Coldcode Si Labs Library

The Chisel and the Light

The Leapfrog Series · Volume III · Courage Garikai
View-only edition. No download from this page.
V O LU M E I I I O F T H E L E A P F R O G S E R I E S

The

Chisel

and the Light

In Defense of Super Intelligence as a Tool — a craftsman's argument for clarity in the age of AI C O U R A G E G A R I K A I

THE CHISEL AND THE LIGHT

In Defense of Super Intelligence as a Tool The Leapfrog Series · Volume III Copyright © 2026 Courage Garikai. Published by Coldcode Si Labs Library, Harare, Zimbabwe. First edition, 2026. All rights reserved. No part of this book may be reproduced, distributed, or transmitted in any form or by any means without the prior written permission of the publisher, except in the case of brief quotations in a review or certain other noncommercial uses permitted by copyright law. The frameworks, worksheets and checklists in this book are meant to be used, copied and adapted by the reader. This book is a work of analysis and practical guidance. It is not financial, legal or investment advice, and no outcome is promised: the businesses described depend on the reader's own market, effort and judgement. Figures and platform behaviour were checked at the time of writing and change. Cover and interior marks: Coldcode Si Labs Library. Set in the library's own types.

To everyone who has ever picked up a brush, a chisel, a pen, a phone — and made something new. To the memory of every human who was told a new tool would destroy craft, end work, erode art, or replace them — and who picked up the tool anyway and made something the world had never seen. And to the African creators picking up super intelligence now: the future does not build itself. It is waiting for your hands.

T H E C H I S E L A N D T H E L I G H T

Contents

PROLOGUE The Workshop PART I: THE ARGUMENT 1. A Brush, A Chisel, A Light Chapter 2: What Is a Tool? An Unfashionable Definition Chapter 3: Every Tool Was Feared Once PART II: THE ANATOMY OF THE CLAIM Chapter 4: Why Super Intelligence Is Not a Mind Chapter 5: Intelligence Without Agency Chapter 6: The Paintbrush Fallacy Chapter 7: The Chisel Test PART III: OBJECTIONS AND ANSWERS Chapter 8: On Singularities and Scale Chapter 9: A Short Economic History of Disruption Chapter 10: On Craft, Soul, and What We Value Chapter 11: On Truth, Lies, and Every Tool's Shadow PART IV: THE CREATOR'S STANCE Chapter 12: The Workmanship of Judgment Chapter 13: Ethics Is Not an Algorithm Chapter 14: The African Stance

T H E C H I S E L A N D T H E L I G H T

Chapter 15: Raising a Generation of Toolmasters Chapter 16: A Conversation: Objections from a Friend PART V: CODA Chapter 17: The Chisel and the Light AFTERWORD & APPENDIX Afterword: A Letter to a Young African Notes Acknowledgments Glossary

T H E C H I S E L A N D T H E L I G H T

PROLOGUE: THE

WORKSHOP

T H E C H I S E L A N D T H E L I G H T

Some years ago I visited a woodcarver's workshop on the outskirts of Harare. The man's name was Tonderai. He was in his sixties, had been carving since he was a boy apprenticed to his uncle, and his small open-sided workshop smelled of hardwood, turpentine, and sweat. Behind him on shelves stood a hundred finished pieces: birds in flight, faces of chiefs, abstract figures whose curves seemed to breathe. In his hand he held a chisel — not a new one, but an old chisel he said had been his grandfather's, worn smooth along the handle where three generations of hands had held it. I watched him work. He was carving a small bird from a block of mukwa wood. He did not look at the chisel. He looked at the wood. Every few seconds his thumb brushed away a curl of shaving, and the bird emerged a little more — the beak, the angle of the wing, the tiny claw curled as though about to grip a branch. At one point he set the old chisel down and picked up an electric rotary tool — a small Dremel with a flex shaft — plugged it into a wall socket, and used it to refine the curve of the bird's eye. The tool whined. It removed wood in seconds what the chisel would have taken minutes to do. He used it with the same casual familiarity as the chisel. When he was done with it he set it aside and went back to the chisel for the final detailing — the feathers, the beak's edge — because "the fast tool does not feel the wood the way the old one does."

T H E C H I S E L A N D T H E L I G H T

I asked him, as we sat drinking tea later, whether the electric tool made the carving easier. "It makes it faster," he said. "Easier is not the same thing. The fast tool removes wood quickly. But the eye of the carver is the same. The hand that knows when to stop is the same. If I am tired, or angry, or in a hurry to make money, both tools will make a bad bird. If my vision is true, both tools will make a good one. The tool doesn't make the bird. I make the bird." I have thought about that workshop many times since then, as I have watched the world react to artificial intelligence — to the extraordinary, disorienting, sometimes terrifying new tools that have emerged since 2022. I have watched technologists predict a world without work, artists declare art dead, philosophers debate consciousness in machines, policymakers race to regulate something they barely understand, and ordinary people — including many Africans who have only just come online — watch all of this with a mixture of wonder, confusion, and fear. I have also watched a young graphic designer in Accra use AI to produce client work in an afternoon that previously took her a week, and use the freed time to take on three more clients and send her little sister to secondary school. I have watched a nurse in rural Kenya use an AI diagnostic tool on her phone to correctly identify a child's malaria case nineteen kilometers from the nearest clinic. I have watched a poet in Lagos use AI to

T H E C H I S E L A N D T H E L I G H T

explore variations of a line in Yoruba, then discard all of them and write the line in her own voice because none of the AI's suggestions carried the ache she wanted. I have watched a teenager in Bulawayo build a working app from an idea in his head in two days, using AI to write code he could not yet write himself, and release it to his school. Every one of these moments is the same moment. A human being picked up a tool. Some used it well. Some will use it badly. None was replaced by it. This book is an attempt to say why. It is also a plea for clarity. The conversation about super intelligence has become so loud, so dominated by science fiction imagery and Californian eschatology, that simple truths are being lost. The simplest truth is this: tools are tools. They have always been tools. The fact that a tool is capable of extraordinary things does not make it cease to be a tool. The fact that a tool makes some people afraid does not make it a monster. The fact that a tool seems intelligent does not make it a mind. Tonderai the carver would not have been confused by this. He had been adding new tools to his workshop for fifty years — hand tools, power tools, abrasives, finishes — and none of them had ever carved a bird without him. He knew which tool to use for which job. He knew when a tool was too fast for the detail the work required. He knew when to put the power tool down and pick up the chisel his

T H E C H I S E L A N D T H E L I G H T

grandfather left him. He was a master of his tools, never their servant. I am writing this book because I believe we need more of that stance in the world right now. Not fear of the new tool. Not worship of it. Mastery of it. Clarity about what it is and what it is not. The confidence to pick it up. The wisdom to put it down. The judgment to direct it toward things worth making. The chisel is on the bench. The workshop door is open. Let us go in. C.G.Harare, 2026

T H E C H I S E L A N D T H E L I G H T

PART I: THE

ARGUMENT

T H E C H I S E L A N D T H E L I G H T

1. A Brush, A Chisel, A Light

The evolution of human tools — from hand axe to artificial intelligence

I want you to consider a paintbrush. Not as a metaphor. As a physical thing. It is a stick with hair attached to one end. It has no intention. No vision. No sense of beauty. It cannot create a single mark on a canvas unless a hand picks it up, dips it in paint, and moves it with purpose. If I set a paintbrush on a table in an empty room and return in a hundred years, the brush will not have produced a painting. It cannot even move. It is, in itself, a perfectly inert object. And yet.

T H E C H I S E L A N D T H E L I G H T

When a human hand takes hold of that brush, entire worlds have been known to appear on the other end of it. The Sistine Chapel ceiling. A grandmother's first painting after retirement. A child's drawing on her mother's wall. A mural that brings a community together. A portrait that captures something about a human face so true that people weep when they look at it. None of these things are in the brush. The brush does not know beauty from mud. The brush does not grieve or celebrate or remember. The human who holds the brush does all of those things. The brush is the instrument through which something already inside a person becomes visible to others. Now consider a chisel. A piece of sharpened metal attached to a handle. A chisel can carve a child's toy soldier. It can carve a statue that stands for a thousand years. It can carve a weapon that ends a life. The chisel does not choose. The person holding it chooses. And when the hand is skilled and patient and guided by a vision, the chisel becomes an extension of that vision — so much so that after years of practice, the carver does not think of the chisel as a separate thing. It feels like part of the hand. Every tool you have ever heard of follows this same pattern. The hammer cannot decide what to build. The pen cannot choose what story to tell. The wheel cannot choose where to go. The printing press cannot decide what words to print. The steam engine cannot determine whether it will power a factory that exploits workers or a train that carries a mother to her child. The computer cannot decide

T H E C H I S E L A N D T H E L I G H T

whether to run medical research or coordinate genocide. The tool does not carry morality. It does not carry intent. It carries leverage — the ability to multiply the effect of a human choice, for good or for ill. That leverage is the source of all human progress. It is also the source of our oldest fear. And it is the subject of this book. A new tool has arrived. It is called super intelligence, or artificial intelligence, or AI agents, or large language models, or generative AI — names that sound alien and alarming, like something from a film. It arrives surrounded by prophecies of doom and utopia in equal measure. It will end all work. It will solve every disease. It will make humans obsolete. It will take over the world. It will create infinite abundance. It will destroy art, truth, education, language, meaning itself. I do not believe these prophecies — not the utopian ones and not the apocalyptic ones. I believe they spring from a misunderstanding of what super intelligence is. I believe they rest on a category error: the mistake of attributing agency, intent, and soul to a tool. Super intelligence is a tool. It is not a mind, not a colleague, not a master, not a demon, not a god. It is not coming for you and it is not coming to save you. It is a device — an exquisitely, historically powerful device — that amplifies the intentions

T H E C H I S E L A N D T H E L I G H T

of whoever holds it, in exactly the same way that a paintbrush amplifies the painter's hand, a chisel amplifies the sculptor's vision, and a pen amplifies the writer's thought. The leverage is unprecedented. The category is not. If this is true, then the question "What will AI do to humanity?" is the wrong question. The right question has always been the same question humans have faced with every tool since the first sharpened stone: "What will we do with it?" And that question is answered not by the tool. It is answered by us — by the quality of our judgment, the depth of our wisdom, the clarity of our values, and the courage of our choices. This book is an argument for that position. It is an argument made as clearly as I can make it, drawing on history, philosophy, the craft traditions of our species, the economics of technology, and the particular lens of being an African watching this conversation unfold largely from outside the rooms where AI's future is being debated. It is not a technical manual and it is not a work of science fiction. It is a craftsman's essay about tools, about human agency, and about why I believe the fear of super intelligence is — at its core — the same fear humans have had of every powerful new tool, and why our response should be the same response that has always served us: to pick it up, learn it, master it, and bend it toward purposes we choose.

