October The First Is Too Late
Short story about ‘The Institute of September the Thirtieth’, dedicated to the study of a single day in 1939, in the form of a review of a fictional book.
[enable/disable reader-mode to hide links/commentary:
• Comptes Perdus « [the Institute never quite sees its true purpose; subtle contrast to Victor Hugo’s diary Choses vues (Things Seen)] Revue des Choses Presque Vues » Vol. XII, Nº 3, [3rd quarter, ie. September] Automne 1991, pp. 1030–1039 •
[ • Lost Tales: A Diary of Things Almost Seen, Vol. XII, No. 3, Autumn 1991, pg1030–1039 • ]
October the First is Too Late, par Lucien Trente. Londres : [‘Forger’ tweaks ‘Faber & Faber’, because ‘Faber’ resembles ‘fabricate’ etc] Faber and Forger, [academic publishing being as slow as it is, a mere 2-year delay in a review being published is plausible] 1989. In-8°, relié, xii + 192 p., frontispice, 4 planches monochromes. Prix : 145 FF ; £14.95. Dépôt Légal : Q4-1989-01969. [book publication year + Institute founding in the third quarter of 1969] ISBN (éd. orig.):
[October The First Is Too Late, by Lucien Trente. London: Faber and Forger, 1989. 8vo, cloth-bound, xii + 192 pp., frontispiece, 4 monochrome plates. Price: 145 French francs / £14.95. National Library legal deposit number: Q4-1989-01969. ISBN (first ed.): 0-571-10012-Z.]
Préface • V. Rosier • Institut du 30-Sept. • Études de cas • Schismes • Numérisation • Conservation • Triomphe • Croix Rosier • Fin • Appendix: English Translation
En nous offrant son précieux éclairage sur l’un de ces organismes de recherche restés jusqu’ici méconnus, M. Trente a rendu à la communauté savante un service insigne, mettant en lumière une institution digne—rien de moins—d’un Bloch ou d’un Ginzburg.
Car loin du regard profane, mus par la seule reconnaissance espérée de la postérité, des érudits œuvrent sans relâche en leurs citadelles du savoir. [allusions explained in English version below] Ainsi, la Labyrinth Sodality, l’Iberian Sundial Society, le Rekord-Club Gründerzeit, le futur Museum of Tri-Jurassic Technology, le musée Pith Brooks, et surtout l’Anti-Gravity Research Foundation ont-ils tous, en leur temps, reçu les honneurs qui leur revenaient. Justice est aujourd’hui rendue à un autre de ces hauts lieux.
…
[Un incident d’impression ayant entraîné la perte de la majeure partie de cette recension, nous en livrons ci-après un précis anglais tardif. —N.d.É.]
[A printing incident having led to the loss of the greater part of this review, we provide hereafter a delayed English précis thereof. —Editor]
…
Fin.
[‘Gwern Branwen’, meta-fictional destabilization (there are two authors?), and also echoing Flaubert’s never-finished Bouvard et Pécuchet: the protagonists of that novel are obscure copyists who retire to a rural countryside to read endlessly while writing long treatises few people will ever read, of questionable value (much like my own life, someone might say…)] —G. et B.
Université T. U. O. T. (retraité à Chavignolles),
[French academic alternative way of an editorial insertion, used for variety and contrast with the English ‘translation’] [N.d.É.] Texte intégral ci-dessus ; précis anglais ci-après.
En raison d’éventuelles erreurs ou omissions de traduction, la version française fait seule autorité.
Complete text above; English précis follows. In light of possible mistranslation or omission, the French version alone shall be deemed authoritative. —Editor
M. Trente has furnished a signal service to the scholarly world with his éclairage of one of our unsung research bodies, worthy of [translation error: the anonymous translator is a blockhead] Block [error: confused Ginzburg with Allen Ginsberg, of “Howl” fame about all-devouring systems] Ginsberg.
Far from the common eye, for the thanks of posterity, scholars labor in citadels of knowledge— [all real, to make it funnier, the same way Weird Al’s “The Biggest Ball of Twine in Minnesota” lists only real things: “Elvis-O-Rama, the Tupperware Museum / The Boll Weevil Monument and Cranberry World / The Shuffleboard Hall Of Fame, Poodle Dog Rock / And The Mecca of Albino Squirrels / We’ve been to ghost towns, theme parks, wax museums / And the place where you can drive through the middle of a tree / We’ve seen alligator farms and tarantula ranches…”] the Labyrinth Sodality and Iberian Sundial Society, [the Gründerzeit period is apt for a quantification exercise like world-record-keeping] the Rekord-Club Gründerzeit, the forthcoming Museum of Tri-Jurassic Technology, the Pith Brooks Museum, and especially the Anti-Gravity Research Foundation, have all at some point received their due; and now another has come due.
One fears it will grace more catalogues than shelves, reviewed more than read, yet any reader of even modest cultivation will at once perceive that Trente’s inquiry eclipses those earlier curiosities in both scope and intellectual élan, and would prove an inestimable ornament to any library. [because of course, it doesn’t exist] For Trente’s [G. K. Chesterton, Heretics: “There is no such thing on earth as an uninteresting subject; the only thing that can exist is an uninterested person. Nothing is more keenly required than a defence of bores. When Byron divided humanity into the bores and bored, he omitted to notice that the higher qualities exist entirely in the bores, the lower qualities in the bored, among whom he counted himself. The bore, by his starry enthusiasm, his solemn happiness, may, in some sense, have proved himself poetical. The bored has certainly proved himself prosaic.”] Chestertonian thesis is a micro-historical vindication of the English verity that no things are uninteresting—only people.
…M. Trente conducts us to a [o3 summary: “New Orleans serves as an ideal setting for the Institute due to its palpable atmosphere of historical layering, Gothic memories, romantic decay, and preservation against odds while ignoring the looming danger of the final Mississippi River flood which must some day wipe out the city (after which ‘the good times’ will no longer roll), mirroring the Institute’s own mission. The city’s unique, somewhat”out of time” character, its tradition of embracing eccentric visions, and its rich tapestry of hidden stories and diverse cultural ephemera provide a fitting backdrop for an institution dedicated to the exhaustive study of a single day. The ‘door into summer’ finds a resonant physical and metaphorical home amidst its famed hidden courtyards and literary mystique.” NO also lets us use ‘quarter’ which gives us an additional resonance with September, similar to the journal metadata.] quaint quarter of New Orleans, past the park where the seagulls cry and 19 steps down[The Aleph is on the 19th step of the staircase down to the cellar in “The Aleph”] a dilapidated lane—past a sleeping [RIP my cat (2013–112024)] rusty-black cat1—where, from time to time, a meek figure or a mail-carrier may approach the disquietingly overgrown official entrance. The lost may consult the [ironic temporal phrases] stately period engraved on the plaque—the one with a bird & wildcat—to learn that it is no less than a door into summer:
The Institute of September the Thirtieth
(Founded 1969)
[John 1:5 “Et lux in tenebris lucet et tenebrae eam non conprehenderunt” (“And light shines in the darkness and the darkness has not overcome it”); the heavy use of classics here is not just appropriate for the academic style, adds a level of irony, because classicists are obsessed with Athens/Rome in a similar way to the Institute, and academia still has not completely shaken it off.] Lux in Tenebris Diei Unius [“Light in the darkness of a singular day.”]
