The Students of Technology

I. The Flying Car and the Phone

Open a magazine from 1958, one of the glossy ones with a chrome-nosed future on the cover, and you will find the flying car. It hovers over a split-level house with a wife in the doorway and a dog on the lawn (the dog, for reasons the illustrators never felt obliged to explain, is nearly always a collie), and the caption promises that by 1975, or 1985 at the very latest, every family will own one. The flying car is not a fringe fantasy. It is the consensus forecast of engineers, admen, and government pamphleteers alike, the thing everybody with a subscription and an income expected to be driving, or rather flying, well within their own lifetime. It never arrived. What arrived instead, decades later than promised and in a shape nobody had quite drawn, was a slab of glass that fits in a coat pocket and holds, among other things, a map of the entire world, a library larger than Alexandria's, a photograph of every face you have ever loved, and a way to summon a stranger in a car to your exact location within minutes. Nobody's cover illustration anticipated that. The flying car, obvious and confidently dated, failed to show up. The phone, unimagined in that particular form by anyone drawing the future in 1958, showed up anyway and then some.

The contrast is not a cute irony to be enjoyed and set aside. Studied as its own subject, rather than as an instance of economics or an occasion for philosophy, technology shows one pattern above all others: over the past half century, progress stalled in atoms while it raced in bits. We cross the Atlantic no faster than our grandparents did, while the machinery of information has compounded past anything anyone thought to promise them. And the stall was not a law of nature. It was a choice, made by identifiable people, in identifiable institutions, for identifiable reasons, and a choice, unlike a law of nature, can be revisited. Most public conversation never gets near this pattern, because it asks a different question. We ask whether technology is good or bad, whether it will save us or replace us, whether the latest product announcement should make us cheer or panic, and the question of what technology actually is and where exactly it has stalled goes begging. Ask a reasonably informed person whether technology is accelerating and they will say yes, gesturing at the phone, at the speed with which a smartphone in 2026 outstrips one from 2016, at the fluency of a language model answering a question that would have required a research librarian a decade ago. Ask the same person why we do not have flying cars, why intercontinental flight takes about as long today as it did in 1970, why nobody has walked on the moon since 1972, why a new nuclear reactor takes longer to build now than the entire Manhattan Project took to invent the bomb, and the answer trails off. Something about safety. Something about the environment. Something about there being no money in it. A handful of writers have refused to settle for the trailing off. They sat down and tried to answer with rigour rather than sentiment, asking what technology, considered as a thing in itself, with its own patterns and its own stalls, actually does, and what studying it closely, rather than cheering or fearing it from the sidelines, teaches a person about hope.

The writers who have taken this up are an unlikely committee. An engineer who builds molecular machines and keeps a slide rule's worth of contempt for regulators. A former hippie who wandered Asia with a camera and ended up editing a countercultural magazine and then founding a technology one. A parasitologist and a cartoonist who spent three years interviewing physicists about fusion reactors that are always, reliably, thirty years away. A wartime research administrator who, weeks before Hiroshima, wrote an essay imagining a desk that would let a person's mind roam freely across the whole of recorded knowledge. None of them agree on much. What unites them is a discipline, and it is rarer than it sounds: they treat technology itself as the object of study. The economist treats technology as a residual, the unexplained remainder in a growth equation after capital and labour have been counted. The philosopher tends to treat it as a symptom, an occasion for an argument about alienation or authenticity that was going to happen anyway. These writers study it the way a naturalist studies a genus of beetle: as a thing with its own anatomy, its own habits, its own stalls and migrations, deserving accurate description before anyone rushes to a verdict. They do not ask whether technology is good. They ask what it is, where exactly it has stalled, where it is heading as opposed to where the brochures say it is heading, and they do this with enough technical literacy and enough patience that their conclusions, when they arrive at hope, arrive at it the way a bridge inspector arrives at confidence in a bridge: by testing the rivets rather than admiring the view.

II. The Curve That Broke

J. Storrs Hall is not a professional futurist in the sense of someone who makes a living giving keynote speeches about disruption. He is an engineer who has spent his career close to the actual machinery of the future he writes about: founding chief scientist of a company called Nanorex that modelled the behaviour of molecular machinery before such machinery existed to be modelled, former president of the Foresight Institute, the oldest organisation dedicated to nanotechnology, and a research fellow at the Institute for Molecular Manufacturing. When Hall writes about flying cars, in other words, he is not being whimsical. He believes we could have had them, has done the back-of-envelope engineering to support the claim, and wants to know, with real irritation, why we do not. The flying car is by now less a vehicle than a genre of grievance, and Hall is the genre's most technically qualified practitioner.

His book, self-published in 2018 under the title Where Is My Flying Car? A Memoir of Future Past and picked up by Stripe Press in 2021 for a wider readership, opens with a piece of historical arithmetic that is easy to state and hard to forget. Henry Adams, grandson and great-grandson of presidents, wrote in his own autobiography, The Education of Henry Adams, about the rate at which raw energy available to an ordinary American had grown across the nineteenth century. Adams noticed that this growth was not merely steady but compounding, doubling roughly every decade for a period spanning generations, a rate Hall calculates at something close to seven percent a year sustained for over a century. Adams treated this as close to a law of history, an engine driving everything else, wealth, mobility, the whole apparatus of modern life, forward at a fixed and apparently permanent rate. Hall names this trajectory the Henry Adams curve, an honour Adams, a man who wrote his entire autobiography in the third person, would presumably have accepted without protest, and then does something Adams never lived to do: he traces it forward into the twentieth century and shows the reader the precise moment it breaks.

That moment is around 1970. Per capita energy consumption in the United States and other advanced economies, having climbed for a century, flattens. It does not collapse. It simply stops climbing at the old rate, settling into something close to a plateau that persists for decades. Everything downstream of energy availability, in Hall's account, flattens with it: the growth in physical capability, in mobility, in the kind of raw material abundance that had defined the previous hundred years. Run the old curve forward as a thought experiment, ignore the flattening, and the household income and physical capability of an ordinary family today would look almost embarrassingly generous by comparison with what actually happened. Hall is careful, and this is the crux of his whole argument, to insist that nothing in physics required the curve to break. There was no wall of thermodynamics, no resource exhaustion, no law of nature standing in the way of continued growth in the energy available to ordinary people. What changed was not the physical world. What changed was a set of institutional and cultural decisions, and decisions, unlike physical law, can be revisited.

Hall performs something close to an autopsy on the broken curve, and three bodies turn up. The first is nuclear power, abandoned in spirit if not entirely in fact after the accident at Three Mile Island in 1979, a partial meltdown that killed nobody and released, by most official estimates, a dose of radiation to the surrounding population smaller than a chest X-ray, but that triggered a collapse in public confidence utterly disproportionate, in Hall's accounting, to the actual physical risk involved. Reactor orders, which had been running at a rate of dozens a year through the 1960s and early 1970s, essentially stopped. The technology did not become less safe after 1979; it became, if anything, considerably safer, as later designs incorporated passive safety systems that made a repeat of Three Mile Island's particular failure mode nearly impossible. What stopped was the willingness to build.

The second body is cold fusion, a field so thoroughly discredited by the 1989 announcement from Stanley Pons and Martin Fleischmann, and the humiliating failure of other laboratories to replicate their claimed results, that the entire subject became professionally toxic. Hall's complaint here is not that Pons and Fleischmann were vindicated; he does not believe they were. His complaint is that the debacle foreclosed an entire category of legitimate low-energy nuclear research for a generation, the scientific equivalent of a courtroom where one defendant's guilt gets the whole neighbourhood evicted.

The contrast with nuclear power's earlier, forgotten optimism carries half of Hall's argument. In the 1950s the mood surrounding the atom was not caution but a kind of untroubled swagger. Lewis Strauss, chairman of the Atomic Energy Commission, told a gathering of science writers in 1954 that nuclear electricity would one day be "too cheap to meter," a phrase that has been quoted derisively ever since but that captured, accurately enough, the genuine engineering expectation of the moment. President Eisenhower's Atoms for Peace programme, launched the same year, aimed to spread civilian reactor technology around the world as a counterweight to the bomb's terror, and for two decades the industry more or less delivered on the promise, bringing reactor after reactor online at a pace that made nuclear power, briefly, the fastest-scaling source of electricity in industrial history. The plants ordered in the 1965 to 1975 window took, by most accounts, an average of a little over six years from order to operation. A plant ordered in the United States today, on the rare occasion one is ordered at all, routinely takes twice that, not because the engineering has become harder but because the licensing has become, in Hall's word, thicker. Thicker is not a metaphor. It is one of the few claims in the regulatory literature that can be verified with a ruler.

