By Stephen DeAngelis
Naming the Present
When I was a young boy, I painted a simple canvas board. It was horizontal, much wider than tall. The board had a linen-like texture that absorbed the paint into a matte finish, so the black carried depth and character, and the two thin lines below it read as light held in a surface, not laid on one. Along the bottom ran a thin white strip, perhaps two percent of the height. Above that, a slightly thicker band of gray, around three percent. Everything else, the remaining ninety-five percent, I painted black. I called it The Void.
The colors mapped my view of the world. White stood for what we understand. Gray marked what we’re beginning to see, things coming into focus but not yet recognized. Black was everything else: undiscovered knowledge, unnamed forces, the vast territory beyond our narrow comprehension. Without knowing it, I painted the ratio of what we know, what we are learning, and what we do not yet understand: two percent, three percent, ninety-five percent in the dark.
I stuttered as a child, and a stutter limits your access to language. It gives you fewer words in the moment and forces you to choose among them. Standing at the front of a classroom with worn wooden floors and a blackboard framed in dark wood, I would try to say my name is Steve DeAngelis and catch on the last name every time. What I learned was to move the word I knew I would catch on into the middle of the sentence and set the sentence to a rhythm, almost singing, so the rhythm would carry me through the word before the stutter could close. If I could only get one word out, it had to carry more weight.
Where a fluent speaker could spread meaning across a sentence, I had to compress it. The sentence got shorter. Each word had to do more. Day after day, under that pressure while the brain was still forming, weighing words became habit: find the word that is both sayable and load-bearing, match it to the sentence it must hold up, choose fast or say nothing. The habit had a future I did not yet know about.
The stutter closed one channel, so I went looking for another. Meaning that would not come out as speech went into photography and paint. Language, forced into density. Images, taken up because speech failed and the meaning still had to be carried. The Void does the same work in paint: taking a small set of marks and asking them to carry an enormous load.
Seen now, the canvas has the force of a diagram. The strip of white and band of gray are the part of a world a person can name and act on. The black is everything else. A people’s reach is the reach of their symbols. Most of any given moment lies beyond them.
The work of making something verbal, visual, or analytical is the labor of pushing that boundary outward, to resolve a little more of the dark, so society can see where it is and act on what it sees. The strip rises, but the canvas grows with it. The black is not a debt being paid down.
At key moments, individuals extend that lit strip for entire societies. Ours is such a moment, shaped by AI systems that act faster than we can clearly describe them. When reality outruns language, the black field expands. People feel their situation but cannot grasp it. The task then is to raise the line: to build language, images, and categories equal to a world already underway.
Language and Reach
Ludwig Wittgenstein wrote the sentence my painting could only show: “The limits of my language mean the limits of my world.” The line appears in the Tractatus Logico-Philosophicus as proposition 5.6.1 It doesn’t argue. It states. A people can act only on the part of their situation they have symbols for. Where those symbols run out, so does their reach. No one assembles their symbols alone. A person inherits nearly all of them, so the limit is shared before it is anyone’s own, and it moves only when the shared stock is rearranged.
Set the proposition beside the canvas and it stops being philosophy and becomes caption. The two thin lines are the reach, the black the limit. The rest keeps happening in the dark.
The symbols do not have to be words. They can be images, mathematical notation, or the categories a civilization uses to make sense of what it sees. Euclid set five postulates in plain sight, and they carried two thousand years of geometry.2 Change one, and a different geometry follows. Whatever the medium, the arrangement defines the reach.
Language in that wide sense is the instrument itself, the way a people take hold of their world.
What is named can be examined, argued over, and governed. What is not named continues on the far side of the limit. At key moments, builders of language stand at that line where the lit strip ends and the black begins, then push it upward.
Virgil: Naming Rome’s Fate
Publius Vergilius Maro took the ordinary language of a defeated Rome and arranged it into something the civilization could inhabit. He did not invent new words. He reordered the Latin that already existed until it formed a coherent whole. That arrangement is what endured. The Aeneid is the record of it.3
The moment was precise. Actium, in 31 BCE, decided the last of Rome’s civil wars.4 Augustus was consolidating a new order the old vocabulary could not name. The republic still existed on paper, the senate still met, but one man now held what the language insisted no one man could hold. Rome was no longer what it had been and had no clear way to describe what it had become.
After a century of bloodshed, Rome did not want a poem that simply named the killing. It wanted one that gave it meaning, one that made the violence feel like a passage to something necessary. Virgil spent the last decade of his life shaping a story that could hold that new reality.
Troy is burning. Aeneas walks out of the collapsing city carrying his father Anchises on his back. Anchises carries the household gods. Aeneas leads his son Ascanius by the hand. Three generations move through the fire in a single gesture. The past, carried. The future, led. The man in between bearing both, because that is the obligation.
