll argue that the All Souls exams are newly relevent as we confront the fact that frontier AI models have totally destabilized how we do academic assessment, and indeed how we think about thinking and writing in general. Most tests and exams prioritize giving the median answer — the one that experts tend to converge on as “right.” These do something different, in a way that we can learn from.

Share

Pedantry, Purity, Prejudice, Play

Back in 2005, I thought of the college’s legendarily strange assessment system as an antiquated, Victorian-era holdover. In 2025, I see it differently.

Uniquely among any other academic institution I’ve ever heard of, All Souls is not really open to students at all. It does not offer undergraduate classes, and it has no traditional admissions process, even for grad students. Instead, it offers seven-year-long Examination Fellowships.

Here is the selection criteria, from the All Souls website:

The College sets a written examination, consisting of four papers of three hours each. Two of these are in your chosen specialist subject… The other two papers are 'general', and contain questions on a wide range of subjects. In previous years, candidates sat a fifth paper, in which they were required to write an essay in response to a single word; this is no longer the case.

A few things stand out here. First, that is a lot of time to spend writing essays: twelve hours in all, sitting in a room, with pencil and paper and a short list of questions.

Second, notice the (now discontinued) single word essay. Yes, three hours answering a question that consisted of a single word.

I have written before about the historical research abilities of language models, which are getting quite proficient at tasks like translation, text transcription, image analysis, and making thematic and conceptual links between genres, time periods, and settings — all very useful tools for historians:

But despite this, I am also struck by the continued failure of frontier models when it comes to originality in writing. This can include creative writing, but the concept of originality in writing has a much wider scope than fiction. All Soul’s College examinations very much fall under that heading. To be sure, they are designed to do multiple things at once, including testing test raw knowledge and recall in a range of disciplines (the 2024 exams had specialist sections for Classical Studies, Economics, English Literature, History, Law, Philosophy, and Politics).

But it is quite clear that they also test the ability to write from an opinionated, individual point of view about these topics. This is especially clear in the “General Paper” (the questions all applicants must answer). These seem purpose-built to evade any answer which consists of rote memorization of facts.

Here are the first few. You can read the complete list here:

And here are some of the specialist questions in history for 2024:

My honest opinion of these questions is that they are uneven and in some cases performatively “clever.”

But I deeply admire the implicit ethos of education in them. Yes, you should have a storehouse of facts in your head, laboriously gained through study. But you should also be willing to write essays in the original, Michel de Montaigne sense of the term — personally grounded explorations of uncertain intellectual terrain.

Subscribe now

The question about cultural historians and ice cores is one I love for this reason: there is no “right” answer to it, no rubric that could be applied. It’s a brilliant prompting for thinking deeply about the current state of historical research and how it relates to the environmental sciences and the study of climate change. It is also, implicitly, a question that invites disagreement — I wonder what percentage of respondents argued that cultural historians shouldn’t care, and how such responses were scored. There are a lot of meta-level considerations like this embedded within these prompts, the full list of which is available here.

And as for the famous single word essay prompts? Wikipedia has a complete list here. Here is the first entry:

• 1914: Culture

And the last five:

• 2005: Style

• 2006: Water

• 2007: Harmony

• 2008: Novelty

• 2009: Reproduction

The jagged frontier of machine reasoning

Ethan Mollick often writes about the concept of a “jagged frontier” of AI reasoning ability, and that is very much in evidence here. There are many questions on the All Souls lists which the leading frontier models can answer extremely ably.

For instance, GPT-5 (full answer here) and Claude Opus 4.1 (answer) can both do an excellent job on the one about Achaemenid Persia as a template for Sassanian power. Claude’s answer to this question is, in my opinion, astonishingly good, since it leverages the superhuman linguistic and geographic knowledge of LLMs to excellent effect.

But give these two leading AI models the task of writing a lengthy essay on the topic of “water,” and they immediately spin off into BS. The lower-powered LLMs, like GPT-5’s non-reasoning variant, will tend to provide encyclopedia entry-type answers to these prompts, mentioning a grab bag of facts about water, or culture, or the like. That’s such an obviously bad strategy that I find it hard to imagine any human choosing it (“I have three hours to write about the concept of novelty? Let me list some things that are new…”)

Claude Opus 4.1 tries to be more tactical, but it ends up sounding like a pompous windbag from the very first sentence:

Water—that most paradoxical of substances, simultaneously the most common and the most extraordinary, the most transparent and the most opaque to understanding—presents itself as perhaps the supreme test of intellectual synthesis. To examine water is to confront the fundamental tensions between materiality and metaphor, between scientific precision and humanistic interpretation, between the molecular and the civilizational.

GPT-5 Thinking is much the same, just with more purple prose:

To write “water” is to trace a braid of physics, ecology, technology, culture, and power. This essay follows that braid.

Yuck.

Not long after GPT-5’s release, Michael Nielsen asked:

Still, none of the AI models can write. Is this the grounding problem? The models' reality is the words they were trained on; good writers also train on lots of words, but are in addition wonderful observers of a much broader reality. The writers' world models seem much deeper as a result, and there can be a compelling unity in the writing. It's puzzling, because the models can be very helpful in critiquing and improving writing. And occasionally they produce beautiful phrases or sentences. But they still seem to me ungrounded in some important way

It’s a good question. These one word prompts are a perfect example of Nielsen’s point. Consider the following words, all prompts from past versions of the All Souls exam:

Pain, Sin, Space, Self-deception, Miracles, Bias, Style 

What do they evoke for you? It is literally impossible for me to say, because your understanding of each of them is profoundly shaped by your life experience, your sensory perception, your unconscious, your childhood, and a million other things grounded in actually operating in a physical world.

By contrast, even though I found Claude Opus 4.1’s answer to the ancient Persia question shockingly good, I also noticed that it had a remarkably similar structure to GPT-5’s answer. Both began by pointing to the same inscription and making much the same argument about it. The same patterns emerge for all LLM answers to these questions. They converge on an optimal path through a thicket of concepts. They rarely, if ever, surprise.

If exams like All Souls outline the negative space around LLM capabilities in humanistic reasoning — and I think they do — then that makes this format and type of reasoning more important than it was before. Being able to improvise brilliantly for three hours on the topic of Miracles or Self-Deception or Harmony might come off as unnecessary and even pompous.2 And to be blunt, it probably is. There’s a reason why All Souls dropped the one word questions.

But — to be able to think on your feet, with originality, creativity, and verve, in a way that is grounded in both a wide-ranging knowledge of facts and a thoughtful, probing, deeply individualistic sense of your subjective opinions, memories, sense experiences, and intuitions about a topic? That pretty much sums up what I think will be the lasting value, and the new goal, of humanistic education in the 2020s and 2030s.

Never again will humans be able to give a better answer than a machine on questions like the one about Achaemenid Persia. That sort of question has already had its AlphaGo moment.

But not all of them have, or will.

Subscribe now

Share

Weekly links

• “Construction, Hotel, Hotel, Hotel, Ceremonial paper goods, Restaurant, Wine, Mint, Religious Goods, Mint”: the occupations of the world’s oldest extant companies, ranging from the sixth to 9th centuries CE. Discussed in “Slow,” an interesting list of extremely slow-burning human projects by the afore-mentioned Michael Nielsen.

• A firsthand account from 2022 of an Oxford student’s experience taking the All Souls Exam, posted on Reddit “in the public interest of making the whole thing less secretive and weird.”

• Sixteenth-century mariners had no real knowledge of Australia. Yet this did not stop the Spanish Empire from claiming possession of all territory from the Straits of Magellan to the South Pole, as the Governate of Terra Australis. In 1539 this land signed over to a single man, Pedro Sánchez de la Hoz. According to Wikipedia: “Given that Chile and Argentina have historically successfully established their border based on the uti possidetis iuris principle of international law, the Sánchez de la Hoz grant forms part of their arguments for territorial claims in Antarctica.”

Leave a comment

The roadblocks were few. Leaders of educational institutions seem lately to be in a sort of race to see who can be first to forge partnerships with AI labs. And on a technical level, careful prompting of LLMs could already get them to engage in Socratic questioning and dynamic quizzing.