The third body, closest to Hall's own working life, is molecular nanotechnology: the prospect of manufacturing with atomic precision, building physical objects the way a chemist builds a molecule, one bond at a time, which Hall believes could do for the manipulation of matter something like what the transistor did for the manipulation of information. The intellectual seed of this idea is usually traced to a 1959 talk by the physicist Richard Feynman, "There's Plenty of Room at the Bottom," in which Feynman mused, half as a party trick and half in earnest, that there was no fundamental physical law preventing engineers from eventually building machines that manipulated individual atoms. The idea lay dormant for decades until K. Eric Drexler, a colleague of Hall's in the same small community of nanotechnology researchers, gave it a fuller and more technical treatment in his 1986 book Engines of Creation, arguing that molecular assemblers, machines capable of building other machines atom by atom, were a coherent engineering target rather than a fantasy. Hall watched this field, in his own account, starved of serious funding and mocked into irrelevance by a combination of scientific gatekeeping, most visibly a public dispute in the pages of Chemical & Engineering News between Drexler and the Nobel laureate Richard Smalley, and popular anxiety about grey goo, a science-fiction scenario in which self-replicating nanomachines consume the biosphere, a scenario Hall regards as roughly as physically plausible as werewolves but considerably more damaging to funding committees.

Behind all three case studies sits a mechanism Hall names, with the flat precision of an engineer labelling a part, the regulatory ratchet. Rules, once written, are rarely removed. Review processes, once added, tend to accumulate rather than streamline. Liability exposure grows with every incident, real or merely alleged, and almost never shrinks, because no regulator has ever been fired for being too cautious, while plenty have been pilloried for being insufficiently so, which makes caution the one resource the energy sector has never run short of. The result, generation after generation, is that each new cohort of engineers inherits a thicker layer of institutional friction than the one before it, entirely independent of whether the underlying risks have actually grown. A nuclear plant proposed today must clear regulatory hurdles that did not exist in 1965, most of which accumulated not because reactors got more dangerous but because the ratchet only turns one way. Hall pairs this mechanical account with a cultural one: a shift, beginning in the 1960s counterculture and never fully reversing, away from enthusiasm for large technological projects and toward a reflexive suspicion of them, a suspicion that supplied the political appetite the ratchet needed to keep turning. He has a coinage for the acute form of this suspicion, ergophobia, a learned fear of energy itself, and he frames the whole diagnosis with Arthur C. Clarke's old distinction between failures of nerve and failures of imagination. The twentieth century, on Hall's reading, did not suffer a failure of imagination; the flying car had been imagined down to the upholstery. It suffered a failure of nerve about the energy required to lift it.

Hall extends the diagnosis, in one of the book's more provocative passages, to the aircraft a passenger boards at an ordinary airport today. A Boeing 707, the workhorse jet that opened the era of mass commercial air travel in 1958, cruised at roughly the same speed as the aircraft that eventually replaced it decades later, and in some respects, cabin comfort, ease of boarding, actual door-to-door travel time once security screening is added back in, the experience of flying commercially has if anything regressed rather than improved. The supersonic Concorde, which briefly promised to cut transatlantic flight time by more than half, was retired in 2003 and never replaced, a rare case of an advanced capability being withdrawn from service entirely rather than merely failing to improve further. Humanity had supersonic passenger flight, tried it for a generation, and handed it back. Hall reads this not as evidence that supersonic passenger flight was uneconomical in some fixed, permanent sense, but as a second data point for the same pattern he traces in energy: a technology that worked, that people who used it loved, allowed to lapse under a combination of noise regulation, fuel cost, and a general cultural mood that had stopped expecting speed to keep increasing and had therefore stopped demanding it. The Henry Adams curve, in other words, was never confined to the electricity meter. It shows up, on Hall's reading, in aviation, in construction, in the general pace at which an ordinary citizen can move a heavy object from one place to another, all of it bending toward the same flat plateau after 1970, all of it for reasons that trace back to the same loss of institutional nerve. The image Hall reaches for, precise in the way only an engineer's image can be, is of a rocket that reaches the edge of the atmosphere and is then ordered to throttle back, not because it has run out of fuel but because someone on the ground has lost their nerve.

Hall reads better against the objections than without them. The first is that the book grieves for a future that was never actually on offer. The flying car of the 1958 magazine cover was a marketing image, not an engineering plan, and its obstacles were never primarily regulatory. Vertical take-off is brutally expensive in energy no matter who signs the permits; a thousand ducted-fan commuters over a suburb would produce a noise no community has ever volunteered to live under; and the median driver, alarming enough in two dimensions, would in three be a public hazard of an entirely new order. Hall has answers to some of this, autopilots for the piloting problem, decades of private aviation data for the safety problem, but the suspicion lingers that the emblem of his argument is the weakest instance of it. The second objection cuts deeper: the regulatory story is too monocausal. The energy flatline around 1970 coincides with the 1973 oil shock, with the drift of advanced economies from steel-making toward services, and with real gains in efficiency, so that some of the flatness in energy per person reflects doing more with less rather than doing less. Economists such as Robert Gordon and Tyler Cowen have offered a rival account of the same stagnation, in which the low-hanging fruit of the second industrial revolution, electrification, plumbing, the internal combustion engine, had simply been picked, and no quantity of regulatory courage was going to produce a second harvest of comparable size. And nuclear construction costs rose over time even in France, which built dozens of reactors on a state-driven schedule through the very decades America stood still, which suggests that something in the economics of vast concrete projects, and not only the American ratchet, was working against the atom. Hall's reply, implicit throughout the book, is that France is precisely his point: a different institutional choice produced a different curve. The reply is good but not complete. The fairest verdict is that the ratchet is real and demonstrable, and that Hall promotes it from a cause to the cause, which is the occupational disease of anyone who has found a mechanism nobody else is looking at. Even trimmed back to one cause among several, though, his claim keeps its falsifiable edge: if any substantial part of the stagnation was policy, then a shift in policy should bend the curve back, visibly, within years rather than generations.

III. The Curve, Bending Back

It is being tested now, and not for the reasons Hall or anyone else quite expected. The revival of large-scale interest in nuclear power over the past several years has come not from a public reconciliation with the atom, not from a Greenpeace conversion, not from a sudden mass rereading of the Three Mile Island data, but from something far less romantic: the electricity bill of the artificial intelligence industry. Training and running the large models that now answer questions, write code, and generate images requires an amount of continuous, reliable electrical power that renewable sources, constrained as they are by weather and storage, struggle to guarantee at the scale and steadiness these data centres demand. The technology companies that need this power discovered, in effect, what Hall had been arguing for years: that nuclear energy remains, by a wide margin, the most energy-dense, most reliable source of continuous electricity available to an industrial economy, and that the obstacles to building more of it were never really physical.

By 2026 the shift had become concrete rather than rhetorical. Something on the order of fifteen reactors are under active construction or subject to firm financial commitment in the United States, a figure that would have seemed fanciful as recently as 2020, when the American nuclear industry was widely regarded as a mature technology in terminal decline, kept alive mainly by the slow amortisation of plants built in the 1970s and 1980s. A parallel and much larger buildout is under way in China, where the state has never fully absorbed the Western cultural anxiety Hall blames for the ratchet and has continued, with only modest interruption, to bring new reactors online through the very decades the American programme stood essentially still. Alongside the large plants has come a wave of ventures dedicated to small modular reactors, factory-built units small enough to be trucked to a site and assembled rather than poured in place over a decade (the accompanying licensing has so far declined to miniaturise on the same schedule), financed in no small part by the same technology companies whose data centres are driving the demand in the first place. It is a strange kind of closing of the loop: the industry that promised, in the 1990s and 2000s, to dematerialise the economy into pure information turns out to need, more than almost any industry before it, staggering quantities of reliable base-load electricity, and has gone looking for it in the one place the previous half-century had declared off limits.