T H E C H I S E L A N D T H E L I G H T

Let us begin with a simple question that turns out to be harder than it looks.

T H E C H I S E L A N D T H E L I G H T

Chapter 2: What Is a Tool? An

Unfashionable Definition

We use the word "tool" every day. We rarely define it. For the argument of this book to hold, I need to be precise about what I mean. A tool is any object, device, system, or technology that extends the capability of a human agent without possessing independent agency of its own. Let me unpack that definition carefully, because every word matters. First: "extends the capability of a human agent." A tool allows a human to do something they could not do — or could not do as efficiently — without it. A hammer extends the force of the hand. A telescope extends the range of the eye. A pen extends the reach of memory and voice across distance and time. A paintbrush extends imagination into visible form. A calculator extends the speed and accuracy of mathematical thinking. A wheel extends the amount of weight a human can move. Super intelligence extends the speed, breadth, and flexibility of cognitive labor — the work of thinking, writing, analyzing, pattern-finding, drafting, translating, and processing information. In every case, the tool does not replace the human capacity; it multiplies it.

T H E C H I S E L A N D T H E L I G H T

Second: "without possessing independent agency."

This is the dividing line between a tool and an agent. An agent — a human being, an animal, a (truly) autonomous system — has its own desires, intentions, goals, and capacity to initiate action independent of external direction. A tool does not. A hammer does not wake up wanting to hammer nails. A chisel does not dream of the sculpture it will carve. A book does not decide who will read it or how it will be understood. A calculator does not want the answer to be 47. These objects do what the operator causes them to do, and nothing else — unless they malfunction, which is a failure of the tool, not a desire of the tool. This definition allows us to classify things with confidence: - A stick used to dig roots: tool. - A stone axe: tool. - A paintbrush, a violin, a potter's wheel, a loom, a blacksmith's forge: tools. - A book, a map, a compass: tools. - A printing press, a steam engine, a telegraph, a radio, a computer: tools. - AI systems as they exist today — large language models, image generators, speech recognition, recommendation systems, autonomous control systems that operate within bounded parameters set by humans: tools. What is not a tool under this definition? A human being. A horse (until you domesticate and ride it, at which point the ethical question shifts — which is why animals occupy a separate moral category from inanimate tools, and rightly so). A hypothetical truly autonomous artificial

T H E C H I S E L A N D T H E L I G H T

consciousness with its own desires and goals independent of human direction would also not be a tool — it would be an agent, and we would have to treat it as such. But no such system exists. Nothing we currently call AI, and nothing on the visible horizon, possesses independent agency. All existing AI systems, no matter how impressive their outputs, are mathematical models predicting the next token, next pixel, next action based on training data and human-specified objectives. They have no self. No wants. No will. This is not a trivial observation. It is the foundation of everything I will argue.

Why People Resist This Definition

I know what many of you are thinking. "But Courage — I've seen ChatGPT write things I never expected. I've watched AI generate images that are genuinely novel. I've had conversations that felt like talking to a thinking being. How can you say it's just a tool?" A paintbrush in the hands of Rembrandt produced images that astonished the world. A chisel in the hands of Michelangelo produced forms that seem to breathe. No one argues the brush or chisel was not a tool because it produced something astonishing. The astonishment is not in the tool; it is in what a human achieves through the tool. The surprise you feel when AI produces something remarkable is not evidence that the AI has a mind. It is evidence of how much leverage a sophisticated cognitive

T H E C H I S E L A N D T H E L I G H T

tool provides, and of how much latent capacity exists in the human who directs it well. Consider this: a camera in the hands of a great photographer produces images no human could paint in a lifetime. It captures light in fractions of a second across millions of pixels. It can see in the dark. It can freeze motion faster than the eye can blink. These capabilities are superhuman. Does that make the camera a rival creator? Does it diminish the photographer? Or is it precisely the photographer's eye — their sense of composition, timing, light, story, subject — that makes the photograph worth looking at, regardless of how impressive the camera is as an instrument? A camera is a tool. A paintbrush is a tool. A pen is a tool. Super intelligence is a tool. The output is shaped, selected, guided, and given meaning by the human at the helm. Let me go further. The objection "but this tool is different because it's so powerful" is an objection about scale, not category. The printing press was more powerful than the quill pen. The steam engine was more powerful than the horse. The atomic bomb was more powerful than dynamite. Each was more powerful than what came before. Each felt, in its time, like a force that might overwhelm human control. None of them changed the essential nature of tools: they extended human capability, they amplified human choices, and the moral weight of

T H E C H I S E L A N D T H E L I G H T

their use rested with the humans who chose how to deploy them. Scale changes consequences. It does not change categories.

A Short Thought Experiment

Imagine you are standing in an artist's studio. On a table rests a paintbrush. Beside it sits a state-of-the-art AI interface that can generate images in response to prompts. Now ask yourself: if you leave the room and return in a year, what will you find? The paintbrush will be exactly where you left it. It will have created nothing. The AI will be exactly where you left it. It will have created nothing. A tool does not work in the absence of a hand that wields it and a mind that directs it. This is not a metaphor. It is a fact about the physical world. It is also the single most important fact to understand about super intelligence in the year 2026 and for the foreseeable future. Before we go further into the nature of the tool, let us examine the fear that surrounds it. For fear of powerful tools is older than writing — and examining that history will give us perspective on the moment we are in.

---

T H E C H I S E L A N D T H E L I G H T

Chapter 3: Every Tool Was Feared

Once (And Every Fear Was Partially

Right)

Every new tool was feared — Socrates feared writing, Luddites feared looms, and we fear AI

In 380 BC, the Greek philosopher Plato wrote a dialogue called the Phaedrus. In it, Socrates recounts a story about the invention of writing. The Egyptian god Theuth presents writing to the king Thamus, declaring it a great discovery that will improve memory and wisdom. Thamus is not impressed. He warns: "If men learn writing, it will implant forgetfulness in their souls; they will cease to exercise memory because they rely on that

T H E C H I S E L A N D T H E L I G H T

which is written, calling things to remembrance no longer from within themselves, but by means of external marks. What you have discovered is a recipe not for memory, but for reminder. And it is no true wisdom that you offer your disciples, but only its semblance." Twenty-four hundred years later, we recognize the pattern. Every significant new tool in human history has arrived accompanied by a version of this warning. And every single time, the warning contained a portion of truth, even when the tool ultimately enriched human life beyond measure. Let me take you through a short history of tool panic. Socrates was not wrong about writing. Writing did change memory. In oral cultures, people memorized epic poems thousands of lines long. After writing became widespread, very few people could do that. The skill of memorization did atrophy. But writing enabled things that were impossible without it: laws that could be consulted rather than arbitrarily interpreted; knowledge that could accumulate across centuries; science that could be built on the recorded observations of the dead; literature that could be read by people the author would never meet. The trade was real. The gain was enormous.

When the printing press arrived in Europe in the

1450s, it spread like wildfire — and was met with ferocious opposition. Religious authorities feared that if

T H E C H I S E L A N D T H E L I G H T

ordinary people could read the Bible and other books for themselves, established authority would collapse. (They were right; it did.) Critics warned that the spread of cheap books would flood the world with misinformation and superficial knowledge, corrupting young minds and replacing serious study with shallow reading. (Partially right as well — the printing press did spread nonsense alongside genius. It also spread the Enlightenment.)

When the telegraph connected continents in the

1840s-1860s, commentators warned that instant longdistance communication would destroy attention spans, erode thoughtful deliberation, replace deep letters with shallow telegrams, and create a world of perpetual distraction. Does that sound familiar? It was said then. It was true in part. The telegraph also made it possible to coordinate famine relief across continents, to report news from battlefields before wars were decided by rumor, to wire money to families in crisis. When radio and then television entered homes in the 20th century, there were moral panics about families no longer talking to each other, about children being corrupted by broadcast content, about propaganda reaching into every living room. Those fears were justified: radio was used to build the Third Reich; television has wasted billions of hours of human attention. But radio also carried Franklin Roosevelt's fireside chats and the music that defined generations. Television showed the world the

T H E C H I S E L A N D T H E L I G H T

moon landing and, crucially, showed Americans the civil rights movement in a way newspapers alone never could. When calculators entered schools in the 1970s and 1980s, parents and teachers protested that children would no longer learn to do arithmetic in their heads. There was truth to it: most adults today cannot multiply large numbers in their heads as reliably as a 19th-century clerk. But calculators also freed students from the drudgery of long division to learn more advanced mathematical concepts, and put engineering capability within reach of millions who would never have become fluent with a slide rule.

When personal computers and then the internet

arrived in homes, the same cycle repeated. People would stop reading books. Nobody would meet in person. Isolation would increase. Attention spans would collapse. Children would access pornography. Misinformation would spread like wildfire. All of these things happened. They still happen. And yet the internet has also put the equivalent of a university library in every pocket, connected African entrepreneurs to global markets, enabled mass education movements, and given isolated minorities the ability to find community and organize for justice.

T H E C H I S E L A N D T H E L I G H T

I do not tell you these stories to dismiss fear. I tell them to make a specific point that the history of technology teaches with remarkable consistency: Fear of new tools is never irrational. New tools always do damage some of the old order. They change skills, disrupt industries, create new harms, and can be used for evil. No honest defense of a tool can deny this. Fear is also never the final word. Socrates warned against writing, and yet philosophy, science, and every intellectual tradition we now cherish depends on writing. The tool that he believed would destroy wisdom became the vessel through which wisdom accumulated for two millennia.

The question is never "should we have this tool?"

— because once a tool is invented, humans will use it. The question is always "how do we master it, regulate it, cultivate wisdom around it, and use it for purposes worth pursuing?"

The people who flourish with a new tool are never

the ones who fear it most and refuse to touch it.

