The Outliers and Their Shoulders

Newton wrote about standing on the shoulders of giants. This piece is about the giants. It argues that the most popular version of civilisational history — the lone exceptional individual who transformed a field — is a systematic misreading of how transformative work actually gets done, and that the mistake matters because it produces institutional designs that damage the very conditions the outliers depend on.

Isaac Newton wrote about standing on the shoulders of giants. This is a piece about the giants, and about what happens to societies that build for the person on top and forget the pile underneath.

There is a durable and popular version of civilisational history in which progress happens because a handful of exceptional individuals, working against the grain of their societies, force through insights or inventions that their contemporaries were too conventional to reach. It is a satisfying story. It gives history a set of protagonists, which makes it easier to remember and to teach. It flatters the individualist temperament of the culture that most enthusiastically propagates it. And it happens to be, in almost every specific case it is applied to, a serious misreading of what actually happened.

This is not a piece against outliers. There are outliers, and they matter. It is a piece against the version of the outlier story that removes the conditions of the outlier’s work and pretends the outlier is doing something the culture around them was not doing. That version is close to universal in the popular telling. And when institutions are designed around it, they tend to damage the very conditions the outliers depend on, in ways the outliers themselves usually recognise and try to protest against, without being heard.

Turing at Bletchley

The canonical modern example of the lone genius is Alan Turing. Turing is remembered as the mathematician who cracked Enigma and effectively invented the computer, working alone against an unimaginative establishment. The film version emphasises his isolation, his awkwardness, his fights with his colleagues, the singular quality of his mind.

The film version is wrong in specifiable ways. Bletchley Park at its peak employed over ten thousand people. Two-thirds of them were women, running the machines, doing the pattern-matching, indexing the intercepts, feeding the possibilities Turing’s theoretical framework needed in order to run. The Bombe machines Turing designed relied on a prior technology — the Polish Bomba, developed by Marian Rejewski, Jerzy Różycki, and Henryk Zygalski in the 1930s and handed to British and French intelligence in 1939, when the Poles could no longer safely continue the work. Turing’s theoretical breakthroughs at Bletchley were embedded in an interpretive tradition and a set of technical practices that had already been developed by his Polish predecessors and were being extended by his colleagues Gordon Welchman, Hugh Alexander, Max Newman, Tommy Flowers, and hundreds of others whose names we do not know.

The Colossus computer, often described as one of Turing’s inventions, was designed and built by Newman and Flowers while Turing was in the United States. Turing’s contribution to Colossus was to have identified Flowers as the person capable of building it. The direct engineering was other people’s. This does not diminish Turing — the identification was itself a critical piece of judgement — but it is a significantly different picture from the lone genius version. Turing was, in the strict sense, standing on shoulders, and the specific shoulders were Marian Rejewski, ten thousand people at Bletchley, and the network of mathematicians who had built the theoretical apparatus he inherited.

The point is not to reduce Turing. Turing was extraordinary; the extraordinariness was real; the specific things he did were things very few other people could have done. But the extraordinariness worked because it landed in a context that could support it. Bletchley Park had an unusually deep bench, an unusually good management culture (Alastair Denniston’s willingness to hire misfits), and an unusually clear objective. In a shallower context — the Turing of 1926 rather than 1940, or a Turing in a country without the Bletchley infrastructure, or a Turing without Rejewski to build on — the same person would have produced very different results. Not necessarily worse ones; probably worse ones; possibly none of consequence at all.

The general shape

Newton, whose “shoulders of giants” line is the epigraph of this piece, gave the most famous articulation of the point about his own work. Modern scholarship has established the substance of what he was pointing at. Newton’s laws of motion were built on the astronomical observations of Tycho Brahe, systematised by Johannes Kepler, given the algebra to describe them by Islamic mathematicians who transmitted and elaborated Greek geometry, situated in a physics developed by Galileo, and made possible by an intellectual culture — the Royal Society, the correspondence networks of the seventeenth century, the Latinate community of scholars — that allowed him to know what others had done. He read everything. He built on the accumulated stock. He was extraordinary at building; the stock itself was the shoulders.

Einstein famously credited a decade of Zurich physics, an obscure patent-clerk job that gave him time to think, the intellectual company of Marcel Grossmann and Mileva Marić, and above all the mathematical apparatus of Riemann and Minkowski, without which general relativity was not writable. Darwin had Malthus, Lyell, the ships of the Royal Navy, and the immense apparatus of nineteenth-century natural history. Marie Curie had the Sorbonne, the Paris polytechnical culture, and Pierre. Every case that gets cited as the lone genius, when you look closely, turns out to be a person of extraordinary capability doing extraordinary work inside a context of extraordinary supporting conditions.

This pattern is not incidental. It is causal. Deep contexts produce more exceptional work than shallow contexts, and the exceptional work of a period tends to cluster in the deep contexts, not because the geniuses were born there but because the geniuses could do their work there. Renaissance Florence had disproportionate output for a small city not because Florentine genes were exceptional but because the patronage, the apprenticeship traditions, the accumulated craft knowledge, and the intellectual culture made the work possible. Victorian Britain, mid-century American physics, twentieth-century Vienna, Bletchley Park, Silicon Valley in the 1970s and again in the 2010s — each of them was a deep context, and each produced far more outliers than the same population would have produced in a shallow one.

What the misreading costs

Getting this wrong is not a harmless error of intellectual history. It changes what institutions design for.