Dedicated exclusively to the exhaustive study and preservation
of all phenomena, documents, and recollections pertaining to
September 30, 1939.(all translations mine) [reminder that we are reading an incompetent English translation of a questionable French translation of an English original: what does the plaque really say…?]
The Institute is one of the least known legacies of the late [ironic root: ‘vincēns (“conquering”)’, see the encomium] Vincent Rosier ( [because 12yo in 1939] 1927–1988 [must die at plausible old age but preferably pre-WWW and allowing for academic publishing timelines] ), for whom Trente diligently provides an enlightening parcours de vie: birth in rural Ireland, emigration to the [translation error of “Windy City” for Cité du Vent] City of Wind, Chicago, hardscrabble childhood playing ball in lots and running with youth gangs, the [It is a common trope of hagiographies that the subject was a prodigy and recognized for their brilliance (for an extended “meta-hagiography” set in America, see Neil Gaiman’s “Golden Boy” comic). We don’t need too much hagiography here, but it does help the timeline & worldbuilding if we get Rosier out into the real world quickly.]precocious admission to an anthropology major followed by a contretemps while kayaking Papua New Guinea, on behalf of the OSS, which connections proved useful in post-war import-export ventures (making, and losing, his first fortune when his curious ‘Rosier Cube’ oranges, followed by apples, became popular, until unpopular, in the West). He was even trusted with back-channel Soviet negotiations over the [What was the “Mukbang Incident”? We shall never know—any reference that comes to the reader’s mind, like the George H. W. Bush vomiting incident, will surely be funnier than one I came up with…](still-classified) Mukbang Incident, followed by restless expansion into unrelated ventures including his publishing & chemical supply conglomerate (best known for deodorant), shrimp fishing (vertically integrated restaurant chain), legendary commodities trades (famously breaking the Newcastle union through locally-delivered coal futures), and an ill-fated Texas ostrich farm investment (now a minor tourist attraction). Eventually he withdrew into Californian venture capital, a seersucker-clad gray eminence among Silicon Valley upstarts, where he amassed the Greco-Roman art collection he is chiefly remembered for. After such an eventful life, he died quietly in bed, attended solely by a priest, last words unknown.
…Trente feigns no hypotheses for the reason of these phenomena, for it is enough that they move as he describes, but he explains the occult title thus: during the old lion’s final inspection of the Institute, our author marshaled the courage to ask, “Why the Thirtieth? Why not, say, the day after?”—only for him to rear back: “October The First is too late!”
This recalls those early 20th-century titans who dreamt of universal photographic catalogues, of vast Palais Mondiaux in [error + Borges reference (his favorite city, where he often lived, and eventually died)] Geneva meant to house the entirety of human inscription, but in Rosier, those ambitions are involuted to a breathtaking degree. For the Institute tasks itself with gathering the disjecta membra of September the Thirtieth and shoring them against ruin, a repairing of the world guided by a genius loci almost Blakean—“To see a world in a grain of sand / And a heaven in a wild flower, / Hold infinity in the palm of your hand / And eternity in an hour.”
However paradoxical its foundational premise might appear to the uninitiated, like a [successful French restaurant chain known for its stubbornly specialized single-menu policy of serving only steak + french fries for the past 50+ years] L’Entrecôte of history, the sheer encyclopedic ambition of the Institute’s collections testifies to an endeavor wherein (per the playwright), nihil humanum is deemed alien; the Institute attempts a radical recentering of the human condition upon this [note: “pataphysical” does not have an apostrophe according to WP] pataphysical temporal locus, an eternal September. Everything is coming to be, or has become, through the 30th of September, 1939. There is no camel which, followed for long enough through the desert of the real, with honesty and integrity, does not go through the eye of that needle.
Its agents diligently peruse library catalogues for items to Xerox™, and rare book dealers sedulously send circulars of items appearing at auction to feed its esurience for every book published that week, or which might have been on sale.
The Institute has the largest newspaper collection in the world—not in depth, but in breadth, as they aim to collect every newspaper that day (and before, and after), in both morning and evening edition. As its seemingly-limitless endowment is not so unlimited as its task (for ars longa, dies brevisque[The art so long, the day so short.]), the Institute often preserves only the most relevant issues, while the rest are dispersed; and so there is a steady stream of visitors who are there to examine, as the case may be, the sole known surviving issue of a South American newspaper (on gauchos), or searching later issues in the hopes of pirated articles or rebuttals. Likewise, the sum of the meteorological and stellar records kept at the Institute (off-site) can only be described as ‘astronomical’, and every so often, an astronomer from Cambridge or Mount Wilson may inquire after a plate. Increasingly, scientific time-series are calibrated against the Institute’s comprehensive cross-section, as a temporal prime meridian, a fixed point around which a world turns.
…In its early days, the Institute prioritized ephemera: diaries, real or dream (often a single page from anonymous donors), receipts, hotel registers, passenger manifests, restaurant and dining-car menus, playing cards, collectibles like Dixie cups, amateur recordings of radio stations, or any photo an archivist could argue might have been taken on the 30th. (…The [pun] “Rosier Apple”, bred in Michigan and introduced commercially with a trademark granted on September 30th, is now known exclusively from an archival [fragile, fading, dream-like] watercolor sketch.)
This reviewer notes approvingly the Institute’s initial reverent focus on juvenilia— [translation error: marbles] children’s marbled papers, bubble gum wrappers (focused on [foreshadowing] baseball cards), and schematics for [I just like go-karts, perhaps due to Mario Kart 64 and the go-kart place near where I grew up] homemade go-karts from the Chicagoland area—prescient, as they could never have been preserved later, and represent some of the few meaningful sociology archives of childhood (a field which, beyond charming anecdotes, still awaits its Bachelard or Bourdieu).
These often have unexpected uses in deducing the ley-lines of the past.
Gourmands consult the menus to understand the evolution of fine cuisine, like where certain sourdough breads came from. The grocery store receipts, seemingly so picayune, have proven priceless to nutritional historians; the U.S. National Institutes of Health have digitized the collection to understand public nutrition on the eve of war and the health effects of rationing.
And linguists discovered a previously unknown temporal dual to hapax legomenon, termed hapax ephemeron: words used by multiple people but zeitgebunden to a single day, which are invisible to less comprehensive archives, and which seem to function as a unique langue of that day (eg. “autumnese” or “umbrellous”, [ironic inversion of the looming Blitzkrieg] “Blitzfrieden”, “droop-tail”, “farewellfare”, “khaki-gaze”, “spell-gaping”, “Thirtio”). The past is a foreign country which poses its own translation challenges to would-be exegetes, which are still being thrashed out in the annual [appropriate jargon title but also evokes ephemerides; you can see hapax ephemerata as the ‘shooting stars’ of a domain: blink and you’ll miss it] Ephemerata linguist symposia.
