The only channel that made the update pathway its object of design. Two movements, from Thales to the steam engine and from the transistor to the installed integral, and the question the frozen integral poses.

In November 2013 The Atlantic asked a panel of twelve experts, among them the economic historian Joel Mokyr, to name the fifty greatest breakthroughs since the wheel. I scored each of the fifty myself against the framework's three axes (integration, resonance, imagination), so what follows is my own coding of their list, not a finding of the panel's. More than four in five score on integration. Nearly three in five on resonance. Somewhat over a quarter on imagination. All of them score on at least one, which is weak: with three broad axes, near-universal coverage is close to what chance would give. The informative number is the subset scoring on all three, and it arrives at the end of the first movement. The innovations that top every ranking score on all three at once. Understanding where they came from requires understanding what science actually is.

What follows is, at one level, a history of how science and technology developed. At another level it is the story of how the information environment changed (what was available to be stored, transmitted and accumulated) and how each change rebuilt what ended up in individual minds. The archive grew. The transmission channels multiplied. Civilisation, though, lives only in the parameter sets of the people who carry it. What technology does to minds is the question underneath the history.

Movement One: From Thales to the Steam Engine

Science occupies a singular place among the channels. Every other one has some way of updating what a civilisation carries. Religion updates through revelation, law through precedent, education through the curriculum. All are slow, resistant to anything that threatens existing power, and without a systematic method for telling what is true from what is merely resonant. Science is the exception. It is how the multiple integral forces a civilisation to update against reality. Alone among the channels, it made the update pathway itself its primary object of design.

The Greek philosophical tradition was the first in the Western record to make the explicit revisability of natural explanation a cultural value. Thales proposed that the fundamental substrate of reality was water, not because the gods had said so but because he thought it the best available natural explanation of the evidence. The specifics were wrong. The move was revolutionary. Anaximander disagreed and said so, and within a generation a tradition of explicit, adversarial, public argument about natural explanation had been established on the Ionian coast. The Socratic method was the same tradition applied to ethics: argument, rather than revelation or authority.

Three civilisational paths accumulated scientific knowledge in parallel across the first millennium CE. Indian mathematics produced the decimal positional number system and the formal treatment of zero, the single most consequential mathematical invention in history, without which arithmetic, algebra, calculus, computing and modern science do not exist. Chinese technology was for most of this period the most practically advanced in the world: paper, printing, gunpowder, the compass. Needham's question, why China with the technology did not produce the Scientific Revolution, has an answer in Mokyr's framework. China lacked the institutional equivalent of the Republic of Letters. The examination system produced extraordinary conformity within the Confucian corpus; there was no competitive fragmentation of states to force heterodox inquiry, and no culture of openness and replication to convert settled technical knowledge into self-sustaining propositional compounding. The information was present. The open network for processing it was not. The Islamic Golden Age was the node that connected the Greek, Indian and Chinese contributions into a single accelerating tradition: al-Khwarizmi's algebra, ibn al-Haytham's optics, ibn Sina's medicine, al-Biruni's comparative anthropology. It also supplied the physical medium that made European accumulation possible. Chinese paper reached Baghdad in the decades after the Battle of Talas in 751 CE (the tidy story of captured papermakers teaching the craft is probably romanticised; that cheap paper reached the Abbasid world in those years is certain). Without cheap paper there is no translation movement and no Luther. The chain from the Scientific Revolution through the Enlightenment runs through a battle in Central Asia that most Western histories do not record.

Bacon's Novum Organum of 1620 was a cultural entrepreneurship event of the first order. Classical authority, however settled, was an obstacle to knowledge rather than a repository of it; true knowledge came from systematic observation and inductive inference. What Bacon added was the declaration that this was not one approach among others but the only one that produces knowledge at all. Newton's Principia was the exemplar: mathematical law extracted from observation, in a form that could be tested, extended and applied by any competent practitioner anywhere. The Royal Society, founded in 1660, was the institution that sustained the method. Peer review, replication, public disclosure. Nullius in verba: take nobody's word for it.

Mokyr's distinction between propositional and prescriptive knowledge is the account of what happened next. For most of history, knowledge of why and knowledge of how had been separate. The savant understood principles; the craftsman mastered technique. The Industrial Revolution was different because the two were brought into sustained positive feedback for the first time. Knowing why the steam engine worked made it improvable almost without limit. Watt did not invent the engine. He made it efficient enough to be useful, and his 1769 patent for the separate condenser was a prescriptive intervention in an existing propositional understanding. The result could drive a factory, pump a mine or propel a locomotive. This is the clearest case of science crossing into the world: propositional knowledge generating prescriptive capability, which generated new physical infrastructure, which generated new social structures, which generated new propositional questions, which generated Carnot, Clausius and the formal science of thermodynamics. The loop ran in both directions, and it ran for two centuries.

The railway extended the moment across geography. The telegraph applied it to information. The factory applied it to labour. The industrial city is what happens when all three arrive in the same place at once. Each is not merely a technology but a new structure within which parameter sets accumulate, resonate and imagine differently than before. Engineering does not merely serve the existing civilisation. It rebuilds the conditions within which integration occurs.

Science crossed into the world through engineering. Engineering crossed into culture through technology. The distinction is worth keeping. A steam engine is engineering: it does work that previously took muscle. A phonograph is technology: it does what no amount of muscle could do, which is store a moment of sound and replay it after the moment has passed. Engineering amplifies human capacity. Technology extends it into dimensions where human capacity had no reach.

Edison's phonograph of 1877 was the first technology to alter permanently the relationship between human beings and time. Sound had always been the most perishable of experiences. It existed in the moment of its production and was gone. The phonograph made it persistent: a performance separated from its performer and delivered to receiving parameter sets across time and distance with no carrier network in between.