The individual deals read, to anyone who remembers the industry's long torpor, almost like satire of themselves. Microsoft signed an agreement to restart Unit 1 of the Three Mile Island plant, the very site whose Unit 2 meltdown in 1979 supplies the opening scene of Hall's autopsy, rebranding it the Crane Clean Energy Center and committing to buy its output for two decades to power its data centres. Amazon and Google, separately, struck agreements with small modular reactor developers and with existing nuclear operators to secure power directly, bypassing the ordinary utility procurement process entirely, the corporate equivalent of a household deciding it would rather dig its own well than wait for the water board. None of this activity is being driven by an environmental ministry or a climate summit. It is being driven by engineers at hyperscale computing companies doing the same arithmetic Hall did in his book: that a data centre running around the clock needs power that does not depend on whether the wind is blowing, and that nuclear, whatever its reputation, remains the only source dense and steady enough to supply it at the scale required. The historical irony is almost too neat, and Hall, to his credit, would likely find it more satisfying than surprising: it was never physics that stood in nuclear power's way, and now that a large enough buyer wanted the electricity badly enough to absorb the ratchet's remaining friction, the friction began, however slowly, to give.

The sources that track this most carefully supply a caveat. A piece in the Bulletin of the Atomic Scientists in 2026 cautioned that a meaningful share of the announced nuclear capacity tied to data centre demand remains, for now, closer to a press release than to poured concrete: memoranda of understanding, letters of intent, ambitious slide decks from small modular reactor start-ups that have yet to pour a foundation. And yet it is not really a rebuttal of Hall's thesis so much as a confirmation of its finer print. Hall's argument was never that the regulatory ratchet had vanished. It was that the ratchet was a human artefact, not a law of nature, and human artefacts, once the underlying incentives shift hard enough, eventually get dismantled, however slowly and however much grinding of gears accompanies the process. The gap between the fifteen reactors under firm commitment and the larger number merely announced is precisely the ratchet still turning, friction that has not yet been overcome so much as outspent. That the outspending is happening at all, driven by the unglamorous economics of a data centre's power bill rather than by any public romance with the atom, is itself a kind of vindication, arriving by the side door of commercial necessity rather than the front door Hall might have preferred, of public persuasion and a change of heart.

There is a fitting footnote here, one Hall himself would likely enjoy, involving Stewart Brand, a man who manages to turn up on both sides of this story. Brand, publisher of the Whole Earth Catalog and one of the most influential voices of the very 1960s counterculture Hall blames for supplying the ratchet's political fuel, reversed himself in his 2009 book Whole Earth Discipline, arguing that the environmental movement's inherited fear of nuclear power had cost the planet decades of clean energy it could ill afford to lose. It is not every day that a movement's own patron saint disowns one of its founding anxieties. Brand's conversion did not, on its own, build a single reactor. But it marked, a decade and a half before the data centre boom made the argument commercially urgent, that the cultural tide Hall diagnosed was not monolithic even at its source, and that the same generation which had once recoiled from the atom contained within it the seeds of its own reconsideration.

None of this proves Hall's larger project, the recovery of the whole Henry Adams curve through nanotechnology paired with abundant energy, is on track. Nanotechnology in Hall's atomically precise sense remains largely undelivered, a promise still mostly on paper.

IV. The Technium and What It Wants

Where Hall sees a curve broken by identifiable human decisions, Kevin Kelly sees something much larger and, in his telling, much harder to stop. Kelly's path to this view was not an engineer's path. He left college to wander Asia for years with little more than a camera, a luggage philosophy he has never seen persuasive grounds to revise, and it was Stewart Brand, that same recurring figure, who in 1983 hired him to edit the later editions of the Whole Earth Catalog and then, from 1984 to 1990, the Whole Earth Review. Kelly went on in 1993 to become the founding executive editor of Wired, and later served as co-chair of Brand's Long Now Foundation, an organisation dedicated, fittingly, to thinking in terms of centuries rather than quarters. This is a résumé built entirely inside the technological counterculture and its respectable descendants, which makes Kelly's eventual argument, that technology is not merely a tool humans wield but something closer to a living system with wants of its own, feel less like an outsider's theory and more like a field report from deep inside the thing being described.

Kelly's 2010 book What Technology Wants introduces a word he needed to invent because no existing word did the job: the technium. The technium is the sum total of everything humans have made, not just machines and gadgets but techniques, institutions, art, language, law, the entire accumulated apparatus of human invention, considered not as a pile of separate objects but as a single, coherent, massively interconnected system, what Kelly calls, with a phrase that has stuck, "the greater, global, massively interconnected system of technology vibrating around us." His claim is not merely poetic. He argues the technium behaves, in important structural respects, like a living organism: it grows, it has something resembling a metabolism, it exhibits patterns of development that recur across wildly different domains, and it emerged from biological evolution as a direct extension of tendencies already present in life itself, rather than as evolution's opposite or rival.

The verb in the title is doing careful work, not sloppy anthropomorphism. When Kelly says technology wants something, he means it roughly the way a biologist means it when saying a plant wants sunlight: not that the plant has formed a conscious preference and voted on it, but that across billions of individual growth decisions made by billions of individual cells, a statistical pull toward the light asserts itself so reliably that it becomes useful, even necessary, to describe it as a want. Technology, in Kelly's account, exhibits the same kind of aggregate pull across billions of small human choices, inventions, adoptions, abandonments, none of them individually decisive, all of them together bending the technium in certain directions and not others. He lists the tendencies this pull favours: toward increasing efficiency, increasing complexity, increasing diversity, increasing specialisation, increasing freedom, increasing mutualism between different technologies and their users, increasing beauty, and, more speculatively, increasing sentience, a list on which nothing decreases. He gathers all of this under a coinage of his own, exotropy, an outward-turning counterforce to entropy, the tendency of the universe to run down into disorder. Where entropy scatters, exotropy, in Kelly's usage, gathers and complicates and builds.

There is a further distinction Kelly draws that is easy to miss but matters a great deal to the shape of his argument: technology, he says, is best understood as an extension of mind rather than of genome. A pair of shoes extends the skin. A wheel extends the foot. A telescope extends the eye. These are useful extensions, but they are, in a sense, conservative ones, magnifying what a body could already do in some crude form. Writing and mathematics are different in kind. They extend the reach of a single mind not merely across space but across time, into minds not yet born, allowing a thought formed by one person in one century to be picked up, verified, and extended by a stranger in a century that person will never see. This is, in Kelly's telling, the technium's most significant trick, and it reappears later in Vannevar Bush's very different but strikingly compatible image of the associative trail.

The most persuasive part of What Technology Wants, and the part least often quoted by people who have only skimmed the book for its more sweeping claims, is Kelly's extended portrait of the Amish. It would be easy to cast the Amish as technology's opponents, the people who said no and meant it, and Kelly's whole point is that this casting is wrong. The Amish are not opponents of the technium; they are, in his account, unusually deliberate participants in it, testing each new tool not against a blanket ideological rule but against its specific likely effect on family cohesion and congregational life. A community might accept a single shared telephone at the end of a farm lane, used for calling a doctor or a supplier, while refusing a telephone in every kitchen, on the judgment that the first strengthens the community's boundary and the second dissolves it. A community might adopt battery-powered tools for trade with the outside world while refusing a television, on the judgment that a power tool extends a farmer's hands while a television colonises a family's evenings. This is not rejection. It is curation, conducted with a rigour that most non-Amish households, scrolling reflexively through whatever a screen serves up, would do well to envy. The Amish are, on this reading, the only technology reviewers in America whose reviews take a decade to complete and are then actually binding. Kelly's line, "we are the reproductive organs of technology," sounds provocative out of context, almost sinister, but set beside the Amish example it reads differently: humans are not passive vessels the technium moves through, but active, discriminating participants, capable, at least in principle, of an Amish-level deliberateness about which extensions of mind and body they choose to let inside the door.