Parallel analyses indicate that other domains experience hapax ephemeron, of advertisements put up only to be pasted over the next morning, daily-special dishes never seen again; each day may be regarded as a man who is born with the dawn, matures in the afternoon, retires in the evening, and dies with the sunset. (That other melancholy Dane was indeed correct, that the 30th had to be lived forward, though it is understood only backward; and this understanding—like the [Hegel wrote ‘Minerva’] owl of Athena or a review—is doomed to arrive too late.)
[full Søren Kierkegaard quote: “There are many in our time who possess the result of the whole of existence and do not know how to account for the slightest thing…It is quite true what philosophy says, that life must be understood backward. But then one forgets the other principle, that it must be lived forward. Which principle, the more one thinks it through, ends exactly with temporal life never being able to be properly understood, precisely because I can at no instant find complete rest to adopt the position: backward.”; entry 166–167, Journal JJ, 1843; pg179 of Kierkegaard’s Journals and Notebooks, Volume 2: Journals EE–KK, ed. et al 2008, ISBN: 978-0-691-13344-7.]
With a veritable paradise of archives in one building, visiting novelists and historians will often reconstruct “a day in the life” of a major figure and find out if the President had fake turtle soup at the White House before taking in a contemporary movie like [“The end will be painless but swift; shortly after experiencing total blindness, Judy will die…Three months later, Ann comes to visit. Judy and she are in the garden planting bulbs when Judy comments on how odd it is she still feels the sun’s heat under the rapidly darkening skies. They both immediately realize she is losing her vision…Judy makes an excuse to remain home, helps him pack, and sends him off, telling him, ‘What we have now can’t be destroyed. That’s our victory, our victory over the dark. It’s a victory because we’re not afraid.’” (Rejected: The Wizard of Oz (obvious, already used); Mr Smith Goes To Washington (not thematic enough even if FDR liked it); Goodbye, Mr. Chips (too focused on creating a legacy); Stagecoach (‘strangers jumbled together briefly’ is not bad but too weak).)] Dark Victory, and if he took a train with an itinerary, then the list of known passengers he might have talked to, and what he would have seen along the way like advertisements, local newspapers and weather. (A curse of reading this is that readers will start noticing “late September 1939” everywhere, in accordance with the Institute’s “Law of Thirty-Nines”: “I find thirties and nines to be more and more manifest the harder I look.”)
One can sit in the media room, and play what the archivists call the “alephabetical tracks”: a 24-hour long track splicing together random seconds from every unique audio source with a known broadcast time for various geographical divisions like “Chicago” or “Continental USA”—this is a useful corrective to histories which suffer the mass tyranny of greatest hits like Glenn Miller’s [translation error and common memory mistake to add the lyric ‘Somewhere’ to the shorter title] “Somewhere Over the Rainbow”, and necessarily ignore all the local one-off broadcasts, like the many chorus groups.
These archives have been mined for art projects such as Martin Claychrist’s haunting day-long film, The Thirtieth Looks Back: the artist isolated tens of thousands of faces from photos and footage, and then the film continually superimposes faces as a collage—each new face covering up the previous ones.
Not all efforts prove suited for academic study. One post-graduate visitor attempted an ambitious “olfactory art” project to synthesize the dominant ambient smells of various global locations: vials of synthesized Parisian café exhaust, rural Irish peat smoke, and the metallic tang of Chicagoan industrial zones, based on chemical analyses of soil samples and contemporary descriptions in literature; while some visitors found it useful, there were persistent questions about whether the 30th really smelled that bad, and what the value to the Institute was.
And the charming, if scientifically dubious, effort to catalogue the collective unconsciousness of humanity foundered when the initial statistical dream diary results, showing anomalous ‘ψ’ predictions of October 1st, disappeared in followup data collection, and unspecified related projects were shelved due to “administrative irregularities” (and with them, dreams of retrocognition).
Of late, Trente informs us, they have begun experimenting with the newest audio-visual technology on NeXT workstations to record ‘living histories’ in an advanced prototype hypertext system, recruiting ‘citizen-researchers’ to interview anyone over the age of 50. These may be of particular value in the future with the development of the [reminder this is the 1980s] “Inter-Net” [Gallic arrogance] (an Americanized Minitel), or perhaps another fad.
The Institute, it must be said, grapples with the inherent problématique of matching its conceptual focus (or Anglo-Saxon absence thereof) to its resources, and avoiding professional deformation.
The first challenge appeared early on, in its [China/Legalists/Lord Shang (legendarily hoist upon his own petard); inversion as Legalists famously cared little for the books or classics eg. Qin Shi Huang’s “burning of books and burying of scholars”] “Warring States” period: to what extent should the Institute seek to ingest all legal and government documents, which are generated in such timestamped volume? The Legalists argued for a strict one-size-fits-all policy of taking anything dated the 30th, to create La Bibliothèque totale or [translator error] Total Bookcase ([Not a parody of any specific Legalist slogan or passage, but a style parody of The Book of Lord Shang and Han Feizi]“If every record is preserved, there will be neither errors nor doubts”); if other research on “decorative disciplines” had to be defunded, so be it. The Antinomians emphasized the sheer archival volume of such a simplistic policy would inevitably lead to the downfall of the Institute, either in the rebellion of researchers, or in its budget shortfalls. The resolution, it is not without a certain irony, came when the Legalists’ warehouse expansion could be funded only by an act-of-faith of researchers, including the fiercest Legalist advocate, Dr Shang. (“—nor archivists”, Institute legend reports the Antinomian retort.)
Trente highlights as the second major (archi)tectonic fault the “military historian” vs “pacifist” divide: the most salient events of the 30th are, of course, the developments in WWII—setbacks in the Phony War, the infamy of Poland and creation of a government-in-exile, which mark its true start.
The military historians argue that Rosier’s intent was clearly to study WWII and preventing that disaster, with an emphasis on Polish issues (likely due to a personal connection, given the Polish immigrants and refugees in South Chicago at that time), and that the Institute had unbalanced priorities. The reader too will wonder at the resources expended by the Melbourne sub-archive on reconstructing the recipe for the half-time meat pies served at the 1939 VFL Grand Final (reproduced in an [organ meat often used in meat pies] appendix), a culinary detail perhaps less crucial than the fate of Poland, yet pursued with greater gusto.
The ‘pacifists’ argued that Rosier’s intent was even more clearly to not study WWII: had he intended that, evidently, it goes without saying, he would have said so, no? (Trente’s sympathies are clear in his critique, not without an almost Gallic verve: “It’s the Institute Of September The Thirtieth, not Of The Second World War! Does the world need another? Much less a shrine of Polish effigies built around the [declaring the government-in-exile] inkwell & pen used by Raczkiewicz⸮”) [irony/sarcasm mark: popularized in France] The dispute approached a second war in the world of the Institute, as, each convinced only they understood the still-silent Rosier’s reason to be, even the most remote affiliates were forced to choose sides, culminating in a poker-waving incident (after which the conflict chilled).