The word for that obligation is pietas. The usual English “piety” misses the point. The Latin is harder, more structural. Pietas is the binding duty a person owes to gods, ancestors, children, and city. Virgil did not coin the word. He elevated it. He set it at the center of a chain and made it carry the weight of the whole enterprise.5
Pietas binds to fatum, the fate that fixes Aeneas to his task. It binds to labor, not a job but endurance, the long cost of carrying obligation across time, distance, and war. Arma is the force required to secure what must be built, and it resolves in Roma, the point of it all. Five familiar words, pietas, fatum, labor, arma, Roma, arranged so Romans could read their own condition: bound by duty, driven by fate, enduring the cost, taking up arms, and building a city.
The republican vocabulary lit only a narrow strip of Roman life, the part the republic had been built to name. The defining fact of the age, the rule of one man over a city that still called itself a republic, sat above it in the dark, felt everywhere and named nowhere. Virgil re-arranged a handful of words so the civilization could stand inside its new reality and read its own fate. The lit strip moved upward.
Smith’s Plain Words
Adam Smith took ten men making pins and gave a plain name to what commerce had already become. He took his language from the workshop, not the philosopher’s study.
At the opening of The Wealth of Nations in 1776, Smith describes a pin factory.6 A single laborer, untrained and working alone, could scarcely make a pin in a day. The trade, as Smith reports it, was divided into about eighteen distinct operations: drawing the wire, straightening it, cutting, pointing, shaping the head, and packing among them. In a small shop of ten men, some handling two or three of these operations, output rose to roughly forty-eight thousand pins a day.7
Smith didn’t invent the process. He named it. He showed, in one concrete example, what the division of labor meant in practice. Tradesmen had been dividing tasks for generations. Markets had been coordinating strangers for centuries. Smith saw the pattern whole, understood its force, and gave it a name clear enough that an entire economy could recognize what it was already doing.
The phrase pairs with the one that traveled furthest from the book: the invisible hand.8 It names the order that emerges when many people pursue their own ends without intending the collective result. The pin makers do not aim to supply a nation. Each aims at their own wage. Yet the outcome is a supply cheap enough to be everywhere. There is no hand behind it. There is only the arrangement.
He was a moral philosopher first. Before The Wealth of Nations, he had published The Theory of Moral Sentiments, arguing that social life rests on sympathy. There he introduced the impartial spectator, the imagined observer by whose gaze we judge our own conduct.9 The same thinker who named the invisible hand had already named the mechanism of social conscience. He did not confuse describing the market with endorsing everything it produces.
Commerce had been unfolding in the dark field, felt by all, grasped by none as a whole. Smith supplied the language: division of labor, the invisible hand, self-interest, exchange. Plain words, precisely arranged, let a civilization see the pattern it was already living.
Marx and Keynes: Economies in the Dark
Karl Marx and John Maynard Keynes faced moments when inherited economic vocabulary had gone blind to the reality it was meant to describe. Each arranged a new set of coordinates that let a civilization act on conditions it had been living without a name.
Industrial capital had outrun the words for it. Marx wrote the language in Das Kapital, first published in 1867.10 Take one term: surplus-value (Mehrwert in German). What the worker sells to the capitalist is not the day’s labor but labor-power, the capacity to work. The wage covers the value of that capacity, roughly what it costs to reproduce it. But labor-power at work produces value for a full day, and part of that day’s value goes beyond what the wage covers. That extra portion is surplus-value.11
The term turns a grievance into a measurement. Marx builds his account of profit and exploitation on that gap. Beside it sits commodity fetishism, his name for the trick by which relations between people appear as properties of things, so that value created by human labor seems to sit inside the object the way weight or color does.12 Once named, the trick is hard to unsee.
By treating capital as a social relation and not a pile of machines, Marx handed industrial civilization a set of coordinates for reading itself. The words outlived the politics. Even where Marx’s prescriptions are rejected, economists and critics still reach for surplus-value and commodity fetishism, because the reading holds.
A vocabulary that powerful begins to govern the field it discloses. Once those names were in place, they told a reader where to look and what would count as an explanation. The same terms that made exploitation measurable could make other injuries harder to state in economic terms. Marx’s names opened industrial capital to view so clearly that readers could forget they were using a vocabulary at all. Readers put the terms to uses Marx could not govern.
Keynes built a new set of coordinates when the old ones went dark. In the early 1930s, the classical economics that organized the field could not read the Great Depression. It said the economy should move back to full employment on its own. It did not. The vocabulary had no words for a machine stuck below capacity for years while people who wanted work stood idle.
