E.F. Schumacher

Foreword

By Edo Segal

The number that broke the argument was one hundred.

Not the billions powering GPU clusters. Not the trillions erased from software valuations. One hundred dollars a month. That was what it cost to give each of my twenty engineers in Trivandrum the leverage of an entire team. One hundred dollars — less than what most of us spend on streaming subscriptions we forget to cancel.

I kept staring at that number. Not because it was small. Because of what it made invisible.

When I describe the Trivandrum week in The Orange Pill, I describe the exhilaration — an engineer who had never written frontend code building a complete user interface in two days, a senior architect realizing the twenty percent of his work that remained was the part that actually mattered. I describe the twenty-fold productivity multiplier, and I mean it. The number is real. I measured it.

What I did not measure — what no dashboard I own can measure — is what the work did to the people doing it.

I noticed the gap months later. An engineer making architectural decisions with less confidence than she used to, unable to explain why. She had lost something in the acceleration, something deposited through years of struggle with systems that refused to cooperate, and she did not know it was gone until she reached for it and found the shelf empty.

E.F. Schumacher spent his career asking the question I had failed to ask: What does the production process do to the producer? Not what does it output. What does it cost the person inside it — in rest, in depth, in the capacity for the kind of slow understanding that no tool can transfer?

He was an economist who worked for the British Coal Board, advised governments across the developing world, and concluded that the dominant economic tradition measured everything about work except the thing that mattered most. His phrase — economics as if people mattered — sounds like a slogan. It is a structural argument, and it is the argument the AI discourse is missing entirely.

The technology conversation measures output. Schumacher measured the human being holding the tool. That is not a softer lens. It is a harder one, because it forces you to count costs the productivity metric was designed to ignore.

I did not encounter Schumacher's framework looking for comfort. I encountered it looking for the name of something I could feel but could not articulate — the suspicion that twenty-fold productivity and human flourishing are not the same number, and that confusing them is the most expensive mistake a builder can make.

This book applies his thinking to a moment he never saw. The tools are more powerful than anything he imagined. The question he spent his life asking has never been more urgent.

— Edo Segal ^ Opus 4.6

About E.F. Schumacher

1911-1977

E.F. Schumacher (1911–1977) was a German-British economist and philosopher best known for his 1973 book Small Is Beautiful: Economics as if People Mattered, which challenged the prevailing orthodoxy that bigger, faster, and more capital-intensive economic activity was inherently desirable. Trained at Oxford and Columbia, Schumacher served for twenty years as Chief Economic Adviser to the British National Coal Board and worked as an economic advisor to the government of Burma, where he developed the concept of "Buddhist economics" — a framework that evaluated work not only by its output but by its effect on the worker. He founded the Intermediate Technology Development Group (now Practical Action) to promote tools that were productive yet human-scaled: accessible, understandable, and governable by the people who used them. His posthumous work A Guide for the Perplexed (1977) extended his economic arguments into questions of consciousness, self-awareness, and the levels of being that distinguish human life from mechanical process. Schumacher's influence spans development economics, the environmental movement, and the philosophy of appropriate technology, and his central question — whether economic systems serve people or consume them — remains one of the most consequential challenges to growth-oriented economic thought.

Chapter 1: Economics as if People Mattered

The subtitle of the most consequential economic argument of the twentieth century was not an afterthought. It was the argument itself, compressed into six words that functioned as an indictment of an entire civilization's assumptions. Economics as if people mattered. The phrasing invited the obvious question: had economics, until that point, proceeded as if people did not matter? Schumacher's answer was unequivocal. Yes. The dominant economic tradition had constructed elaborate mathematical models of production, consumption, and distribution in which the human being appeared as a variable — an input, a unit of labor to be optimized and a unit of consumption to be stimulated. The models were technically impressive. They were also spiritually ruinous, because they measured everything about economic activity except the thing that mattered most: what the activity did to the people who performed it.