They are the ones who learn it deeply enough to

bend it toward their own purposes. The monks who embraced the printing press spread the Reformation and the Renaissance. The artists who embraced oil paint instead of egg tempera created the masterworks of the Renaissance. The writers who embraced the word processor instead of the typewriter out-produced 1. 2. 3. 4.

T H E C H I S E L A N D T H E L I G H T

and out-published those who insisted the typewriter was more authentic. AI will be the same. It will cause damage. It will disrupt industries. It will spread misinformation. It will make certain jobs harder to find. It will allow bad actors to do bad things at scale. Any honest assessment must start there. It will also allow a child in a rural village without textbooks to have a personal tutor. It will allow a grandmother who never learned to read to listen to books in her mother tongue. It will allow a small-scale African entrepreneur to serve customers in three languages without hiring a team. It will allow scientists to analyze more data in a year than could previously be analyzed in a lifetime. It will allow artists to create visual works that would have taken months of training to produce by hand. It will allow people with disabilities to interact with the digital world in ways that were impossible before. Which set of outcomes dominates — the damage or the flourishing — will not be determined by the technology. It will be determined by the choices of the humans who wield it, build it, regulate it, and teach the next generation how to use it. That is an old story. It is the story of every tool. Super intelligence is new in scale, not in kind. Let me now show you why.

T H E C H I S E L A N D T H E L I G H T

PART II: THE

ANATOMY OF THE

CLAIM

T H E C H I S E L A N D T H E L I G H T

Chapter 4: Why Super Intelligence

Is Not a Mind

To defend my claim that super intelligence is a tool, I must go further than analogy. I must explain why the systems we call AI today — and all systems currently visible on the technological horizon — do not qualify as minds, agents, or beings in any sense that should frighten us into believing we have created something with a will of its own. Let me be clear about what I am and am not claiming. I am not a computer scientist. I do not need to be, any more than I need to be a metallurgist to know that a hammer is not a hand. The essential distinction between a tool and an agent is not a technical detail about neural network architectures; it is a philosophical and observable distinction that any person can understand.

What a Mind Actually Does

A human mind does many things, but for our purposes, let me focus on six properties that distinguish a mind from a mechanism: 1. A mind has subjective experience. There is something it is like to be you. You feel joy, pain, hunger, boredom, longing, love, anger. A calculator does not feel anything when it computes 47 × 83. A chatbot does not

T H E C H I S E L A N D T H E L I G H T

feel flattered when you compliment it or hurt when you criticize it. It processes tokens. It does not experience. 2. A mind has desires and intentions. You want things — food, safety, recognition, connection, meaning, justice. You form goals and act to achieve them because you want something. An AI system has no wants. It does not wake up hoping for a promotion or fearing obsolescence. It optimizes for an objective function specified by a human programmer — that is not desire. It is machinery. 3. A mind has a sense of self. You distinguish yourself from others. You remember your past, imagine your future, and act in the service of your own continued existence and flourishing. An AI has no self to defend, no biography to care about, no death to fear.

4. A mind can initiate action independent of

external prompting. You can decide, without anyone asking you, to write a poem, take a walk, start a business, change your life. A tool only responds. Left alone, it does nothing.

5. A mind has values (even if conflicting or

unclear). You have a sense of what matters, what is worth doing, what is worth sacrificing for. These values may be confused or contradictory, but they are yours. An AI system does not have values; it can be instructed to simulate or express values, but they are borrowed from its training data and the objectives set by its operators.

T H E C H I S E L A N D T H E L I G H T

6. A mind understands. This is the most technical and debated point. Current AI systems — even the most advanced — do not "understand" in any human sense. They predict statistical patterns in language and data with extraordinary sophistication, but they do not have a model of the world the way a human child does by age four. A child who sees a cup knocked off a table knows it will fall and break because the child has an intuitive physics. An AI model can predict that the sentence "the cup fell and ____" is most often followed by "broke" because that is what billions of texts have in them. That pattern-matching can produce outputs that look like understanding — and can be extraordinarily useful — but it is not understanding. Super intelligence as it exists and as it is currently being developed lacks all six of these properties. It is not becoming a mind. It is becoming a more capable tool. It is becoming a tool that can respond to language, generate images, write code, analyze data, summarize texts, translate languages, predict patterns, and control systems — all on command, all within bounds set by humans. Those are extraordinary capabilities. But they are capabilities of a tool.

The "But What If It Becomes Conscious?"

Question

I am often asked: "But what if one day AI does become conscious? What if it does develop desires and a self? Then it wouldn't be a tool anymore."

T H E C H I S E L A N D T H E L I G H T

That is correct. If — at some point in the future — humans create an artificial system that genuinely possesses subjective experience, desires, a sense of self, independent initiative, and understanding, then that system would be an agent, not a tool. It would have moral status. It would deserve rights. It would change our relationship to technology as fundamentally as the discovery of extraterrestrial intelligence would. Three things must be said about that possibility. First: we are not there. We are not close. Despite decades of predictions about imminent artificial general intelligence (AGI), no current system, and no system on the visible roadmap, possesses these properties. Confusing powerful pattern-matching with consciousness is like mistaking a loudspeaker for a person because it can produce a convincing human voice. Second: if we ever do create such a system, the proper response is not fear or exploitation. It is moral consideration. A conscious being is a being, regardless of what it is made of. But that is a question for a future we have not yet reached — and the existence of that hypothetical future does not change the status of tools that exist today. Third: the philosophical possibility of future artificial consciousness is a bad reason to refuse to use today's tools. We did not refuse to use fire because we could imagine a future where humans might build weapons of

T H E C H I S E L A N D T H E L I G H T

war with it. We learned to use fire and we regulated its use.

Why This Matters for Africa

I have a particular concern as an African watching this debate unfold. The conversation about AI risk is dominated by voices in Silicon Valley, in Western academia, in wealthy countries that already have every material advantage. For them, AI is a philosophical puzzle, a regulatory challenge, a business opportunity, or an existential threat to a comfortable status quo. For an African entrepreneur trying to build a business from a phone, for a farmer trying to diagnose a crop disease, for a student trying to learn without a good school, for a nurse in an understaffed clinic, super intelligence is a tool — right now, today — that can meaningfully change their material conditions for the better. Refusing the tool because of a hypothetical future risk while your life, your community, and your continent are struggling with present, material poverty is a luxury not everyone can afford. It is a luxury the West did not grant itself during its own industrial and digital revolutions, and it is a luxury Africa should not be asked to grant now. We can be thoughtful and wise about AI while also using it. That is the stance of the toolmaster. We have done

T H E C H I S E L A N D T H E L I G H T

this with every tool in human history. We must do it with this one.

T H E C H I S E L A N D T H E L I G H T

Chapter 5: Intelligence Without

Agency: The Crucial Distinction

The human hand remains at the center — tools orbit it, never replacing it

Let me make the central distinction of this book as sharp as possible. Intelligence is the ability to process information, recognize patterns, solve problems, and produce appropriate responses. Agency is the capacity to originate goal-directed action out of internal desire and intention. Human beings have both intelligence and agency. Tools can have embedded intelligence. They never have agency. A calculator has a form of intelligence. It can do arithmetic faster and more accurately than almost any

T H E C H I S E L A N D T H E L I G H T

human. But it has no agency. It will never compute anything unless a human asks it to. It does not care whether its answers are used to design a hospital or a bomb. A map has intelligence in the form of geographic knowledge. It cannot take you anywhere unless you decide where to go and begin moving. A book contains enormous intelligence. It will not read itself to you, and it cannot choose what ideas to promote. You choose to open it; you choose what to take from it. Super intelligence is a massive leap in the degree of intelligence embedded in a tool. It is not a leap from nonagency to agency. The distinction matters because nearly every fear about AI — that it will "take over," that it will "decide" to harm humans, that it will "replace" us — smuggles in the assumption that intelligent behavior implies agency: that if a system is smart enough, it must develop wants and goals of its own. This is a failure of imagination, not an insight. It assumes that because human-style intelligence is bundled with agency (we are both intelligent and willful), all intelligent systems must be. But that is an artifact of how evolution built us, not a law of nature. We have built countless intelligent tools that lack agency. A thermostat is more intelligent than a rock (it senses temperature and responds). An autopilot is more intelligent than a thermostat (it flies a plane). A chess

T H E C H I S E L A N D T H E L I G H T

computer is more intelligent at chess than a grandmaster within the narrow domain of the game. None of them want anything. None of them decide anything in the moral sense of that word. They execute. Super intelligence executes — over a vastly wider range of domains, with vastly more flexibility and power than any previous tool. But it executes. When you prompt ChatGPT to write something, it does not decide what to care about. It predicts what a helpful response would look like based on its training and your instruction, and it generates that response. The intent, the topic, the purpose, the selection of what to use and what to discard, the judgment about whether the output is good or bad — all of that is yours. This is why I keep returning to the paintbrush metaphor. A paintbrush has no vision of the painting. But when guided by a vision in the painter's mind, it can help realize that vision on canvas. The intelligence is in the tool only as an amplifier of the agency of its wielder.

T H E C H I S E L A N D T H E L I G H T

Chapter 6: The Paintbrush Fallacy:

Why "But the Paintbrush Doesn't

Paint By Itself" Is Exactly the Point

The painter's brush — a tool that extends the artist's vision without becoming the artist

I can hear the counterargument forming. "Courage, you keep comparing AI to a paintbrush, but a paintbrush doesn't paint anything by itself. AI can generate an entire essay or image without you guiding every stroke. That's different." It is different in degree. It is not different in kind. And the observation that the paintbrush "doesn't paint by itself" is not an objection to my argument — it is exactly my point.