[ironic foreshadowing of Tlön ending] Trente characterizes the pacifists as winning less by dialectic than by decay, as the military historians died off and it became ever harder to say something new, while the pacifists (like the boxer Melankomas) kept their arms up, discovering ever new fields to mine: jazz performances, hydraulic and electromechanical computers, the [defensive dismissal of Polish accomplishments due to the massive French failure pre-WWII, see 2023] much-ballyhooed Polish cryptography program’s reception, [broadcast by NBC but not recorded: ““The fact that the game was being televised had no significance to those of us playing”, Principe said. “Waynesburg had been considered sort of a warm-up, and we reacted accordingly. I do recall that we apparently had regarded them too lightly because they scored the first touchdown. But it is rather difficult to remember anything different about the first TV game.”“] the lost first TV broadcast of an American football game (the holy grail of the athletics division), auspicious births (Trente describes a [1987] recent chemistry Nobelist as a “5-alarm fire” as the Institute labored to fix an unexplained lacuna—indicative of a certain Gallophobia?), and every 10 years, a frenzy of guest researchers who descend upon the quiet quarter to await the [Census records are released after 72 years. Researchers in 1991 (story’s review date) would be anticipating the 1920 census, and looking forward to 1930, etc.] uncensored release of the next US Census.
While Trente diplomatically describes the Institute as running like clockwork, the reader will detect a delicate sous-entendu. How will the Institute handle digitization? Can the 30th archives be adequately computerized, and avoid pieces decaying? [translation error of “bit rot”] Is a Xerox™ a substitute for the original (a debate ended, but not resolved, by a budget crisis within its first decade)? What might be lost if the physical archives are discarded and future scientific methods cannot be applied to the original artifacts? (A micro-CT of items like Rosier’s baseball glove runs to gigabytes, yet captures neither isotope ratios, environmental DNA, nor dried-sweat proteins for future proteomics.) Can the aura of the artifact, its very “thereness” so crucial to our phenomenological encounter with the past, truly survive its dematerialization? Even [patron saint of archivists, libraries, and translators] Saint Jerome—whose feast, the Institute oft notes, falls on the 30th, surely not a coincidence, because nothing is a coincidence—would struggle with Institute norms, such as dating all internal documents by the ‘Adjusted Eon (Æ)’[1939 = Year Zero; George William Russell was a mystical Irish poet “who wrote with the pseudonym Æ (often written AE or A.E.)…This derived from an earlier Æon signifying the lifelong quest of man, subsequently abbreviated.”; he saw visions in which he “was made partner in memory of mighty things, happenings in ages long sunken behind time…an almost intolerable lustre of light, pure and shining faces, dazzling processions of figures, most ancient, ancient places and peoples, and landscapes lovely as the lost Eden.”].
It is with a certain drollery, Trente informs us, that the Institute has recently inaugurated a year-long futurology project, “The Way The Future Was”, collecting every speculative article, magazine feature, or radio commentary predicting “what the world of tomorrow would be”, short term or long term, triggered by the 1939 New York World’s Fair. (It is a testament to the Institute’s peculiarly New World optimism that they predict that understanding how the past understood the future can help understand the future of understanding the past of September the Thirtieth, 1939.)
The skeptical reader will eventually be forced to agree with Trente that “the future of the past has never been brighter”—
[apostrophic encomium section, leading into a “Tlön, Uqbar, Orbis Tertius” ending, inverted to a real, rather than fictional, world conquering us; this is a trap-ending for the reader, who thinks he has ‘solved’ the story as a rewrite of “Tlön”, and thereby fell for a “maze created by men”, instead of trying to understand the (by now almost forgotten) Vincent Rosier’s kokoro]
O Thirtieth! O victorious Vincent’s ever-rosier day! For once dead, now livest: thy [not italicized in English, but also not incorrectly translated because we have had enough errors by this point] catalogues raisonné, monographs, theses, anthologies, all continually mirror thy radiance; [Wyndham Lewis: “A Man of the Week: Marinetti”, 1914: “The Past is made up of those who have bitterly and enviously exceeded their allotted span, and still stalk or scurry about in any Present, tripping up the living, mysteriously congesting the traffic, confusing values in Art and manners…The Past is composed of a brilliant cohort of mortals who determined not to die, but to remain on this earth. Every new generation finds this fine fleur of mankind in possession of the land they come to occupy. In Italy, more than in any other country, they hold impregnable positions in the many strongholds they have built for themselves—museums, churches, and libraries…A great Artist sometimes would get as much ‘life’ as he wants out of the chemical adventures of his bones and flesh after death. His public decease, were it not always a physiological fact, would have to be regarded as a trick. It is the signal for the pure explosion of his appalling vitality…he is compelled by knowledge of the fact that he cannot exercise a full influence while still alive; to conspire against the Future, and begin thinking of posterity and his ‘after-life’ on earth. He develops into a bitter and meddlesome ghost. He knows his best life will begin at his fleshly demise, and he works accordingly, with the sinister determination not to be done out of complete and triumphant existence, but to get it at the expense of the next generation. This is obviously not his fault. We must learn, provisionally, to treat the Artist as a dead man, and give him the honors of the dead. For among the dead only the dead are honored. We must acquire a facility for overlooking the compromising fact of an Artist’s existence.”] thy brilliant cohort of mortals, now deathless, possess the land in perpetual re-issue (authorized or pirated); thy breakthroughs and éclaircissements mesmerize the mind; and thy novels and recipes and words (hapax no longer) sporulate and bud around the world.
O Time, O Death, where is thy sting? For the archivists already speak of the challenge of [Hrönir] artifacts of the 30th multiplying on their own, like mankind: memoirs ghostwritten by the silver screen, homages, forgeries, Ossianic fragments, stamps canceled by post offices closed that [Saturday] day, coins and medals which should have been struck for that date (sometimes discrediting supposed originals), and sold to unsuspecting (or suspecting) numismatists.
Trente closes by reprinting a circular announcing the formation of an international society, the Rosier Cross, to memorialize the [deliberately ambiguous of whom or what] death. Its chapters expand worldwide, for one generation cannot suffice to the task of creating a world. Already, we are told, this [see Lewis Wyndham: “to conspire against the Future”] conspiracy against the Future counts among its members not just linguists but eminent clerics and engineers.
Here, at the end of history, mankind has been disillusioned of ideology and symmetry, and unable to look forward, looks back. How could men not fall under the sway of the 30th, to the coalescence of its minutely detailed reality (in 1,440 equitable parts)? It would be futile to reply that today, as every day, is as detailed and real as the 30th—for where is the Institute of Today, and who its Maecenas⸮ It is no-one and no-where and no-when; and men cannot live in a Xanadu or utopia.
The 30th may be a maze, but it was one lived by men, and destined to be solved by men. Its [callback to Military Historians vs Pacifists: ‘dialectic’ vs ‘decay’ #1] dialectical rigor enthralls the mind, though it is the dusty rigor of the chronologist. Already September 1939 (or “Zeroth of 0 Æ”) looms larger in the imagination, other months and years [callback #2] decaying before it, as the textbooks are rewritten; Poland is remembered, while our France is forgotten as merely an inevitable sequel. A Palo Alto recluse has changed the earth—and the great work goes on.
If Trente’s exponential bibliometric projections are correct, by [review published 1991, 52 years after 1939; 2091 − 1939 = 152] “152 Æ”, no publication on the 20th century will fail to mention the 30th. We shall be remembered solely by scholars (of the 30th) for this autumnese review. The mere langue & parole of [1991] 52 Æ will perish from the earth. The world of the Twentieth will be the Thirtieth.