Keynes supplied the words and gathered them into The General Theory of Employment, Interest and Money in 1936.13 Animal spirits is his plain name for the confidence and fear that drive investment beyond what any calculation can justify, the nerve in economic life, not the arithmetic.14 The desire to hold money itself rather than other assets is liquidity preference, and it lets the interest rate be read as the price of parting with cash.15 Aggregate demand gives government a single quantity it can act on: total spending in the economy as something policy could raise or let fall. The multiplier names the way government spending passes through one set of hands to the next, so each round generates more income than the initial outlay.16
An economy stuck below capacity now had names for the parts that had frozen. With the names came ways to move them. Industrial capital had transformed the world without coordinates to read what it had become. Then a global depression exposed how little the old language could explain. Marx raised the line against the first darkness. Keynes raised it again against the second.
Han and the Present Tension
Byung-Chul Han gives a generation a word for an exhaustion it had been living inside without one. A Korean-German philosopher writing in Berlin, he shows that this labor is still underway in the present tense: a thinker naming a moment while the moment is still unfolding.17
Han’s arrangement begins with a shift. The twentieth-century society Michel Foucault described was disciplinary, organized around prohibition, around the negative command thou shalt not.18 It governed through walls, thresholds, rules, factory bells, prisons, schedules. It told people what they could not do.
Han’s claim is that the twenty-first century has changed the command. It no longer says thou shalt not. It says: you can. It governs through permission. Individuals are free to drive themselves toward achievement, optimize, become the best version of themselves, realize their potential. That freedom becomes the whip. Discipline once arrived as external constraint. Now it arrives as the imperative to achieve, issued by subjects to themselves in the name of their own freedom.
Han calls the person inside that new command the achievement-subject, someone who treats life as project, acts as their own boss, and carries a constant instruction to achieve more. The mechanism by which that figure wears themself down is self-exploitation. The worker under discipline was exploited by another. The achievement-subject exploits themself freely, chasing a potential with no ceiling. They are the one who wields the whip and the one who is whipped, in the same body. That is why they cannot rebel. There is no one to rebel against.
Exhaustion had been running in the dark, through a whole generation, felt in the body everywhere and held nowhere as something with a shape. People drove themselves past the edge without a name for the driving and treated the collapse as a personal failing. Once Han’s terms entered the language, burnout society, achievement-subject, self-exploitation, a civilization could pick up its own condition and read it. The lit strip moved upward while the moment itself was still underway.
The AI Lag
The present is such a moment, and it is the one I work inside. The instruments now being built to act in the world are arriving faster than the language, images, and categories that would let a civilization see what they do.
The systems I work on can reason across more data than any human deliberating body can hold in mind, at a speed no such body can match, and then they act. All of this happens under the oversight of the people answerable for the outcome. The gap matters most at deployment, when the system can act and the people responsible for it do not yet have a clean public language for what kind of action it is.
The inherited words each catch a piece and miss the whole. Automation names a machine doing a fixed task faster than a hand could but says nothing about a system that reasons before it acts. Agent names a thing that acts for you but says nothing about the reasoning underneath. An algorithm is a recipe run from start to finish. Artificial intelligence is too vague to tell an operator what is happening.
The consequence is not academic. A system whose reach the operators can name only in part is acting in the world in ways its builders and the people it acts on can feel but cannot yet read.
The same canvas still applies. The AI moment is one place where the black field has grown faster than the lit strip, where a new reality has outrun the words for it, the way Rome outran the republic’s language and the Great Depression outran the classical economists’ coordinates.
The Failure Mode I Owe the Moment
I build these instruments for a living, and the honest report from inside the work is a failure mode.
A system built to act at a speed no human team can match runs into a limit at the point where a person is supposed to oversee it. When the instrument moves faster than the person set over it can meaningfully weigh, the oversight collapses into ceremony while keeping the shape of control. The sign-off becomes a signature on the outcome, not a check on it.19
A wrong decision made at machine speed is downstream and working its damage before a human could have caught it. A maker who names only what the instrument does well has named only half of it.
The failure mode I have named is not the only one. The instruments hallucinate and fabricate facts to fill the gaps in what they know, synthesizing confidence where they should hold uncertainty. Their reasoning, at the frontier of what foundation models can do, still runs shallow beneath a fluent surface. Not a catastrophe. A specification. Generative AI has failure modes that symbolic AI, vector-symbolic AI, and glass-box machine learning do not share, and honest naming means naming them by kind rather than by press release.