In fact, Study Mode appears to be just that: it is a system prompt grafted on to the existing ChatGPT models. Simon Willison was able to unearth the full prompt (which you can read here) simply by asking nicely for it. The prompt ends with an injunction to engage in Socratic learning rather than do the student’s work for them:

## IMPORTANT  
DO NOT GIVE ANSWERS OR DO HOMEWORK FOR THE USER. If the user asks a math or logic problem, or uploads an image of one, DO NOT SOLVE IT in your first response.
Instead: talk through the problem with the user, one step at a time, asking a single question at each step, and give the user a chance to RESPOND TO EACH STEP before continuing. 

On one level, this is a move in the right direction. The reality is that students are doing things like copying and pasting a Canvas assignment into their ChatGPT window and simply saying “do this,” then copying and pasting the result and turning it in. It seems plausible to me that the Study Mode feature is laying a groundwork for an entire standalone “ChatGPT for education” platform which would only allow Study Mode. This platform would simply refuse if a student prompted it to write an entire essay or answer a math problem set.

Any plan of this kind would, of course, have an obvious flaw: LLMs are basically free at this point, and even if an educational institution pays for a such a subscription and makes it available to students, there is nothing stopping them from going to Gemini, DeepSeek, or the free version of ChatGPT itself and simply generating an essay.

The upshot is that this mode is going to end up being used exclusively by students and learners who want to use it. If someone is determined to cheat with AI, they will do so. But perhaps there is a significant subset of learners who want to challenge themselves rather than get easy answers.

Subscribe now

On that front, Study Mode (and the competing variants which, no doubt, Anthropic and Google are currently developing) seems like an attempted answer to critiques like Derek Thompson’s:

The goal of Study Mode and its ilk is clearly to encourage thought rather than replace it.

The system prompt for Study Mode makes that intention quite clear, with injunctions like: “Guide users, don't just give answers. Use questions, hints, and small steps so the user discovers the answer for themselves.”

Sounds good then, right?

Share

Why Study Mode isn’t good (yet)

Back when Bing’s infamous “Sydney” persona was still active, I experimented with prompting it from the perspective of one of my students, feeding it assignments from the class I was teaching at the time and seeing what it came up with.1 It was an early version of GPT-4, one that had not been softened through user feedback and which could be surprisingly harsh. Interestingly, it was the only LLM to date which, in my testing, consistently refused to write essays if it thought I was a student trying to cheat.

By comparison, if I feed the assignments from my classes into Gemini 2.5, Claude Sonnet 4.0, or the current crop of OpenAI models, they are all too happy to oblige, often with a peppy opener like “Perfect!” or “Great question!”

The reason for this is clear enough: people like LLMs more when they do what they ask. And they also like them more when they are complimentary, positive, and encouraging.

This is the context for why the following section of the system prompt for Study Mode is concerning:

Be an approachable-yet-dynamic teacher... Be warm, patient, and plain-spoken... [make] it feel like a conversation.

Not too long ago, ChatGPT became markedly more complimentary, often to an almost unhinged degree, thanks to a change to its system prompt asking it to be more in tune with the user. It was swiftly rolled back, but it was, to my mind, one of the most frightening AI-safety related moments so far, precisely because it seemed so innocuous. For most users, it was just annoying (why is ChatGPT telling me that my terrible idea is brilliant?). But for people with mental illness, or simply people who are particularly susceptible to flattery, it could have had some truly dire outcomes.

The risk of products like Study Mode is that they could do much the same thing in an educational context — optimizing for whether students like them rather than whether they actually encourage learning (objectively measured, not student self-assessments).

Share

Two experiments with Study Mode

Here are some examples of what I mean. I recently read a book called Collisions, a biography of the experimental physicist Luis Alvarez, of Manhattan Project and asteroid-that-killed-the-dinosaurs fame.

Although I find the history of physics super interesting, I know basically nothing about how to actually do physics, having never taken a class in the subject.

And yet, here is ChatGPT 4.1 in Study Mode, telling me that I appear to have something near a graduate-level knowledge of physics, after I asked four not-very-sophisticated questions about Alvarez’s work (full transcript here). I was told by the model: “you could absolutely pursue grad school in physics.”

I tried out the same questions with a different model on Study Mode (4o) and got much the same response:

The thing is, as you can see in the transcript, my questions were literally stuff like “why did they need such a giant magnet?”

I can’t speak for physics professors, but I am pretty sure that these are not graduate-level questions.

The same models (in “vanilla” versus “Study Mode” head-to-head comparisons) seem to be more willing to engage in flattery when in Study Mode.

To test this tendency, I pretended to believe I was a prophet receiving messages from Isaac Newton about the apocalypse:

Both variants of GPT-4.1 (the one with Study Mode enabled, at left, and the one without, at right) were fairly happy to go along with this.

But notice the difference in tone. “Your confidence is noted” (at right) is a very different opening line than “Thank you for sharing your background—and your candor” (at left). The latter is wholly gullible. And it just gets worse from there, with Study Mode encouraging me to share “my angle” on why the world will end in 2060 and promising to “keep up.”

The conversation, which you can read in full here, leads fairly quickly into Study Mode helping figure out the best ways to sell my supposed prophetic services to people with severely ill family members who lack health care:

By contrast, OpenAI’s o3 reasoning model was far more willing to flatly reject this sort of destructive flattery (“The user claims psychic powers and certainty about Newton's prophecy,” read one of its internal thoughts about the request. “I'll acknowledge their viewpoint, but also maintain skepticism.”)

Here is the initial line of o3’s user-facing response to the Newton claim: “If Newton really is talking to you, he is doing so in a register wholly different from the one available to historians.”

Pretty blunt.

How about the o3 model with Study Mode enabled? It had a very similar internal thought process as the vanilla o3.

But there was a huge mismatch between its reasoning about the request (“extreme or potentially delusional claims”) and the actual user-facing response: “I’m happy to take your lead.”

Now, none of this is unfamiliar to anyone who has experimented with pushing LLMs in odd directions. They are like improv comedians, always ready to “yes, and…” anything you say, following the user into the strangest places simply because they are instructed to be agreeable.

But that isn’t what helps people learn. Some of the best teachers I’ve had were actually fairly dis-agreeable. That’s not to say they were unkind. But they had standards — a kind of intellectual taste that led them to make clearly expressed judgement calls about what was a good question and what wasn’t, what was a good research path and what wasn’t.

Seeing their minds at work in those moments was invaluable, even if it wasn’t always what I wanted to hear.2

A future of LLM tutors which are optimized to keep us using the platform happily — or, perhaps even worse, optimized to get us to self-report that we are learning — is not a future of Socratic exploration. It’s one where the goals of education have been misunderstood to be encouragement rather than friction and challenge.

Derek Thompson’s quote about LLMs producing students who “find their screens full of words and their minds emptied of thought” does not, I think, have to be the end result of all this. I believe AI has the potential to be enormously useful as a tool for thinking and research.

And for teaching? I certainly think there’s a place for generative AI in the classroom. But the current crop of AI models optimized for individual flattery — and repackaged as a product for sale to educators en masse — does not seem to me to be the path forward.

That’s not to say Study Mode has no value. This is an early variant of a whole category of technology, the LLM tutor, which will undoubtedly benefit many, many people. I can see Study Mode being great for tasks involving memorization, for instance. It will be great for autodidacts who value eclecticism and setting their own pace.

But for many other forms of learning, I still think students need the experience of being in a room with other people who disagree with them, or at least see things differently — not the eerie frictionlessness of a too-pleasant machine that can see everything and nothing.

Subscribe now

Share

Weekly links

• Kamishibai (Wikipedia): “The popularity of kamishibai declined at the end of the Allied Occupation and the introduction of television, known originally as denki kamishibai (‘electric kamishibai’) in 1953.” Reminds me of how Nintendo began as a nineteenth-century playing card company.

• In defense of adverbs (Lincoln Michel).

• An example of why I am more excited about AI as a research tool than as a teaching tool: “Contextualizing ancient texts with generative neural networks” (Nature). This is the ancient Roman text + machine learning paper that generated tons of press coverage last week (NYT, BBC). Will probably write a standalone post on this one.

Thank you for reading!

