In 1958, Leonard Read published an essay titled "I, Pencil." It is written in the voice of a pencil, describing its own creation. The pencil traces its family tree: cedar from Oregon, felled by loggers using saws made of steel alloyed with nickel from Canada. Graphite from Sri Lanka, mined by workers using equipment powered by diesel refined from oil drilled in the Middle East. Clay from Mississippi. Candelilla wax from Mexico. Rapeseed oil from Indonesia. Pumice from Italy. The lead is not lead at all — it is graphite mixed with clay, baked, treated with a mixture of fats and waxes. The eraser is rubber from Malaysia, vulcanized with sulfur, bonded with a brass ferrule made of zinc and copper. The paint is six coats of lacquer. The label is a film of carbon black pressed into the wood.
"Simple? Yet, not a single person on the face of this earth knows how to make me."
That is the sentence. The pencil is simple. A child can use it. It costs a few cents. Nobody knows how to make one.
"Actually, millions of human beings have had a hand in my creation, no one of whom even knows more than a very few of the others. There isn't a single person in all these millions, including the president of the pencil company, who contributes more than a tiny, infinitesimal bit of know-how."
The knowledge is distributed. No master mind coordinates it. No central planner designs the supply chain. The loggers, miners, chemists, machinists, and factory workers do not collaborate. Most have never met. Most do not know their work contributes to pencils. They exchange their tiny know-how for wages and trade those wages for goods they need. The pencil emerges from the exchange. It is a product of the market — of millions of specialized know-hows coordinated not by command but by price.
"There is a fact still more astounding: the absence of a master mind, of anyone dictating or forcibly directing these countless actions which bring me into being. No trace of such a person can be found. Instead, we find the Invisible Hand at work."
"The lesson I have to teach is this: Leave all creative energies uninhibited. Permit these creative know-hows freely to flow. Have faith that free men and women will respond to the Invisible Hand."
When ideas have sex
In 2010, Matt Ridley stood on the TED stage and gave the talk that extends Read's pencil into a general theory of innovation. The title: "When Ideas Have Sex."
Ridley's argument is that human progress is not driven by individual intelligence. It is driven by the recombination of ideas through exchange. Biological evolution accelerates when sex mixes genes from different lineages. Cultural evolution accelerates when trade mixes ideas from different minds. The computer mouse is not one idea. It is a confection of plastic, lasers, transistors, ergonomics, injection molding, USB protocols — each the product of thousands of specialized minds, none of whom could make a mouse alone.
"It's not important how clever individuals are; what really matters is how smart the collective brain is."
"There is nobody on the planet who knows how to make a computer mouse. I mean this quite seriously."
The pencil and the mouse are the same argument at different technological scales. The pencil required global supply chains of raw materials. The mouse requires global supply chains of ideas. The principle is identical. Distributed knowledge, coordinated by exchange, produces objects no individual understands. The collective brain is smarter than any of its neurons. The neuron's job is not to know everything. It is to connect.
Ridley contrasts two objects of similar size and shape, both designed to fit the human hand. The Acheulean hand axe was made by Homo erectus to an unchanging design for 30,000 generations — a million years of hitting things with the same rock. The computer mouse is obsolete within five years. The difference is not individual intelligence. Homo erectus had brains. They had no exchange networks. Their ideas did not have sex. Human ideas do.
"It's the interchange of ideas, the meeting and mating of ideas between them, that is causing technological progress, incrementally, bit by bit."
Ridley's darkest example: Tasmania. When rising sea levels isolated Tasmania from mainland Australia 10,000 years ago, its population of 4,000 was too small to sustain the specialization needed to maintain bone tools, fishing equipment, and cold-weather clothing. They lost technologies their ancestors had possessed for millennia. Not because they became less intelligent. Because the network was too small. The collective brain atrophied. Tierra del Fuego, connected by a land bridge, kept its technology. The difference was not genetics. It was exchange.
The software pencil
Every non-trivial software system is a pencil. Nobody knows how to make it.
The Linux kernel: 30 million lines of code, contributed by thousands of developers, running on hardware designed by thousands of engineers, compiled by toolchains built by hundreds of contributors, deployed on servers manufactured across global supply chains. The kernel maintainer does not know how to make a server. The server manufacturer does not know how to write a scheduler. The knowledge is distributed. The system emerges from the exchange — not of money, in the kernel's case, but of patches, reviews, and trust. Open source is the pencil argument applied to software. No master mind. Distributed know-how. Emergent order.
"The lesson I have to teach is this: Leave all creative energies uninhibited."
The open source movement learned Read's lesson instinctively. Permit contributions to flow. Have faith that distributed developers will respond to something — not the Invisible Hand of price, but the visible hand of reputation, utility, and the satisfaction of solving hard problems with smart people. The mechanism is different. The principle is the same. The knowledge is distributed. The coordination is emergent. The product exceeds any individual's understanding.
The economics of software specialization
Read and Ridley describe a world where specialization and exchange produce objects of extraordinary complexity with no central coordination. Software engineering has not internalized this. The dominant model of software production is still hierarchical: architects design, developers implement, managers coordinate. The organization chart is the architecture. Conway's Law guarantees it.