The lone-genius reading suggests that what you want in a research institution is a small number of exceptional individuals given maximum autonomy and minimum bureaucratic overhead. That is close to the standard playbook of the modern research-funding model: identify the promising individuals, give them grants, remove the friction, and let their brilliance produce the outputs. This is a workable model in shallow ways — it produces papers, careers, and the appearance of activity — but it fails to build the shoulders. The apprenticeship traditions, the deep-bench technical staff, the culture of technical criticism, the accumulated craft knowledge held by people who are not themselves stars — these are the substrate the outliers actually depend on, and none of them are individually promotable, individually fundable, or individually rewarded by the metrics the lone-genius model uses.

Bletchley Park, on the lone-genius model, would have been a small research grant to Alan Turing. On the actual model, it was a five-thousand-strong operation with a deep managerial and technical culture, of which Turing’s theoretical work was one central but not sufficient component. The Bombe would not have worked without the specific women who ran it and became, over the war, the world’s most experienced operators of it. The wider culture of code-breaking would not have existed without hundreds of linguists, indexers, and analysts who between them noticed the patterns Turing’s framework mathematised. If Britain had approached the Enigma problem the way modern research funding approaches promising individuals, it would have given Turing a grant and lost the war.

Silicon Valley in its productive periods has been closer to Bletchley than to the standard research model. It has deep technical bench (the accumulated engineering culture of Fairchild, Intel, Xerox PARC, Bell Labs), deep management culture (the Andy Grove/Bob Noyce tradition), deep informal knowledge networks, and specific institutions (the Homebrew Club, the venture-capital feedback loop) that support the outliers who happen. The unusual founders arise there not because unusual people are born there but because unusual people can do their work there. The moment the shoulders start to erode — senior engineering talent priced out, apprenticeship replaced by six-month bootcamps, the tacit knowledge of the previous generation lost to retirement without being transferred — the outliers become less common, not because there are fewer geniuses but because there is less for the geniuses to stand on.

The same argument applies at national scale. Countries that lose their manufacturing base do not just lose factories; they lose the population of skilled machinists, process engineers, and industrial designers who were the shoulders on which the next generation of hardware innovators would have stood. The exceptional hardware founder in Shenzhen has an easier problem than the exceptional hardware founder in Cleveland, not because Shenzhen produces different people but because Shenzhen’s shoulders are still there. Cleveland’s aren’t.

The negative version

There is an inversion of the great-man theory that is worth taking seriously. It comes from a 2020 paper in the physics literature (Boyd et al., The problem of scientific greatness and the role of ordinary scientists), which argues that if you take the great-man theory seriously, you must also take seriously the possibility of the great negative man: the exceptionally influential individual whose influence delays, distorts, or blocks progress rather than accelerating it. The paper cites Lord Kelvin’s late-career campaign against Darwinian evolution and against radioactive dating of the Earth (both of which Kelvin was confidently wrong about, and both of which were held back for years by the weight of his authority). It cites Niels Bohr’s influence on the interpretation of quantum mechanics, which some physicists argue delayed the exploration of alternative interpretations by a generation.

The negative-great-man observation is important because it tells you something about the failure mode of designing institutions around exceptional individuals. If the individuals are correct, the concentration of authority they command accelerates the field. If they are wrong, the same concentration retards it. Since being right and being commanding are only weakly correlated in intellectual history, a system that concentrates authority on the assumption that the person at the centre is right is a system with a large downside risk.

The shoulders-of-giants model, by contrast, produces a distributed error-correction mechanism. When Turing was wrong about something — and he was, sometimes — there were Newman and Welchman and Flowers to notice, argue, and adjust. Deep contexts contain the correction; shallow contexts leave the lone genius’s mistakes uncorrected until the mistakes become the received wisdom. The lone-genius model does not only fail to build shoulders. It fails to build the correction mechanisms that let the field recover from its own mistakes.

What follows

The uncomfortable practical implication is that institutions serious about producing transformative work should be building shoulders, not hunting outliers. This is not intuitively obvious to institutions, because it is much harder to measure whether you are building shoulders than whether you have identified an outlier, and the funding models that support institutions reward what is measurable. It is possible to build a career on discovering the next Turing. It is much harder to build a career on running a well-managed apprenticeship programme for a hundred technicians none of whom will individually be famous.

But the outliers happen in the places where the shoulders were built. The correlation is not incidental. The insight worth taking from a century of civilisational history is that societies serious about their own future should be putting most of their institutional design energy into the substrate, not the star. The stars are the visible output; the substrate is the causal input. Confusing the two produces institutions that hunt for what is already hard to find and neglect what would make it easier to find.

None of this is against the outliers. The outliers are real, and their contribution is real. The argument is that the outlier is the tip of an iceberg, and mistaking the tip for the iceberg produces institutional designs that carve up the ice underneath, and then wonder why fewer tips emerge.

What would change my view

Two things.

One: a rigorous statistical analysis of the correlation between contextual depth (as measured by, for instance, the size and age of the field infrastructure a person is embedded in) and the frequency of extraordinary outputs, controlling for the individual-level variables the great-man theory emphasises. If the correlation turned out to be weak — if outliers turned out to arise as often in shallow contexts as deep ones once you control for the observability bias — the argument would need substantial revision.

Two: a clean historical case of a lone individual producing transformative work with no meaningful contextual support, that survives the standard forensic biographical treatment. The candidates that get cited (Ramanujan is a common one) tend on closer inspection to show substantial supporting conditions that the popular version omits. If a case survived that scrutiny, the shoulders argument would become an empirical average with real exceptions, rather than the near-universal pattern the piece claims.

The publication’s working position is that the outlier narrative is not merely popular but institutionally destructive, and that the correction — building shoulders — is one of the two or three most important pieces of civilisational infrastructure. The publication does not consider this position beyond dispute.