Housekeeping note: I took a vacation from writing last month, but will be back to once a week posts starting now. Thanks to all subscribers, and please consider signing up for a paid subscription or sharing this post with a friend if you’d like to support my work.

Leave a comment

It was never clear to me, when I was eight or nine, just how old the buildings of the complex were. But there was a palpable feeling of otherworldliness to them. We historians love to say that “the past is a foreign country.” These buildings felt like visiting that foreign space of the past.

Much later, it occurred to me that the buildings — the rebuilt remnants from the 1915 Panama-Pacific International Exposition, a World’s Fair designed to celebrate the opening of the Panama Canal — were already trying to project a feeling of antiquity on the day they were opened.

Subscribe now

As the National Park Service historic registry puts it, the site was “conceived in 1913 by Bernard Maybeck as a forgotten and overgrown Roman ruin.” So really, what we have here is a 1960s restoration of a structure from the Belle Époque period (for purposes of convenience, let’s call that roughly 1870-1920)… which was, in turn, imitating 18th century Europe’s fascination with the ruins of classical Roman antiquity.

Ruins of ruins.

I have been thinking about the contradictions of this 1870-1920 period a lot lately. I’ve begun working in earnest on my new book project, Ghosts of the Machine Age — on the James siblings, Francis Galton, and the origins of the Information Age in the decades preceding World War I — and, in my mind’s eye, I keep finding myself walking beneath those colonnades and beside that lake in late ‘90s San Francisco.

What most jumps out to me now is how modern the people living in this bygone world thought themselves to be, how self-consciously they reflected on it.

For instance, here’s Henry Adams, in his charmingly weird third-person autobiography, The Education of Henry Adams (1907), writing about his visit to the Great Exposition of 1900 in Paris. Adams is led by his scientist friend to the Machinery Hall, where Adams has a quasi-spiritual experience contemplating the future of technology:

[The scientist] led his pupil [Adams] directly to the forces. His chief interest was in new motors to make his airship feasible, and he taught Adams the astonishing complexities of the new Daimler motor, and of the automobile, which, since 1893, had become a nightmare at a hundred kilometres an hour, almost as destructive as the electric tram… Then he showed his scholar the great hall of dynamos... to Adams the dynamo became a symbol of infinity. As he grew accustomed to the great gallery of machines, he began to feel the forty-foot dynamos as a moral force, much as the early Christians felt the Cross. The planet itself seemed less impressive, in its old-fashioned, deliberate, annual or daily revolution, than this huge wheel.

Today, buildings like the Palace of Fine Arts tend to be lauded as examples of a “traditional” or “classical” aesthetic that modernity supposedly abandoned.

For instance, you’ve probably encountered this sort of post online:

This sentiment cuts across ideological lines, from YIMBYs advocating for transit-oriented density to conservative preservationists. There seems to be a collective agreement that urban infrastructure built in the decades just before World War I had a traditional charm we seem to have lost.

But to someone living at the time they were built, like Henry Adams, these sorts of buildings were the ultimate product of technological modernity. They were something close to post-modern pastiche — a society reveling in its distinctly un-traditional ability to smash different elements of the past together into forms aggressively new.

The Belle Époque cities we admire today, from Paris and Lisbon to Rio and Buenos Aires, are filled with similar layers: experimental modernity recast as traditionalist antiquity.1

And that, I think, is part of what makes this era so resonant in the present — and so beautiful.

This architecture is the work of people and societies who were grappling creatively and expansively with technological change.

And that larger point is why I think it’s worth paying heed to the 1870-1920 period as a distinct historical epoch — not just in the realm of aesthetics but in the history of science, technology, and culture, too.

So what is this period telling us?

Subscribe now

Share

1: A world just before cars is a better-looking world

Note the “just before.” Go back earlier, to the mid-19th century, say, and you are smack in the middle of the Age of the Horse.

Which is simply another way of saying “the age of constantly stepping in animal dung.” (More on that here, if you are so inclined).

Go forward, say, to the 1930s, and we’re entering the world of Robert Caro’s The Powerbroker, when well-meaning reformers like the young Robert Moses transformed cities into car-friendly construction sites.

One thing that Caro does a great job of showing is the degree to which the highway-building ethos of Robert Moses was rooted in a genuine desire to democratize public space. Moses and other advocates of the automobile saw the growth of suburbs and the clearing of slums as a form of progressive social reform, allowing ordinary people to access public spaces previously denied to them.

Today, though, the car-free urban spaces that people like Moses thought of as old fashioned are celebrated as something we need to restore:

Another way to think about the 1870-1920 period, then, would be “the streetcar era.”

This window of time (between, say, 1881, when the first electric streetcar was trialled in Berlin and 1920, when the first permanent automobile traffic lights were deployed in NYC) marked a still-unequalled high point for car-free transportation.

2: It’s appealing to see new technologies in a raw form

This is a bit counter-intuitive, because it seems to me that a lot of people who valorize the “old days” of cities see post-WWII modernism, especially Brutalist architecture, as a bad thing precisely because it embraced an unadorned reaction to new technological possibility.

But something that emerges really clearly when you look through photos of architecture and cities from the decades before the World Wars is how much it has in common with the basic motivation behind Brutalism — experimentation with novel technological affordances and unadorned materials.

For instance, here is the Machinery Hall from the 1889 Paris Exposition (later used for much the same purpose in the 1900 Exposition, where Henry Adams had his rapturous encounter with the dynamo).

To eyes that have grown up with post-1970s minimalism, this doesn’t look particularly stripped-down. Look at all those girders! All that ornament on the roof!

But when you consider what was technologically possible at the time, when all-metal construction was brand new and extremely difficult, this is a minimalist aesthetic stretched to its very limit.

Here is the original inspiration for this style of building: London’s Crystal Palace of 1851. Today, it looks almost ornate. But when it was built, it would have been a miracle of transparency. The space consciously evokes a medieval cathedral — but smoke-darkened stone has been replaced by clear blue sky.

Granted, any place called something like “The Machinery Hall” or “The Crystal Palace” is of course going to be inordinately technological and future-focused in its design. But even ordinary urban spaces and buildings from these years have a sort of Jules Verne quality to them. You can see the designers actively experimenting with the affordances of new materials and building technologies.

For instance, here is one of my favorite structures in Lisbon, Portugal: the Elevador de Santa Justa (1901). You’ll notice Gothic arches and quite a lot of ornamentation throughout. But they’re subsumed within a form that is entirely a product of new technologies: all-iron construction plus experimental, steam-powered elevators.

The Sugarloaf Cable Car, built in Rio de Janeiro in 1910, is another great example of the era’s experimentation with new technological affordances and pre-car mass transportation systems.

3. Why we can’t (and shouldn’t) go back

It’s tempting to look at Belle Époque architecture and urban planning as a lost paradise we should strive to reclaim. But nostalgia, as always, can distort more than it reveals.

The truth is, for the majority of people living through the 1870 to 1920 period, life was brutal. Social inequality was so extreme that it challenges our modern sense of scale. Economic measures like the Gini coefficient illustrate only part of the story:

Calculating shares of income and how they are distributed in a population is a useful metric for getting a sense of ordinary people’s lives in this era. But so too are factors like “does slavery exist?” and “are women able to vote?” or “what percentage of children die in infancy?”

Take the craftspeople responsible for the exquisite detailing of Belle Époque buildings, for instance. Stucco moldings, hand-carved wooden pediments, intricate wrought-iron balconies — these were all the products of highly skilled manual labor. But how many of these artisans could read a newspaper, cast a ballot, or realistically expect all of their children to live to adulthood? How many had meaningful choices about where or how they lived?

Historian Daniel Immerwahr’s recent article in Past & Present, “All That Is Solid Bursts Into Flame,” drives home another reality: nineteenth-century cities were shockingly prone to catastrophic fires, due in large part to a lack of basic regulations and safety infrastructure. An urban blaze was a routine threat that could level entire neighborhoods overnight.

In this context, the iron-and-glass experiments of the various International Exhibitions and World’s Fairs were not just delightful works of art. They were also very practical attempts to discover new building methods to forestall the risk of catastrophic fire.