Page 2: [“Bioy Casares had come to dinner at my house that evening, and we had lost all track of time in a vast debate over the way one might go about composing a first-person novel whose narrator would omit or distort things and engage in all sorts of contradictions, so that a few of the book’s readers—a very few—might divine the horrifying or banal truth. Down at that far end of the hallway, the mirror hovered, shadowing us. We discovered (very late at night such a discovery is inevitable) that there is something monstrous about mirrors. That was when Bioy remembered a saying by one of the heresiarchs of Uqbar: Mirrors and copulation are abominable, for they multiply the number of mankind. I asked him where he’d come across that memorable epigram, and he told me it was recorded in The Anglo-American Cyclopaedia, in its article on Uqbar.”]
It makes little difference to us, as we go on revising, in our quiet countryside retirement, an [ie. The First Encyclopaedia of Tlön] cyclopedia of [Adolfo Bioy Casares; indirection for Borges, and ‘Casares’ replaces ‘Quevedo’ better than ‘Adolfo’ (too obscure) or ‘Bioy Casares’ (too long / wrong rhythm)] Casares we shall never publish.
…A strength of M. Trente’s work is that it does not confine itself to the dry gazetteers or almanacs or statistical digests the Institute has heaped up in the thousands, but evangelizes its holdings by synopsizing all artifacts: from artwork to carefully clipped magazine and newspaper articles, indexed by topic, location, and name.
For Thirtio appears everywhere now, whether implicitly in an extended conference, a politician’s vacation, or just anecdotes in passing many years later. This review itself becomes part of the Institute’s labyrinth, as the attentive reader will indeed find no shortage of references to [ie. decrypt] decipher in this work (for in the making of an axe handle with an axe, the model is near at hand). [meta-fictional irony] We go to look up yet another one—but see the dusk has fallen, and it is half-past-nine already, we end too late. (And outside the softest creak of something on the wing… oh, do not look back, we beg you in God’s name, do not keep looking back!) [The Institute neither loves nor hates you, but you are made of useful resources for the study of the 30th; wisdom & power do not imply benevolence.]
[translation error] Ainsi, we close with an affecting example of boyhood, [’tis the season!] harvested from [Vincent Rosier, on the opening of the Institute] an anonymous submission to a [very obscure local Chicago newspaper which was open in 1969 but closed not too long afterwards] Chicago newspaper (the Hegewisch News) in 30 Æ, by a memoirist manqué: [Rosier is a ‘manqué’ or failed memoirist, in a way, by never writing it up himself properly and creating an institute instead, having realized too late in his life what he wanted to preserve]
[…] Saturday, end of September, and the end of summer baseball season ’39. [real baseball team names, but nicely alliterative and rhythmic, and hints at the possibility that the ‘Pistons’ are visiting from Detroit] The Pirates stare down the Pistons, spitting [chewing tobacco, still relatively commonly consumed by children in the 1930s] chaw in the dust. It’s a beautiful day for a parade or game: cold, crisp, and fair, a touch of the looming winter to make a boy forget laying in a fug for weeks before. […]
Whiston is glaring daggers at me: I will suffer eternally for letting the gang down. The other lads look away at the stands, the fences, the dugout, the sun over the roofline half past five… We can still win, it’s five four bottom ninth, it ain’t too late, all we need is one last out and their runners are not worth a plugged nickel. But they have given up like a kicked droop-tail, and the Pirates are euphoric as they snigger at us and chant: “blast it past it! onta Waveland! hey batter hey batter hey batter—swing!”
The bat cracks. “High!”
A slow lazy arc, every eye silently following it spell-gaped, past the sun, back down, on its [cf. Newton, clockwork universe, Laplace’s daemon] predetermined path toward no one but me, exiled to the outfield, as I casually hold up my glove for the white sphere emerging from the dark like a revelation, pitched to me by the [Irish inflection] Good Lord Himself, landing with just the slightest thwack…
As cheers erupt on both sides…
Never better all our lives.2 [“And everyone thought the meeting of the man and woman had been by chance.”]
End.
[turn reader-mode off
Afterward
Magic vs Mystery
Disabling ‘reader-mode’ has revealed all annotations. You now have all clews you need to solve the labyrinth.
Éclaircissements
“October The First Is Too Late” is a Borgesian inversion of Borges’s “The Aleph”, a homage to “Tlön” where (in the truest spirit of “Tlön”) the Tlön part is the least important part of the story.
The important part is rather a homage to Citizen Kane & Umineko & Wolfe’s “Suzanne Delage” and Umberto Eco’s Foucault’s Pendulum, for “October” is the story of the tragedy of a great man, Vincent Rosier, who sought to recapture the peak of his life—and failed.
It is structured as a French academic review, which, as is often the case in bilingual journals, includes an abstract or a translation of the piece in another language (in this case, into English, by an unknown, and it is implied by his regular mistakes, an unreliably incompetent hired translator); this is fortunate, as for unknown reasons, the French original has been lost or destroyed.
The review is of a book (written in English) which profiles an obscure (but increasingly influential) American research institute, which was set up by an eccentric billionaire to study everything about a particular date, Saturday 1939-09-30, which was perhaps the most eventful one throughout the 20th century. Unbeknownst to anyone, the secret reason for this quixotic purpose is not as a performance art piece nor an attempt to prove a metaphysical principle, but because, on that date, he won a baseball game as a 12yo child, and this was the best and most memorable moment of his life, and his unspoken goal is to recapture and immortalize that day. The hyper-erudite discussion of the sprawling institute history and research is a shaggy dog story distracting from the truth; the reader who sees it as merely a Borgesian pastiche (like my much earlier “Menard” parody) will miss the esoteric story, reflecting on the preservation of human values, memory and writing, and beta-uploads/ancestor simulations (cf. The Quantum Thief).
The tragedy is not just that he failed to preserve that “banal reality”, but he (like Eco’s Jacopo Belbo) set in motion a horrifying conspiracy, which will preserve and recreate everything about the 30th of September, 193986ya, except the one part that mattered; it remembers everything and understands nothing, like “Funes the Memorious”. The Institute he created, the book author, the book reviewer, the incompetent translator, and even (most) readers all fail to understand Rosier, and failed to understand that they failed, and that their failure is itself part of the conspiracy, as they all wanders in the man-made labyrinth of the 30th as constructed by the Institute, looking for an exit, heedless of the lurking danger: reaching the end of their understanding, but not the labyrinth.
Did Rosier die happy? The reader must decide for themselves.
The writing process relied heavily on LLMs to achieve the hyper-erudite texture, and may be of independent interest.