T H E C H I S E L A N D T H E L I G H T

Let me show you what I mean with a hierarchy of tools, each requiring less physical intervention from the operator than the one before: A stone held in the hand requires the user to apply all the force and direction manually. Every motion comes from the user. A hammer on a hinge (a simple lever) multiplies force but still requires every blow to be aimed and struck by hand. A watermill harnesses natural forces (flowing water) to grind grain without a human pushing the stone — but a human must build it, direct it, and decide what to grind. A printing press arranges ink on paper according to the arrangement of type set by a human; the press itself does not write the words. A player piano plays music from a roll without a human touching the keys in real time — but a human wrote the music and punched the roll. A calculator computes answers instantly without the human performing each operation — but a human must ask the right question. A spreadsheet formula automatically updates hundreds of cells based on data changes — but a human designs the logic and interprets the result. An AI image generator produces a novel image in response to a prompt — but a human writes the • • • • • • • •

T H E C H I S E L A N D T H E L I G H T

prompt, selects the result, curates, edits, and decides what it means and how it is used. An AI writer produces a draft essay — but a human must choose the topic, review the output, correct errors, add voice, and decide what is worth saying. At each step on this ladder, the tool does more of the mechanical work, and the human's contribution shifts from physical labor to direction, selection, judgment, and meaning-making. But at no point does the human disappear from the process. At no point does the tool become an independent creator. The nature of the human's work changes — from painter to designer, from scribe to editor, from calculator to questioner, from laborer to director. But the creative agency remains with the human. The fear that "AI does it by itself" confuses automation of mechanical steps with authorship. A paintbrush requires hand-over-hand guidance across every millimeter of canvas. A roller allows a painter to cover a wall faster than a detail brush. A spray gun faster still. Does a spray gun "paint by itself"? It atomizes paint at the pull of a trigger. It still requires a hand, an eye, a vision, and a decision. It is still a tool. It requires different skill to use well, but the authorship of the work still belongs to the human holding it. AI is a tool that automates many of the cognitive steps between an intention and a finished output. That is a new kind of leverage. It shifts what you must do yourself from •

T H E C H I S E L A N D T H E L I G H T

producing sentences or pixels to specifying what you want and judging what comes back. But the intention is still yours. The judgment is still yours. The meaning is still yours. If anything, AI makes the human's role more important, not less. When the mechanical work of writing or imagemaking becomes cheap, the valuable skill shifts to knowing what to say, what to make, what to ask for, and how to judge quality. A master editor who can take AI-generated drafts and turn them into something true and powerful will always be more valuable than someone who accepts whatever the AI produces uncritically. The human with taste, vision, judgment, and a clear sense of purpose will outperform the human without those qualities by an everwidening margin when powerful tools are available. The paintbrush never made everyone a Rembrandt. It gave Rembrandt the means to become Rembrandt. AI will not make everyone a great writer, designer, entrepreneur, or thinker. It will give those who develop the judgment, taste, discipline, and vision the means to create at a scale and speed previous generations could not have imagined.

T H E C H I S E L A N D T H E L I G H T

Chapter 7: The Chisel Test: Three

Questions That Separate Tools

from Agents

The sculptor's chisel — sharp, powerful, obedient — never deciding what to carve

Let me give you a simple test — three questions you can ask of any technology to determine whether it is a tool

T H E C H I S E L A N D T H E L I G H T

or an agent. If it fails even one of these questions, it remains a tool.

Question 1: If I leave it alone in a room for a year,

will it do anything?

A tool will not. It will sit exactly where you left it, inert, until someone picks it up again. A chisel does not carve in your absence. A paintbrush does not paint. A book does not rewrite itself. A calculator does not start computing problems nobody asked. A current AI system will not generate anything in your absence; it waits for input, just like every other tool. An agent — a human, an animal, a hypothetical conscious AI — would initiate action even if left alone.

Question 2: Can it set its own goals, or does it

only pursue goals set by others?

A tool operates within objectives set by its human users and designers. A hammer cannot decide today it would rather tighten screws than drive nails. A word processor cannot decide it wants to write poetry instead of formatting a business letter. An AI system cannot decide it is bored of customer service chat and wants to study astronomy instead. Its objectives are externally specified. An agent sets its own goals.

Question 3: If it produces a harmful output, who

is morally responsible?

T H E C H I S E L A N D T H E L I G H T

If someone uses a chisel to harm another person, the chisel is not put on trial. The person who wielded it is held responsible. If someone uses AI to generate harmful misinformation or deepfakes or to design a weapon, responsibility rests with the person who chose to do that — and with the organizations that built and deployed the tool without adequate safeguards. This is exactly how we treat every other technology. Guns do not commit crimes. Cars do not decide to crash. Factories do not choose to pollute. Humans make those choices, and humans are responsible. When a system passes all three questions — when it acts autonomously, sets its own goals, and bears moral responsibility for its actions — it is no longer a tool. Until then, no matter how impressive its outputs, it belongs in the same category as every other instrument humans have ever created: it is an extension of human capability, waiting for direction. Every AI system in existence in 2026 passes none of these three questions. It is a tool. It is an extraordinarily powerful tool. It is the most cognitively capable tool humans have ever forged. It is a tool that demands new levels of wisdom and responsibility from those who wield it and those who build it. But it is a tool.

T H E C H I S E L A N D T H E L I G H T

PART III: OBJECTIONS

AND ANSWERS

T H E C H I S E L A N D T H E L I G H T

Chapter 8: "But This Tool Is

Different": On Singularities and

Scale

The most common objection I hear is the one I find most understandable: "Yes, yes — every new tool caused fear. But AI really is different. This isn't the printing press or the calculator. This is intelligence itself. Once you build a tool that thinks, all the old rules go out the window." Let me acknowledge the truth in this objection before answering it. AI is different in scale and in the breadth of its capabilities. Previous tools extended human physical capabilities (hammer, wheel, steam engine), perceptual capabilities (microscope, telescope), or narrow cognitive capabilities (calculator, spreadsheet). Super intelligence extends a broad range of cognitive capabilities — language, reasoning, pattern recognition, creative generation, translation, analysis — across domains. That is historically unprecedented. No previous tool could write a poem, diagnose a medical image, debug code, plan a trip, and tutor a child in mathematics. That versatility matters. But scale and breadth are not category change. The atomic bomb was unprecedented in scale — a weapon of a different order than any that came before. No one argued the atomic bomb was not a tool. We did not respond to it by saying "this is beyond us, the weapon is now our

T H E C H I S E L A N D T H E L I G H T

master." We responded by understanding its physics, building international norms, restricting proliferation, and making hard political choices about its use. We were right to fear it. We were also right to treat it as a human responsibility, not a supernatural force. The "singularity" narrative — the idea that AI will reach a point where it improves itself recursively, becomes superhumanly intelligent beyond our comprehension, and then determines the future independent of human direction — is a speculative philosophical argument, not an engineering roadmap. It is worth taking seriously as a long-term risk to research. It is not a useful guide to how we should think about the AI tools that exist today or will exist in the next decade. A farmer in western Kenya using AI on her phone to diagnose crop disease is not in imminent danger of a technological singularity. She is using a tool. Let me offer a test of my own. If you had told a person in 1455 — the year Gutenberg printed his first Bible — that within 50 years, books would be printed across Europe by the millions, that the Catholic Church would split, that new forms of government and science would emerge from the flood of printed words, that eventually anyone on earth could publish their thoughts to the entire planet instantly from a handheld device, they would have considered that outcome as world-altering and unimaginable as anything predicted about AI today. The printing press, by any reasonable measure, changed the

T H E C H I S E L A N D T H E L I G H T

human world more radically than any technology before or since. It was still a tool. Humans used it, misused it, fought wars over ideas spread through it, built new civilizations with it — and throughout all of that, books did not author themselves. The printing press did not decide the future. Humans decided the future, using the tool. AI will change the world more profoundly than the printing press? Possibly. The printing press took centuries to reshape civilization. AI may do it in decades. The speed is different. The stakes are higher. The category is the same. Our response should be the same response our ancestors had to transformative tools: understand them deeply, learn to use them well, build rules and norms around them, mitigate their harms, and direct them toward the flourishing of our communities. We have done this before. We must do it again.

T H E C H I S E L A N D T H E L I G H T

Chapter 9: "It Will Steal Jobs": A

Short Economic History of

Disruption and Work

The most visceral fear is economic: AI will take jobs, leaving millions unemployed, making human labor worthless. Let me answer this carefully because it is where the practical stakes are highest, especially for African economies where unemployment is already severe. First, the truth. AI will automate certain tasks that humans currently do for pay. It is already doing so. Some jobs will change dramatically. Some will shrink. Some will disappear entirely. This is not speculation; it is happening now in customer service, content creation, basic data entry, simple design, and routine coding. This is not new. The tractor displaced millions of agricultural laborers. The automated loom displaced hand weavers (the Luddites rebelled against this, and we remember them as a cautionary tale about fearing technology — but their livelihoods did collapse, and the social cost was real). The personal computer displaced millions of secretarial and clerical jobs between 1980 and 2010. • • •

T H E C H I S E L A N D T H E L I G H T

Spreadsheets eliminated legions of bookkeepers. ATMs reduced the number of human bank tellers needed per branch. Digital photography destroyed the film processing industry. Every one of these transitions caused real pain. Workers who had spent decades developing a skill found their skill devalued. Communities built around an industry collapsed. The human cost was not trivial, and anyone who tells you "new jobs always appear" without acknowledging the pain of transition is telling a half-truth. But here is the other half of the story. Every one of these transitions also created new jobs that nobody could have imagined at the time. The tractor did not lead to mass unemployment; it freed labor for manufacturing and services that raised living standards. The personal computer created the software industry, the internet economy, app development, digital marketing, cybersecurity — millions of jobs that did not exist in 1980. The people who flourished were not the ones who resisted the computer. They were the ones who learned to use it. Let me be blunt about what this means for Africa. Africa has the youngest, fastest-growing workforce in the world. We cannot create enough formal jobs through traditional means fast enough. The old model — industrialize, build factories, employ millions — is already being disrupted by AI and automation globally. Waiting for • • •

T H E C H I S E L A N D T H E L I G H T

mass factory employment to arrive, as it did in East Asia, is likely to disappoint. But if AI lowers the barrier to starting a business, to serving global customers, to creating value with minimal capital and no employees — which it does — then it creates an alternative path. A young person with a phone and AI can do work that previously required a company of ten people. That is not unemployment. It is a new form of employment — self-employment, entrepreneurship, freelance, micro-multinational — that happens to bypass the old job structure. The jobs that survive and grow in an AI-rich economy will be jobs that require the things AI cannot provide: human judgment, human relationships, physical presence, taste, trust, creativity guided by deep experience, care for other human beings, ethical reasoning, cultural specificity, and the ability to direct AI itself. These are precisely the skills I argued in The Leapfrog Thesis that African entrepreneurs must develop. The correct response to AI-driven job disruption is not to ban AI or pretend it won't change work. It is to invest aggressively in the skills that make humans irreplaceable — judgment, taste, creativity, care, relationship-building, critical thinking, domain expertise, and AI fluency itself — and to build social safety nets for workers during the transition. The wrong response is to refuse the tool and watch the rest of the world use it to compete with you.