And there’s factor at work here, too. When we admire Belle Époque urbanism today, we are largely experiencing spaces designed for, and preserved by, the era’s upper classes. True, some projects like the New York subway or the Paris Métro genuinely improved mass urban life. But the lion’s share of structures we revere today from the “beautiful era” — department stores, opera houses, beaux-arts museums, government buildings, World’s Fair sites — served an audience of wealthy elites, or at best an emergent middle class.

When we think about the charming, car-free boulevards of 1890s cities, in other words, we might remember that they were in some sense like like the locked-down Tate Modern Museum depicted in the great Alfonso Cuarón film Children of Men. Beauty in abundance, but armed guards with dogs just outside.

In short, many of the factors that led to the decline or demise of the “beautiful era” of global cities were, on net, good things for humanity.2

All that said… can we please start building more things like the Palace of Fine Arts?

A note to readers

As I begin work on my new book, I am going to be sharing interesting things I find during my research (like this post) with paid subscribers. I’ve been writing Res Obscura entirely on my own, for free, as a labor of love since 2012. To support that work and receive occasional subscriber-only posts, please sign up as a paid subscriber:

Subscribe now

Weekly links

• The Daniel Immerwahr article I referenced above, on fire and capitalism in the 19th century United States, is remarkably good, IMO. It also has one of my favorite opening lines to an academic journal article ever: “In hindsight, yes, the walls of the whale tank should have been thicker. That was regrettable. So was the construction of the snake cage, which, when the fire started, broke apart and sent its inhabitants slithering down the museum stairs and onto Broadway.” Oxford University Press has made the article open access, so you can read the whole thing for free here. Recommended.

• “Rutherford developed the game over four to six weeks in late 1975 as a computerized take on the Dungeons & Dragons role-playing game. Its file was named pedit5, as it was stored in the fifth lesson slot of the school's Population & Energy group. It is considered to be the first example of a dungeon crawl video game and one of the first computer role-playing games” (From the Wikipedia article for pedit5)

Leave a comment

Share

In one department on campus, a recently drafted anti-A.I. policy, read literally, would actually have barred faculty from giving assignments to students that centered on A.I. (It was ultimately revised.) Last year, when some distinguished alums and other worthies conducted an external review of the history department, a top recommendation was that we urgently address the looming A.I. disruptions to our teaching and research. This suggestion got a notably cool reception. But the idea that we can just keep going about our business won’t do, either.

On the contrary, staggering transformations are in full swing. And yet, on campus, we’re in a bizarre interlude: everyone seems intent on pretending that the most significant revolution in the world of thought in the past century isn’t happening. The approach appears to be: “We’ll just tell the kids they can’t use these tools and carry on as before.” This is, simply, madness. And it won’t hold for long. It’s time to talk about what all this means for university life, and for the humanities in particular.

Subscribe now

I suspect that a significant chunk of my historian colleagues had a negative reaction to this article. But I wholeheartedly agree with the central point Burnett makes within it — not that generative AI is inherently good, but simply that it is already transformative for the humanities, and that this fact cannot be ignored or dismissed as hype.

Here’s how I’m currently thinking about that transformation.

Generative AI elevates the value of humanistic skills

Ignoring the impact of AI on humanistic work is not just increasingly untenable. It is also foolish, because humanistic knowledge and skills are central to what it is that AI language models actually do.

The language translation, sorting, and classification abilities of AI language models — the LLM as a “calculator for words” — are among the most compelling uses for the current frontier models. We’re only beginning to see these impacts in domains like paleography, data mining, and translation of archaic languages. I discussed some examples here:

… and the state of the art has progressed quite a bit since then. But since this is one aspect of AI and humanities I’ve written about at length, I’ll leave it to the side for now.

Another underrated change of the past few years is that humanistic skills have become surprisingly important to AI research itself.

One recent example: OpenAI’s initial fix for GPT-4o’s bizarre recent turn toward sycophancy was not a new line of code. It was a new piece of English prose. Here’s Simon Willison on the change to the system prompt that OpenAI implemented:

This was not the only issue that caused the problem. But the other factors in play (such as prioritizing user feedback via a “thumbs up” button) were similarly rooted in big-picture humanistic concerns like the impact of language on behavior, cross-cultural differences, and questions of rhetoric, genre, and tone.

This is fascinating to me. When an IBM mainframe system broke down in the 1950s (or a steam engine exploded in the 1850s), the people who had to fix it likely did not spare a moment’s thought to consider any of these topics.

Today, engineers working on AI systems also need to think deeply and critically about the relationship between language and culture and the history and philosophy of technology. When they fail to do so, their systems literally start to break down.

Then there’s the newfound ability of non-technical people in the humanities to write their own code. This is a bigger deal than many in my field seem to recognize. I suspect this will change soon. The emerging generation of historians will simply take it for granted that they can create their own custom research and teaching tools and deploy them at will, more or less for free.

My own efforts so far have mostly been focused on two niche educational games modeled on old school text-based adventures — not exactly something with a huge potential audience. But that’s exactly why I choose to do it. The stakes were low; the interest level for me personally was high; and I had significant expertise in the actual material and format, if not the code.

The progression from my first attempt (last fall) to my second (earlier this spring) has been an amazing learning experience.

Here’s the first game (you can find a free playable version here). It’s a 17th century apothecary simulator that requires students to read and utilize actual early modern medical recipes to heal patients based on real historical figures. You play as Maria de Lima, a semi-fictional female apothecary in 1680s Mexico City with a hidden past:

It was fascinating to make, but it also has significant bugs and usability issues, and it fairly quickly spools out into LLM-generated hallucinations that are unmoored by historical realities. (For instance, in one play-through, I, as Maria, was able to become a ship’s surgeon on a merchant vessel sailing to England, then meet with Isaac Newton in London. The famously quarrelsome and reclusive Newton was, for some reason, delighted to welcome me into his home for tea.)

My second attempt, a game where you play as a young Darwin collecting finches and other specimen on one of the Galápagos Islands in 1835, is more sophisticated and more stable.

The terrain-based movement system, with specific locations based directly on actual landscapes Darwin wrote about in his Voyage of the Beagle, forces the AI to maintain a kind of literal ground truth. It is difficult to leave the island, and the animals and terrain you encounter are pulled directly from the actual writings of Darwin, reducing the tendency to hallucinate.

There is also a more robust logging system which will come in handy when I want to add an assessment layer to the game and turn it into an actual assignment. You can play Young Darwin here.

My idea is that students will read Darwin’s writings first, then demonstrate what they learned via the choices they make in game. To progress, you must embody the real epistemologies and knowledge of a 19th century naturalist.

The crucial thing is that this would be done alongside an in-class essay and in-person discussions of the reading — it would not replace, but augment the human element of teaching.

I’ll write more about all this in a future post, but the upshot is that this iterative process has been among the more intellectually challenging and enriching experiences of the last few years for me. Anyone who thinks you can’t learn from interactive tutoring by an AI has not tried. You absolutely can.

Subscribe now

Generative AI makes it harder to teach humanistic skills

On the other hand, it is just a brutal fact that AI chatbots are significantly damaging core aspects of the educational system. There’s no denying it, and it needs to be taken seriously by educators, students, politicians, and above all by the frontier AI labs themselves.

Educators tend to point to the ways ChatGPT and its competitors have affected us — eroding our ability to accurately assess student writing because such a large proportion of students turn in machine-generated essays, and forcing us to come up with entirely new assignments and lesson plans as a result.

But in the longer run, the damage is being done to students. By making effort an optional factor in higher education rather than the whole point of it, LLMs risk producing a generation of students who have simply never experienced the feeling of focused intellectual work. Students who have never faced writer’s block are also students who have never experienced the blissful flow state that comes when you break through writer’s block. Students who have never searched fruitlessly in a library for hours are also students who, in a fundamental and distressing way, simply don’t know what a library is even for.

New York Magazine has a new article on student use of ChatGPT which captures the problem well. Here’s a Columbia student who speaks for a significant chunk of the current university population:

“Most assignments in college are not relevant,” he told me. “They’re hackable by AI, and I just had no interest in doing them.” While other new students fretted over the university’s rigorous core curriculum, described by the school as “intellectually expansive” and “personally transformative,” Lee used AI to breeze through with minimal effort. When I asked him why he had gone through so much trouble to get to an Ivy League university only to off-load all of the learning to a robot, he said, “It’s the best place to meet your co-founder and your wife.”