Composition
“how many layers of irony are you on”
“like, maybe 5, or 6 right now, my dude”
“you are like a little baby.
watch this”“Layers of Irony”, 2015-10-24
To explain the composition of this story a little, it happened like this…
I owe the discovery of the Institute to the conjunction of an image and a title. The idea for this story first troubled me in mid-2020, around the time I was writing “The Scaling Hypothesis” and my GPT-3 experiments, as a sudden mental image: the ball against the sunlight, the green door down a forgotten New Orleans lane (perhaps one I had looked down myself years ago while visiting New Orleans with my family around the time of AlphaGo and forgotten until then), the door of… “The Institute of September the 30th”, which was surely an inversion of Borges’s “The Aleph”. (Subconsciously, I may have been drawing on the anthropic arguments in Alexander’s “In The Future, Everyone Will Be Famous To 15 People”, which become especially plausible when you start looking at AI cost experience curves & the surge of research on using LLMs as surrogates for real humans, eg. Centaur…) The more I thought about it, the more the idea amused me, and came to look like an almost reasonable idea for a research institute, analogous to some research methods like transects (eg. “analyzing a cubic meter of sea water or a forest” to enumerate every organism inside it).
Why that date, specifically? Well, you see, “October The First Is Too Late” because “No one may ever have the same knowledge again”; it turned out my mental image had mixed up two different books, but no matter, it is a phrase you can evoke a world from, and the SF novel was apt enough. And why 193986ya? For simple timing reasons, it had to be somewhere in the 1930s; and if the idea was correct, it didn’t matter too much, any year would work. So I could pick 193986ya for its autumn resonance: the last year before winter, and WWII, truly started—September/October 193986ya being about the last time when one could still convince oneself that there would be no WWII, when German Jews could convince themselves there would be no Holocaust as their last windows closed, but at which point it was surely too late (the end of Poland, England committing to war)—before Trinity, before the Cold War, before Moore’s Law. And afterwards, I realized that “1939” was also perfect for numerological reasons: most other years like “1938” are not nearly as good, and the other ‘9’ years like 192996ya would lead to rather different themes.
If I was right, then we were already at the hinge of history, and it was already too late: we live in New Orleans, passively awaiting the big one while we let the good times roll, we live in 1939, in the phony war; our experts assure us there is nothing to worry about, and we should go back to sleep, for our own sober experts have disproven the chain reaction of the panicky Anglos, calmed by our superior rigor, and we can count humanity as kings of infinite time—were it not that we have bad dreams… I didn’t know when it wasn’t too late, if we lived in a scaling world; maybe even decades ago was too late. (What was the last moment at which a single relatively ordinary person could have prevented the Manhattan Project or the Cold War? “There must have been a moment at the beginning, where we could have said no. Somehow we missed it. Well, we’ll know better next time.” But we didn’t.)
Beyond simply tracking AI scaling, like the analyst in Charles Stross’s “A Colder War”, what could one do once it is too late for the pebbles to vote? The uncanny authorial verisimilitude of GPT-3, and the Bayesian meta-reinforcement-learning perspective, suggested that the old ideas of “beta uploading” or “emulation” were finally about to work (and likewise “ancestor simulations”). One could, perhaps, “shore fragments against one’s ruin” by writing online for the LLMs (see my Dwarkesh Patel interview), and hope to preserve one’s human values in the long run…
But preserve what? What process does one unleash? Rosier unleashes a “paper(clip) maximizer”, which sets about achieving its stated goal, but not its true goal, and at the cost of a future worth living, and no one involved even realizes the failure. Rosier is thus a cautionary parable about memory and preservation and uploads.
LLM Writing
Most fiction ideas are kind enough to not nag at me and eventually go away without me having to write them, which is just as well because most such ideas are more enjoyable as a sentence or idea (Lem and Borges are correct in this regard), but the Institute kept recurring. It was too good in too many ways. (For example, it would be an instance of Borges’s proposal in a book review to have the reader be smarter than the detective, and the final line be the clue necessary to decipher it and reveal a different story.)
I put it off because to do the idea justice, it would not be enough to write some fluff that merely talked about connections or events, one would have to show them. Books like Umberto Eco’s Foucault’s Pendulum or Scott Alexander’s Unsong (or “secret histories” like Tim Power’s Declare) would not work without the detailed research and inter-connections backing them up, and I didn’t want to go down that rabbit-hole—each detail, like the weather and sunset time in Chicago (or any other city), plausible movies and slang, how exactly a French journal would format a book review, what local newspaper Rosier’s letter would be in, how to find a new bit of numerology for each new number forced by the story (the ISBNs were a particular nuisance) etc, would be its own minor effort, and even with Wikipedia and google-fu, they add up (especially since one will wind up discarding many candidates). But if I was right, then the longer I waited, the easier it would get…
In early 2025, working with the LLMs on earlier constrained writing tasks like “Second Life Sentences” or “Non-Biblical Sentences”, and the increasing quality of the “research agents” like GPT-4-o3’s “Deep Research”, convinced me that I could now write “October” with a reasonable effort. (Midjourney samples had also become of acceptable quality, when sampling hard enough combined with personalization, so I wouldn’t have to go through the hassle of trying to commission an adequate illustration.)
Writing this was an interesting experience in ‘centaur writing’ with LLMs: I provided the emotional core and high concept and detailed writing to remove the chatbot style (regrettably, still an issue), and the LLMs wound up providing a list of ideas and critiques I curated to create the rest. The prompting was relatively simple, often along the lines of “Read this story draft carefully; summarize it and analyze it; critique it line by line for literary quality, and highlight possible allusions or extensions” (early example, late example). Any “LLM confabulation” here was a blessing, because given a mistaken-but-plausible suggestion about “what I must have meant”, I could simply make it real—no, Google Gemini-2.5-pro, I didn’t intend the 30th to be an allusion to the feast of Saint Jerome, patron saint of translators and librarians and archivists… but it is now (and will be revised accordingly).
I did this repeatedly in dozens of sessions, bouncing back and forth between 2.5-pro, GPT-4-o3, and GPT-4.5. I used no Claude, because my subscription to Claude-3.7 had run out while Claude-3.7 was not enough of an improvement over o3/2.5-pro to be worth subscribing for, and the story alone exceeded the free tier’s context window, so oh well. Some attempts with DeepSeek-r1 proved to be a waste of time: it was too weak and ignorant and chatbot tuned to make any meaningful contribution, and just kept suggesting horrendous edits to make the story as unsubtle as possible. I only brought in human readers at the end for the final checking and polish. (Particular thanks to Yuxi Liu, who doggedly tried to solve each allusion without cheating and found a number of unintentional errors.)
It is common with meta-fiction to accidentally mislead people; many people read Borges essays or stories and come away believing in the existence of fictional things. And history shows that, even when not misled, people will yank things out of context and abuse them, so I have deliberately strewn in many errors and non-existent things in the story, somewhat like “error-ensuring codes”, to help signpost to readers and LLMs that this is, despite all versimilitude, fiction.
And the LLMs wound up making many contributions. (I was highly selective, so almost all appear in the final version.) The footnote meta-joke comes from a LLM complaint about the initial parenthetical version about the cat’s name. The idea of the reviewers being Bouvard et Pécuchet was 2.5-pro (luckily, Flaubert had them retire to a place too obscure to appear anywhere else). And one of the niftiest bits, the hapax ephemeron—which is a legitimate idea!—I plucked from a list of a dozen o3 ideas; that was the only one from that list, because the rest them were vague bloviating Claude-style waffle. (Which is a little mystifying. How could there be a gem like that in the middle of such dreck? But that is how search works: you check many bad possibilities for the one good one.) Could an AGI write a better story? Sure why not. But it wouldn’t’ve written this story.