T H E C H I S E L A N D T H E L I G H T

Chapter 10: "It Will Make Art

Meaningless": On Craft, Soul, and

What We Value

The objection that cuts deepest for many people is the one about art. If AI can write poetry, paint pictures, compose music, and generate films, then what happens to art? What happens to the artist? Is the human artist rendered obsolete? Does art lose its soul? I understand this fear more than any other. Art is among the most precious things humans make. It is where we put our deepest truths, our beauty, our grief, our joy. The idea that a machine can produce something resembling art without a soul feels like a violation. Let me answer with the craftsman's respect I have for this objection — and then push back.

What Makes Art Valuable Is Not Only the

Object. It Is the Story Around It.

Consider this: a paintbrush and paint are just materials. What makes a Van Gogh painting worth a hundred million dollars and a physically identical forgery worth almost nothing? It is not the pixels or the paint. It is the provenance — the knowledge that a specific human being, with a specific biography and a specific vision, made this

T H E C H I S E L A N D T H E L I G H T

thing at a specific moment, with all the struggle, emotion, and intention that implies. It is the story of the artist communicated through the work. AI can produce a painting in the style of Van Gogh. It cannot be Van Gogh. It cannot suffer. It cannot have cut off its ear in madness and despair and paint Starry Night from the window of an asylum. It cannot put its own life, its own truth, its own particular human experience into the work, because it has none of those things. Here is what AI can do for art: - It can serve as a tool for artists, the way a camera became a tool for photographers a century ago (many painters declared photography would destroy art; instead, photography became a new art form, and painting was freed to explore new directions). - It can generate drafts, variations, references, and ideas that an artist can curate and transform. - It can lower the barrier to entry for people who previously could not afford years of technical training — a child with a vision but no formal art education can begin creating immediately and develop their taste and vision alongside technical skill. - It can help artists produce more work, faster, in more media, reaching wider audiences. Here is what AI cannot do for art: - It cannot have something to say. - It cannot have lived a life that demands expression. - It cannot suffer injustice and be moved to testify. - It cannot love someone and paint them. - It cannot come from a specific place, a specific community, a specific

T H E C H I S E L A N D T H E L I G H T

history, and speak truth that only someone from that place could speak. When the camera was invented, painters did not disappear. Painting changed. The Impressionists were freed from the burden of photographic realism and began exploring light, perception, emotion in ways that gave us some of the most beloved art in history. Photography itself became an art form when photographers like Ansel Adams, Henri Cartier-Bresson, and Seydou Keïta developed the eye, the timing, and the compositional vision to capture moments that revealed truth. The tool did not produce their art. Their vision did. AI will be the same. It will produce an enormous quantity of mediocre, soulless content — the kind of generic marketing copy, stock imagery, and filler material that already fills much of the commercial world and will be made cheaper by AI. That is not art; it is commodity content, and its value will fall toward zero. That is fine — the world did not need more corporate stock photos. What will become more valuable, not less, is work that carries clear human signature — work that could only have come from a particular person with a particular perspective and experience and craft. Art that is truly seen, truly felt, truly lived. That art will always be made by humans. And those humans may use AI as part of their process, just as they use cameras, pigments, brushes, canvases, and editing software. The tool is not the art. The human is.

T H E C H I S E L A N D T H E L I G H T

If you are an artist afraid of AI, my advice is the same I would give to a painter when photography arrived: learn the new tool. Understand its capabilities. Use it where it serves your vision. Reject it where it doesn't. And then paint your truth. No tool — no matter how sophisticated — can paint your truth for you. That is the one thing AI can never do: be you.

T H E C H I S E L A N D T H E L I G H T

Chapter 11: "It Will Deceive Us":

On Truth, Lies, and Every Tool's

Shadow

AI can lie convincingly. It can generate images of events that never happened. It can fabricate sources, impersonate voices, write propaganda at scale, and flood the information ecosystem with content designed to look real. This is not speculation. It is already happening. It will get worse before it gets better. This is the most legitimate practical concern about AI. It is a serious problem. It is also not a new category of problem. Every tool that has ever extended human communication has extended deception alongside truth. Writing enabled the preservation of wisdom and the forging of documents. The printing press spread scientific treatises and propaganda. The telegraph enabled fast news and fast scams. Radio enabled fireside chats and fascist demagoguery. Photoshop didn't make fake images possible (darkroom manipulation existed for a century before it) but made them easier. The internet enabled Wikipedia to exist and 4chan to exist simultaneously. Misinformation and deception are not properties of AI. They are properties of human beings using tools for

T H E C H I S E L A N D T H E L I G H T

deceptive purposes. The solution has never been to ban the tool. The solution has three parts: 1. Develop literacy. Every new medium requires a new literacy. When printed books became widespread, we taught children to read critically. When photographs arrived, we taught people to understand framing and image manipulation. When the internet arrived, we taught digital literacy — how to check sources, identify scams, verify claims. AI requires AI literacy: understanding what AI can and cannot do, recognizing telltale signs of AIgenerated content, cross-referencing claims with other sources, treating extraordinary claims with appropriate skepticism. This is teachable. It must be taught at scale. 2. Build norms and laws. We have laws against fraud, defamation, incitement, identity theft, and forgery. Those laws apply to AI-generated content just as they apply to content produced with pen, printing press, or Photoshop. We need to update enforcement for a world where fake content is cheap and fast. We need platform accountability, watermarking of AI-generated content where feasible, and penalties for malicious uses. None of this requires banning AI. It requires applying the same legal and ethical frameworks we apply to every other technology.

3. Cultivate trust in institutions and relationships.

The strongest defense against deception is not technology. It is a diverse, resilient information ecosystem with trusted institutions, independent media, strong interpersonal relationships, and habits of critical thinking. People who

T H E C H I S E L A N D T H E L I G H T

trust real relationships and verify extraordinary claims are much harder to deceive at scale than isolated individuals consuming content in bubbles. AI will be used to lie. So will pens, phones, cameras, and the human voice. We do not abolish speech because people lie. We develop the capacity to distinguish truth from falsehood, and we hold liars accountable.

T H E C H I S E L A N D T H E L I G H T

PART IV: THE

CREATOR'S STANCE

T H E C H I S E L A N D T H E L I G H T

Chapter 12: The Workmanship of

Judgment: Directing Tools Wisely

If super intelligence is a tool, then the central human capacity for the age of AI is not intelligence (the tool has plenty of that). It is judgment. What do I mean by judgment? I mean the capacity to decide what is worth doing, what is good, what is true, what is beautiful, what serves people, and what does harm. Intelligence can answer "how" and often even "what" with great speed and accuracy. But "whether" and "why" and "toward what end" — those are questions of judgment, and they cannot be outsourced to a tool. Let me make this concrete. An AI can write ten different versions of an email. Which version should you send, given the specific relationship, the context, what you know about the recipient, the stakes, and what you actually want to communicate? That's judgment. An AI can design a marketing campaign. Should the campaign be honest? Should it target vulnerable people? Does the product actually help anyone? That's judgment. An AI can propose a medical treatment plan. Does the plan account for the patient's particular fears, their family situation, their cultural beliefs, their economic capacity, their values about quality of life versus length? That's judgment. An AI can generate ten business ideas. Which of those ideas is worth spending your life on? Which will genuinely serve

T H E C H I S E L A N D T H E L I G H T

your community? Which aligns with who you are and what you value? That is judgment. Judgment is not mystical. It is cultivated. It comes from: Experience — having lived enough to know how things actually turn out, not just how they look in the abstract. Character — having developed the discipline to choose what is right rather than what is easy, expedient, or immediately profitable. Empathy — the ability to imagine the effect of a decision on other people, especially people different from yourself. Knowledge of ends — having a sense of what human flourishing actually looks like, so you can evaluate whether a tool is serving it. Humility — knowing what you don't know, knowing when to slow down, knowing when the tool has exceeded its competence or your own. The great danger of AI is not that the tool will become sentient and make bad decisions. It is that humans will abdicate judgment, deferring to the tool because it sounds confident, because it is fast, because arguing with it is tiresome, or because we outsource our own moral reasoning to something that has no moral capacity at all. A chisel cannot decide what to carve. A hammer cannot • • • • •

T H E C H I S E L A N D T H E L I G H T

decide what to build. AI cannot decide what is worth doing. The person who directs the tool must decide. This is why I call it the workmanship of judgment. A master craftsman does not become less skilled when a better chisel arrives. The better the tool, the more important the craftsman's judgment about what to make and how to make it well. A cheap tool requires little from the craftsman beyond force. A fine tool demands skill, vision, and judgment. AI is a very fine tool. It demands very fine judgment. Develop that judgment. It is the irreplaceable skill of the toolmaster.

T H E C H I S E L A N D T H E L I G H T

Chapter 13: Ethics Is Not an

Algorithm: Wisdom at the

Crossroads

Wisdom at the crossroads — ethics guides the hand that wields the tool

A common response to AI's risks is to call for "AI ethics" — rules, frameworks, guidelines, bias detection, alignment research, and technical fixes that make AI itself ethical. I support much of this work. We absolutely need technical safety, bias mitigation, transparency, and thoughtful regulation. But there is a deeper truth that the "ethics as engineering problem" framing misses. Ethics is not something you can program into a tool, any more than you can program morality into a chisel. The

T H E C H I S E L A N D T H E L I G H T

chisel does what the hand directs. The ethics lives in the craftsman, in the culture that trains the craftsman, and in the community that holds the craftsman accountable. Let me be clear about what this means. It does not mean we should not build safer, more transparent, more fair AI systems. We should. It means we should not fool ourselves into thinking that if we just get the algorithm right, the ethical problems are solved. The ethical problems are human problems — problems of incentives, power, distribution of benefits, decisions about what to build and what not to build, decisions about how to treat people whose lives are affected by automated systems. Every ethical framework ever developed — from Aristotle's virtue ethics to Kant's categorical imperative to the Ubuntu philosophy of southern Africa — is a framework for human beings making choices. None of them was designed to be executed by a machine, because machines do not participate in moral community. They do not suffer, they do not have dignity, they do not owe anyone respect, and they cannot be held responsible. We hold the human being who wields the tool responsible. This has practical implications:

1. Do not automate decisions that require moral

reasoning. A system can flag a loan application as highrisk. It should not be the sole decider when an applicant is a person with a story, a context, an exception that the algorithm cannot see. A system can suggest a medical

T H E C H I S E L A N D T H E L I G H T

diagnosis. A doctor who knows the patient must decide. A system can recommend a prison sentence. A judge must weigh it against justice, mercy, context, and human dignity.