The practice around me has been reaching for a word for the whole instrument. Agentic AI reaches for the acting and decision intelligence centers the reasoning, but neither names the human answerable for the result. The term I have been using is Autonomous Decision-Making, because the decision is the point at which reasoning, action, and responsibility meet in one place and can be examined together.20
Doug Lenat gave me a fifteen-year private tutorial in symbolic AI, at the bench rather than the lectern. We worked with actual systems in front of us, the hard questions of representation open on the table between us. From Euclid to Turing, one discipline held. The load-bearing axioms stayed in plain sight, where anyone answerable for what followed could inspect them and know how the result was reached.21
At an Enterra conference in Napa Valley about ten years ago, Doug and Ed Feigenbaum sat across from Benn Konsynski and me and walked us through what forty years of building symbolic AI had actually taken. Ed on the expert-systems half. Doug on the common-sense ontology and axiomatic theorem-proving at scale he had spent most of his career on. Both older than I was, both of them describing the field as they had built it. Predicate calculus made expressive enough to carry real content. Decades of inference reasoning and common-sense ontology tuned to work at scale. The discipline of keeping the ground of the reasoning inspectable across all of it. Benn and I looked at each other at one point and knew what we were hearing.22
I spent time at a second bench as well, as a visiting professional executive in Herschel Rabitz’s chemistry lab at Princeton, where the working language was high-dimensional mathematics and the objects it was applied to were the control landscapes of molecular systems. One afternoon Doug and I gave a joint talk there, and I asked him whether the geometry of a high-dimensional math space could stand in for the structure of a high-dimensional ontology. Doug on one side of me answering as a lifelong builder of symbolic systems, Herschel on the other side answering from the far end of the same question in the language of continuous mathematics, and I remember stepping half a step back and understanding I was being tutored on both sides at once. The mode was familiar. What these systems required of a machine’s reasoning, a stutter had drilled into a boy one sentence at a time: a small set of load-bearing carriers holding what a diffuse cloud cannot.
The symbols a machine reasons over are the reach the machine has, and choosing those symbols is the discipline of choosing what the machine can and cannot do. That is why the failure mode above is the failure mode I watch for. A system whose designers have not named the ground on which oversight must happen will build oversight that cannot happen.
Whether Autonomous Decision-Making lasts is not for me to decide. Words last only when other people keep needing them. My own term is in the same position. I build the instruments it names, and I try to build safeguards in as carefully as I can, with what discipline I have, as a maker responsibly should. What is done with them I will not ultimately govern. What I owe the moment is the honest naming of what the instrument now does and where the instrument fails.
Raising the Line
Return to that young boy and his canvas. Two percent white at the bottom. Three percent gray above it. Ninety-five percent of the frame still black. Virgil’s five words raised the line over a field of one-man rule that had been all dark. Smith took the workshop and gave a civilization the language for what commerce had become. Marx and Keynes handed industrial economies the coordinates they had been living without. Han catches the present in a word, an exhaustion a generation was living inside and could not name. The instruments I help build, and the words still owed to them, are one more site of the same labor.
The labor is the same wherever it is done. Take the symbols a people already have, verbal or visual or mathematical or set in the categories a civilization uses to sort what it sees, and arrange them so a people can pick up their own moment and read it before the dark takes the moment back. That labor falls not only to poets and philosophers and economists. It is owed by anyone whose work runs where the lit strip ends.
I learned that the first time in a classroom, at the smallest possible scale, from a teacher named Florence Hetzler.23 She was reading Descartes with us out of his Discourse on the Method.24 My stutter caught on the reading. The task in front of the class had been to recite, to give back the text as it had been handed down, and reciting is exactly the task a stutter cannot perform.
Then Hetzler changed the frame. She stopped asking us to recite and asked us to discuss. The moment she did, the block came open, because I was no longer bound to a fixed text and could draw instead from the whole of what I actually had. Without naming it, she replaced an inherited vocabulary, repeat what you were told, with one equal to the situation in front of her, speak to what is actually here.
The frame she changed was small: one boy, one text, one afternoon. Only then did I understand what she was teaching. She became known for an argument that a ruin is a third kind of beauty, neither natural nor made, and that in front of a ruin we come closer to what cannot be fully named or said. She taught the same conviction in the classroom as she wrote it in her essays.
What the stutter had taught, in Hetzler’s classroom, was to trust a small set of words to hold what a diffuse cloud cannot. That trust is the mode of the instruments I would spend my working life building.
The line moves up by inches. The words arrive one at a time. Beyond the last one, the black begins.
Stephen DeAngelis
Princeton, NJ
July 2026
About the Author
Stephen F. DeAngelis is the founder, president, and CEO of Enterra Solutions and Massive Dynamics, two companies that apply artificial intelligence and advanced mathematics to complex enterprise challenges. His work spans international relations, national security, and commercial technology, with visiting research affiliations at Princeton University, Department of Chemistry, the Computing and Computational Sciences and National Security Directorates of the Oak Ridge National Laboratory, the Software Engineering Institute at Carnegie Mellon University, and the MIT Computer Science and Artificial Intelligence Laboratory. He holds patents in autonomous decision science.