Kelly does not pretend the ledger is uncomplicated, whatever the caricature of him as an uncritical booster suggests. He offers his own rough estimate, and it is exactly that, a rough estimate rather than a formula, and a range wide enough to accommodate most of the disagreement about it, that something on the order of sixty to eighty percent of what the technium produces improves the human condition on balance. It is a strikingly modest number for a man sometimes filed under techno-optimist. Sixty to eighty percent leaves a substantial remainder, twenty to forty percent of everything the technium produces, that does not improve the human condition on balance: species driven extinct by habitat conversion, ways of life dissolved by economic disruption, forms of attention and community lost to devices that promised to enhance them and instead colonised them. Kelly's accounting comes out on the side of hope, but it is hope arrived at by facing a mixed ledger and doing the arithmetic, not hope asserted as a mood.

The case against Kelly is serious, and he has heard every word of it. It comes in two parts. The first is the charge of technological determinism. If the technium wants what it wants, if its large directions are inevitable, then the company profiting from ubiquitous tracking is not an agent making choices but a midwife assisting a birth that was always going to happen, and inevitability-talk becomes a solvent for responsibility, converting somebody's business plan into a law of nature. Language of this kind flatters exactly the people who need flattering least, and it is no accident that the people fondest of quoting Kelly on inevitability tend to be the ones whose quarterly plans it absolves. The second charge aims at the concept itself: that the technium is mysticism dressed as analysis. To say technology grows more complex because it is exotropic sits uncomfortably close to the physician in Molière who explains that opium causes sleep because of its dormitive virtue; the coinage renames the pattern without explaining it. The biological analogy, on this reading, smuggles in the conclusion: a plant wants light because natural selection built phototropism into its lineage, while the technium has no genome, no reproduction, and no selection mechanism apart from the very human choices Kelly's framing pushes offstage. These are strong objections, and Kelly's defence, scattered through the book rather than assembled in one place, is subtler than his critics usually allow. He distinguishes the general form of a technology, which he does hold to be close to inevitable, from its particular character, which he insists is chosen: some kind of telephone was coming once wires and electricity existed, but nothing dictated the advertisement-funded pocket version we actually received, and the gap between forms is exactly where politics, regulation, and taste do their work. He also went looking for his strongest critic in person, devoting a substantial section of the book to Theodore Kaczynski's manifesto, conceding that its analysis of technology as a self-perpetuating, interlocking system was largely correct while rejecting its conclusion, on the ground that the system so described has also been the greatest expander of human choice in history. Whether any of this fully answers the dormitive-virtue objection is less certain. It may be that the technium works best not as a theory that explains but as an instrument that reorients, forcing the reader to see the sum of human invention as one connected system rather than a heap of gadgets, which is a real service even if the organism talk is finally a metaphor.

The Inevitable, Kelly's 2016 follow-up, narrows the aperture from the whole technium across all of history to the coming several decades, and it does something formally clever: rather than forecasting nouns, specific devices or products that will exist, which is a game every futurist eventually loses to the phone in every 1958 magazine's blind spot, Kelly forecasts verbs, twelve present-tense participles describing directions of change he expects to intensify regardless of which particular gadgets end up embodying them: Becoming, Cognifying, Flowing, Screening, Accessing, Sharing, Filtering, Remixing, Interacting, Tracking, Questioning, and Beginning. The wisdom of this choice becomes clearer with every year that passes; a forecast of "everyone will carry a personal supercomputer" ages better than a forecast of "everyone will own a specific model of flip phone," because the first names a direction and the second bets on a shape.

Two of these twelve forecasts have aged into the mid-2020s with particular visibility, one largely vindicated and one far more contested than Kelly's original framing suggested. Cognifying, Kelly's word for the embedding of cheap, ambient intelligence into ordinary objects and processes, arriving the way electrification once did, house by house and socket by socket, reads today less like speculation and more like description. Kelly imagined this as a workaday utility rather than a single dramatic arrival, closer to electrification than to the sudden appearance of a science-fiction superintelligence, and that is very much the shape the last several years have taken: intelligence folded into search boxes, into email drafts, into customer service chats, into the coding tools programmers now use without a second thought, arriving piecemeal and mostly boring rather than as a single cinematic threshold crossed on a specific Tuesday. Accessing, Kelly's second big bet, the claim that ownership would increasingly give way to rented or streamed access, has proven true for media and, in more contested and consequential fashion, is now being tested for something Kelly could not quite have anticipated in 2016: a rented relationship with intelligence itself, in which millions of people and companies do not own a model but subscribe to the use of one, a shift with implications for dependency and control that Kelly's original chapter, focused mainly on music streaming and car-sharing, did not fully anticipate.

Kelly's own coinage for what these subscription intelligences have become, in his more recent writing, is disarmingly homely: universal personal interns. It is a very Kelly move, reaching for the domestic and unthreatening noun rather than the grand one, an intern rather than an oracle, something eager, useful, occasionally wrong, and decidedly junior rather than something to be feared or worshipped. He has also, in recent commentary, floated an arresting piece of intellectual humility: that people looking back from several decades hence may conclude there was no real artificial intelligence in the mid-2020s at all, only its clumsy childhood, a reframing that costs him nothing and buys considerable protection against the fate of every futurist who ever declared a threshold crossed a year or two too early.

Two further concepts from Kelly's broader work do real philosophical labour. The first is protopia, coined in a 2011 blog essay of the same name and threaded through both books: a state of continuous, marginal, unglamorous betterment, "better today than yesterday, although it might be only a little better," standing in deliberate contrast to both utopia, a static, finished perfection that Kelly regards as not merely unattainable but actively undesirable, since a finished world would be a dead one, and dystopia, whose only real function, in Kelly's view, is to induce despair and thereby excuse inaction. Protopia is a harder sell than either, because it offers no dramatic ending, no triumphant arrival, no cathartic collapse either, just the daily grind of things being, on the whole, slightly better than they were, with enough exceptions and enough new problems generated along the way to keep anyone from getting smug about it. The second concept is Questioning, one of the twelve verbs, and arguably the sleeper claim of the whole book: that the internet's central historical achievement has not been better answers so much as radically cheaper questions, and that a civilisation able to pose a hundred questions for every one an earlier civilisation could afford to ask, because asking used to require a trip to a library, a letter to an expert, a research budget, will simply out-innovate a civilisation still paying the old price per question. It is a claim about the economics of curiosity, and a better one than it first appears.

The Inevitable closes with the twelfth force, Beginning, and the sentence Kelly reaches for there, "we are not late, we are early," rewards a slower reading than an upbeat final line usually gets. It is a direct rebuttal to a specific and very common anxiety, one that Kelly returns to and develops at greater length elsewhere, that the technologies of any given present moment, however impressive they seem while you are living through them, will look primitive and embarrassing thirty years hence, the way a 1990s mobile phone the size of a brick looks primitive today, having been, at the time, the most desirable object in any room it entered, and that this is not a reason for despair but the entire point. At every stage of the internet's short history the insiders of the moment believed the frontier had just closed, and at every stage they were wrong. If you are early, there is more still to build, and more still to build is, on Kelly's accounting, the best news a species that likes making things could possibly receive.

V. The Arithmetic of Optimism

Kelly's case for optimism, developed most fully in a 2021 TED talk delivered at TEDMonterey under the title "The Future Will Be Shaped by Optimists" and later published as an essay that became, by Warp News's own account, the most-read piece the outlet has ever run, rests on a single, almost absurdly simple piece of arithmetic, and its simplicity is the whole point. Imagine, Kelly says, a civilisation that in any given year creates only one percent more than it destroys. Not ten percent. Not a heroic, revolutionary leap. One percent, an amount so small it would be difficult to detect in the noise of any single year, easily swamped by a war, a famine, a bad harvest, a stupid decision made by someone in power. But compound that single percentage point, year upon year, across centuries, and you get something close to the entire visible arc of human civilisation: the accumulated cooperation, technology, and material prosperity that separates a hunter-gatherer band from a modern city. Civilisation, in Kelly's phrase, "amplifies and accumulates cooperation between strangers," and the amplification does not require heroic annual leaps forward. It requires only that the ledger tip, ever so slightly and ever so reliably, toward creation rather than destruction, and that the tipping be sustained over a very long time, which is exactly the kind of slow, compounding process human intuition is notoriously bad at feeling in its gut, however easily it can be verified with a calculator.

