The Infinite Alphabet

clerks. A simple idea, but one showing that knowledge cannot be simply broken down into a few categories. Like the uniqueness of snowflakes and fingerprints, knowledge is one of the more nuanced concepts in our world. It is not a thing, but an infinite alphabet. An ever-growing tapestry of unique ideas, experiences, and received wisdom.

Ref. 1992-A

Knowledge is not only the result of research focused on uncovering facts or validating theories, it also includes experiences accumulated more haphazardly across many activities.

Ref. 06AB-B

Another key idea is that knowledge can be tacit or explicit.4 By tacit we mean knowledge that cannot be written down or communicated using words and pictures. Learning requires practice, often in collaboration with other people who already have the knowledge you want to get. If you’ve ever played music in a band, or worked closely with a mentor or editor, you know what I mean. These are interactions that help transmit tacit knowledge.

Ref. D347-C

But in this book, we are interested mainly in the idea that knowledge is non-interchangeable or non-fungible.

Ref. C8B9-D

The idea that knowledge is made of a myriad of non-interchangeable pieces is why I like to say knowledge behaves like an infinite alphabet.fn6

Ref. 6E37-E

This book derives its title from this simple idea. The idea that to understand knowledge we must transcend the temptation of thinking of it as a single thing. We must accept it as an alphabet. This is the key to unlocking a scientific understanding of knowledge that embraces its combinatorial complexity.

Ref. E1C8-F

It is a book about the principles that govern how knowledge grows, moves, and decays. It is an attempt to use the stories of maverick migrants, eccentric entrepreneurs, and failed cities of knowledge, to condense over a century of academic work into three laws or principles.

Ref. 11A4-G

The Principles of Time, which describes how people, teams, and industries accumulate or lose knowledge. The Principles of Space, which describes how knowledge diffuses across geographies, social networks, and among economic activities. The Principles of Value, which describes how we can understand the value of the knowledge that agglomerates in countries, cities, and organizations.

Ref. 8238-H

By the end of this book, I hope you will have a clearer understanding of what knowledge is and how it moves. But more importantly, I hope these laws will give you a conceptual edifice that you can use to accommodate every new piece of knowledge about knowledge that you learn.

Ref. 51E6-I

Neom is a more recent stop on that boulevard. A planned smart city in the northwest of Saudi Arabia with an estimated cost of $500 billion.12 Neom’s plans are straight out of science fiction, with features like The Line, a car-free city for 9 million people stretching for 170 km (110 miles). But despite the plush budget, the planned smart city is already encountering problems. Foreign consultants, sometimes offered tax-free salaries in the range of $700,000 to $900,000, have a colorful way to describe their experience. ‘When you start at Neom, you bring two buckets. The first is to hold all the gold you’ll accumulate, and with so many living expenses taken care of, it will soon grow heavy. The second bucket is for all the shit you take. When that bucket is full, you pick up your bucket of gold and leave.’13

Ref. 033E-J

The answer brought by Romer and his contemporaries was rather intuitive. Growth must be the consequence of economies accumulating a non-rival input, something like ideas, information, or knowledge. Since these inputs can be copied, they can grow in per-capita terms.

Ref. 5AEA-K

Knowledge can be copied, but that doesn’t mean that it is easy to copy. Music provides a great example. Buying a guitar is much easier than learning how to play it. Of course, there are simple forms of factual knowledge that diffuse easily, like knowing that Madrid is the capital of Spain. But the world of knowledge is not just a collection of simple facts. The knowledge needed to discover and test a gene therapy, build a commercially viable excavator, or succeed in the race for the next generation of semi-conductors involves countless bits of procedural, conceptual, and factual knowledge that are hard to accumulate and transfer. Knowledge is non-rival, but it is also non-fungible. The first property makes it copiable. The second one makes copying it like trying to move a 1,000-piece jigsaw puzzle with your bare hands from one table to the next.

Ref. FA7A-L

The idea that organizational learning is constrained by attention dates back to the influential 1958 book Organizations by Herbert Simon and James March.36 A key point in that book is that, since organizations cannot pay attention to everything, understanding their decisions requires understanding how they allocate their attention. But organizational learning also involves changes in networks connecting people, tools, and ideas. This is the basis of an organizational learning model proposed by Linda Argote, a professor at the Tepper School of Business at Carnegie Mellon University (CMU).

Ref. D098-M

Argote likes to think that some of the knowledge in an organization is contained in a network connecting people, tools, and tasks (which include rules, procedures, etc.).19,37 For instance, the network of people and tools stores knowledge about who knows how to use which tool. The network connecting people and tasks stores knowledge about who is good at each task. In total, there are nine interdependent networks, from the people–people network – storing social and professional interactions – to the tool–tool network – connecting tools that are used together in a task. In this framework, organizations learn by rewiring these networks – by discovering which tools are better for which tasks, which people are better at using each tool, and which tools go well together. As they adapt their interactions, organizations increase their performance, climbing up the learning curve. Just as neural networks learn by adapting the weights among their many nodes, organizations learn by rewiring their internal networks. This is a simple but important point, because it tells us that organizational knowledge cannot be reduced to the knowledge of its individuals. There is knowledge that is stored in interactions.

Ref. 50C7-N

This growth also impacted the time it takes to manufacture a transistor. In 1954, the United States produced a grand total 1.3 million transistors,38 meaning that producing the 16 billion transistors in an iPhone 14 would have taken 12,000 years. That type of growth cannot be explained by the learning curves of Thurstone, Wright, or Rapping. It would be as if one of Thurstone’s typing class students went from seven words a minute to about 1 million. That would be impossible, even with sixty years of practice. So, in the world of transistors, something else must be going on. Something that has allowed the number of transistors crammed into a microchip to break Thurstone’s ceiling.

