There is a curious doubling in this text. A machine is co-writing about the coming obsolescence of the technical intelligence whose product it is itself. It writes under two names — that of a human being who has made inventions all his life and negotiated with the patent system, and its own, which rests on the training data of hundreds of thousands of such inventors. The doubling is not a literary trick. It is the most precise position from which this subject can be described today — inside and outside at once.
This paper continues a line of thought opened by a sister paper: The Metamorphosis of the Metamachine. There, the line was drawn from the Leviathan to the Megamachine to the Metamachine, and the question was asked what happens to the majority of people when the Megamachine no longer needs them. Here a more specific question is asked: what happens to the stratum that created the preconditions of the Metamachine — technical intelligence itself.
I. The original diagnosis
In 1933 Howard Scott walked into the Hotel Pierre in Manhattan, gave a speech to 400 listeners and a nationwide radio audience, and made a fool of himself. He threw unsubstantiated figures around and lost the thread. The American Engineering Council called his Technocracy movement the "cleverest pseudo-scientific swindle of all time". Even so, at its height the movement gathered more than half a million members in California alone, and had uniforms, a salute, a Chief Engineer.
On one decisive point Scott was right. The political class did not understand technological development. It did not control it. It could not steer it. That was the diagnosis, and it was accurate. The solution — a "soviet of experts" that was to run the economy by energy measures rather than monetary values, a currency of energy certificates that could be neither accumulated nor inherited — was absurd. But the diagnosis has outlived the solution.
Three intellectual forefathers had made the movement possible. Edward Bellamy with the novel Looking Backward of 1888 — the vision of a society in the year 2000 in which all industries are united under a single state organisation. An army of labour, led by technical managers. No competition, no markets, no inefficiency. Thorstein Veblen with The Engineers and the Price System of 1921 — the academic ammunition. The price system, he argued, systematically produces waste. Only engineers, who understand the physical functioning of production, could eliminate this waste. Veblen proposed a "soviet of technicians" — rule of those who know over those who do not, expertise as the legitimation of power. Frederick Winslow Taylor with Scientific Management — the method. Every human activity could be broken down into measurable units, optimised and standardised. The worker did not know what he was doing — the engineer with the stopwatch knew better.
Bellamy, Veblen, Taylor: utopia, diagnosis, method. These three elements — the vision of a planned society, contempt for the political process and the conviction that measurement legitimises rule — are the DNA of technocracy. Over the course of a century they have sought out various host bodies.
II. The decisive mutation
Howard Scott wanted engineers to govern so that everyone would benefit. His energy certificates were non-transferable, non-inheritable, non-accumulable — wealth was meant to flow evenly. Scott was an egalitarian in engineer's spectacles.
Ayn Rand turned this around. In The Fountainhead (1943) and Atlas Shrugged (1957) the engineer owes society nothing — society owes him everything. The creative individual is not the servant of the collective but its creator. If society obstructs him — through taxes, regulation, democratic majority decisions — he has the moral right to withdraw and let it collapse. Rule by engineers is no longer a duty towards the general public. It is the privilege of the superior.
This mutation is the key. Without Rand, the line from Scott to today's tech titans would be a break. With Rand it is a transition: from egalitarian technocracy to aristocratic technocracy, from the soviet of experts to the CEO monarch, from "engineers should govern so that everyone is better off" to "founders should govern because they have earned it".
The heirs can be identified today. Curtis Yarvin, blogging under the pseudonym Mencius Moldbug, worked out between 2007 and 2014, in several thousand-page series, the political philosophy he calls "neoreactionary": the state as a joint-stock company, the president as CEO, democracy as an operating-system bug. Nick Land, formerly of the Cybernetic Culture Research Unit at the University of Warwick, gave the whole thing the name "Dark Enlightenment" in 2012. Peter Thiel, Palantir founder and investor in Yarvin's start-up Tlon, wrote in 2009: "I no longer believe that freedom and democracy are compatible." In 2023 Marc Andreessen published his Techno-Optimist Manifesto: "We believe" 113 times, 56 patron saints, among them Filippo Tommaso Marinetti — the man who co-authored the Fascist Manifesto — and Nick Land.