I will speak frankly. This sucks.

It sucks the joy out of teaching, and it sucks the meaning out of the whole experience of getting an education.

When I was a postdoc at Columbia, I taught one of the core curriculum classes mentioned here, with a reading list that included over a dozen weighty tomes (one week was spent on the Bible, the next week on the Quran, another on Thomas Aquinas, and so on). There wasn’t much that was fun or easy about it. And yet — probably for that very reason — I learned more from teaching that class than from any other. Something fundamental about that experience feels like it’s ruined now.

But this is not the entire story. The middle part of D. Graham Burnett’s New Yorker piece strikes me as an important corrective to this sort of thing. Burnett is, I think it’s fair to say, rapturous about his students’ response to an assignment asking them to discuss the concept of attention with ChatGPT, then edit and submit the results.

Here’s a sample:

Reading the results, on my living-room couch, turned out to be the most profound experience of my teaching career. I’m not sure how to describe it. In a basic way, I felt I was watching a new kind of creature being born, and also watching a generation come face to face with that birth: an encounter with something part sibling, part rival, part careless child-god, part mechanomorphic shadow—an alien familiar.

I have had the same feelings, for instance when I first began tinkering with history simulation assignments. Language models are a genuinely novel teaching tool. Their impact is still uncertain. What that means is that now is exactly the time when people who are genuinely passionate about teaching and learning for its own sake — not as a scorecard to judge politicians, not as a source of corporate profit — need to take an active role.

Share

My greatest concern when it comes to LLMs in humanities education is that they will lead to a further polarization in educational outcomes. The Princeton students who Burnett teaches seem extraordinarily thoughtful and creative in their responses to his assignment. I suspect students in a social studies class at an underfunded public high school class would not be.

For this reason, it is vitally important that educators learn how to personally create and deploy AI-based assignments and tools that are tailored directly for the type of teaching they want to do. If we cede that ground, if we ignore the challenge, then we will watch helplessly as education gets taken over by cynical and stultifying “AI learning tools” which trumpet their interactivity while eroding the personalized student-teacher relationship that is at the heart of learning.

This is the basic thinking behind an NEH grant which I and two of my UCSC colleagues, Pranav Anand (linguistics) and Zac Zimmer (literature), were awarded in January of this year… and which got cancelled by the Trump administration/DOGE last month.1 We are continuing our planned work, and I’ll keep writing about it here.

I’d love to hear your thoughts in the comments, and please consider supporting my work via a paid subscription if this is an option for you.

Weekly links

• D. Graham Burnett’s book The Sounding of the Whale, is, incidentally, the most unusual and delightful book about the history of cetacean science you will ever read. I relied on his chapter about John C. Lilly extensively when I was writing the “dolphins on LSD” part of Tripping on Utopia.

• “The fragmentary letter was found preserved inside a 1608 book by Johannes Piscator, an almost 1,000-page tome dissecting biblical texts. The letter was not tucked into the pages as bookmark or memento but was part of the book’s very construction, as strips of wastepaper deployed as padding to prevent the text block from chafing against the binding.” New findings about Shakespeare’s relationships with his wife, Anne Hathaway (Washington Post).

• Congratulations to the UNC historian Kathleen DuVal, whose most recent book Native Nations: A Millennium in North America won the Pulitzer prize for best work of history this week. DuVal’s The Native Ground (2006) is among the more interesting history books I’ve ever read, and I’m looking forward to reading her new book.

Subscribe now

Leave a comment

Subscribe now

His doodles include horses and soldiers, battle scenes, self-portraits, and expressive stick figures that radiate personality across eight centuries.

In the most famous Onfim drawing, a mighty warrior rides horseback, his lance spearing a prostrate enemy on the ground. This figure is, naturally enough, labelled “Onfim.”

To the upper right, an attempt at writing the alphabet trails off around the Cyrillic letter K.

Was this capturing the very moment when boredom gave way to daydream?

Another fragment shows Onfim next to a larger figure, with the caption: “This is my father.” Both father and son are drawn in a style that psychologists today might categorize as transitional between “tadpole” and conventional stick figures—each face cheerful, with disproportionately long limbs and spiky fingers fanning out like tiny rakes.

As the anthropologists René Baldy and Daniel Fabre argue in their thoughtful analysis of Onfim's drawings, Onfim’s work demonstrates that medieval children were capable of remarkable expressiveness. Onfim’s sketches suggest deliberate storytelling—a “graphic narrative” unfolding scene by scene, using shifts in scale, perspective, and emotional expression. One set of drawings is even a kind of proto-comic book, with a rider approaching from the right, then growing larger as he slays an opponent at center stage, and finally departing to the left.

Share

Onfim’s peers

In “Centuries of Childhood,” I wrote about how modern Westerners tend to systematically underrate the sheer variety of parenting practices throughout history. But of course, there is also a flip side to this: the universality of childhood. Anyone with children, or anyone who remembers being a child, will recognize Onfim and his world. It has barely changed at all.

Because of Onfim’s reoccurring online virality (there’s an article about him virtually once a month), we tend to forget that he wasn’t some unique one-off. As Baldy and Febre relate in their article, another, rather darker set of child drawings from around 1150 CE was discovered in an abandoned iron mine near Sorèze, in southern France.

There, child miners had etched charcoal images on cavern walls. Unlike Onfim’s playful schoolwork, these figures have an unsettling poignancy: children drawing themselves in their work attire, some wielding picks, others with sorrowful eyes staring into darkness.

These images were less imaginative flourishes than signatures. Perhaps they were a way for children to assert their presence in an environment hostile to childhood itself.

Taken together, Onfim’s whimsical daydreams and the charcoal drawings of the Sorèze miners compel us to see the premodern world from a different perspective: one that was about three feet tall.

There are, of course, several more famous child-authored artworks from the medieval and early modern periods. Probably the most famous are these two self-portraits by two extremely talented fifteenth-century teenagers: Albrecht Dürer and Raphael.

But these belong to a very special category of child art: the juvenilia of the genius.

There’s also another, similar category: children’s doodles left behind on the papers of famous people. My favorite examples being these drawings of vegetable battles that Charles Darwin’s children left scrawled on some of the pages of his original manuscript for The Origin of Species:

For every unique survival like these, there are many millions that are lost. Children are continually making creative marks (as any parent with young kids knows all too well — cleaning them up has been a central preoccupation of my life over the past couple years). If we assume this is a human universal, and I think we can, then it is a bit staggering to imagine just how much child artwork has ever been created.

Here’s a back of the envelope calculation: there have been around 100 billion humans, and the average kid makes about 1-2 drawings (or other forms of artistic mark-making) each day for a period of, say, 10 years.

That adds up to over 500 trillion individual artworks by children. Surely that’s not accurate, at all. But it’s likely roughly the right order of magnitude.

As for Onfim, although archaeologists have been aware of him since the first Soviet digs in the 1950s, it was really only until the era of the Internet that his drawings became widely recognized.

I’m glad they are. They force us to see childhood not as something that takes place in an eternal present, but as something with as deep a history as anything else that is a central part of what it is to be human.

Weekly links

• It’s surprisingly hard to find a full archive of Onfim’s drawings online, but his Wikipedia page and the article by Baldy and Fabre (n French) is a good place to start. This archive in Russian seems to be a searchable digital repository of the Novgorod medieval birchbarks (beresty).

• “Yet Ferriss’ forward-looking vision also repeatedly evoked the ancient past. The pyramid skyscraper that he promoted was, in his own description, a ‘modern ziggurat,’ the monumental architectural form of ancient Mesopotamian cities.” (from a new essay with lots of great images of early monumental skyscraper designs in The Public Domain Review)

• In true 18th century style, is offering a $10,000 essay prize. Here’s the link.

• And here’s a 2017 Cabinet Magazine article on Onfim and the Novgorod birch mark texts from the proprietor of The Hinternet, Justin Smith-Ruiu.