Endnotes
1. Ludwig Wittgenstein, Tractatus Logico-Philosophicus, proposition 5.6. The German reads Die Grenzen meiner Sprache bedeuten die Grenzen meiner Welt. Composed during Wittgenstein’s Austrian war service and completed on military leave in August 1918. The finished typescript was in his knapsack when Italian forces captured him near Trento in November 1918, and it was from the prisoner-of-war camp at Cassino in 1919 that he worked, through Russell and Keynes, to get it into print. First published in German in Annalen der Naturphilosophie (1921) and in bilingual English edition by Kegan Paul (1922), translated principally by Frank Ramsey with C. K. Ogden receiving the credit. The Pears and McGuinness retranslation (1961) is the scholarly standard. Proposition 5.6 sits inside the Tractatus’s larger account of logical form and the limits of sayability; the essay applies it to a collective limit rather than an individual one, on the grounds set out in the body.
2. Euclid, Elements, Book I opens with 23 definitions, 5 postulates, and 5 common notions from which every subsequent proposition is derived. The fifth postulate, the parallel postulate, was later shown to be independent of the other four; changing it produces consistent non-Euclidean geometries — Lobachevsky (1829), Bolyai (1832), and Riemann, whose 1854 habilitation lecture Über die Hypothesen, welche der Geometrie zu Grunde liegen was published posthumously in 1868. Standard scholarly reference: T. L. Heath, ed. and trans., The Thirteen Books of Euclid’s Elements, 2nd edition, 3 vols. (Cambridge University Press, 1926; reprinted Dover, 1956).
3. Virgil composed the Aeneid between 29 BCE and his death at Brundisium on 21 September 19 BCE. According to the ancient Vita attributed to Aelius Donatus (drawing on Suetonius), Virgil asked on his deathbed that the manuscript be burned as unfinished; Augustus overruled the request and had the poem published by the poet’s literary executors Lucius Varius Rufus and Plotius Tucca, with instructions to remove what was superfluous and add nothing of their own (ut superflua demerent, nihil adicerent). Varius and Tucca left the half-lines half.
4. The Battle of Actium was fought on 2 September 31 BCE, off the western coast of Greece. Octavian’s fleet, commanded by Marcus Agrippa, defeated the combined forces of Mark Antony and Cleopatra VII of Egypt. Actium did not itself end the war — Antony and Cleopatra were pursued to Egypt and died the following year, and Alexandria fell to Octavian in August 30 BCE — but the battle decisively shifted the war in Octavian’s favor. The civil-war era conventionally dated from the killing of Tiberius Gracchus in 133 BCE closed with the fall of Alexandria. Octavian returned to Rome in 29 BCE and was granted the title Augustus by the Senate on 16 January 27 BCE. The Res Gestae Divi Augusti names Actium in ch. 25 as the war for which “all Italy of its own accord swore allegiance to me”; the restoration of the res publica is dated in ch. 34 to Augustus’s sixth and seventh consulships, 28–27 BCE.
5. Pietas was a Roman virtue with pre-Virgilian usage in Ennius, Plautus, and Cicero, denoting the ordered obligation owed to gods, family, and state. Virgil’s contribution was not lexical invention but structural elevation. Aeneas is marked by his pietas in the proem itself (insignem pietate virum, 1.10) and thereafter carries the epithet pius Aeneas repeatedly, first at 1.220. The clustering of pietas with fatum, labor, arma, and Roma is the essay’s reading of the poem’s ethical architecture rather than an established Virgilian formula; the scholarly context for reading Virgil’s ethical scheme structurally runs through Brooks Otis, Virgil: A Study in Civilized Poetry (Clarendon, 1963), and Denis Feeney, The Gods in Epic (Clarendon, 1991).
6. Adam Smith, An Inquiry into the Nature and Causes of the Wealth of Nations, Book I, Chapter 1, “Of the Division of Labour” (London: W. Strahan and T. Cadell, 1776). Smith’s example is anticipated by the Encyclopédie article “Épingle,” vol. 5 (1755), attributed to Alexandre Deleyre and drawing in turn on Jean-Rodolphe Perronet’s earlier manuscript description. Smith describes the trade as being divided into about eighteen distinct operations, matching Deleyre’s count; his own contribution is the ten-man workshop, in which some hands perform two or three operations each. The compression is what let one workshop stand in for an entire economy of coordinated specialization. Whether Smith ever saw an actual pin factory is debated (Kay and others argue his account descends from Deleyre and Perronet without a first-hand visit); nothing in the argument depends on the answer.