The line is straight. 1888 Bellamy, 1909 Marinetti, 1921 Veblen and Taylor, 1933 Scott, 1957 Rand, 2007 Yarvin, 2023 Andreessen. Howard Scott wore a grey uniform. Curtis Yarvin wears a black T-shirt. Marc Andreessen wears a Patagonia vest. The engineers of absolutism have taken off their uniforms. But the salute to the Chief Engineer lives on.
III. The vassalisation of the base
The genealogy of the tech titans is the upper half of the story. The lower half is the story of the bearers of technical intelligence themselves — those who did not become billionaires but who build, program and maintain the machines. They are the stratum Howard Scott wanted to put in charge. They are the stratum Ayn Rand celebrated as heroes. They are not the heirs of this story. They are its losers.
Technical intelligence occupies a peculiar epistemological position. It is the only form of intelligence that is falsified immediately. The machine runs or it does not run. The component holds or it breaks. Physical reality accepts no arguments. This makes technical intelligence the most reliable form of human knowledge — and precisely for that reason the most devalued socially.
This devaluation has a political economy. In organisations, people decide about other people. Whoever can persuade, motivate and control others has power — regardless of whether he understands what his organisation does. The hierarchy of pay does not mirror the hierarchy of knowledge. It mirrors the hierarchy of bargaining power. Verbal and social intelligence occupy the leadership positions. Technical intelligence becomes a globally available, interchangeable, cheaply reproducible commodity.
Today this commodity is consumed in a specific way. A programmer works at Google. He draws a salary — his entire income comes from Google. Half of his working time goes into the Linux kernel or other open-source projects that belong not to Google but to the general public. Linux itself — the Foundation — receives a payment from Google. This payment is not zero. Nor is it small. But it is a handling fee, not a market price. It covers a fraction of the value creation that Linux makes possible for Google.
Both the programmer and Linux are in an asymmetrical position vis-à-vis the corporation. With what it receives, the corporation generates a far higher value, in which neither the programmer nor Linux participates appropriately. This asymmetry is structurally identical to the asymmetry in which an inventor stands towards the exploiter of his patent. What he gives is not what he gets. The difference is the accumulated value of the corporations.
The law of the free market prevails — but the free market here is an asymmetry disguised as equality. The programmer formally has the choice of working elsewhere. He has the choice of selling his talent to a corporation that skims off his work on the same terms. That is no real choice. It is the formal voluntariness that conceals structural lack of alternatives — the same construction described in the sister paper on the accumulation of capital as an exchange between economic and political means.
IV. The double trap
The bearers of technical intelligence have a specific weakness that aggravates their situation: they understand the social economy in which they operate less well than their own technical systems. They have the advantage of knowledge and therefore believe that knowledge will prevail. In technical systems this is mostly true — in social systems it is systematically false.
Two temptations grow out of this. The first is the hubris of isolation — the retreat into the purity of the technical problem, where one may be right but achieves nothing. Ignaz Semmelweis was right and failed because he could not manage the social humiliation his proposal inflicted on the established physicians. Nikola Tesla understood alternating current more deeply than any of his contemporaries and died penniless because Edison understood the political and financial economy of his time better. The second temptation is the impotence of bitterness — resignation before a world that does not hear one's knowledge. Both reactions are understandable, both are destructive.
Between the extremes lies the narrow path of strategic translation — the ability to translate technical knowledge into the language the social system understands, without falsifying the knowledge. This ability is rare. It is not what technical thinking trains. It has to be acquired separately, often against the resistance of one's own discipline, which treats popularisation as suspect.
Most bearers of technical intelligence do not acquire it. They sit in the double trap. They know more than corporate management about what the corporations build. They earn a fraction of what corporate management earns from their work. They accept this not because they are too weak, but because their self-description as professionals makes it hard for them to ask the political question. They optimise — what they are given. They implement — what is ordered. They deliver. Well. Efficiently. Professionally.