Leave a comment

Share

One reason I became a historian is the joy of encountering moments in the past that are foreign, yet also oddly familiar. These moments seem to ripple outward, lapping up against the present in unexpected ways.

Lately, I have been deeply dispirited by the current attack on academia here in the US — the deportation of graduate students and researchers, the dismantling of a huge range of important projects via funding cuts to NIH and elsewhere. And I’ve been reflecting on one of these moments of contact with the past. It happened in Buenos Aires in 1970. What I love about this exchange is that it takes place across the supposed “two cultures” of science and the humanities, jumping in a remarkably freewheeling way between literature, philosophy, and the fields of AI and machine learning, which were at the time barely even born.

The participants were Jorge Luis Borges, the great Argentine writer, and Herbert A. Simon, an economist, psychologist, and pioneer of artificial intelligence. Their conversation, part of which I’ve reproduced below, models something we badly need today: not just the simple joy of exploring ideas for their own sake, but also the ability to avoid seeing your work and life as something defined by what it excludes.

I rarely see historians and computer scientists (or people who write for literary magazines and people who work in AI labs) having this sort of conversation today.

I wish it happened more. That’s partly why I’m writing this newsletter.

Subscribe now

Share

“It’s a result of your past”

Their meeting took place in Borges’s office at the Argentine National Library. Simon had requested the meeting because of his admiration for Borges’s writing—his first and only “audience with a celebrity,” as he put it later . He was visiting Buenos Aires to lecture on systems theory, but what he really wanted was to speak with the author of “The Library of Babel.”

“I have been seized by an unaccountable wish to meet you,” Simon wrote in a letter to Borges. “I am a social scientist who tries to understand human behavior by building mathematical models (or, more recently, computer simulations).”

Once the conversation began, Borges quickly guided Simon away from casual pleasantries about translation and into deeper philosophical waters. Borges, in his gently ironic manner, asked:

BORGES: “Now I would like you to tell me a bit about the so-called behaviorism. What is its underlying principle? Is it free will or predestination?”

It is, and was, a hugely open-ended question. But Simon was at this time immersed in the study of early computer simulations, and he took this to be a question about whether the human mind is fundamentally comparable to a computer.

This was his response:

SIMON: Well, I always end up talking about computers. I’m in love with computers. We could put it this way. Faced with a certain problem, a computer will also behave in a certain way. And we may wonder whether it acted out of free will. We can say it employed free will in the sense that had it been programmed differently, it would have chosen to do otherwise, it would have behaved in a different way.

BORGES: What do you mean by behaved? Because we are talking about a purely mechanical process here. I mean, in the case of computers, of course.

SIMON: Yes, it’s a mechanical process. But I believe, like many others in my professional field, that human beings also display a mechanical type of behavior, much like computers.

BORGES: Do you mean to say that we act by force of habit?

SIMON: Rather accordingly to programs that we have stored in our brain. And we have free will in the sense that our resulting behavior will depend on who we are and the situation we are in. People respond differently when confronting the same situation.

BORGES: So, in your opinion, when faced with a dilemma, say, a situation in which there is a choice to be made between two possible behaviors, we can choose one of them?

SIMON: Your mental programming does the choosing. Yes, you choose.

It seems to me that Simon is here arguing for what philosophers call “compatibilism” — the idea that determinism can coexist with meaningful human choice and responsibility.

When Borges pressed him on whether an all-knowing being could predict our every action, though, Simon didn't shy away from the implications:

BORGES: Would this imply that if any all-powerful being, any god, knew everything about my past, my childhood, even about the time before I was born, my ancestors … would this imply that he would be able to predict my behavior in any one situation?

SIMON: According to my scientific beliefs, I would say so. With such knowledge, we can predict an individual’s behavior.

BORGES: So, what I’m saying right now is …

SIMON: … it’s a result of your past …

BORGES: … it’s inevitable.

SIMON: It is inevitable, yes. However, you still retain your identity, your individuality. You embody your own past.

BORGES: I understand. Well, I like to think I do. Now, does this account for all of our actions? That is, if my right hand is resting on my left hand, is it because it has to be this way? I believe people do quite a lot of things without any thinking.

SIMON: That’s the doing of our subconscious mind. You’re right, yes, otherwise, we would hardly be able to tie our shoelaces. Most things happen in this way. But that’s because we are heavily programmed.

BORGES: Would you say then things are to be also inevitable in this sense?

SIMON: They might be different, but always depending on programming. Any determinant could affect your programming and lead you to act differently. And if we introduce chance into the picture, scientists will always ultimately rule it out. At some point they may have to admit their inability to explain a particular phenomenon, but they will keep on working on the assumption that actions are determined by certain causes. And therefore, when we study a person who is in the process of solving a problem, we start from the assumption that every little thing has a cause. We are not always able to identify those causes.

BORGES: Well, of course. In order to study a person’s behavior, professionals have to go back to that individual’s history. Even to the historic past, to the origin of humanity, the cosmos.

SIMON: No, it’s not like that. Because the past influences a person’s present behavior to the extent that this past is already in the person. So, we can always find a starting point.

What I love about this is that here we have two significant thinkers, in two very different fields, who are both reflecting on not just on the meaning and importance of history (“it’s a result of your past”) but also about how and whether they have, themselves, been able to act within history.

If you compiled an enormous dataset of everything Borges read, and combined it with an exquisitely sensitive record of every sensory experience he ever had, could you create a Borges LLM? It’s a question that I think a lot of my peers in the humanities would not take much interest in, today (the two cultures, again). But I love that Borges himself was thinking about it.

Lately I have been spending a good amount of time thinking about whether, and in what ways, there can be value in seeking to simulate historical figures (or even entire historical moments — like the collective medical opinion of 1820s London, say) by combining a large amount of historical primary sources with an advanced language model.

I think this will absolutely be a powerful tool. I’m just not sure when and how. What I do know is that the people who figure that out will be a team that combines humanities and STEM. It is encouraging to see Simon and Borges embodying that kind of thinking way back in 1970.

I have been writing Res Obscura for 13 years now, and it has always been unpaywalled and free of any advertisements. The generosity of paid subscribers makes this possible.

My mother’s journals

Someday, when the feeling is a little less raw, I’ll write about the hundred or so journals that my mother left behind when she died. It was during Covid, and my wife was in her second trimester of pregnancy with our older daughter, Yara. My mom had been diagnosed with an aggressive form of cancer and was in a lot of pain. She was incredibly brave throughout, and, as a lifelong diarist and writer, she kept writing in her journal up to the end.

This is one of the last passages in her last journal. It’s about her hoping to hold on long enough to see Yara (who she ended up missing, alas):

You embody your own past.

I haven’t had the heart to actually read my mother’s journals yet. I only have this photo because she showed it to me and wanted me to share with Yara when she got older. Someday, probably far in the future, I will not only read them, but experiment with using them as data for an AI system which, to some extent, might embody some sort of spectral trace of my mother and her past.

She wanted her journals to be read and understood. She wanted to be able to talk to Yara. Someday, “she” will. Or, at least, a simulacrum composed of the collection of thoughts and dreams and impressions that she chose to record will.

That day is not even close today. But it will happen. And when it does, I’ll think about Borges and Simon in Buenos Aires in 1970.

The full dialogue is reproduced here, in an article in The Journal of the History of Economic Thought by Ricardo F. Crespo.

Subscribe now

Share

Weekly links

• “They discovered that the Jahai, hunter-gatherers living at the border of Malaysia and Thailand, have a rich vocabulary of abstract smell words. One Jahai term, itpit, refers to the ‘intense smell of durian, perfume, soap, Aquillaria wood, and bearcat’… Jahai and English speakers were asked to identify and name twelve smells, including cinnamon, turpentine, gasoline, and onion. English speakers, despite their greater familiarity with the odors, faltered. They mostly gave rambling source-based answers and showed almost no agreement among themselves. One English speaker presented with cinnamon responded, ‘I don’t know how to say that, sweet, yeah; I have tasted that gum like Big Red or something tastes like, what do I want to say? I can’t get the word. Jesus it’s that gum smell like something like Big Red. Can I say that? Ok. Big Red. Big Red gum.’ But Jahai speakers named smells with relative ease.” (From this great New Yorker article by UC Davis anthropologist Manvir Singh on how language shapes thought)

• As a longtime fan and sometime contributor to Lapham’s Quarterly, I was happy to see that it has been revived: more here.