Kelly draws three grounds for this optimism, and none of them require pretending the past was gentle. The first is simply the historical record, read plainly and over a long enough span: life expectancy, literacy, the proportion of humanity living in extreme poverty, measures of violent death, all of these have moved, unevenly and with real reversals along the way, in the direction most people would call improvement, when measured across centuries rather than news cycles. The second ground is that the capacity to solve problems itself expands with each generation, so that a problem which would have been intractable to solve in 1900, contained a disease, a famine averted by better agronomy, is not merely solved once but solved with tools that then remain available for the next problem, compounding in the manner of Kelly's own percentage point. The third ground, and the most quietly radical, is a reframing rather than a fact: problems, in Kelly's view, are better understood as raw material than as obstacles, opportunities that have not yet been correctly shaped, a habit of mind that turns a discouraging list of the world's difficulties into something closer to an inventory of unfinished business.

Compounding is a treacherous thing to reason about informally, and the whole force of Kelly's argument depends on getting it right. A single percentage point of net creation sustained across a decade barely moves the needle, which is exactly why any one generation, looking only at its own lifetime, can be forgiven for concluding that nothing much is really changing, that the world of their middle age looks depressingly similar to the world of their youth. Sustained across a century, one percent compounding produces a civilisation roughly two and three quarter times more capable than it started; sustained across a millennium, the multiple passes twenty thousand. This is simply how compound growth behaves, whether the thing compounding is a bank balance, a bacterial colony, or, in Kelly's argument, the net creative output of a species.

Kelly goes further than most writers are willing to go and states plainly that, given this arithmetic, optimism is close to a moral duty. The reasoning is not sentimental. Every great and difficult undertaking in human history, he argues, from a cathedral that would not be finished within the lifetime of the mason laying its first stone to a vaccine programme aimed at eradicating a disease across an entire continent, required a supply of optimism sufficient to attempt it before there was any proof the attempt would succeed. Pessimism, whatever its intellectual credentials, has never laid a single stone of a cathedral. It can be right about a great many things, and often is, but it has never once been the thing that got a difficult, multi-generational project off the ground, because a person who is confident the project will fail does not show up to lay the first stone. He is available afterwards, to explain why it fell.

The obstacle Kelly is most worried about is not despair in its dramatic form, the kind that makes headlines and generates think pieces, but something quieter and, in his diagnosis, far more corrosive: future-blindness, the modern inability to picture, with any specificity or appetite, a future one would actually want to live in. Ask most people to describe a good future for the year 2076 and they will struggle, reaching either for warmed-over science fiction or for a vague absence of the current decade's specific anxieties, which is not the same as a vision. Kelly's claim, and it is an uncomfortable one to sit with, is that a civilisation which cannot picture a desirable future is at real risk of failing to build one, not because building it is impossible but because nobody is rowing toward a destination nobody can see. Fear at least points somewhere, even if the direction is away. Future-blindness does not point anywhere at all.

Notice how thoroughly this diagnosis fits the popular culture that surrounds it. Ask a room of adults to name a film or novel depicting a future they would actually choose to live in, as opposed to one they would spend the running time trying to escape, and the room tends to go quiet for an uncomfortably long time. The dominant imaginative furniture of the last several decades, from dystopian teenage fiction to prestige television's crumbling near-futures, supplies vivid, detailed, well-lit pictures of collapse and almost nothing in the way of vivid, detailed, well-lit pictures of a future worth wanting. This is not merely an aesthetic complaint. Kelly's point is closer to an engineering one: an organisation, a family, or a civilisation that spends its imaginative energy rehearsing catastrophe in high definition while leaving its pictures of success blurry and vague is, in a quite practical sense, rehearsing for the wrong outcome, the way an athlete who only ever visualises losing is unlikely to perform well regardless of talent. Optimism, in this reading, is not a personality trait laid over the facts. It is closer to a discipline of attention, a decision about which future to render in detail, and Kelly's insistence that this decision carries moral weight follows directly from noticing how much of ordinary cultural life quietly makes the opposite one by default.

Read his blog essays and his books together and Kelly does not come across as a man who has decided in advance that everything will be fine and is now hunting for evidence. He comes across as a man who has spent four decades interviewing his own optimism, a witness he clearly finds credible but keeps recalling to the stand anyway: he did the sum, found it came out slightly positive, and decided that a civilisation which has run a slight surplus for ten thousand years owes itself the benefit of the doubt on the ten thousand and first.

VI. Trust Enthusiasm Proportional to Specificity

If Kelly supplies the altitude, sweeping across the whole technium and the whole arc of civilisation, Kelly Weinersmith and Zach Weinersmith supply the ballast. Soonish, published in 2017, is a far more modest book in scope and a far more useful one as a corrective, and its authors bring an unusual combined competence to the task. Kelly Weinersmith is a parasitologist and behavioural ecologist, trained to be suspicious of any claim that has not survived a peer reviewer with a grudge. Zach Weinersmith is the cartoonist behind the webcomic Saturday Morning Breakfast Cereal, a strip whose humour runs on deflation: the grand claim marched, politely and without mercy, into contact with what it actually rests on. Together, over nearly three years of interviewing working researchers, they produced a hybrid book, prose chapters interleaved with Zach's line drawings, that reads less like a manifesto and more like a series of extremely well-informed pub conversations conducted with people who actually build the things being discussed. The authors give every impression that they would be pleased to discover their own premise was wrong, and would report the discovery in a footnote, with a diagram.

The book surveys ten fields: cheap access to space, asteroid mining, fusion power, programmable matter, robotic construction, augmented reality, synthetic biology, precision medicine, bioprinting, and brain-computer interfaces. The subtitle promises these ten technologies will "improve and/or ruin everything," and the slash is not a marketing flourish but the book's entire method compressed into two characters. Every chapter weighs plausible benefit against plausible harm with equal seriousness, refusing either the breathless certainty of the futurist keynote or the reflexive gloom of the technology critic. Progress, in the Weinersmiths' telling, is neither a headwind to be resented nor a tailwind to be assumed. It is contingent, effortful, and expensive, achieved by specific people solving specific problems, none of it guaranteed by any law of history.

The chapter on cheap access to space is the one field in the book that has since moved fastest, and rereading it now is an unusually direct test of the Weinersmiths' own method. Written before reusable orbital rockets had become routine, the chapter treats the prospect of a booster landing itself upright and flying again within weeks as an open engineering question, one more idea competing against cost curves that had barely moved since the 1970s. It is exactly the kind of claim the book's discipline would have flagged as low on specificity at the time, interesting in principle but unproven in the metal. What has happened since, repeated landings becoming so routine they no longer make the news, a launch cadence measured in days rather than months, costs per kilogram to orbit falling by something like an order of magnitude, is a rare case of a technology arriving faster and more completely than even a careful, sceptical treatment assumed likely. The Weinersmiths do not get to claim credit for predicting it, and to their credit they did not try to; the chapter's honesty about uncertainty is precisely what makes its later, partial vindication interesting rather than merely lucky.

Asteroid mining, treated in the following chapter with the same rigour, has aged rather differently, and the contrast is instructive. The economic logic is not in question: a single modest metallic asteroid could contain more platinum-group metal than has ever been mined on Earth, and the orbital mechanics of reaching some near-Earth objects require less energy than reaching the lunar surface. What the Weinersmiths correctly identified as the binding constraint was never the geology but the cost of getting there and back cheaply enough to matter, precisely the cost curve the space-launch chapter was busy watching fall. Several well-funded asteroid mining ventures from the 2010s folded or pivoted entirely before extracting a single gram of anything, undone less by bad astronomy than by the ordinary difficulty of financing a venture whose revenue lies a decade or more in the future. The platinum is still there. It is a tidy illustration of the book's warning that a technically sound idea can still be strangled by an unsound business model, a distinction that pure enthusiasm, unmoored from specificity, tends to blur. Held together, the ten chapters do not read as ten separate bets so much as ten applications of one consistent method, and the method is the actual subject of the book far more than any single technology is.