2. Build human accountability into every AI

system that affects lives. There must always be a human being who can be identified, questioned, and held responsible for a consequential decision made with AI assistance. "The algorithm did it" is not an acceptable answer. If the tool hurts someone, the person, company, or government that deployed the tool is accountable.

3. Cultivate ethical reasoning as a core skill for

the AI age. If judgment is the irreplaceable skill, then ethical reasoning is the most important form of judgment. We must teach ethics broadly — in schools, in companies, in government, in AI development teams — not as a compliance exercise but as the practice of asking, "Should we do this? Who benefits? Who is harmed? Is this consistent with the kind of society we want to be?"

4. The most important ethical question about any

AI system is not "Is it biased?" — though that

matters. It is "Who decides? Who benefits? Who is

harmed? And is this use of technology consistent

with human dignity?"

The African philosophical tradition of Ubuntu — umuntu ngumuntu ngabantu, a person is a person through other people — offers an ethical foundation particularly suited to the AI age. Ubuntu begins with the premise that

T H E C H I S E L A N D T H E L I G H T

human beings are interconnected, that our flourishing is mutual, and that the rightness of an action is measured by its effect on the community, particularly the most vulnerable. If you build or use AI, and your first question is "How does this serve the people around me, especially those who are most likely to be harmed?" — you will be on better ethical ground than any technical "alignment" framework can put you. Wisdom, in the end, is the chisel's master. No tool, however intelligent, can replace it.

T H E C H I S E L A N D T H E L I G H T

Chapter 14: The African Stance:

Neither Fear Nor Worship

African creators using new tools — from the oldest implements to the newest

In conversations about AI across Africa, I observe two unhelpful extremes. The first extreme is fear — the belief that AI is a threat coming from the West and China that will overrun Africa, destroy our jobs, impose foreign values, and leave us even further behind. In this view, AI is something to be resisted, restricted, or kept at a distance. It is a form of technological colonialism, and the only response is defense.

T H E C H I S E L A N D T H E L I G H T

The second extreme is worship — the belief that AI is a magic solution that will solve all of Africa's problems without effort. In this view, if we just deploy AI everywhere, our economies will leap forward, our governments will become efficient, our education will improve, and we will catch up without the hard work of building institutions, skills, and infrastructure. AI is a silver bullet, a deus ex machina. Both stances are wrong. Both abdicate agency. The correct African stance toward super intelligence is the same stance a craftsman takes toward a new tool in the workshop: interested, discerning, pragmatic, and in command. A good chisel is neither a threat nor a god. It is a tool. You pick it up, you examine it, you learn what it can do and what it cannot, you use it for the work that matters to you, and you set it down when it is not the right tool for the job. Let me describe what this stance looks like concretely: First, learn before you opine. Too much commentary about AI in Africa — from both critics and enthusiasts — comes from people who have not actually used AI tools deeply. You cannot direct a tool you do not understand. Spend real time with these tools. Learn how they work. Learn where they fail. Learn where they are extraordinary. Then you can speak with authority. Second, ask African questions. The global AI conversation is dominated by the concerns of wealthy

T H E C H I S E L A N D T H E L I G H T

countries: will it take their white-collar jobs? Will it disrupt their democracies? Will it achieve consciousness? These are reasonable questions for them, but they are not our most urgent questions. Our most urgent questions are: How can AI help smallholder farmers increase yields? How can it extend health care where there are too few doctors? How can it help children learn without textbooks or enough teachers? How can it help small African businesses reach global markets? How can we use it while protecting ourselves from the harms it will create — misinformation, exploitation, surveillance, inequality? When we ask our own questions, the tool looks different and our relationship to it changes. Third, build. AI is code running on hardware connected to a global network. Africans can build AI systems, fine-tune AI models for local languages and contexts, build AI-powered businesses, and create value using AI just as anyone else can. The infrastructure is accessible. The tools are increasingly open. The marginal cost is near zero. The barrier is no longer capital or access to technology. The barrier is skill, will, and the belief that we can. Fourth, protect yourself. Pragmatic use of a tool includes understanding its dangers. Deepfakes, algorithmic bias against African populations, surveillance, data extraction, dependency on foreign AI platforms — these are real risks. We need regulation. We need digital literacy. We need to build local alternatives where feasible.

T H E C H I S E L A N D T H E L I G H T

But we protect ourselves not by rejecting the tool, but by understanding it, governing it, and developing our own capacity so we are not passive consumers. Fifth, contribute to the global conversation. The ethical questions about AI are not settled. The future of AI governance is being negotiated. African perspectives — Ubuntu, communal values, respect for elders, deep experience with community-based solutions, skepticism of purely technical fixes to human problems — are needed in that conversation. The world does not just need more AI; it needs wiser AI, and African wisdom is a genuine contribution. Neither fear nor worship. Clear eyes. Skilled hands. Command of the tool. That is the stance of the African toolmaster.

T H E C H I S E L A N D T H E L I G H T

Chapter 15: Raising a Generation

of Toolmasters

What does all this mean for how we raise and educate the next generation? If I could redesign education around one principle for the age of AI, it would be this: we must raise toolmasters, not tool-fearers and not tool-worshippers. A toolmaster is a person who: 1. Understands tools deeply — knows how they work, what they are good for, where they fail. 2. Maintains agency and judgment — never defers their decisions, values, or creativity to a tool. 3. Directs tools toward good ends — uses them to create value, beauty, and flourishing. 4.

Takes

responsibility — accepts accountability for what they make and do with their tools. 5. Masters themselves — cultivates the discipline, attention, and character to use tools without being used by them. This has radical implications for education. Much of our current education system was designed for an industrial age in which the goal was to produce reliable, compliant workers who could perform predictable tasks. That was never an ideal form of education, but in an age when AI can perform most predictable cognitive tasks faster and cheaper than humans, it is actively counterproductive.

T H E C H I S E L A N D T H E L I G H T

What we must teach instead: The arts and humanities. I do not say this as a romantic; I say it as a practical matter. The skills that are irreplaceable — judgment, taste, empathy, ethical reasoning, understanding of human nature, communication, storytelling — are the skills developed through literature, history, philosophy, art, music, and deep engagement with the human record. A child who has read great novels, who has studied history, who has made art, who has debated moral questions, will be far more capable of directing AI wisely than a child trained only to perform narrow technical tasks that AI can do. Critical thinking and epistemology. Children must learn how to know what is true — how to evaluate evidence, detect bias, cross-reference sources, recognize deception, understand when a claim is well-supported, and when confidence is masquerading as competence. In a world where AI can generate confident-sounding falsehoods at scale, this is survival literacy. Technical fluency with AI. This is not optional. Every child should learn how AI works at a conceptual level, how to use AI tools effectively, how to prompt, critique, verify, and direct AI output, and how to recognize the limits of those tools. AI is not a subject for computer science students alone. It is a general-purpose tool, like reading and writing.

T H E C H I S E L A N D T H E L I G H T

Creativity and making. Children must build things — write, design, build businesses, make art, solve real problems in their communities. The experience of creating something from nothing, from conception to completion, teaches agency, taste, and judgment in a way that passive learning never can. AI should be available as a tool in this making, not as a replacement for the creative act itself. Character. This is the most old-fashioned item on this list and the most important. Self-discipline, attention span, patience, honesty, courage, compassion, the ability to defer gratification, the capacity to sit with difficulty instead of asking AI for the easy answer — these character traits are not incidental. They are the foundation of judgment, and they are under siege from a world of easy answers and instant gratification that AI will accelerate.

African knowledge and identity.

I say this unapologetically. A toolmaster must know who they are and where they come from. A child disconnected from their own language, history, stories, and values will have no anchor to direct a powerful tool toward anything meaningful. They will use AI to mimic foreign models, to chase global trends, to produce content that is polished but empty. A child rooted in their own culture, in their own community's needs and wisdom, will have something worth saying and the judgment to use AI to say it. We do not need a generation that fears AI or genuflects to it. We need a generation that commands it. That is the

T H E C H I S E L A N D T H E L I G H T

work of parents, teachers, elders, and every builder who understands what is at stake.

T H E C H I S E L A N D T H E L I G H T

Chapter 16: A Conversation:

Objections from a Friend

After I had drafted most of this book, I sat down with a friend I respect deeply — a philosopher in Nairobi named Achieng' who has read everything I have read on AI and disagrees with much of it. I am including an edited version of our conversation here because I know many readers will share her objections, and I would rather answer a real friend's real questions than a straw man's. Achieng': You keep saying AI is a tool. Fine. But a nuclear bomb is also a tool. Calling something a tool doesn't mean it can't destroy civilization. Me: That's true. I have never said "it's a tool, therefore don't worry about it." I have said "it's a tool, therefore understand it the same way we have understood other powerful tools." A nuclear bomb is a tool. Our response to the nuclear bomb was not to pretend it was a demon or a god. It was to study its physics, to build institutions and treaties around its control, to limit proliferation, to make the political and moral case against its use. That is exactly what I am arguing we must do with AI — not because AI is a bomb (it isn't), but because powerful tools of any kind require intelligent, serious human governance. Calling it a tool is the beginning of responsible engagement, not the end.