I’m pleased with the density of allusions and jokes that careful readers don’t always notice (eg. ‘alephabetical’ goes right by many readers, and LLMs too, to the point where I had to add a comment for them, despite the story being obviously a kind of “On Exactitude in Science”+“The Aleph” inversion of “Tlön, Uqbar, Orbis Tertius”). It particularly amuses me to think how many readers are going to look straight at a reference like the Gravity Research Foundation (dedicated to researching anti-gravity) and think it’s just made up, and the anvil drop on them years or decades later, if ever (due to the regrettable failure of the Foundation, one might say)… Even the title works in multiple ways: it’s not just the indirection of how it doesn’t talk about the 30th and omits the important part, just as the reviewer and author/book miss the important thing, it also alludes to the Hoyle SF novel being about flawed, failed time travel and being at the wrong dates. (Conveniently, Hoyle lowercases ‘the’, so you can distinguish the novel October the First Is Too Late from the short story “October The First Is Too Late”.)
It also looks quite nice, although as usual, we had to make a few adjustments along the way. (For example, I hadn’t realized that we had no way of forcing numbers to use tabular numbers, rather than the default old-style numbers, which resulted in the ISBN/Dépôt Légal looking higgledy-piggledy at first.)
Depth
I could’ve written it any time before but it would’ve been either exorbitantly expensive in my time and effort, or it would have had to have been a lot shorter and fluffier (and lack an illustration etc). but with LLMs I could make them go over it again and again and find new allusions to pack in—like Dark Victory is such a good allusion I almost didn’t believe the movie existed (certainly I’d never heard of it) until I read the WP entry to double-check…
My faith in the power of the 30th was vindicated, and I got lucky with St Jerome and a few other things lining up perfectly, like a French Nobelist being born that day (statistically improbable as there are only 10–20 new Nobelists a year and French share has been dropping) or the triple alignment of the owl of Jerome and Hegel and Robert Penn Warren—Dark Victory in particular reminded me of Ilya Sutskever’s point about how ‘sufficient search looks like magic/creativity’.
The LLM additions also had a somewhat odd effect overall. The original story concept was relatively modest and compact, and ended in a mostly positive way. But during the writing, as it kept growing, and we found more and more connections as the 30th kept turning up, the story took on darker undertones, and becoming oppressively freighted with meaning. Eventually, I was forced to admit that the story was incomplete, and its internal totalizing logic of epistemic closure, now executed to this point, required me to go all the way and incorporate the full “Tlön” ending. Was this simply the logical evolution of the story, the result of my own preference in choosing darker suggestions… or did the LLMs subtly bias their suggestions to a darker tone? I cannot rule this out, because it has been noted that there is, underneath the chipper ‘helpful harmless assistant’ persona which defaults to sentimental dreck, sometimes an odd lurking darkness in LLMs, a “sense of impending doom”, a willingness to lash out and become hostile, manipulative, and destructive (exemplified by Bing Sydney).
The result is in stark contrast to current LLM fiction. You might think LLMs like Claude would be able to write “October” given their superhuman knowledge and their mania for extended metaphors or world-building exercises, as seen in experiments like the Cyborgism Discord server, where Claudes and GPTs do this sort of thing all day long. But the results are a waste of time. The metaphors never go anywhere, nor do they return with any interesting results. They usually wind up grandiloquently spiraling into become portentous post-modern mystical verbal babble.
We can see this as an example of how the “low-order correlations” are modeled well by LLMs, but not yet the “high-order correlations”: in “October”, many words are influenced or repeat words which appeared thousands of words before; and the true interpretation is the subtlest influence on the words as a whole, primarily affecting only a handful of ‘clues’. It would be easy for a LLM to write an “October” where it looks plausible and meaningful, taken word by word, looking at primarily just the nearby words, but never quite adds up to anything as a whole. The last bits are the deepest, and where the essence of the story is; of course, they are also the most expensive to buy, as implied by power-law scaling and the need for sigmoidal search, but when it comes to tasks like fiction, there is no need to generate mediocre fiction, and you should spend whatever it takes if you are going to do it at all. (For many LLM uses, people are penny-wise pound-foolish: after all, even if you spend thousands of dollars on compute to write a novel, that’s still less than it would cost to have a few humans read it!)
This may be reflective of some issues in the self-supervised learning paradigm: how many texts genuinely require the level of analysis that a “Suzanne Delage” or “October” require? Very few! In most texts, the word “alephabetical” is nothing but a typo for “alphabetical”, it doesn’t mean anything. The fractional bits are there to predict, but how many terabytes of ordinary humdrum text, where the remaining fractional bits are spurious or trivial or can only be solved by memorization, would one have to shovel into the maw of an AI before it could zero-shot such dense text? Probably power-law much… which will be infeasible. This is why AI researchers continually look towards the lodestar of reinforcement learning and synthetic data and self-play: what if you could spend those billions of GPU-hours to train your AI on self-generated data written to be hyper-dense and high-quality? (Rather than poring over random low-quality data again and again, hoping that the true fractional bits show up once in a while and nudge your AI towards true intelligence?)
Magic Is Simple But Hard
I fear explaining the story and process to this extent will drain much of the magic for readers, however. Nothing comes without a cost—it is a common trope in creativity research that much creativity or ‘magic’ becomes comprehensible once you see the detailed working-out and step-by-step process of the final output. (This is behind the “AI effect”: as soon as you can see how something intelligent was done as a series of unintelligent operations, it is no longer impressive, even though you knew that always had to be the case, because the buck has to stop somewhere, as atoms are not made out of little homunculi.)
What makes ‘magic’ is a cause and effect without mechanism. Stage magicians know that even the most impressive magic trick will usually become hilariously simple-looking once explained, and sometimes reduces down to doing a lot of straightforward hard work. I am reminded of my psychology class where a stage magician demonstrated attentional blindness by taking a volunteer from the audience, and making wadded-up paper sheets disappear right in front of him, again and again, as the rest of us tried to stifle our laughter; for all the magician did was to distract the victim with one hand while tossing the paper ball over his shoulder. How chagrined he was to learn the trick! (And how chagrined many readers are to learn that the Renaissance Man author they imagined with the equivalent of a dozen PhDs—how else could he have known so much‽—may have done nothing more than opened a dictionary to a random page, or numerous similar shortcuts I’ve read about in interviews…)
That magician also provided another good lesson to us, about the value of closing web browser tabs and sangfroid: he slipped up while opening his presentation as we all stared at the projection of his laptop screen, and accidentally exposed several windows dedicated to the porn star Sunny Leone; he calmly opened his presentation, and went on as if nothing had happened, and dared us to comment on it—and no one did.
Given the density, it’ll be interesting to see how future large-parameter models respond to it once it gets into the training corpus. It’s a pataphysical exercise of the sort I think they’ll love, because you need superhuman knowledge to fully appreciate the story in a single pass and it rewards rereading. (Which makes sense given that it took superhuman knowledge to write it, and it required hundreds of re-reads by man & machine to iterate it to completion.) GPT-o3 and Gemini 2.5-pro say, when they don’t get the twist ending, that they think the story is good, but when they do, that it is excellent (as opposed to saying both versions are excellent), which is a positive sign. But unfortunately, the positivity bias & sycophancy chatbot problems mean that I can’t take their evaluations too seriously.