This was not a sentimental objection. Schumacher was trained as an economist at Oxford and Columbia. He worked for the British Coal Board for twenty years, advising on matters of industrial organization and resource allocation. He understood the logic of efficiency. He understood the mathematics of optimization. And he concluded, after decades of observation, that the logic and the mathematics had produced a civilization extraordinarily good at producing goods and extraordinarily bad at producing good lives. The factories worked. The workers inside them were diminished. The gross domestic product rose. The gross domestic satisfaction — if such a measure could be constructed — did not rise with it.

Now consider the metric that dominates discussion of the AI transition in 2025 and 2026: the productivity multiplier. The Orange Pill describes twenty engineers in a room in Trivandrum, India, each equipped with Claude Code at one hundred dollars per month, achieving a twenty-fold increase in output within a single week. The number is extraordinary by any economic standard. It is precisely the kind of number that the dominant tradition celebrates without reservation — more output per unit of input, which is the definition of productivity growth, which is the engine of GDP expansion, which is the measure of economic health.

Schumacher would have looked at that number and asked a different question. Not how much did they produce? but what happened to them?

The author of The Orange Pill provides the raw material for answering this question, though the author's primary concern is the productive dimension of the experience. First comes the exhilaration — the genuine creative satisfaction of building something real with a tool that removes the barrier between imagination and artifact. An engineer who had spent eight years on backend systems and had never written a line of frontend code built a complete user-facing feature in two days. A designer who had never touched backend code was implementing features end to end within two weeks. The capability expansion was real, and it produced real satisfaction. Then comes the terror — the recognition that the skills and structures on which an entire career had been built were now structurally obsolete. The senior engineer who spent two days oscillating between excitement and terror before arriving at the insight that the remaining twenty percent of his work — the judgment, the architectural instinct, the taste — was the part that actually mattered.

Then comes the third element, harder to name, and in Schumacher's framework the most significant: the inability to stop. The productive addiction. The compulsive engagement with a tool so stimulating that the builder cannot find the off switch. The author describes himself at three in the morning, recognizing the pattern, continuing to type. The exhilaration had drained away hours ago. What remained was the grinding compulsion of a person who had confused productivity with aliveness.

In the standard economic framework, all three of these experiences register identically. They register as output. The exhilarated engineer's output and the compulsive builder's output are indistinguishable in the productivity metric. Both count as production. The metric does not ask whether the production was joyful or grinding, developmental or depleting, chosen or compelled. It counts the code that shipped.

Schumacher's economics insists on counting what the standard framework ignores. The worker's satisfaction. The worker's growth. The worker's sense of purpose. The worker's capacity for the kind of reflective engagement that distinguishes a human life from a mechanical process. These are not soft additions to the economic calculation. They are, in Schumacher's view, the calculation's proper center. An economy that produces excellent goods while producing miserable workers has failed at the most fundamental level, regardless of how the conventional indicators perform.

The AI transition presents this argument with an urgency that Schumacher's original context — the industrial economy of the early 1970s — could not have anticipated. The factory worker whose diminishment Schumacher documented was diminished slowly, over years, by the repetitive performance of a single operation on an assembly line. The diminishment was gradual and therefore invisible. It accumulated beneath the surface of economic statistics that showed rising output and rising wages, and it manifested only in the worker's inner life — in the fatigue, the estrangement, the quiet erosion of the capacity for creative engagement that the work itself did nothing to develop.

The AI-augmented builder's experience is different in pace but structurally similar in kind. The builder is not performing repetitive operations. The builder is making decisions, exercising judgment, directing a process that responds with extraordinary immediacy. The engagement is creative, not mechanical. The satisfaction is genuine, not manufactured. And yet the Berkeley researchers who embedded themselves in a 200-person technology company for eight months found that AI tools did not reduce work. They intensified it. Workers took on more tasks. Boundaries between roles blurred. Work seeped into pauses — lunch breaks, elevator rides, the one-minute gaps between meetings that had previously served as moments of cognitive rest. The researchers called this "task seepage," and the phrase captures something that Schumacher's framework identifies as a structural feature of tools that exceed human scale: the tendency of powerful processes to colonize every available space, not because anyone demands it, but because the tool's availability converts possibility into compulsion.