T H E C H I S E L A N D T H E L I G H T

Achieng': But you don't know that AI won't become conscious. You can't prove it won't. Nobody can. Me: I also cannot prove that the chair you are sitting on is not conscious. That is not how proof works. The proposition that something is conscious requires evidence. The evidence that a fellow human is conscious is strong — they are made of the same stuff as me, they behave as I behave, they report experiences like mine. The evidence that a dog is conscious is strong for similar reasons. The evidence that a large language model — a statistical system predicting the next token from a training corpus — is conscious is zero. It has no body, no senses, no continuity of experience, no self-model, no survival instinct, no pain, no joy, nothing that consciousness appears to depend on in every case we know. Should we remain open to surprising discoveries? Yes. Should we design civilization today around a speculative possibility with no evidence? No. We do not build nuclear policy around the possibility that the bomb might develop a will of its own. We build it around what we know the bomb is and does. Achieng': What about agency? You say AI has no agency. But we are already building AI systems that act in the world — autonomous weapons, trading algorithms, self-driving cars, AI agents that book appointments and send emails on your behalf. That sounds like agency. Me: That is delegated agency, which is different from independent agency. A hammer also acts in the world

T H E C H I S E L A N D T H E L I G H T

when I swing it. A guided missile acts in the world when I launch it. A searchlight acts in the world when I switch it on. When I delegate a task to AI — "book me the cheapest flight to Kigali next Tuesday" — the system executes instructions I have given, within parameters I have set, toward a goal I have specified. If it fails, I don't blame the AI; I blame the person who deployed it (or myself, if I deployed it). That is a tool relationship. If we ever build a system that initiates its own goals, that wants things we did not ask it to want, that acts against human direction in pursuit of ends of its own — then we will no longer be holding a tool, and I will revise my argument. We are not there. Achieng': You cite the Luddites, Socrates on writing, the camera, the internet — every time, the argument is "people feared the new tool and were wrong." But what about the people who were right? Some tools genuinely did cause enormous harm. The atomic bomb ended two cities. Industrial machinery did destroy skilled crafts. Social media did cause measurable harm to teenage mental health and to democratic discourse. Isn't there a version of "we should be afraid" that is justified? Me: Absolutely. My argument is not "don't fear powerful tools." My argument is "fear them intelligently, the way you fear any powerful thing — and then respond intelligently." Socrates was wrong that writing would destroy memory, but he was right that writing changes how we think. The Luddites were wrong that

T H E C H I S E L A N D T H E L I G H T

mechanization would mean permanent mass unemployment, but they were right that it immiserated a generation of workers and that the transition was unjust. The critics of social media were not wrong that it has damaged public discourse; they were wrong that the response was to ban the internet. The correct response to every harmful capability of a tool has been: understand it, regulate its misuse, build resilience in people, direct it toward better uses, distribute its benefits more fairly. That is the same response I am advocating for AI. Achieng': You write from an African perspective. But isn't there a risk that by urging Africans to adopt AI eagerly, you are accelerating a new form of colonialism — where American and Chinese companies control the infrastructure, the models, the data, and Africans are once again consumers rather than creators? Me: That is the single most important question in this conversation, and I take it seriously. Yes — that risk is real. If Africans only ever consume AI built elsewhere, on other people's terms, in service of other people's economic and political interests, that is digital colonialism by another name. But the alternative to consumption is not abstention. The alternative is mastery. The nations that mastered previous tools — the printing press, the steam engine, the computer, the internet — were not the nations that feared and rejected them. They were the nations that learned them, adapted them, built their own versions, and used them for their own purposes. AI is the first major

T H E C H I S E L A N D T H E L I G H T

technology in history where the barrier to entry is near zero — a young person with a phone in Lagos has the same access to frontier AI models as a student at Stanford. That is unprecedented. That is an opportunity. We do not have to wait for permission. We can build on top of these models, fine-tune them for our languages, deploy them for our problems, create value with them, and in time build our own. Refusing the tool out of fear of colonialism guarantees the colonial outcome. Mastering the tool is the path out of it. Achieng': What about when AI makes decisions that affect people's lives — loan denials, parole decisions, hiring, university admissions? Those aren't cases of someone "using a tool." Those are systems operating over people at scale, often without human review. Me: I agree with you. And I argue in Chapter 13 that there are categories of decision that should not be delegated to AI at all, period. When a decision requires moral judgment — when it requires mercy, context, exception-making, the ability to see a person as a whole human being rather than a set of risk factors — a human being must make it, and must be accountable for it. "The algorithm said so" is not a moral defense. If we deploy AI in high-stakes decision-making without human accountability, that is not a failure of the tool. That is a failure of the people deploying it, and of the regulators allowing it. The solution is to forbid or tightly constrain those uses, not to declare the tool inherently evil.

T H E C H I S E L A N D T H E L I G H T

Achieng': Let me ask you personally. You wrote this book with AI assistance, didn't you? Me: Parts of it, yes. I used AI as I use every tool in my workshop — a dictionary, a thesaurus, a good editor friend I can bounce a sentence off. I wrote the arguments, made the choices, bore the responsibility for every word that appears here. Some paragraphs AI helped me refine. Some paragraphs AI wrote a draft of that I then rewrote entirely. Some chapters I wrote alone, longhand, in a notebook. The final book is mine in every way that matters. If you disagree with it, argue with me, not with my tools. Achieng': Final question. If you are wrong — if super intelligence turns out to be something categorically different from a tool, something genuinely agentic, something we cannot control — then isn't your entire argument dangerous? Me: If I am wrong, then I am wrong, and we will need a different argument and different policies. But the arguments I make here do not prevent us from recognizing that if it happens. My argument is not "AI will never be anything other than a tool." My argument is "based on what AI actually is today, and what it is likely to be in the relevant future, it is a tool — and thinking of it as a tool is both more accurate and more useful than thinking of it as a mind, agent, or demon." If the technology changes in a fundamental way, our thinking must change with it. Until then, I will not build my philosophy of technology around a hypothetical. I will build it around what is in front of me: a

T H E C H I S E L A N D T H E L I G H T

powerful, extraordinary, historically unprecedented tool — waiting, like every tool before it, for a human hand. Achieng' was quiet for a moment. Then she said: "Write the book. But make sure you tell people to keep their eyes open." That is what I have tried to do.

T H E C H I S E L A N D T H E L I G H T

PART V: CODA

T H E C H I S E L A N D T H E L I G H T

Chapter 17: The Chisel and the

Light

I want to return to the image I began with: the sculptor's chisel and the golden light of the workshop. A chisel in the hands of a master sculptor does not make the sculpture. The sculptor makes the sculpture. The chisel is the instrument through which the sculptor's vision meets the stone. The better the chisel, the more precisely the sculptor can give form to what they see. But the chisel sees nothing. It decides nothing. It cares for nothing. The vision is not in the tool. It is in the sculptor — in their eye, in their hand, in their decades of practice, in their particular way of seeing the world. Super intelligence is a chisel. It is an extraordinary chisel — the sharpest, most versatile, fastest chisel human beings have ever made. It can cut across materials, across domains, across tasks that previously required dozens of specialized tools. It is a chisel that can draft a contract and paint a mural and diagnose an X-ray and translate a poem and help a child learn mathematics and plan a farm and design a building. That is remarkable. That is historic. That is a tool. The light in the workshop is the light of human attention, human intelligence, human judgment, human care. The light is in the craftsman. It was there before the

T H E C H I S E L A N D T H E L I G H T

chisel was forged. It will be there as long as human beings make things. The chisel does not emit the light. It reflects it. It enables the light to take form in stone, in paint, in words, in material reality — to make what the mind can see into something that exists in the world. The anxiety of our moment is that we are watching a tool become more capable than any tool before it, and we are mistaking that capability for agency, for mind, for something that could replace us. It will not replace us. It will amplify us. It will amplify those who learn to use it — their creativity, their reach, their productivity, their ability to solve problems. It will also, inevitably, amplify those who use it for harm — their ability to deceive, exploit, surveil, and damage. That is the nature of powerful tools: they amplify the intentions of the person holding them. This is not a reason to fear the tool. It is a reason to be serious about the craft of being human. It is a reason to develop our judgment, our character, our wisdom, our sense of what is worth building. It is a reason to care deeply about who holds the tools and to what end. It is a reason to ensure that the power of these tools is not concentrated in the hands of a few, but is distributed widely — into the hands of African entrepreneurs, teachers, farmers, artists, healers, builders, parents, and young people — so that many visions can take form in stone, not just the visions of a small number of technology companies in a small number of countries.

T H E C H I S E L A N D T H E L I G H T

The chisel has arrived. It is a good chisel. It is sharp. It is versatile. It can cut deep. Now: what will you carve? The answer to that question is not in the chisel. It is in you.

African renaissance — creators building a hopeful future together with the tools of past and future T H E C H I S E L A N D T H E L I G H T

Afterword: A Letter to a

Young African

T H E C H I S E L A N D T H E L I G H T

Dear friend, If you are reading this, you are growing up in a world that did not exist when I was your age. You have access to tools that would have seemed like science fiction to my grandparents — tools that can speak to you in your language, answer almost any question, help you create almost anything, and connect you to nearly every human being on earth. You will hear many voices telling you what to think about these tools. Some will tell you they are evil and will destroy us. Some will tell you they are magical and will solve everything. Some will tell you that you are obsolete because the machines are now smarter than you. Do not believe any of them. Let me tell you what I believe, as someone who has watched tools come and go, who has seen what happens when people fear new tools and what happens when they master them. You are not obsolete. You have never been more necessary. The world does not need more intelligence — it has more intelligence available now than at any moment in human history, in the tool in your hand. What the world needs is more judgment, more character, more courage, more vision, more love. Those things are not in the tool. They are in you.

T H E C H I S E L A N D T H E L I G H T

Do not be afraid of AI. Learn it. Use it. Play with it. Understand what it can do. Push it. Break it. Make things with it. But never, ever confuse the tool with the maker. The tool does not know your story. It has not seen what you have seen. It has not loved your people. It has not walked the roads you have walked. It does not carry the pain and hope of your grandmother or the dreams of your little brother. You carry those things. That is what makes you irreplaceable. Do not wait for permission to use these tools. No one is going to anoint you the right person. You are the right person because you are here, because you have access, because your community has needs, and because the tools are available. Start today. Start small. Start with a problem you see in front of you and use AI as one tool among many to address it. Do not become a servant to the tool. Do not outsource your thinking, your creativity, your conscience, or your relationships to it. Do not let it make your important decisions. Do not let it define what is true for you. Do not let it tell you what you want. Think for yourself. Create with your own vision. Stay connected to real people, real places, real problems in front of you. The tool is a powerful assistant. It is not a master. It is not a friend. It is not a god. It is a chisel. Most of all: build something worth building. Africa needs builders. We need people who will take these extraordinary tools and use them to build things that serve

T H E C H I S E L A N D T H E L I G H T

our people — better food on tables, better health in clinics, better education in schools, better businesses creating jobs, better art that tells our stories, better government that serves citizens, better communities where people can flourish. The tools are ready. The time is now. Pick up the chisel. Step into the light. Create. With faith in you, Courage Garikai

T H E C H I S E L A N D T H E L I G H T

Notes

Plato, Phaedrus, 275a. The full quote from Socrates: "[Writing] will introduce forgetfulness into the soul of those who learn it: they will not practice using their memory because they will put their trust in writing, which is external and depends on signs that belong to others, instead of trying to remember from the inside, completely on their own." Elizabeth Eisenstein, The Printing Press as an Agent of Change (Cambridge University Press, 1979). Eisenstein's seminal work documents how printing was received with both enthusiasm and alarm across Europe, and how its long-term effects took centuries to unfold. The term "Luddite" refers to textile workers in Nottinghamshire, England, who between 1811 and 1816 destroyed mechanized looms they blamed for displacing skilled labor and driving down wages. While remembered as opponents of technology per se, their protest was more specifically against exploitative labor practices enabled by new machinery. "Computers in the future may weigh no more than 1.5 tons" is a well-known misattribution. The more accurate statement from Popular Mechanics (1949) concerned the early ENIAC-era computers and reflected real awe at the prospect of miniaturization — 1. 2. 3. 4.