The chapter on fusion power is the book's emotional and argumentative centre. Fusion has been famous, for roughly seventy years now, for being thirty years away, a joke so old within the physics community that researchers make it about themselves before anyone else can. The Weinersmiths do not treat this as evidence the physicists are charlatans or fools. They treat it, correctly, as a case study in how badly a hard engineering problem can be underestimated by people extrapolating too confidently from early success. Confining a plasma hot enough to fuse hydrogen isotopes, hotter than the core of the sun, inside a magnetic bottle stable enough to hold it there for useful lengths of time turned out to be a problem of a completely different order than the 1950s pioneers, flush with the recent, world-altering success of fission, assumed it would be. The book's sympathy for the physicists, rather than mockery of the perpetually receding deadline, is one of its most valuable contributions, a reminder that "thirty years away" is sometimes a symptom of institutional overpromising and sometimes, more charitably and more often than critics assume, a symptom of a problem being harder than anyone outside the field can easily appreciate.

The book's discipline, and its most quotable line, doubles as its title's own justification: trust enthusiasm proportional to specificity. A researcher who tells you, in broad strokes, that a technology will change everything within a decade is offering you very little to actually evaluate. A researcher who tells you precisely which physical bottleneck remains unsolved, what has been demonstrated in a laboratory versus what remains theoretical, and what specific engineering milestone would need to be cleared before the technology leaves the lab, is offering you something you can actually check, return to, and hold them accountable for later. The Weinersmiths apply this test relentlessly across all ten chapters, and it produces a book that manages to be excited about, say, programmable matter, tiny robotic units capable of reconfiguring themselves into different physical shapes and functions on demand, something that sounds like pure science fiction, while being scrupulously clear about exactly which parts of that vision rest on real laboratory demonstrations already achieved and which parts remain, for now, extrapolation.

One image from the book does more argumentative work in a single panel than several paragraphs of prose could manage: a joke about a robot capable of building an entire house, framing it, wiring it, roofing it, but incapable of finding the house's own front door. It sounds like a throwaway gag, and Zach Weinersmith's comic instincts make sure it lands as one, but it is quietly making a serious point about the difference between narrow competence and the general capability popular imagination tends to assume follows automatically from any sufficiently impressive demonstration. A robot arm that can weld a car chassis with superhuman precision has not thereby become a robot that can navigate a cluttered garage, and a language model that can write fluent prose about quantum mechanics has not thereby become a system that reliably understands what it is saying. The gap between a narrow, spectacular skill and the broad, unremarkable competence humans deploy without thinking about it, walking into a room and finding the light switch, is one of the most persistently underestimated gaps in any conversation about technological progress, and the Weinersmiths' joke names it more economically than most academic papers manage.

Beneath all ten chapters runs a darker-sounding lesson, that the history of technology is in large part a history of delay, and the Weinersmiths plainly do not intend it as pessimism. It is a statement about the shape of effort rather than a verdict on its worth. Every technology that eventually arrives, arrives later than its earliest champions promised, dragged through longer development cycles, more failed prototypes, and more institutional friction than the initial enthusiasm ever accounted for. Nothing has ever arrived early. This is simply what building difficult things looks like from the inside, as opposed to how it looks from a magazine cover, and a reader who has internalised the lesson is inoculated against both the con artist promising imminent transformation and the cynic declaring, the moment a promised deadline slips, that the whole project was fraudulent from the start.

Positioned against Hall, Kelly, and Bush, Soonish occupies a useful middle register. It shares Hall's core observation that physical, hands-on-hardware innovation slowed markedly across the second half of the twentieth century, that the sheer velocity of change in atoms, as opposed to bits, has not kept pace with either its own earlier trajectory or with the runaway pace of computing. But where Hall reads that slowdown mainly as stagnation, a curve broken by identifiable and in his view largely avoidable human failure, the Weinersmiths read it more as slow, ongoing, and largely honest effort: not a rocket ordered to throttle back, in Hall's phrase, but a great many rockets still being painstakingly assembled, later than their engineers hoped, by people working as hard as they know how against problems that turned out to be far harder than 1958's magazine covers ever let on. And against the more confident tradition of technological forecasting exemplified by writers who assign specific calendar years to specific future capabilities, a tradition whose natural home is with the philosophers of progress rather than the students of technology, Soonish functions as a standing, good-humoured corrective: not a rejection of optimism, but an insistence that optimism, like everything else in the book, be held to the same test as everything else the Weinersmiths examine. Trust it in proportion to its specificity.

VII. The Desk That Thinks by Association

If Hall diagnoses a stall and Kelly describes a system with tendencies of its own, Vannevar Bush supplies something closer to a blueprint, drawn up decades before the technology existed to build it, for the one domain in which the acceleration nobody quite predicted actually happened. Bush wrote "As We May Think" from an unusually authoritative vantage point. He was a dean at MIT and, during the war, director of the Office of Scientific Research and Development, the position from which he coordinated something on the order of six thousand American scientists working on projects that included radar, the proximity fuze, credited with materially shortening the war in the Pacific, and, at one remove through his oversight of the broader research apparatus, the Manhattan Project itself. He published the essay in The Atlantic in July 1945, an abridged version following in Life magazine that September, which means Bush wrote it in the weeks immediately before Hiroshima and Nagasaki, a fact that gives its closing paragraphs a weight the essay's more famous technical passages can sometimes obscure.

The essay did not spring from nowhere. Bush had rehearsed its central worry six years earlier, in a shorter and less remembered 1939 piece called "Mechanization and the Record," and had described the memex idea, under a different name, in private correspondence before the war intervened and put him in charge of coordinating the American scientific effort instead of writing about it. The interruption mattered. He returned to his prewar worry about scientific information with the practical experience of a man who had just spent years managing thousands of scientists and had seen, firsthand, both how much a coordinated research effort could achieve and how badly even that effort strained under the sheer volume of what it was producing.

Bush's starting diagnosis is not really about machines at all. It is about the sheer, drowning volume of scientific publication, which by the mid-1940s had grown so large that no individual researcher could plausibly keep up with even a fraction of the work relevant to their own narrow specialty. He worried, with the specificity of a man who had just spent years watching thousands of researchers try and fail to stay abreast of each other's work, that genuine discoveries could be made and then effectively lost, buried in a journal nobody in the right adjacent field happened to read, for want of the correct index or the correct reader crossing paths with the correct paper at the correct time. His own worked example, Gregor Mendel's laws of heredity, formulated in a monastery garden in the 1860s and then essentially ignored by the scientific mainstream for thirty-five years until independently rediscovered around 1900, a stretch during which the machinery for circulating scientific results performed exactly as designed, was for Bush not a historical curiosity but a warning: if a discovery as fundamental as the basic mechanism of inheritance could vanish for a generation in a scientific culture producing a comparative trickle of publications, what was likely to happen as the trickle became the flood he could already see building around him.

His proposed answer was not a single dramatic invention but a combination of instruments already sitting, mostly underused, in laboratories and offices around him: microfilm, which could compress a small library into a container the size of a matchbox; the photocell and the thermionic vacuum tube, which made rapid electronic processing and switching possible; mechanical systems for sorting punched cards, the era's closest approximation of a database; facsimile transmission, the ancestor of the fax machine, for moving documents instantly across distance; and dry photography, a then-recent innovation allowing an image to be developed without the wet chemical processes that had made photographic reproduction slow and messy. None of these was new by 1945. Bush's genuine originality lay in seeing that their combination, properly assembled, would let the human mind master its own accumulating inheritance rather than drown beneath it.

The device Bush imagined to house this combination he called the memex, a name he never quite explains beyond its evident debt to "memory" and "index," and its precision is what makes it remarkable rather than merely visionary in the vague sense that word usually implies. The memex was, in Bush's description, a desk, an actual piece of furniture, with translucent screens set into its surface on which pages of text and image could be projected for reading. Its interior held a microfilm library sufficient to store a lifetime's worth of books, correspondence, photographs, and personal notes, all instantly retrievable through a keyboard and a set of levers, without the delay of walking to a shelf or waiting for a librarian. Bush specifies all of this, levers included, with the unblinking calm of a man describing a filing cabinet, which, as far as he was concerned, is more or less what he was doing. This much, on its own, would have been a considerable achievement, essentially a personal, instantly searchable library compressed onto a desktop, decades before anyone had a practical way to build one.