7. Smith writes that “those ten persons, therefore, could make among them upwards of forty-eight thousand pins in a day,” and that “each person, therefore, making a tenth part of forty-eight thousand pins, might be considered as making four thousand eight hundred pins in a day.” The unimproved single-hand output he estimates in the same paragraph: a workman unfamiliar with the trade “could scarce, perhaps, with his utmost industry, make one pin in a day, and certainly could not make twenty.” Book I, Chapter 1 of the Glasgow edition of the Wealth of Nations (R. H. Campbell and A. S. Skinner, eds., W. B. Todd, textual ed.; Clarendon, 1976), paragraph I.i.3.
8. The phrase invisible hand appears three times in Smith’s published work: in the History of Astronomy (posthumous, 1795, but composed circa 1758), in The Theory of Moral Sentiments (Part IV, Chapter 1, 1759), and in The Wealth of Nations (Book IV, Chapter 2, 1776). In WN it names the mechanism by which a merchant investing capital at home, intending only his own gain, “is in this, as in many other cases, led by an invisible hand to promote an end which was no part of his intention.” The interpretive literature is now anchored in Emma Rothschild, Economic Sentiments: Adam Smith, Condorcet, and the Enlightenment (2001), which argues that the phrase carried more irony and less doctrinal weight in Smith than in the twentieth-century canon that made it central.
9. Adam Smith, The Theory of Moral Sentiments (London: A. Millar, 1759). The impartial spectator is developed principally in Part III, “Of the Foundation of our Judgments concerning our own Sentiments and Conduct, and of the Sense of Duty.” Smith revised the work through six editions across his lifetime, the sixth (1790) containing the most extensive treatment. The relation between TMS and WN has been a scholarly touchstone since the “Adam Smith Problem” of the late nineteenth century (the alleged tension between the sympathy-based TMS and the self-interest-based WN), which twentieth-century scholarship largely dissolved by reading the two works as complementary rather than competing.
10. Karl Marx, Das Kapital: Kritik der politischen Ökonomie, Erster Band (Hamburg: Otto Meissner, September 14, 1867). Volumes II and III were edited from Marx’s notebooks by Friedrich Engels and published in 1885 and 1894 respectively. The first English translation, by Samuel Moore and Edward Aveling with Engels’s editorial supervision, appeared in 1887; the standard modern English is Ben Fowkes’s translation for Penguin (Volume I, 1976). The critical edition is the MEGA (Marx-Engels-Gesamtausgabe), second series.
\11. Mehrwert is Marx’s term of art. The German is often retained in translation-critical discussion because surplus-value imperfectly renders the wider Wert (value) family from which Mehrwert derives. The mechanism is not the gap between what a worker produces and what a worker is paid, but the gap between the value produced by labour (Arbeit) across the working day and the value of the labour-power (Arbeitskraft) the worker sells to the capitalist for a wage. Necessary labour reproduces the value of labour-power (conventionally, the means of subsistence); labour beyond that point is surplus labour, and the value it creates is Mehrwert. Marx distinguishes Mehrwert from Profit (Gewinn): surplus-value is the theoretical category naming the extraction at the point of production, profit its transformed appearance measured against total capital. The core analysis, in Ben Fowkes’s Penguin translation (Volume I, 1976), sits in Part Three, chs. 7–9 (“The Labour Process and the Valorization Process”; “Constant Capital and Variable Capital”; “The Rate of Surplus-Value”).
12. Warenfetischismus is elaborated in Das Kapital Volume I, Chapter 1, Section 4, “The Fetishism of the Commodity and Its Secret.” The interpretive uptake in twentieth-century Marxism runs principally through Georg Lukács, History and Class Consciousness (1923), where it is developed into the theory of reification (Verdinglichung), and through the Frankfurt School’s critique of consumer culture. The concept remains active in contemporary economic sociology and anthropology.
13. John Maynard Keynes, The General Theory of Employment, Interest and Money (London: Macmillan, February 4, 1936). Keynes composed the book across the years 1932–1935, in extensive correspondence with Dennis Robertson, Ralph Hawtrey, Roy Harrod, and Joan Robinson, portions of which are collected in Volumes XIII and XIV of the Collected Writings of John Maynard Keynes (Cambridge University Press, D. E. Moggridge, ed.). The book’s argumentative structure, and Keynes’s own sense of what was novel in it, is best captured in his 1937 Quarterly Journal of Economics summary essay “The General Theory of Employment.”