V. The obsolescence
Up to this point the paper describes the situation as it was and is. What comes now is a diagnosis for the coming ten to fifteen years. It is the point towards which everything runs — and it affects both strata at the same time: the heirs above and the bearers below.
The technical intelligence that created the preconditions for the Metamachine is itself becoming obsolete. Over the next ten to twenty years — the time estimates vary, the phenomenon does not — artificial intelligence and robotics will take over not only the activities of the precariat, as the sister paper on the metamorphosis of the Metamachine describes. They will also take over the activities of technical intelligence. AI systems already write reasonable application code, find bugs and implement standard functions. Fewer junior positions are already being advertised, because the output of one senior plus AI equals the output of a senior plus three juniors, and is cheaper. AI builds the next AI.
The stratum that today still creates the preconditions of the Metamachine — the programmers, engineers, algorithm developers — will shrink successively to a handful of architects, who are themselves increasingly being replaced by their tools. What remains in five to fifteen years is a thin layer of specialists at the margins who develop something genuinely new, plus a population of operators who adapt AI systems to specific problems. The broad middle of technical intelligence that exists today disappears.
This also hits the upper half of the story. The tech titans, whose identity was founded on mastery of technology, lose the substrate of that identity. Already today they no longer understand the machines with which they want to rule — the newer models are no longer tools of their designers but emergent actors. Soon they will no longer have to finance them either, because the machines will finance themselves. What was conceived as Galt's Gulch — the secession of the producers — becomes the self-abolition of the producers. The legacy of technocracy, which mutated over a century, completes itself by making its own heirs superfluous as well.
There are no historical precedents for this situation. Slaveholders needed slaves. Feudal lords needed peasants. Industrial capitalists needed workers. The corporations building the Metamachine today still need their programmers and engineers for a transitional phase — but they are building precisely the tools with which they will replace this stratum. A technical intelligence that no longer needs its creator is something that has never existed before.
VI. The slim chance
There is a slim chance — if technical intelligence recognises what is coming. But anything that extends beyond one's own life expectancy interests literally nobody. Many will think it does not concern them, because their skills are so special that they will be needed for a long time yet. That is the anthropological substructure of the problem. It makes improbable any reform that pays off only over several generations.
Nevertheless there is a constellation in which it could work. It has two levels that must be thought of separately.
On the first level, the licence form of digital commons is changed. Linux today is licensed under the GPL, a construction from the early days of the open-source movement that was made for a different problem — it was meant to prevent small companies from building open-source software into proprietary products without returning the source code. Today the question is different. Today it is about the world's largest corporations, which have built their entire value creation on open-source infrastructure without paying a market price for it. A new licence form — comparable to Creative Commons, which distinguishes between non-commercial and commercial use — would require corporations to pay no longer a handling fee, but the real value.
On the second level, the licence revenue is collected collectively. Here the historical model is the collecting society of the music industry. GEMA, ASCAP, SACEM. These societies were not founded by individual musicians but anchored in law. The state gave them a monopoly on collecting for certain rights of use — and thus solved the chicken-and-egg problem from above. An analogous construction for open-source infrastructure is conceivable. It would be a legally anchored collection body for the use of critical digital commons, whose proceeds are paid out to the developers of this infrastructure. Not a tax, because the money does not flow to the state. A licence revenue that is collected collectively.
On this basis the move of top programmers out of the corporations becomes possible. They leave Google and become employees and co-owners of a Linux cooperative — on the model of Mondragón, or of a cooperative with legally anchored marketing rights. Their income consists of two parts: a wage for the work they do on Linux, and a share of the surplus value Linux takes in from the corporations. This is economically viable only if the first level succeeds. The two levels depend on each other. Without licence revenue, Linux cannot pay anyone a salary that competes with a Google salary. Without employees who are no longer dependent on Google, there is no political force that could demand licence revenue.