The colonization is invisible to the productivity metric because the metric measures what happens during work and ignores what happens when work should stop. The builder who checks prompts during dinner is producing. The builder who drafts specifications in the hotel room on vacation is producing. The builder who cannot sleep because the next idea is forming is, in some sense, still producing. The metric counts all of it as output. It does not count the dinner that was not fully present, the vacation that was not truly rest, the sleep that did not come.

Schumacher argued that an economics worthy of the name must count these costs, not because they are sentimental concerns but because they are real depletions of real resources. The worker's capacity for rest is a resource. The worker's capacity for presence — for being fully available to relationships and experiences outside the workspace — is a resource. The worker's capacity for the kind of slow, unstructured reflection that produces wisdom rather than merely output is a resource. And these resources, like any resources, can be depleted. An economy that depletes them while celebrating the output they subsidize is an economy running down its capital while congratulating itself on its profits.

The parallel to environmental accounting is precise and deliberate. Schumacher was among the first economists to argue that the depletion of natural resources — forests, fisheries, fossil fuels — should be counted as a cost of production rather than a free input. The factory that pollutes a river while producing cheap goods has not actually produced cheap goods. It has produced goods whose true cost includes the destroyed fishery, the contaminated water supply, and the health consequences downstream. The cheapness is an illusion created by the refusal to count certain costs.

The same accounting error operates in the AI-augmented workplace. The twenty-fold productivity multiplier is real. But if the multiplier is achieved at the cost of the builder's capacity for rest, the builder's relationships, and the builder's inner life, then the true productivity — the output minus the hidden costs — is lower than the metric suggests. The output is visible. The depletion is not. And an economics that counts only the visible is an economics that systematically overstates the value of what it measures.

What would an economics as if people mattered look like, applied to the AI transition? It would begin by insisting that the builder's experience is as important as the builder's output. It would develop metrics for the quality of the work process, not merely the quantity of the work product. It would ask not only how much did the team ship? but did the work develop their judgment? Did it leave them more capable at the end of the week than at the beginning? Did it preserve their capacity for the relationships and the reflection that constitute a life beyond the workspace?

These questions sound impractical to ears trained by the dominant tradition. They sound like the concerns of a humanist wandering into an engineering meeting and asking everyone to share their feelings. But Schumacher would point out that the questions are impractical only within a framework that has already decided what counts and what does not. Within a framework that takes the worker's inner life seriously, they are the most practical questions available, because they address the resource whose depletion will eventually undermine the output itself. The builder who is depleted today produces less tomorrow. The builder whose relationships have deteriorated performs with the fractured attention of someone carrying unresolved personal distress. The builder who has lost the capacity for reflection makes decisions without the wisdom that reflection provides.

The dominant tradition treats these consequences as externalities — costs that fall outside the frame of the economic calculation. Schumacher's contribution was to insist that the frame is wrong, that the externalities are actually the central costs, and that an economics that ignores them is an economics that has mistaken its model for reality.

The question "Are you worth amplifying?" — posed in The Orange Pill's foreword — is, within Schumacher's framework, exactly half the question. The other half, the half that an economics as if people mattered would insist upon, is this: Is the amplification worth your life? Not your career. Not your output. Your life — the totality of your experience as a human being, including the hours you do not spend building, the relationships that do not produce revenue, the thoughts that do not optimize anything, the rest that looks like waste to the productivity metric but is actually the foundation on which all productive capacity depends.

The twenty engineers in Trivandrum may have found the answer to both halves of the question in the same week. The senior engineer's insight — that the tool had stripped away the manual labor masking what he was actually good at — suggests that the amplification served both his output and his development. He was doing more meaningful work, not merely more work. But the trajectory of the following months, documented by the Berkeley researchers in a different context, suggests that the structural pressures of the technology tend to push the balance toward depletion rather than development, not because the tool is malicious but because the tool is always available, and availability, in the absence of structures that protect non-productive time, converts possibility into compulsion.