But Bush's decisive insight lay one layer deeper than storage and retrieval, in how the memex's owner would be able to navigate what was stored. Existing libraries, he observed, organise knowledge by classification, a tree-like hierarchy of subjects and subcategories that requires a reader to know, in advance, which branch of the tree a piece of information lives on, which is the one thing a reader looking for it is unlikely to know. Bush thought this was a poor match for how a mind actually works. "The human mind does not work that way," he wrote, in the essay's most quoted and most quietly radical sentence. "It operates by association. With one item in its grasp, it snaps instantly to the next that is suggested by the association of thoughts, in accordance with some intricate web of trails carried by the cells of the brain." The memex, accordingly, would let its owner build what Bush called an associative trail: a chain of documents linked not by their formal classification but by whatever connection of thought had occurred to the reader while moving between them, a trail that could be given a name, revisited later, and even passed along to a friend, who could splice it into a trail of their own.

Bush illustrates this with a worked example precise enough to function almost as a piece of software documentation written decades before there was software to document. He imagines the owner of a memex wondering why the short bow used by Turkish archers during the Crusades had proven superior in some respects to the longer English longbow, and following the question outward: from an encyclopedia entry on the Crusades, sideways to a textbook on the elasticity of materials, back again to a historical account, the reader building, link by link, a trail that connects military history to materials science in a way no library's shelving system would ever have suggested, because no cataloguer organising books by subject would have thought to shelve the physics of bow construction next to the history of medieval warfare. Read today, this passage does not describe an obscure historical curiosity so much as it describes, with startling exactness, the experience of falling down a chain of hyperlinks on a modern encyclopedia site, following a footnote from one article into an entirely different discipline and back again, an hour vanished, six browser tabs open, none of them things you set out that evening to learn about. Bush had no working prototype and no electronic computer sophisticated enough to build one. What he had was the correct description of an experience the technology to deliver it would not exist for another forty-five years.

Bush closes the essay on a considerably graver note than its engineering passages might lead a reader to expect, and the timing, those final weeks before Hiroshima, makes the gravity impossible to read as accidental. The same science, he wrote, that had just enabled humanity "to throw masses of people against one another with cruel weapons" might yet, if pursued with equal seriousness in peace, "allow him truly to encompass the great record and to grow in the wisdom of race experience." He warned that abandoning the effort now, in the flush of a victory purchased at such cost, would be "a singularly unfortunate stage at which to terminate the process." It is a remarkable thing for a man who had just spent years directing the development of weapons of unprecedented lethality to write: that the very scientific apparatus capable of producing the bomb was equally capable, redirected, of producing something that might help humanity think its way out of the kinds of catastrophic misjudgement that produce wars in the first place. Bush was not naive about the danger. He was making a bet, stated as plainly as the essay's technical passages, that the same tools could cut in either direction, and that the answer to a dangerous capability was not to abandon capability itself but to aim a comparable portion of it at wisdom.

The essay's afterlife is, by the standards of technical writing, close to unprecedented. Douglas Engelbart, a young engineer who encountered "As We May Think" shortly after the war while stationed in the Philippines as a radar technician, has described the memex as a formative influence he carried through decades of subsequent work at the Stanford Research Institute, work that eventually produced the computer mouse, practical word processing, video conferencing, and a working, if primitive, implementation of the hyperlink itself, demonstrated to a stunned audience of a thousand computer professionals in San Francisco in December 1968 in a ninety-minute live presentation now remembered simply as "the mother of all demos," a session that packed into a single hour and a half more of the technology most people would spend the rest of the century catching up to than almost any other public demonstration in the history of computing. Ted Nelson's later coinage of the word hypertext, introduced in 1965, and his ambitious, only partially realised Xanadu project, an attempt to build a universal, non-proprietary system of linked documents complete with a mechanism for compensating original authors whenever their words were quoted elsewhere, descends from the same lineage, an attempt to build, in software rather than microfilm, something close to Bush's associative trail.

The line from Bush's memex to the ordinary act of reading a Wikipedia article today, clicking from one blue underlined word to the next, runs very nearly straight. Bush even anticipated the encyclopedia itself would need to change shape, writing that the future would bring "wholly new forms of encyclopedias" that arrived pre-linked, their associative trails already built by an editor and ready to be extended by any reader's own trails threaded through them, which is a strikingly exact description of what a hyperlinked encyclopedia edited by millions of volunteers turned out to look like, decades before anyone had a network capable of delivering one.It is also, not incidentally, a rebuttal in miniature to the problem of the flying car and the phone. Bush got the domain of information exactly right, decades early, while the domain of energy, the one Hall spends a whole book grieving, stalled out almost precisely when Bush's own domain was quietly beginning its ascent.

"As We May Think" was not the only document Bush published in the summer of 1945, and the other one, less loved and more consequential, is where the strongest objection to him lives. In November 1944 President Roosevelt had asked Bush how the wartime research machine might be turned to peacetime ends; the answer, delivered to President Truman in July 1945 under the title Science: The Endless Frontier, argued that basic research, performed in Bush's words "without thought of practical ends," is the pacemaker of technological progress, that new products and new industries are founded on new scientific knowledge, and that the federal government should therefore fund university science permanently and at scale, a recommendation that, after five years of political wrangling over who would control the money (an instructive wait for a man who had got radar from blackboard to battlefield in rather less), produced the National Science Foundation in 1950. It is difficult to overstate how completely this report set the terms of postwar science policy in the United States. It is equally difficult, by now, to find a historian of science who accepts its central model. The picture embedded in the report, a pipeline running one way from basic research through applied research and development to products, came to be called the linear model of innovation, and half a century of scholarship has largely dismantled it. The traffic between science and technology runs in both directions and frequently backwards: the steam engine preceded and provoked thermodynamics rather than the reverse, a fact compressed into the quip, usually attributed to the biochemist L. J. Henderson, that science owes more to the steam engine than the steam engine owes to science. Donald Stokes argued in Pasteur's Quadrant that the most fertile research has usually been fundamental and use-inspired at once, a category the linear model cannot even represent. And the most awkward witness against the report is its author's own war. Radar, the proximity fuze, and mass-produced penicillin did not flow down a pipeline from disinterested theory; they were mission-driven engineering programmes that reached back into science whenever they needed it. Bush the administrator, in other words, ran his shop on a model that Bush the theorist then wrote out of the official story.

Why the wrong model survived so long is a question with an institutional answer rather than an intellectual one. The linear model was, whatever its descriptive faults, a superb funding argument: it promised politicians that money spent on undirected research would return as industry and medicine without anyone having to specify how, and it promised scientists autonomy as a condition of the bargain rather than a concession wrung from it. A model can be wrong and useful at the same time, and this one was both for decades. The objection to The Endless Frontier barely grazes "As We May Think," and seeing why sharpens the case for the second document. The report is a theory of where technology comes from, and it is wrong in the way most single-arrow theories are wrong. The essay is not a theory at all. It is a mechanism, specified part by part, microfilm reel by photocell, with the traffic between mind and machine running explicitly in both directions along the associative trail. Bush forecasting the institutions of science overrated the tidiness of a process he had spent the war improvising. Bush describing a machine he could almost build was on his own ground, and the difference in accuracy between the two documents, published weeks apart by the same man, is itself a small argument for the Weinersmiths' rule: trust enthusiasm in proportion to its specificity.