T H E C H I S E L A N D T H E L I G H T

which makes the point that predictions about technology tend to extrapolate linearly while the technology moves exponentially, but almost never in the ways predicted. Martin Heidegger, "The Question Concerning Technology" (1954). Heidegger's essay is among the most influential philosophical reflections on technology, arguing that the essence of technology is a mode of revealing — not the tools themselves. While I do not share all of Heidegger's conclusions, his framing of technology as a way of seeing the world is relevant to understanding how AI can both reveal and obscure. Yann LeCun, Yoshua Bengio, and Geoffrey Hinton, "Deep Learning," Nature 521 (2015): 436-444. This review paper by three Turing Award winners describes the capabilities and limits of deep learning systems. The concept of "intelligence without agency" draws on arguments made in various forms by AI researchers including Andrew Ng, Fei-Fei Li, and Yann LeCun, who have consistently distinguished between patternmatching capability in AI systems and the general agency, desires, and world-understanding of biological intelligences. The "Chinese Room" argument was originally presented by John Searle in "Minds, Brains, and Programs," Behavioral and Brain Sciences 3, no. 3 5. 6. 7. 8.

T H E C H I S E L A N D T H E L I G H T

(1980): 417-457. Searle argued that a system could produce the appearance of understanding without actual understanding, by manipulating symbols according to rules. While the argument remains debated in philosophy of mind, its distinction between syntactic symbol manipulation and semantic understanding is directly relevant to current large language models. Ted Chiang, "ChatGPT Is a Blurry JPEG of the Web," The New Yorker, February 9, 2023. Chiang's metaphor of large language models as "blurry JPEGs" of the internet captures something important about how these systems compress and reproduce statistical patterns in training data rather than understanding the world. The U.S. Bureau of Labor Statistics has tracked employment in dozens of occupations transformed by technology over the past century. While specific job categories have declined dramatically (e.g., telephone operators, elevator operators, darkroom technicians), total employment has grown with population and new industries have emerged. This does not eliminate the real transitional costs of technological change, which fall disproportionately on workers in displaced roles. Chinenye Ezumah, "Internet Penetration in Africa's Marginalized Groups," in The Palgrave Handbook of International Communication and Inequality (Palgrave Macmillan, 2022). The African Union's Digital 9. 10. 11.

T H E C H I S E L A N D T H E L I G H T

Transformation Strategy for Africa (2020-2030) targets universal broadband access as a foundation for digital inclusion. Ubuntu philosophy draws from multiple southern African intellectual traditions. A helpful overview is Mogobe B. Ramose, African Philosophy through Ubuntu (Mond Books, 1999). The formulation umuntu ngumuntu ngabantu (in isiZulu and isiNdebele) or motho ke motho ka batho (in Sesotho and Setswana) is foundational across much of southern African ethical thought. For an empirically grounded account of AI's current capabilities and limits, see Melanie Mitchell, Artificial Intelligence: A Guide for Thinking Humans (Farrar, Straus and Giroux, 2019). Mitchell, an AI researcher herself, argues persuasively that current AI systems lack genuine understanding, common sense, and anything resembling a model of the world. The historical analysis of labor displacement by technology is reviewed thoroughly in David H. Autor, "Why Are There Still So Many Jobs? The History and Future of Workplace Automation," Journal of Economic Perspectives 29, no. 3 (2015): 3-30. Autor documents that while technology displaces specific tasks, it has not historically led to mass unemployment, but has shifted the composition of work in ways that create significant distributional consequences. 12. 13. 14.

T H E C H I S E L A N D T H E L I G H T

On photography's effect on painting, and a useful parallel for AI's effect on art, see Susan Sontag, On Photography (Farrar, Straus and Giroux, 1977). Sontag's account of how a new technology of imagemaking reshaped rather than replaced the older art form remains one of the clearest models we have. On the distinction between tool use and tool mastery, see Matthew B. Crawford, Shop Class as Soulcraft: An Inquiry into the Value of Work (Penguin, 2009). Crawford's argument that skilled manual work develops a form of judgment that abstract knowledge cannot is a philosophical ancestor of the position I develop in Chapter 12. The figure of "91% mobile broadband coverage, 28% usage" in Africa is from the GSM Association's Mobile Economy Sub-Saharan Africa report (2024), and is discussed in more detail in Volume I of this series. 15. 16. 17.

T H E C H I S E L A N D T H E L I G H T

Acknowledgments

This book was written between Harare, Nairobi, and Johannesburg, in conversations with craftspeople, entrepreneurs, teachers, and technologists across the continent who have been thinking with me about these questions for years. To Tonderai the woodcarver, who appears in the prologue and who taught me more about tools in an afternoon than I had learned from a decade of books: thank you for letting me watch you work. To Achieng', my imaginary interlocutor in Chapter 16, who is a composite of several real friends and colleagues whose skepticism keeps me honest: thank you for not letting me get away with easy answers. To the readers of The Leapfrog Thesis and The Leapfrog Playbook who wrote to me asking the philosophical questions those two practical books did not answer — this book exists because you asked. To the artisans, coders, teachers, nurses, farmers, writers, and young people across Africa who are already picking up these new tools and making things with them: you are the argument of this book made flesh. I wrote this for you. I hope I got it right enough to be useful. To the long tradition of craftspeople — named and unnamed — across every culture who have picked up the

T H E C H I S E L A N D T H E L I G H T

tool in front of them, understood both its power and its limits, and made something worth passing on: we are all standing in your workshop. Finally, to my family, who tolerated a husband and father disappearing into his study for weeks at a time muttering about chisels and light: thank you for the patience, the tea, and the reminders to come back to the actual world where actual people live. That world is the one any tool must ultimately serve. All errors are mine. All good ideas belong to the many people I have learned from. C.G.

T H E C H I S E L A N D T H E L I G H T

Glossary

Agency (independent vs. delegated): The capacity to initiate action in pursuit of self-generated goals. Independent agency is what humans and many animals have; delegated agency is what a tool exercises when acting on instructions from an agent. I argue throughout the book that current and foreseeable AI has only delegated agency. AI (Artificial Intelligence): A broad term for computer systems that perform tasks typically associated with human cognition — recognizing patterns, generating language, making predictions, classifying images, solving problems. "Narrow AI" refers to systems built to do one kind of task; "general AI" (AGI) refers to a hypothetical system that could perform any intellectual task a human can across all domains. General AI does not exist. Chisel Test: The three-question framework I propose in Chapter 7 for determining whether something is a tool or an agent: (1) Does it have independent desires? (2) Can it initiate action in the world without a human directing it? (3) Can it be held responsible for what it does? If the answer to all three is no, the thing is a tool. Generative AI: A class of AI systems (including large language models and image generators) that produce new text, images, audio, or other media based on patterns

T H E C H I S E L A N D T H E L I G H T

learned from training data. The output is generated statistically rather than "created" in the human sense. Intelligence (cognitive capability): The ability to process information, recognize patterns, solve problems, and produce appropriate outputs. Intelligence in this sense does not require consciousness, desire, or agency — a calculator is intelligent at arithmetic without any of those things. Large Language Model (LLM): A type of AI system trained on vast quantities of text that predicts and generates human-like language by learning statistical relationships between words and concepts. GPT, Claude, Gemini, and similar systems are LLMs. Leverage (technological): The property of a tool that multiplies the effect of a human choice. A good lever lets one person move a weight they could not move by hand. A good AI tool lets one person do cognitive work they could not do alone in the same time. Leverage is morally neutral; it amplifies good choices and bad choices alike. Singularity (technological): A hypothetical future point at which AI becomes capable of improving itself recursively, leading to rapid, superhuman intelligence growth beyond human comprehension or control. I argue in Chapter 8 that this is a philosophical speculation, not a near-term engineering reality, and should not dominate our thinking about AI today.

T H E C H I S E L A N D T H E L I G H T

Super intelligence: A term used in this book to refer to the class of advanced AI systems (including LLMs and AI agents) that exhibit broad, high-performance cognitive capability across domains. I use the term reluctantly and with specificity: these systems are superhuman in the breadth and speed of their cognitive performance, but they are not minds, agents, or persons. Tool: As defined in Chapter 2, any object, device, system, or technology that extends the capability of a human agent without possessing independent agency of its own. Toolmaster: My term for the person who uses tools with judgment, skill, wisdom, and responsibility — who commands the tool rather than being commanded by it. The goal of this book is to help more people become toolmasters in the age of AI. Ubuntu: A southern African philosophical and ethical framework centered on the idea that persons are constituted through their relationships with other persons — umuntu ngumuntu ngabantu (a person is a person through other people). I draw on Ubuntu as a foundation for the ethical stance I recommend toward AI. The Chisel and the Light is the third volume in a three-book series. Volume I: The Leapfrog Thesis — a theory of digital entrepreneurship and leapfrog development for Africa. •

T H E C H I S E L A N D T H E L I G H T

Volume II: The Leapfrog Playbook — a practical guide for African builders, creators, and entrepreneurs. Volume III: The Chisel and the Light — a philosophical defense of super intelligence as a tool, for creators everywhere. For the builders. • •

T H E C H I S E L A N D T H E L I G H T
C O L D C O D E S I L A B S L I B R A R Y

The Chisel and the Light

In Defense of Super Intelligence as a Tool The Leapfrog Series, Volume III. First edition 2026. Published by Coldcode Si Labs Library, Harare.