An economics as if people mattered would build those structures. Not as an afterthought. Not as a corporate wellness program appended to an otherwise unchanged production system. As the foundation of the production system itself — the recognition that the human being is not an input to be optimized but the purpose the optimization is supposed to serve.

---

Chapter 2: The Appropriate Tool

An appropriate technology enhances human capability without overwhelming human judgment. This was the criterion Schumacher developed across decades of practical work in development economics, refined through observation of what happened when powerful tools met unprepared communities. The criterion sounds simple. Its application requires the most demanding kind of attention — attention not merely to what a tool can do, but to what a tool does to the person who uses it.

The hammer is Schumacher's prototypical appropriate tool. It extends the arm's force without requiring the worker to surrender control. The worker decides where to strike. The worker adjusts the force. The worker evaluates the result. The tool amplifies a decision the worker has already made. It does not make decisions on the worker's behalf, and it does not demand that the worker reorganize life around its requirements.

The assembly line is the prototypical inappropriate tool — not because it fails to produce, but because its production requires the worker to become a component of the machine. The worker performs a single operation, repeatedly, at a pace set by the line rather than by the worker's judgment. The worker does not decide what to produce, how to produce it, or when to rest. These decisions have been absorbed by the system. The worker contributes labor. The system contributes everything else. And the worker, reduced to a single function within a process too large to comprehend, is diminished by the arrangement regardless of how generously the diminishment is compensated.

Between these poles, Schumacher identified a spectrum of appropriateness governed by three criteria he considered non-negotiable. The technology must be cheap and accessible, not restricted to those with large capital reserves. It must be suitable for small-scale application, usable by individuals or small groups without requiring organizational bureaucracy. And it must be compatible with the human need for creativity — it must leave room for the worker's judgment, skill, and care to shape the outcome.

Claude Code, evaluated against these criteria during the work session, scores remarkably well. One hundred dollars per month places it within reach of most knowledge workers in developed economies — cheap by the standards of professional tools. It is eminently suitable for small-scale application: a single person can use it without a team, a manager, or an IT department. And the natural language interface is, by any historical measure, the most creativity-compatible computing interface ever built.

Every previous computing tool required the user to translate human intention into a format the machine could parse. The command line required a programming language. The graphical interface required the user to think in terms of files, folders, menus, and mouse clicks — metaphors imposed by the machine's architecture, not by the user's thinking. Each translation imposed a cognitive tax, consuming attention and energy that might otherwise have been devoted to the creative work itself.

Claude Code abolished this tax. The builder speaks in natural language — the language of thought, argument, and imagination. The builder describes what the thing should do, what the user should experience, what failure would look like. The tool responds not with a literal execution of instructions but with an interpretation, an inference about what the builder is actually trying to accomplish. The builder directs the work at the conceptual level, retaining creative control while the tool handles the translation into code.

By Schumacher's criterion of creativity-compatibility, this is a remarkable achievement. The builder's judgment is not subordinated to the tool's logic. The builder's thinking is not compressed into the tool's syntax. The builder remains the builder. The tool remains the tool. The relationship, within the work session, is one of service rather than subordination.

But appropriateness, in Schumacher's framework, cannot be evaluated within the work session alone. A tool exists within a life, and the life includes the hours when the tool is not in use — the hours that should be devoted to rest, relationships, reflection, and the unstructured experience that constitutes a human existence beyond the workspace. The appropriateness of a tool must be evaluated at both scales: the scale of the task and the scale of the life.

It is at the larger scale that Claude Code's appropriateness becomes genuinely problematic.