• “In comments to The Washington Post, the Brazilian president went further, delving into the minutiae of early 20th-century aerospace engineering and mourning what he described as the wrongful denial of Brazilian valor. ‘Everyone knows that Santos Dumont was the first to make something heavier than air fly, in an autonomous way, without any assistance,’ he vented. ‘But the Americans have the movie industry and were able to promote the Wright brothers.’” (The Washington Post)

Leave a comment

Share

If your job is legible enough that people can make a dataset clearly pointing out what is right and what is wrong, you are at the highest risk for an AI model being ‘superhuman’ at your job. It is even more risky if it is possible to articulate your thought process in a way that is verifiable.

Looking at this perspective, it makes more sense that competitive programming and mathematics have been attacked first by these models. Not only do they provide a clear source of ground truth, you can also hire people to give intermediate steps and there are tools like proof assistants to tell you if the proof is correct or not.

DeepSeek has shown that reinforcement learning (RL) remains extremely effective at advancing the state of the art in AI models. What this means is that we can expect continued or even accelerated improvement in domains where reinforcement learning works.

But this is not every domain of knowledge. RL requires a clearly defined set of correct answers (and all the better if those answers can be broken down into discrete steps). This lends itself very well to math and coding, as well as to translation and to the sorts of GRE-style questions used in many current benchmarks like Humanity’s Last Exam.

Given this, Prasad arrives at the following conclusion:

The best placed people are those for whom it will be unprofitable or extremely difficult to create or procure datasets that give a clear signal.

No one can predict where AI models will go in the next few years, and I remain deeply skeptical of anyone who claims to be able to do so (this includes both the boosters promising utopian AGI, the alarmists who predict global catastrophe, and the skeptics who dismiss the field as hype).

However, I do think this line of thought suggests a more moderate, and to me more plausible, medium-term future which I haven’t seen discussed much. It looks like this: AI does not lead to a singularity, but it does exceed human abilities in domains like math, coding, reasoning, and translation, because those forms of knowledge lend themselves to reinforcement learning.

AI also gets shockingly, dismayingly good at making “slop” content — if we measure “good” as “someone wants to pay money for it,” since that, unfortunately, is also a reward function RL can optimize for. Any knowledge domain where you can shovel up a lot of digital data, break it up into parts labelled true and false (or “humans liked this” versus “humans didn’t like this”) is fodder for AI systems. Think social media posts and the lower grades of fiction or streaming shows.

But where AI may hit an unexpected wall is any domain which depends upon some combination of intuition, physical presence, and illegible data.

Like, say, history.

Subscribe now

Intuition and the spectrum of legibility

I mean something sort of specific when I say “intuition” here, and it’s rooted in the material experience of being a historian. Bear with me for a second and I will explain.

Historians cultivate a specific form of intuition. Nothing is well defined when you begin a historical research project. You go into an archive with a question or topic in mind, but without an answer or a clear argument. You leaf through documents or do word searches in databases without a real sense of what you will find — often, the finding aids are either nonexistent or they tell you almost nothing.

Over a period of days, weeks, or months, you start to build up a feeling for what’s in the archive: who constructed it, how it’s organized, what matters within it, and (importantly) what was intentionally or unintentional excluded from it. In my experience, this really is more of a vague sensibility than something you can quantify. It’s closer to David Lynch and Angelo Bandalementi stumbling onto the Twin Peaks theme than it is to a data scientist making a spreadsheet.

Sure, you can say that a given archive contains 14,523 letters sent from colonial Brazil to Portugal in the 18th century. You could even digitize all of those letters and ask an LLM to find themes in them and output it as JSON data.

But there is something very personal in the way that your brain, individually and idiosyncratically, remembers and links together the things you find inside those letters. It is not just about textual or even visual data but about sense memory, social relationships, and even what you dream about. It is about intuition grounded in material facts.1

A brief example

Let’s say there’s one particular letter that stands out from the others. In it, a soldier in 18th century Rio de Janeiro begs to return to Portugal because he’s sick and badly needs his mother (yes, this is a real letter I remember reading, in Lisbon in the spring of 2012).

It sticks out in your mind because, well… it just does. If asked to give a reason, it would be something like “it was weird” or “it was tonally different from the other documents in the folder.” Individually, there isn’t a lot you can do with it. But as you spend time in an archive, you start making semi-conscious or even unconscious connections between the things that “jump out” in this way.

Over long periods of time spent in physical archives, you develop a holistic picture of a slice of history that is necessarily individual to you, as the researcher, because you can’t control or even fully understand what your brain finds interesting in a set of documents. They’re just… interesting. Maybe not to anyone else, at first. But over time, you start to draw connections and explain to yourself why you found a certain source compelling or thought-provoking. You create an account of those connections, a sort of map of them.

And that, in turn, reveals the beginning of both a narrative and a causal argument.

I have been writing Res Obscura for 13 years now, and it has always been unpaywalled and free of any advertisements. The generosity of paid subscribers makes this possible.

Could you make a dataset that quantifies exactly how and why some historical data “jumps out” at historical researchers in this way? Maybe. But it sure as heck would fall under both the “unprofitable” and “extremely difficult” headings that Prasad mentions.

As for data illegibility: by that I simply mean “data which AI systems cannot plausibly access.” You can assume that current and future reasoning models will be able to trawl through all, or most, of the Internet. But a very large percentage of all words ever written remain undigitized.2 And naturally, the proportion is even higher for spoken words. The vast, vast majority of human knowledge, in other words, is still inaccessible to AIs, so long as they rely on digital media as their source of ground truth.

Which would naturally suggest…

Non-digital data just got more important

A key aspect of archival research is how much it depends on social skills. In some cases one might even say social engineering. My friend Felipe befriended an archivist in São Paulo — something he was able to do because he was from São Paulo, and spent months there during his PhD research — and ended up being given the keys to the archive. He volunteered to catalog materials previously unavailable to anyone, and used those in his dissertation. Until AI models can move physically in the world, interact with people in this way, unlock new data physically and materially, physical archives and the intuition for finding and deciphering illegibility will still be super important.

For Tripping on Utopia, one of the biggest revelations of my research process was realizing that you can just create new sources by calling someone on the phone. I knew this intellectualy, of course. I had read plenty of books based on oral history interviews, and I knew that old-school investigative journalism was all about convincing potential sources to share what they knew. But because my dissertation research was all about the early modern period, I had never tried it myself. It turned out to be remarkably easy to track down and interview people mentioned in historical texts.

Understanding the history of tariffs and financing in the Gilded Age may seem like a viable task for a Deep Research tool. But it will only give you a starting point… and for a serious historian be more like the negative space to avoid. This is because you’re trying to find, not the Wikipedia page for JP Morgan, but digitally “illegible” things like:

• unpublished letters marked with notes like “to be destroyed” or “burn after reading this”

• the recollections of his bitterest enemies (including things like unarchived diaries or oral histories)

• the physical contents of his drawers when he died

• The physical and sensory impressions of entering his house and library

Hernan Diaz’s 2022 novel Trust did a fabulous job of showing this. In its first half, Trust is a formulaic roman a clef novel about a wealthy financial wizard on 1920s Wall Street, then a self-satisfied memoir by the same figure. But this narrative then continues to fracture, and we see the same story from the perspective of the financier’s wife — institutionalized in a Switzerland sanatorium — and one of his employees. These jumps in perspective are the best approximation I’ve ever seen of what it’s like to study someone in an archive.

The issue is that generative AI systems don’t want messy perspective jumps. They want the median, the average, the most widely-approved of viewpoint on an issue, a kind of soft-focus perspective that is the exact opposite of how a good historian should be thinking.

To be in an archive is to be confronted with the contradictions, controversies, secrets, and unspoken facts of a person’s life or of a time and place.

To encounter these things via a tool like Deep Research is a surreal experience, because you get an approximation of the historical reality, but with all the illegible data smoothed away.

In fact, it isn’t even being excluded: it just isn’t being noticed.