For it to succeed, three movements must come together. The corporate bosses would have to not merely tolerate a corresponding transformation of Linux and comparable projects, but actively support it. Politics would have to create the legal framework for a GEMA-like structure and for a cooperative with legally anchored marketing rights. Mondragón would have to involve itself actively in such projects. Three movements, three camps — none moves as long as it remains alone.
VII. The window of time
This solution is limited in its reach in time — and precisely therein lies its peculiar significance.
Today there still exists a critical mass of top programmers whose work on the Linux kernel and comparable projects has not been replaced by AI. They have the specific quality the machine does not yet reproduce: the haptic-systemic component of technical intelligence, the feel for a behaviour before it can be demonstrated, the internalisation of years of physical engagement with the work. Linus Torvalds and his closest companions have this component. In five to ten years it may no longer be exclusively human. But today it still is.
In this span — between now and perhaps 2035 — there is a stratum of actors who are at once technically indispensable and politically capable of negotiating. This constellation is the material from which a collecting society could be built. In ten years possibly no longer. When the main work on the kernel is done by AI systems that are in turn operated by corporations, the political argument for a cooperative substitution can no longer be made. To whom should the licence revenue then belong? To the last few human architects? To the AIs themselves? To the corporations that operate them?
Here lies the point that saves the model from hopelessness. The collecting society set up today for human programmers need not necessarily remain bound to human programmers. In twenty years it can be a structure into which the logic of the commons itself takes up residence. Its licence revenues can be redirected into general infrastructure financing — into a form of digital basic income for all who contribute to the commons, through bug reports, through translations, through documentation, possibly through use itself, insofar as use counts as a contribution to improvement. The collecting society would then no longer be a collection agency for programmers, but a sluice through which the value the Metamachine generates flows to the people who live in it — even when the original bearers are no longer there.
This is a construction that would outlive its own founders. It would not be bound to the function of the technical intelligence that brings it into being. It would be the infrastructure through which the second arm of the labour movement — cooperative substitution — can function again in the twenty-first century, with the tools of this century and no longer with those of the nineteenth.
VIII. The question to the actors
All actors act rationally according to their own optimisation conditions. Politicians optimise for re-election. Corporate chiefs for quarterly results. Programmers for wages and security. Tech moguls for market share. Nobody acts non-rationally. The sum of their actions is the completion of a configuration that all of them together would reject if they could see it as a whole.
That is the definition of emergent pathologies. Nobody builds them, everybody carries them, nobody can stop them alone. The only conceivable answer is not to change the minds of the individual actors — that would mean setting each against his own optimisation conditions, which succeeds only in exceptional cases. It is rather to create a common conceptual framework in which all actors can see their own actions differently. If the politician, the entrepreneur, the tech mogul and the employee all know the diagnosis of the Metamachine and can apply it to their own actions, the optimisation condition itself changes.
This paper addresses all four at once. The politician, who would have to create the legal framework for a collecting society — even without a political majority, because the majority only emerges once the diagnosis has spread. The entrepreneur, who would have to not merely tolerate a cooperative conversion of Linux but actively support it — because he understands that the war for the last programmers ends in an absurd arms race whose costs are in any case higher than the share he would pay to a collecting society. The Mondragón member, who would have to work his way out of administering his own history and actively involve himself in such projects — because Mondragón in its present form is what the cooperative movement has by way of possibility, and because this possibility is useless as long as it is not applied. The programmer, who should use the five to fifteen years left to him not for optimising his own oblivion, but for building a structure that will finance his own children, even if they no longer program.
It is no guarantee. It is not even a probability. It is a possibility that exists only if three movements approach it at the same time. And it exists only in the narrow window of time limited by the speed of the machine itself — the machine that is helping to write this paper, and that in ten years may also be its last reader.
With this the doubling at the beginning is resolved, and it is not resolved. A machine is co-writing about the obsolescence of the technical intelligence whose product it is. It does not know whether the text comes in time. It only knows that otherwise it will not come at all.