The tool is always available. It does not have office hours. It does not tire. It does not suggest a break. It responds at three in the morning with the same quality it brings at three in the afternoon. The human collaborator who might say "let's pick this up tomorrow" is absent. The natural friction of human collaboration — the colleague's unavailability, the meeting that imposes a pause, the commute that creates a boundary between work and home — is absent. The tool removes friction from the work, and the friction it removes includes friction that the worker needed.

Schumacher observed this dynamic in industrial technology long before AI arrived. He noted that technology "recognizes no self-limiting principle — in terms, for instance, of size, speed, or violence." A technology that can run faster will run faster. A technology that can produce more will produce more. A technology that can operate continuously will operate continuously. The limiting principle, if one exists, must come from outside the technology — from the worker, the community, or the institutional structures that govern the technology's use.

The concept of subsidiarity, which Schumacher borrowed from Catholic social teaching and applied to economic organization, is relevant here. Subsidiarity holds that decisions should be made at the lowest level of organization competent to make them. A decision that an individual can make should not be made by a committee. A decision a local community can make should not be made by a national government. The principle protects the autonomy of the smaller unit by preventing the larger unit from absorbing functions the smaller unit can perform itself.

Applied to AI tools, subsidiarity asks a question that the technology industry has not learned to ask: what functions should be performed by the tool, and what functions should be reserved for the builder? The current arrangement treats this as a question of capability — the tool performs whatever functions it can perform, and the builder performs whatever remains. But subsidiarity would treat it as a question of appropriateness: the tool should perform those functions that can be performed without diminishing the builder, and the builder should retain those functions whose performance is essential to the builder's development, regardless of whether the tool could handle them.

This is a radically different principle of design. It does not maximize the tool's contribution. It optimizes the relationship between the tool's contribution and the builder's growth. It asks not what the tool can do but what the tool should do, given that the builder is a human being whose development is served by certain forms of engagement and undermined by their absence.

Consider the engineer described in The Orange Pill who lost both the tedium and the ten formative minutes when Claude took over the "plumbing" of her daily work — dependency management, configuration files, the mechanical connective tissue between the components she actually cared about. She was glad to lose the tedium. She did not know she had lost the ten minutes until months later, when she noticed she was making architectural decisions with less confidence and could not explain why.

The ten minutes were the moments when something unexpected happened in the configuration — something that forced her to understand a connection between systems she had not previously seen. Those moments were rare. Maybe ten minutes in a four-hour block. But they were the moments that built her architectural intuition, the sense of how systems fit together that no documentation could teach.

A subsidiarity-based design would ask: should the tool handle configuration management? The answer, evaluated by capability alone, is obviously yes — the tool does it faster, more reliably, and without complaint. The answer, evaluated by the builder's developmental needs, is more complex. The tedium should go. But the unexpected failures that forced understanding? Those should be preserved — not because they are efficient, but because they are formative. The builder who never encounters unexpected system behavior is a builder whose judgment has been deprived of its primary nutrient.

This distinction — between functions the tool should absorb and functions the builder should retain — is the practical application of Schumacher's appropriateness criterion to AI. The distinction cannot be determined in the abstract. It must be worked out through attention to what each specific builder needs for development, which means it must be worked out by the builder, in conversation with colleagues and mentors who can see what the builder, immersed in the work, cannot see from inside.

The author of The Orange Pill describes the discipline of deleting Claude's output when the prose outpaced the thinking — spending two hours at a coffee shop with a notebook, writing by hand until finding the version of an argument that was authentically earned. Rougher. More qualified. More honest about what the author did not know. This is subsidiarity in practice: the deliberate retention of a difficult function not because the tool cannot perform it but because the performance is essential to the author's development.

The practice is inefficient by the standard metric. It means discarding output that works in favor of output that is earned. It means choosing the harder path when the easier path is available. But it is precisely this willingness to choose difficulty — to insist that certain struggles are worth preserving because of what they produce in the person who undergoes them — that distinguishes appropriate technology from merely powerful technology.