But the pencil suggests a different model. Software as emergent from specialized modules, each produced by a team that knows only its own domain, exchanging through stable interfaces. This is microservices done right — not better monoliths, but genuine specialization coordinated by interface contracts rather than organizational hierarchy.
Parnas's information hiding is the software equivalent of market exchange. Each module hides its implementation. Each module exposes only what others need. The modules do not collaborate. They exchange. The interface is the price mechanism — it communicates what is available and what is required without revealing how either is produced. The system emerges from the exchange of specialized know-hows. No module knows how the whole system works. No module needs to.
"There is a fact still more astounding: the absence of a master mind."
A well-architected software system has no master module. No orchestrator that knows everything. The knowledge is distributed across components, each responsible for one hidden decision. This is Parnas's vision, implemented. It is also Read's vision, applied to code. The pencil and the module are the same argument in different media.
Task automation economics: the asset as exchange
The task automation economics paper discussed earlier on this blog argues that the economic unit of software production is not the agent run but the verified automation asset — a released object with explicit specification, evidence, and a defined interface. This is the pencil argument applied to automation.
A verified automation asset is a unit of specialized know-how, encapsulated behind a stable interface, exchangeable across teams. One team produces the asset. Another team consumes it. Neither knows how the other works. The interface is the contract. The exchange is the mechanism. The system's capability grows not through central planning but through the accumulation and recombination of specialized assets. This is Ridley's "ideas having sex" implemented in software infrastructure. Each asset is an idea. Each pipeline that composes assets is an idea mating. The collective brain of the organization grows with each new interface, each new composition, each new recombination of existing capabilities.
The dark factory extends this further. Specs go in. Software comes out. The spec writer does not know how the agents generate the code. The agents do not know what the spec means. The validation layer does not know why the spec was written. Each component knows its fragment. The whole exceeds the parts. The factory is the pencil. No component knows how to make the software. The software emerges anyway.
The limits of the metaphor
Read's pencil has a weakness. It assumes the Invisible Hand produces optimal outcomes without specifying the conditions under which it fails. Markets produce pencils efficiently because the pencil's design is stable, its components are well-specified, and its interfaces — the size of the ferrule, the hardness of the graphite, the diameter of the wood casing — are standardized. When interfaces are unstable, markets fail. The pencil works because the specifications are fixed.
Software interfaces are rarely fixed. They change. They drift. They accumulate assumptions that become invalid. Lehman's Laws apply: E-type systems must evolve, and evolution increases complexity. The market mechanism — stable interfaces, specialized producers, emergent coordination — breaks down when the interfaces themselves are changing. This is why software architecture is harder than pencil manufacturing. The pencil's architecture is stable. The software's architecture is not.
The insight is not that software should be more like pencils. It is that the parts of software that are stable should be modularized behind stable interfaces, and the parts that are volatile should be contained. Parnas again: hide the volatile decisions. Expose the stable ones through interfaces that don't change. Where the interface is stable, the market works. Where the interface shifts, the market fails, and coordination requires something closer to hierarchy. The art of software architecture is knowing which parts are which.
The collective brain of a codebase
Ridley's collective brain is not a metaphor. It is a description of how a well-architected codebase actually works when it works well. Each module contributes specialized knowledge. Each interface enables exchange. Each composition of modules is an idea mating — a recombination of existing capabilities into something new. The system's intelligence is not in any single component. It is in the connections between them.
"It's not important how clever individuals are; what really matters is how smart the collective brain is."
Replace "individuals" with "modules." Replace "collective brain" with "architecture." The sentence remains true. A system of simple, well-specified modules connected by stable interfaces is smarter than a monolith written by geniuses. The monolith is a single mind. The modular system is a collective brain. The collective brain outlives its components. The monolith dies with its authors.
Brooks argued that conceptual integrity requires one mind. Ridley argues that progress requires millions of minds exchanging ideas. The resolution: one mind designs the interfaces. Millions of minds implement behind them. The interfaces are the market. The implementations are the specialized producers. The system has conceptual integrity because one mind controlled the interfaces. The system has distributed intelligence because many minds contributed the implementations. This is the pencil argument synthesized with Brooks. The pencil has conceptual integrity — it is clearly one thing, designed for one purpose. The pencil is also the product of millions of specialized know-hows coordinated by exchange. The pencil is both. Good software is both.
References:
- Leonard E. Read, "I, Pencil: My Family Tree as Told to Leonard E. Read," The Freeman, December 1958. Full text at Econlib.
- Matt Ridley, "When Ideas Have Sex," TEDGlobal 2010. Video.
- Matt Ridley, The Rational Optimist: How Prosperity Evolves, Harper, 2010.
- Related posts: Brooks on Software Design, Parnas's Information Hiding, Lehman's Software Evolution, Task Automation Economics, Software Dark Factories, Disruptive Innovation
Engineering is the through-line. Every topic on this blog — version control, networking, philosophy, economics, AI — connects to the discipline of designing and building systems that work within constraints. The constraint may be compute, attention, time, or complexity. The method is the same: understand the problem, design a solution, verify it works, iterate. The domain provides the specifics. The method is engineering.
No single person knows how to make a pencil. The knowledge is distributed across millions of minds, coordinated by prices, not by a plan. The pencil is the product of the market. The market is the mechanism. The mechanism is the architecture of cooperation without coordination.