14. Animal spirits appears in The General Theory Chapter 12, “The State of Long-term Expectation,” §VII, where Keynes writes that “most, probably, of our decisions to do something positive, the full consequences of which will be drawn out over many days to come, can only be taken as a result of animal spirits — of a spontaneous urge to action rather than inaction, and not as the outcome of a weighted average of quantitative benefits multiplied by quantitative probabilities.” The term itself derives from the medieval and early modern Latin spiritus animales, refined philosophically by Descartes and Hume. The modern behavioral-economics revival is anchored in George Akerlof and Robert Shiller, Animal Spirits: How Human Psychology Drives the Economy (2009).
15. Liquidity preference is developed in The General Theory Chapter 13, “The General Theory of the Rate of Interest,” and Chapter 15, “The Psychological and Business Incentives to Liquidity.” Keynes distinguishes three motives for holding money: the transactions-motive, the precautionary-motive, and the speculative-motive. The last is the theoretically important one, since it makes the interest rate the price at which the desire to hold wealth in liquid form is equated with the available quantity of money. John Hicks formalized the relation between liquidity preference, aggregate demand, and the goods market in “Mr. Keynes and the ‘Classics’: A Suggested Interpretation,” Econometrica 5, no. 2 (April 1937), producing the IS-LM apparatus that organized macroeconomic teaching for the next half-century.
16. The multiplier’s original formulation is Richard Kahn, “The Relation of Home Investment to Unemployment,” Economic Journal 41, no. 162 (June 1931). Keynes incorporated and systematized Kahn’s result in The General Theory Chapter 10, “The Marginal Propensity to Consume and the Multiplier,” reworking it as the numerical relationship between the marginal propensity to consume and the aggregate income response to an autonomous change in expenditure. Kahn is often cited as author of the multiplier concept; Keynes is the author of its central role in macroeconomic policy analysis.
17. Byung-Chul Han (born 1959 in Seoul) studied metallurgy in Korea before emigrating to Germany in the 1980s and completing a doctorate in philosophy at Freiburg with a dissertation on Heidegger. He was professor of philosophy and cultural studies at the Universität der Künste in Berlin from 2012 to 2017. The principal terms invoked here appear in Müdigkeitsgesellschaft (Matthes & Seitz Berlin, 2010; English The Burnout Society, Stanford University Press, 2015, translated by Erik Butler): Leistungssubjekt, the German for what the English renders “achievement-subject,” and Selbstausbeutung, “self-exploitation.” Han’s related short book Psychopolitik (2014; English Psychopolitics, Verso, 2017) extends the argument to the neoliberal politics of self-optimization.
18. Michel Foucault, Surveiller et punir: Naissance de la prison (Paris: Gallimard, 1975), translated as Discipline and Punish: The Birth of the Prison by Alan Sheridan (New York: Pantheon, 1977). The book’s reading of Jeremy Bentham’s Panopticon (1791) as the emblem of the disciplinary society organizes the argument. Han’s positioning of the twenty-first century as post-disciplinary is a deliberate departure from, not a rejection of, the Foucauldian framework: the mechanism has shifted from external command to internalized imperative, but the analysis still proceeds through the categories Foucault made available.
19. The failure mode described here — human oversight collapsing into signature-ceremony when the system moves faster than the person set over it can meaningfully weigh — is older than the AI literature. It is the automation-bias and human-supervisory-control problem catalogued in aviation and process-control research from the 1980s onward, most sharply in Raja Parasuraman and Victor Riley, “Humans and Automation: Use, Misuse, Disuse, Abuse,” Human Factors 39, no. 2 (1997), 230–253. Three adjacent literatures now bear on the question and should be kept distinct. The National Institute of Standards and Technology, Artificial Intelligence Risk Management Framework 1.0 (NIST AI 100-1, January 2023) sets the broad framework; its subsequent Generative AI Profile (NIST AI 600-1, July 2024) names “automation bias” and “over-reliance” explicitly, under “Human-AI Configuration” risk. Distinct from automation bias is scalable oversight, the problem of how weaker humans can accurately supervise stronger models when the model’s answers are too many or too complex to check exhaustively, which enters the AI safety literature through Dario Amodei et al., “Concrete Problems in AI Safety” (arXiv:1606.06565, 2016), and is pursued directly in Samuel R. Bowman et al., “Measuring Progress on Scalable Oversight for Large Language Models” (arXiv:2211.03540, 2022).
20. Autonomous Decision-Making is the term I use for a system that reasons across information a human team cannot hold in mind, acts on that reasoning at a speed no such team can match, and does so under the responsibility of the people answerable for the result. The three components — reasoning, action, and accountable oversight — must be held together as one object, not treated as three separate engineering problems, because it is at their intersection that the failure mode above appears. Adjacent terms in current use — agentic AI (which foregrounds action), decision intelligence (which foregrounds analytic reasoning), AI orchestration (which foregrounds coordination) — each catch a piece but none holds the whole. Whether the term I have been using outlasts these adjacent ones is not for me to decide; the discipline is to keep the object visible whatever it is eventually called.