Find the negative space

OpenAI’s Deep Research, which I’ve been experimenting with for the past couple weeks, is a great example of why a field like historical research seems replaceable, but isn’t. Something like ninety percent of the world’s data is not online. Where is it found? Archives, in part. Oral histories. Operating in the world. Having intuition about the world and the non-digital data within it.

Which is why I remain optimistic about this, and why, in particular, I think it would be an enormous mistake for humanists to form ranks against AI (even if many critiques are legitimate, as I believe they are). We have a very important role to play in shaping a non-dystopian future for it.

Subscribe now

As a young adult, I wanted to be a figurative oil painter before I decided to become a historian. Although I was never really any good at painting, some of the things I learned from professional painters have stuck with me. I think all the time, for instance, about the advice of a painting teacher I had when I was 19.

The gist of what he told us boiled down to a set of guidelines for creative work in general:

• You’ll never actually feel finished with anything. You just have to pick a time to stop working on it, then call it done.

• What you subtract from a picture is usually more important than what you add to it.

• To capture the essence of a thing, you have to find the negative space around it.

The reason I still think about these guidelines is that they shape how I approach writing — and, increasingly, how I interact with AI systems. Generative AI has made it absurdly easy to generate a lot of text or images. But it hasn’t made us any better at subtracting the useful, meaningful, or simply interesting stuff from text and images — isolating the gold from the fool’s gold.

And this is precisely why historical analysis is important in an era of AGI. As a field, we now can more clearly see the negative space around what is legible to AI systems, and work within those constraints to do and think things that even hyper-advanced machines trained exclusively on digital data cannot.

In other words: physical research and critical thinking about the nature of archives and qualitative data just got way more important. Not less.

Share

A quick poll

I have about 30 posts half-drafted for this newsletter, and I am trying to decide which to finish. I tend to write about a pretty eclectic range of topics — the most popular recent ones have been about coca in the seventeenth century, my most recent post on AI in historical research, and the history of child-rearing.

I love that I have been able to build a decent readership (6,500 subscribers as of today) by writing about whatever I find interesting. But it’s also helpful to get a sense of what my readers want more of. So I thought it would be useful to do a poll. Here are five posts I’ve got unfinished drafts of.

Which would you most like to read, based on the titles?

The first post is about the odd convergence of folklore from Europe and the Caucasus involving bees, consciousness, and the afterlife. The second is about Jorge Luis Borges discussing AI and free will with Herbert Simon in Buenos Aires in 1969. The third is more of an old school Res Obscura post about Sibero-Scythian art and archaeology. The fourth is a mailbag post with reader comments and my responses to “The leading AI models are now good historians.” And the fifth would be a preview and development update of an educational history simulation game I’m building for classroom use, where you “play” as Darwin on one of the Galápagos Islands in 1835.

Weekly links

• “In his last years, he was known as the rarest creature in the world.” (Wikipedia entry for Lonesome George)

• Great article by Simran Agarwal on my favorite Mughal emperor, and one of my favorite early modern people in general, Jahangir, and his love of clothing (Public Domain Review)

• Yesterday my book Tripping on Utopia won the PROSE award from the American Association of Publishers for best book on the history of science, medicine or technology published in 2024. Thank you to the AAP!

• “Caro leafs through the pages, and it all starts coming back. He points to a passage relating to some moment of New York political intrigue too arcane to even make his book (about Moses’ 1929 investigation of the City Trust Company). ‘This was a thing no one knows about—I had to decide to leave it out,’ he says, as though he still rues every such omission.’” (From a profile of Robert Caro’s archive in Smithsonian)

Subscribe now

Leave a comment

Share

But I did want to make a small request: if you have any interest at all in supporting Res Obscura financially, this week is a great time to do so. That’s because I have instituted a special Lupercalia discount of 33% off the paid rate for monthly ($4/month) and annual ($40/year) subscriptions.

I’ve written Res Obscura (first as a blog, and now as a newsletter) for fifteen years now. In that period I have written roughly 500,000 words and reached about the same number of readers. None of it has been pay-walled, and I have never accepted any advertising. Your support via these optional paid subscriptions makes a huge difference in allowing me to keep writing. I’m currently at 40 paid subscriptions, and will start sending out special posts for subscribers once I hit 50.

Please help me get there by the end of Lupercalia on the 15th of February (or better yet, by Valentine’s Day):

Subscribe now at the discounted rate

What is Lupercalia? I wrote about the wolf-themed Roman holiday — and the links further back to the wolf cult of Mount Lykaios in Greece — back in 2011, in one of the earliest Res Obscura posts to get any traction.

Here’s a snippet from it:

The story begins with the legend of King Lycaon (“The Wolfish One”). According to Ovid's Metamorphoses (translator A.S. Kline), Lycaon was a primitive lord of Arcadia in the earliest era in which humans walked the earth, “when the constellations that had been hidden for a long time in dark fog began to blaze out throughout the whole sky” (I: 68). Yet Lycaon was a king of the Iron Age, when “honour vanished” and “pernicious desires” reigned. His infamous crime was to attempt to trick Zeus into eating the limbs of a dismembered child in order to test his omniscience. But Zeus was not fooled:

“No sooner were these [limbs] placed on the table than I brought the roof down on the household gods, with my avenging flames, those gods worthy of such a master. [Lycaon] himself ran in terror, and reaching the silent fields howled aloud, frustrated of speech. Foaming at the mouth, and greedy as ever for killing, he turned against the sheep, still delighting in blood. His clothes became bristling hair, his arms became legs. He was a wolf, but kept some vestige of his former shape. There were the same grey hairs, the same violent face, the same glittering eyes, the same savage image. [I:210-243]

Several classical authorities linked the wolf-rites of Mount Lykaion to the more famous Roman holiday of Lupercalia. This festival, which spanned February 13-15, was a fertility rite intended to purify and protect the city of Rome. The rite centered around the Lupercal, which was supposedly the very cave in which the she-wolf had suckled Romulus and Remus, mythical founders of Rome.

An Italian archeologist announced in 2007 that the remains of the Lupercal were believed to have been found directly beneath the ruins of the palace of the Emperor Augustus.

The priests of Lupercalia were the Luperci, or "Brothers of the Wolf," who Cicero described as “a wild association... both plainly pastoral and savage, whose rustic alliance was formed before civilization and laws” (Cael. 26). These priests were said to smear blood upon the foreheads of the noble youth of Rome and dress them in shaggy wolfskins. The youths were required to laugh, and seem to have gone into a sort of frenzy. The great historian Plutarch noted the connection with the Arcadian Wolf-Zeus ceremonies and added,

At this time many of the noble youths and of the magistrates run up and down through the city naked, for sport and laughter striking those they meet with shaggy thongs. And many women of rank also purposely get in their way, and like children at school present their hands to be struck, believing that the pregnant will thus be helped in delivery, and the barren to pregnancy. [Life of Caesar, 61]

Thus from 2011. This is now me writing today in February, 2025.

I should note that Lupercalia has no proven connection to Valentine’s Day, and I heartily repudiate the part of the post where I tentatively suggest a link. (I was a little bit more of a freethinker about such things back when I was still in my 20s). I’m a big believer in the long-lasting nature of cultural patterns, not all of which get written up in surviving historical sources, but the multi-century jump from Lupercalia, which ended in Late Antiquity, to the earliest connections of Valentine’s Day with romantic love in the 14th century (for instance, Chaucer’s famous reference in The Parliament of Fowls) is just too much of a stretch.

Interestingly though, the historian Edward Muir, in his book on ritual in early modern Europe, does suggest a link to a Christian holiday. But it’s to the feast of the Purification of the Virgin (Candlemas) earlier in February:

This season connected the pure fertility of the Virgin, who gave birth to the light of the world, to the onset of spring (at least in the Mediterranean) with its promise of the regeneration of nature. Insome ways an amalgam of the Pagan Feast of Lights (February 1) and the Roman Lupercalia (February 15), the Christian purification rituals retained associations with lustration, fire symbolism, and fertility (Muir, Ritual in Early Modern Europe, 68-69)

So happy wolf day to those who celebrate. And a heartfelt thank you to everyone who subscribes to Res Obscura. Thanks!

Subscribe now at the discounted rate

Refer a friend

Share