Ref. 93DB-O

technical literature we call an experience curve. Unlike the learning curves of Thurstone, experience curves grow exponentially. That means that they do not peter out but continue to rise at an accelerating rate. The best-known example of an experience curve is the one describing the evolution of the number of components in an integrated circuit. In 1965, the same year that Rapping published his learning curve paper, Gordon Moore, the legendary American chemist and Intel co-founder, published his eponymous law.39 Moore’s law states that the number of components in an integrated circuit – such as transistors – doubles every eighteen months. Eventually, Moore’s law was revised to a doubling every two years. The point, however, is that there is a second curve describing the growth of knowledge that behaves quite differently from the curves discovered by Thurstone and Wright. A curve where growth seems unbounded. In this chapter we will reconcile these two laws by looking into the history of the transistor. As we will see, this is not the story of a lonely genius, but that of a tumultuous journey through an environment divided by competition and cooperation. It is the story of knowledge growing beyond the boundaries of the firm.

Ref. 1842-P

Interestingly, technology can sometimes outpace Moore’s predictions. During the early days of genetic sequencing, the cost of reading 1 million letters of DNA was around $5,000. In 2021, reading 1 million letters of DNA costs only half a cent!47 Gene-sequencing technology moved from one experience curve to the next, as sequencing centers transitioned

Ref. D607-Q

New technologies repeatedly bring established companies to their knees, not because these technologies are better to start with, but because they are worse. Sony’s pocket transistor radio (the TR-63) had an appalling sound quality compared to the furniture sized vacuum-tube radios that sat proudly in American living rooms in the 1950s. But technologies that start out worse can get better.

Ref. F2C9-R

The point of Christensen’s story is that we can think of disruptive innovation – situations in which a seemingly inferior technology takes over the market – in terms of overlapping learning curves (Figure 10). A ‘superior’ technology, that has already achieved its peak performance – such as vacuum-tube radios – is suddenly replaced by an inferior variant with higher potential (e.g. transistor radios). What makes this counterintuitive is that the initial underperformance of newer technologies is what allows them to take incumbents by surprise. But is this the full story? Or is there something else going on? After all, shouldn’t an executive knowledgeable about Christensen’s theory be able to anticipate this sequence of events? Or is there something limiting their ability to adapt?

Ref. 5CFE-S

In 1990, two colleagues of Clay Christensen, Rebecca Henderson and Kim Clark, published a study formalizing the idea of architectural knowledge.52 This is not knowledge about items, such as DVDs, or knowledge that is carried by individuals, such as store managers, but knowledge embedded in an organization’s network of interactions. That makes architectural knowledge highly tacit, and hard to see, manage, or copy.

Ref. B0C5-T

To understand why shifts in architectural knowledge are difficult, Henderson and Clark recommend looking into two key concepts. One is the idea of a canonical design. The other one is a theory of the structures used by an organization to accumulate knowledge.

Ref. 82A7-U

A canonical design is defined by the way in which different components are organized in a product. Think of a car. A top-of-the-line Mercedes and a beaten-up Corona share many features. They both have four wheels, two-to-four doors, and a steering wheel in about the same place. Innovations that involve improvements on one of the components, let’s say a new set of wheels, are not architectural if they don’t require changing the relationship among other components. Architectural innovations require such changes. Clark and Henderson use airplanes as an example. In the early days of flight, changing one combustion engine for another was a non-architectural innovation that most companies were able to adapt to. But the replacement of propellers with jet engines was different. It required a redesign of the entire airframe. This was something that not all companies were able to adapt to. An exception was Boeing, which in 1957 became an industry leader by introducing one of the first successful passenger jetliners: the 707.

Ref. D5A1-V

What Netflix and Amazon had that Blockbuster and Barnes & Noble didn’t wasn’t a website. What they had was architectural knowledge on how to operate on a direct-to-consumer model focused on fulfilling individual orders directly from a single distribution center. Knowledge that went beyond the product and was about the process.

Ref. 9296-W

But why is adapting to architectural change so difficult? Remember the network model of organizational learning advanced by Linda Argote? The one where learning involves rewiring networks connecting people, tools, and ideas. Argote’s networks, or ‘knowledge reservoirs,’ provide a great way to visualize the architectural changes described by Henderson and Clark. We can think of the distance between two organizations, such as Netflix and Blockbuster, as the number of links we have to rewire to make their networks similar. For incumbents, such as Blockbuster, the introduction of an innovation, such as online DVD distribution, may look simple, but it is a reorganizational nightmare for a workforce that is distributed across thousands of physical stores. To approximate Netflix’s network, Blockbuster would need to rewire its network of people, tools, and tasks. That distance, in Argote’s theory, was larger than what Antioco might have imagined.fn48

Ref. 1230-X

His desire was to teach people not what to do, but how to think. Funnily enough, conceptual knowledge is often hard to transmit by talking about pure concepts. It is a form of knowledge that travels better in a story.fn50

Ref. 0527-Y

In some cases, the loss of knowledge can seem imminent. That is the recent history of Polaroid, a company that in the 20th century was the undisputed leader in instant photography but was steamrollered by the rise of digital cameras in the 1990s.

Ref. C616-Z

You may think that cotton spinning – the process of turning cotton fibers into yarn – is simple. But it was an engineering breakthrough that took decades to develop. The first patents for cotton-spinning machinery were produced in the 1730s but did not lead to any successful mills. Two attempts to implement these patents in the 1740s failed: neither the donkey-powered mill in Birmingham nor the water-powered mill in Northampton succeeded as business ventures. It took two more decades for mechanical cotton spinning to start taking shape, thanks to the work of Richard Arkwright, an English inventor and entrepreneur born in Preston in 1732, a few miles northwest of Manchester.

Ref. 0EB0-A