VIII. Currents, Not a Current

Set Hall's stalled energy curve beside the computing industry's decades-long adherence to Moore's Law and the central pattern comes into focus. Gordon Moore, then at Fairchild Semiconductor and later a founder of Intel, observed in 1965 that the number of transistors that could be fit onto a silicon chip of a given size was doubling on a schedule he first put at every year and later revised to roughly every two. Moore's Law held, with only minor wobbles, from the mid-1960s clear through to the 2010s, an unbroken run of compounding improvement in one narrow domain of manufacturing precision that happened to coincide almost exactly with the decades Hall spends his book mourning as stagnant. The two trends sat side by side in the same economy, the same regulatory environment, often the same government funding agencies, and yet one compounded at a rate that transformed the world beyond recognition while the other went essentially flat. This is the clearest possible evidence against any theory of technology as a single undifferentiated force accelerating or decelerating in unison. Whatever combination of physics, economics, and institutional culture allowed semiconductor fabrication to compound so relentlessly for fifty years did not extend its protection to nuclear reactor licensing, sitting in the very same decades, subject to many of the very same national governments. No transistor has ever been asked to file an environmental impact statement. The explanation has to be local rather than global: something specific to how chips were regulated, financed, and culturally received, as against something specific to how reactors were regulated, financed, and culturally received, rather than any single dial governing the rate of technological change as such.

The most productive tension in the whole set sits between Hall and Kelly, who are not describing the same kind of process at all. Kelly's technium has, in his own account, tendencies, a statistical pull toward efficiency, complexity, and the rest, operating across such a vast number of individual human choices that it starts to resemble something close to a law of nature, the way the behaviour of a gas emerges from the chaotic motion of individual molecules too numerous to track one by one. Read uncharitably, this can sound like a claim that technological direction is all but inevitable, a current so broad that no single decision, however consequential, can meaningfully redirect it, which is a comforting thought in one sense and a faintly fatalistic one in another. Hall's regulatory ratchet is a direct challenge to exactly this reading. His entire book is an argument that one domain of the technium, energy, did not merely slow down within some larger inevitable current; it was stopped, cold, by specific, nameable human decisions: a regulatory response to Three Mile Island disproportionate to the measured risk, a funding collapse following the Fleischmann-Pons debacle, a cultural mood that made large technological projects suspect rather than admirable. These are not statistical pulls emerging from billions of untraceable choices. They are, in Hall's telling, traceable to specific hearings, specific votes, and specific magazine covers, all of which had names attached to them.

The honest resolution, and it is one Kelly's own sixty-to-eighty-percent ledger already half concedes, is that both things are true at different scales, and the disagreement is less a contradiction than a difference of magnification. Zoom out far enough, across centuries and across the whole technium at once, and something like Kelly's statistical tendencies do seem to hold: technology has, on the whole, trended toward greater complexity, efficiency, and reach for as long as there has been technology to track, and no single bad decade has ever permanently reversed that broader trend. Zoom in to a specific domain across a specific fifty-year window, energy in the advanced economies between 1970 and roughly 2020, and Hall's more granular, more contingent, more human-scale account of a ratchet turning one way is exactly right, and Kelly's larger tendencies offer no comfort at all to the engineer who wanted to build a reactor in 1985 and could not get one licensed. A civilisation can be exotropic, in Kelly's coinage, in its broadest, longest-run tendency while still, within any given fifty-year stretch, stalling one entire current of itself through decisions that were neither inevitable nor irreversible, which is precisely what the recent, halting, still-unproven nuclear revival appears to be demonstrating in real time.

What emerges from holding all of these works in view at once is not a single unified theory of technology, and it would be a betrayal of the writers' own discipline to force one onto them. What emerges instead is a more useful and less comfortable picture: technology is not a force that simply happens to a civilisation from outside, the way weather happens to a harvest, nor is it a force so thoroughly internal to human nature that its direction is fixed and merely awaits discovery. It is closer to a garden with several different beds, tended with wildly uneven care. Some beds, information processing above all, were tended so well, so early, and by such a fortunate combination of wartime urgency and postwar peacetime investment that they raced decades ahead of anyone's 1945 expectations, Bush's memex realised and then wildly exceeded within the span of a single working life. Other beds, energy and physical infrastructure chief among them, were planted with equal early promise and then left to run to weeds by a gardener who, spooked by one bad harvest, decided the whole bed was cursed and stopped watering it, exactly the fate Hall spends a book documenting and the current nuclear revival is now, tentatively, beginning to reverse.

IX. What Patience Teaches

What all five writers share, underneath their real disagreements, is a single discipline: they treat technology as the thing to be studied rather than the occasion for a verdict, and they do the work of finding out rather than the easier work of deciding in advance how they feel and then hunting for confirming anecdotes. Hall did not arrive at his account of the regulatory ratchet by disliking regulators on principle and reaching for evidence afterward; he arrived at it by tracing, reactor order by reactor order, exactly what happened to nuclear licensing after 1979 and asking what, specifically, had changed about the underlying physical risk to justify it. The other four worked the same way. Patience, in this narrow but important sense, is not a virtue opposed to optimism or pessimism. It is the precondition for either one being worth listening to.

The discipline these writers model has a specific, practical shape, and it carries well beyond the books discussed here. It means learning to distinguish a genuine, structural current, the kind Bush identified in 1945 and that then actually arrived, from a mood or a fad dressed up in the vocabulary of inevitability, the kind that fills a magazine cover in 1958 and then quietly vanishes from the historical record within a generation. The tell, across every case examined here, is not enthusiasm, which both the genuine current and the fad generate in identical abundance, but mechanism: whether the person making the claim can explain, in the Weinersmiths' terms, specifically what remains to be solved and specifically why the solution is plausible, or whether the claim rests, when pressed, on nothing sturdier than a general sense that the future ought to look a certain way because the present has been trending in that direction. Bush could explain his memex down to the microfilm reel and the associative trail. The 1958 magazine illustrators, for all their confidence, generally could not explain how a suburban commuter without a pilot's licence was going to safely navigate three-dimensional airspace above a city full of other untrained commuters doing the same, any more than they could explain the collie, and the absence of that explanation was, in retrospect, the whole story.

There is a responsibility bound up in all this that none of the five writers shies away from, and it is perhaps the deepest thing they share, the thing that separates their patient, technically literate optimism from either uncomplicated celebration or uncomplicated dread. If technology's direction were entirely fixed, either by Kelly's tendencies operating as an inexorable law or by some equally deterministic story running the other way, in which every new capability inevitably degrades whatever it touches, there would be nothing left to argue about and nothing left to do except spectate, cheering or booing from a fixed seat as the show unfolds on its own schedule. What Hall's regulatory ratchet, and its current, halting reversal, actually demonstrates is that at least some of technology's direction remains open, actually open, to human choice: a stalled curve can, with sufficient will and sufficient reason, be coaxed back toward its old slope, even if the will in this particular case arrived through the accounts department of a data centre rather than through any change of heart. What Kelly's Amish demonstrate is that deliberate, discriminating adoption, testing each tool against its actual effect on the things a person cares about rather than swallowing or rejecting the whole technium wholesale, remains available to anyone willing to do the work the Amish do, which is more work by far than either uncritical adoption or blanket refusal. What the Weinersmiths demonstrate is that the correct response to an exciting claim is neither applause nor dismissal but a specific, answerable question. And what Bush demonstrated, writing at his desk in the exhausted, apprehensive summer of 1945, weeks before two Japanese cities would be destroyed by the same scientific apparatus he had helped direct, is that even a person who has seen a technology's very worst possible use at close range can still, in the same essay, in the same paragraph almost, choose to describe its best possible use with equal seriousness and equal care, and mean both halves of the sentence.

None of this amounts to a guarantee, and none of the five writers offers one, which is itself part of the discipline. Kelly's own arithmetic requires only a single percentage point of net creation over destruction, sustained, not a landslide, which is a far lower bar than triumphalism demands but a far higher one than despair is willing to credit. Hall's curve may yet fail to bend back; the fifteen reactors under firm commitment may become five by the time ground is actually broken, the Bulletin of the Atomic Scientists's caution proving the more accurate guide than the press releases. Kelly's protopia offers no finish line, only the daily, marginal work of tomorrow being slightly better than today, with no promise that the slight margin holds forever. But the case these five writers make, taken together and read with the patience they themselves modelled, is that the flying car and the phone were never really opposite outcomes of some single technological fortune, one a promise kept and the other a promise unexpectedly exceeded. They were two different currents, tended with two very different degrees of care, by people making choices that, however small and however invisible at the time, turned out, compounded over decades, to matter enormously. That is not a story that guarantees a happy ending. It is a story that insists, against both the easy cheer and the easy dread, that the ending has not been written yet, and that the writing of it remains, stubbornly and inconveniently, up to us.