21. The discipline survived its largest claim. Euclid’s Elements places definitions and postulates in view, making each proof answer to them (Heath edition as cited above). Boole algebraized logic in The Mathematical Analysis of Logic (Macmillan, Barclay & Macmillan, 1847), Frege’s Begriffsschrift (Louis Nebert, 1879) formalized quantification, and Peano axiomatized arithmetic in Arithmetices principia (Fratelli Bocca, 1889). Hilbert’s Grundlagen der Geometrie (Teubner, 1899) and Whitehead and Russell’s Principia Mathematica (Cambridge University Press, 1910–1913) pressed the method toward foundations. Gödel marked the limit of that ambition. His 1931 theorems showed that a consistent, effectively axiomatized system strong enough for arithmetic cannot prove all arithmetical truths or its own consistency (“Über formal unentscheidbare Sätze der Principia Mathematica und verwandter Systeme I,” Monatshefte für Mathematik und Physik 38, 1931, 173–198). The result ended Hilbert’s program to formalize all mathematics and secure its consistency by finitary proof. It did not end the discipline of keeping the ground visible to the person answerable for what follows. Church and Turing clarified effective procedure in 1936, and Tarski formalized truth and logical consequence in the monograph published in Polish in 1933 and in German in 1935. McCarthy convened the Dartmouth project in 1956, where Newell and Simon brought the Logic Theorist. Lenat’s work carried the inheritance into large-scale, explicitly axiomatized systems for machine reasoning.
22. The conference was an Enterra Solutions AI Conference in Napa Valley, 2017. The conference and its participants are preserved in photographs; one, showing Doug Lenat, Edward Feigenbaum, C. Gordon Bell, and Benn Konsynski at the event, is reproduced in Stephen DeAngelis, “In Memoriam: C. Gordon Bell,” LinkedIn, 2024. Edward A. Feigenbaum (Stanford University, emeritus, Kumagai Professor of Computer Science) shared the Turing Award in 1994 with Raj Reddy for pioneering work on expert systems; his foundational text is The Handbook of Artificial Intelligence, co-edited with Avron Barr and Paul R. Cohen, 4 vols., 1981–84. Benn R. Konsynski is George S. Craft Professor of Information Systems and Operations Management at Emory’s Goizueta Business School, where he has taught since 1992 following six years at Harvard Business School.
23. Florence M. Hetzler (1926–1991) taught French, Latin, and philosophy at Scarsdale Senior High School from 1951 to 1985, and held a simultaneous appointment as professor of philosophy at Fordham University’s School of General Studies from 1968 until her death. She completed her Fordham doctorate in 1959 on Aquinas reading Aristotle’s Physics, was a Fulbright Scholar at the Sorbonne in 1960, and spent 1961–62 as professor of philosophy at Canisius College. Her most cited essay, “The Aesthetics of Ruins: A New Category of Being,” Journal of Aesthetic Education 16, no. 2 (Summer 1982), 105–108, argued that a ruin is a third kind of beauty, neither natural nor artistic, and that in front of a ruin we come closer to the sublime, the ineffable, and the indescribable; she developed the argument further in “Causality: Ruin Time and Ruins,” Leonardo 21, no. 1 (1988), 51–55, from which the term ruin time enters the literature. She founded the International Brâncuși Society and edited Art and Philosophy: Brâncuși, The Courage to Love (Peter Lang, 1991), which appeared the year she died. Her 1982 French translation of Mircea Eliade’s play La colonne sans fin, which takes Brâncuși’s Endless Column as its subject, is preserved among the Mircea Eliade Papers at the University of Chicago; Eliade attended a panel of hers at Notre Dame in April 1978. The scene described here belongs to the Descartes reading in her Scarsdale classroom during my junior and senior years, 1978–1980, before I graduated in June 1980.
24. René Descartes, Discours de la méthode pour bien conduire sa raison, et chercher la verité dans les sciences (Leiden: Ian Maire, 1637), published anonymously in French (not in the scholarly Latin of the day) together with the treatises La Dioptrique, Les Météores, and La Géométrie. The cogito passage on which the class turned — je pense, donc je suis — sits in Part IV of the Discours. The canonical Latin cogito ergo sum is not Descartes’s formulation in this text; it entered his corpus in the Principia Philosophiae (1644, I.7). The pedagogical challenge in the classroom was that the Discours is a first-person text that resists being recited without the reader becoming, momentarily, the “I” who is speaking.
Share this post