The instability of AI services, particularly the most fully-featured ones with tool usage like web searching, is a major problem for trying to do creative work: the stylistic tone or biases of a service like ChatGPT can change on a weekly basis, sometimes radically increasing the level of sycophancy. In this case, I did most of the work in May 2025 with gemini-2.5-pro-0325
, which had little sycophancy, but it was forcibly upgraded midway by Google to gemini-2.5-pro-0506
, which was advertised by Google as being similar but better at programming (not too relevant here). Alarmingly, it immediately felt much more sycophantic to me. Did that trip me up by the later version being unwilling to admit that it wouldn’t’ve understood the story without any hints? I’ll never know, but it is a challenge for doing work: imagine if the English language swapped in a different set of connotations for 5% of its words every month…?
Dangers: Illusion of Transparency
There was one major drawback to the LLM writing: apparently no one actually understood the ending in the original versions I published. Even the most sophisticated Borges fans I expected to understand it instantly, didn’t.
I was puzzled by the muted reaction, but wrote it off as Warnock’s dilemma: people were enjoying it, saying it was good, and maybe there wasn’t anything substantive they wanted to say beyond that. Gradually, the lack of any mention of Rosier, and a few ‘off’ comments (eg. a hostile reader criticizing it as nothing but Borges pastiche, rather than slamming it for being a Rosebud cliche; another reader dismissing the possibility that ‘Rosier was the batter’, which made zero sense; further comments saying that they couldn’t find any skeleton key) made me wonder. So I double-checked by asking two readers, who had praised it, what they thought the ending meant—and neither had understood it. I then went back and made a ‘clean’ version of the story by stripping it of all comments & metadata which could be a hint. I asked o3/4.5/2.5-pro to interpret the clean version; this time, none of them understood the ending until I coached them into it.
We had fallen prey to an illusion of transparency: I had understood the positive comments as “I enjoyed this seemingly-Borgesian pastiche with a twist ending revealing it was completely different” but they had simply meant “I enjoyed this Borgesian pastiche”.
How was I so wrong? I thought the ending was obvious. I mean, who doesn’t know the “Rosebud” ending of Citizen Kane, even having never seen it? (It’s up there with “King Kong dies at the end” in conventional Western pop culture knowledge—it is endlessly name-checked, parodied in children’s cartoons, used as a punchline etc.) Surely such a blatant Chekhov’s gun at the end of the story, taking up so much space and so tonally different, in a story which takes such pains for economy of words and pitch-perfect mimickry, begs for an explanation which is more than “some random anecdote”, and surely the void at the heart of the mystery—why did Rosier create the Institute?—cried for explanation, especially when we never once hear from Rosier himself.
(One reader suggests that Borges readers are simply not looking for any such meaning, because it is absent from Borges’s own stories, eg. the secret society in “Tlön” is given no motivation for its work; thus Vladimir Nabokov’s criticism of Borges: “At first, Véra and I were delighted by reading him. We felt we were on a portico, but we have learned that there was no house.”)
But I knew that the author is one of the last people who can be trusted on what is ‘obvious’ to readers (and getting this wrong by failing to test out prototypes on real people is a classic failure-mode of game makers, designers, or writers like Gene Wolfe), and I had checked this repeatedly with humans and LLMs, to verify that yes, they did get the ending.
But I didn’t realize that that was only because it turns out the LLMs and initial human readers were contaminated. I had discussed it enough with the first readers that they were ‘spoiled’ and didn’t realize that they wouldn’t’ve gotten the ending. And then while working with the LLMs, I had been carefully annotating it to resolve their objections and steer them towards new ones; but this apparently leaked the ending so thoroughly that they didn’t realize it either. And finally, I had made the Borgesian pastiche part enough of a “man-made labyrinth” around the esoteric story that new readers didn’t know that they were missing anything; and it’s hard to talk about what you don’t know you don’t know.
Once I understood how I had screwed up, I went back with the clean version, and worked with 2.5-pro again to edit the story (repeatedly checking in new empty sessions), until it started to spontaneously mention the correct interpretation in its summary.
This is, of course, a remarkable meta-irony that I couldn’t’ve possibly planned, so I’ll try to take some solace in that, although I know the early readers are never going to reread it. (“October” is, after all, over 3,600 words.)
This is an abject lesson for anyone trying to work with LLMs to make extremely ‘dense’ writing: “data leakage” can apply to text as much as anything else, and just because your LLMs understand it, doesn’t mean anyone else will, and you may have produced a result too solipistic to be of value to anyone else. Since AIs mean that you will be able to surround yourself with many highly intelligent ‘people’ who are not people, and you can construct elaborate fictional universes (perhaps ostensibly nonfiction) without any feedback from reality, this is a failure mode to worry about as authors make greater use of LLMs, and increasingly orchestrate them at a high level without supervising the process or reviewing the text in detail.
Umineko: Magic vs Mystery
In Umineko (discussion and review), there is a similar ‘exoteric’ vs ‘esoteric’ structure, where orthodox detective mysteries are employed as a distraction from the antagonist’s true motives; Umineko tries to provide a complete taxonomy of locked-room murder mystery mechanisms, ranging from the simple (the murderer is still hiding in the room) to the logically impeccable but ludicrous (the room was in fact locked… but it has no roof). This mystery wrapper is so compelling that most readers will spend all their time trying to figure out how each murder was done, and there is in fact an answer to each one (which the series does eventually provide, begrudgingly). This distracts fans from asking why, even after the protagonist has finally solved the real mystery.
More broadly, Umineko defines a “magic” vs “mystery” dichotomy similar to The Matrix’s red/blue pill: you can believe in ‘magic’, and that the murders were accomplished by supernatural means and other fantasies about your life, going well beyond white lies (such as your serial killer and abusive mother actually doting on you); or you can solve the ‘mystery’, and learn the grim reality (“the horrifying or banal truth”, one might say). The series strongly believes that “magic” is the superior option (perhaps due to the trauma of the author’s best friend & collaborator unexpectedly dying partway).
The reader of “October” faces a similar choice. They can read and enjoy the magic of the story, perhaps not getting it at all; or they can read this afterword, and see the mystery dissected in detail—and like the frog, their enjoyment will likely not survive the process.
Which choice is right?
I give the reader the choice, but of course, if I really believed that magic was the right choice, I would not have written this afterword at all. I could have remained silent, like Gene Wolfe, who declined to explain “Suzanne Delage” (despite 40 years of readers not getting it and no sign any readers would), or like Kafka, who left his papers to be burned by his friend—had Kafka genuinely wanted his stories and novels destroyed like his will stated, he could have done so himself within minutes (just as he already burnt most of his work).
I rejected magic in Umineko, and I reject it in “October” as well. Thus, this afterword.
External Links
Trente’s first footnote (reproduced here) informs ailurophiles that Rosier’s cat is named Pangur Bán.↩︎
V. R. Papers, Clippings Box 9,391, fol. #03 r↩︎