The Media channel takes McLuhan's point, that every medium has its own Jacobian properties, through the whole history of transmission. The deeper point for this channel is what each medium does to the mind that receives it, and Socrates saw it first. In the Phaedrus he has Thamus tell Theuth that writing is a remedy not for memory but for reminding, and that those who learn through it will appear to know much while knowing less, because the knowledge has passed through the eyes rather than through integration. He was right about the process and wrong about the verdict. Writing changed what minds accumulated, and what they accumulated was richer for it. The same question recurs at every shift in the external information environment: manuscript to print, print to broadcast, broadcast to network, network to generative AI. Each changes what is available to be written into the parameter set during the formative window. External storage is not neutral. It is the formation channel wearing a different face.

Science expands integration. Engineering embeds it in physical form. Technology extends resonance and imagination into domains where biology could not reach alone. Seven of the Atlantic panel's fifty score on all three axes at once: the printing press, electricity, semiconductor electronics, the internet, computers, television and photography. These are technologies that integrate new knowledge into a civilisation, open new channels through which resonance runs between people, and expand what human beings can imagine and move towards. The digital stack is the first system in history deliberately engineered to do all three at once, at global scale, in real time.

Movement Two: The Digital Integral

The personal computer (1975 onwards) put general-purpose processing in the hands of individuals rather than institutions. Every previous computer was a shared resource whose output belonged to whoever controlled access. The PC dissolved that control. On the integration axis it extended what an individual parameter set could store, process and model. On the resonance axis its early impact was limited; the PC of the 1980s was a tool, not a node. On the imagination axis the impact was out of proportion to the capability: the image of one person at a screen, producing something from nothing, seeded a generation of parameter sets with a sense of agency that took decades to unfold.

The internet (1995 onwards) completed what the PC had begun. The PC gave every individual a processing node; the internet connected them. The integration effect was immediate: the archive of human knowledge became in principle available to any connected parameter set. The resonance effect proved more consequential: for the first time since the oral tradition, every receiver could also be a transmitter, and the gatekeepers who had managed what entered the channel (editors, broadcasters, regulators) were bypassed. The imagination effect was the most immediate of all. The mid-1990s produced an expansion of what felt possible that is hard to overstate in retrospect.

The smartphone (2007 onwards) removed the last geographical constraint, and the result was a qualitative change on all three axes at once: from intermittent access to continuous coupling on integration, from scheduled to always-on transmission on resonance, from a tool met at a desk to a dimension of experience on imagination. The parameter set was no longer carried internally into a life. The life and the network became continuous with each other.

8086, CUDA and the Installed Integral

Jensen Huang co-founded NVIDIA in 1993 to make graphics accelerators for gaming. The critical moment came around 2006. CUDA was a programming platform that would let NVIDIA's graphics processors be used for general-purpose computation, any parallel numerical workload and not only rendering. Deploying it on GeForce, the high-volume consumer product, would add real cost to every card sold into a market that paid gaming prices and would not pay more. After the decision the company's market capitalisation fell from around eight billion dollars to one and a half.

Huang held. His reasoning was about the installed base. A computing architecture is not defined by its elegance. The x86 was a palimpsest of historical compromises, inelegant by any technical standard, and it is the defining architecture of computing today; beautifully designed RISC architectures largely failed. Developers write software for platforms that reach people, and platforms reach people through installed base. NVIDIA was already selling millions of GeForce cards a year, and many of the researchers and students who would one day need general-purpose GPU computation were gamers who had built their own machines from gaming parts. Put CUDA on GeForce, teach it in universities, give it away, and eventually something remarkable will happen. CUDA is the 8086. The installed base is the integral.

What happened was deep learning. The researchers who found that neural networks trained on large datasets could produce extraordinary capabilities needed hardware to run the training, and CUDA, already resident in university labs on gaming cards, was the tool to hand. NVIDIA had not anticipated deep learning. But Huang had accumulated the prescriptive knowledge, the developer ecosystem and the installed base that made NVIDIA the only organisation able to meet the Jacobian when it arrived. This is Jacobian preparation at civilisational scale: the deliberate upstream management of the conditions that make the next event possible. Bacon performed it for the method that made Newton possible. Huang performed it for the technology community of the twenty-first century.

NVIDIA's internal organisation runs on the same logic. Huang's direct reports number over sixty, a span of control conventional management theory regards as ungovernable. He holds no one-on-one meetings. Every significant problem is brought before the full group and attacked collectively, with the specialists in memory, networking, chip architecture, algorithms, optics and power all in the room, because the problems NVIDIA solves require co-design across every dimension at once. It is, at corporate scale, a deliberately designed resonance event.

The Frozen Integral

In Blade Runner the Voight-Kampff test is designed to detect whether a replicant has an integral of its own or only a simulation of one. The film's answer, delivered through Rachael, is that the distinction has already collapsed: implanted memories that shaped the parameter set are real in every functional sense the Integral recognises. Blade Runner 2049 extends the question. K's arc asks whether a manufactured integral constitutes identity. The film's answer is yes, not because K's origins are resolved but because he acts against his own survival on values he has accumulated rather than on programmed instruction. The integral asserts itself over its initial conditions.

Whether AI systems will develop something equivalent to resonance proper, not its simulation but the architecture that makes an encounter with another integral possible, is not currently answerable. The book's account of the large language model in Part Three says why the test that matters is architectural rather than behavioural. Does the system have something at stake? Is it changed by what it encounters? Does its parameter set accumulate, or does it merely process?

These are questions the current moment is only beginning to ask with the seriousness they deserve.