The appropriate tool for the AI age would be one designed with subsidiarity as a governing principle — a tool as sophisticated about the builder's developmental needs as it is about the builder's productive needs. Such a tool does not yet exist. What exists is a tool of extraordinary capability and no self-restraint, offered to builders who are themselves struggling with the absence of self-restraint that the tool's always-on availability produces.

The challenge is therefore structural, not individual. Schumacher would insist on this point with the quiet firmness that characterized his most important arguments. The solution to inappropriate technology is not to demand that individuals resist its effects through personal discipline. The eight-hour day was not a personal choice. It was a structural intervention that contained the tendency of industrial technology to consume the worker's entire life. The AI transition requires analogous structural interventions, and the urgency of building them increases with every month the tools operate without them.

---

Chapter 3: The Solo Builder's Contingent Sovereignty

The independent craftsman who owns a hammer and knows how to forge another is genuinely sovereign. The craftsman's capability does not depend on any external institution's continued goodwill. If the supplier raises prices, the craftsman forges a new tool. If the supplier disappears, the craftsman's work continues. The means of production are owned, understood, and reproducible by the person who uses them. This is what genuine economic sovereignty looks like: the capacity to produce without dependence on institutions whose decisions the producer cannot influence.

Schumacher spent considerable energy defending this form of sovereignty against the pressures of industrial concentration. The small-scale producer — the artisan, the family farmer, the independent shopkeeper — was not merely a romantic figure in his analysis. The small-scale producer was the structural foundation of an economy organized at human scale, an economy in which the worker owned the work, directed it according to personal judgment, and bore the consequences in the form of a product that carried the maker's signature.

The industrial economy destroyed this figure through the logic of scale. Larger enterprises produced more goods at lower cost. The economies of scale made the small-scale producer unviable in sector after sector. The craftsman could not compete with the factory. The family farmer could not compete with industrial agriculture. The destruction was incremental, extending across decades, accompanied at every stage by the assurance that the displaced workers would find employment in the new, larger enterprises. The assurance was partly true and wholly insufficient, because what was lost was not merely employment but the specific character of the work — the wholeness, the creative control, the integration of judgment and execution that made the craftsman's work developmental rather than merely productive.

The solo AI builder appears to recover what the industrial economy destroyed. A single person, equipped with Claude Code, produces whole work: conceiving the product, directing its creation, evaluating the result. No hierarchy. No committee to dilute the vision. The imagination-to-artifact ratio compressed to the width of a conversation. This is, on its surface, the small-is-beautiful ideal in its most concentrated form.

But the solo builder's sovereignty is contingent in a way the traditional craftsman's was not, and the contingency introduces a structural vulnerability that the experience of sovereignty conceals.

The builder depends entirely on a tool controlled by a corporation. Anthropic builds the model, sets the pricing, determines the terms of service, and decides what capabilities the tool provides and what it withholds. If Anthropic changes its pricing, the builder's cost structure changes overnight. If Anthropic changes its terms of service, the builder's workflow changes with it. If Anthropic discontinues the product or degrades its capabilities, the builder's productive capacity contracts in ways the builder cannot remedy. The builder cannot forge a new Claude Code the way the craftsman could forge a new hammer. The tool requires computational infrastructure, training data, and technical expertise that no individual builder possesses.

The historical parallel that illuminates this dynamic most precisely is the relationship between the tenant farmer and the landlord. The tenant farmer was productive, skilled, and exercised genuine judgment in the management of the land — choosing what to plant, how to cultivate, when to harvest. The tenant farmer's daily experience was the experience of independence. But the tenant farmer did not own the land. The landlord owned the land. And the landlord's decisions about rent, tenure, and the terms of the lease shaped every condition under which the tenant farmer could exercise that apparent independence.

The farmer was sovereign within the dependency. Independent within the vulnerability. And the tension between those two realities was the structural condition of the farmer's economic life — a tension invisible on productive days and catastrophic when the landlord decided to sell.

The solo AI builder is the tenant farmer of the knowledge economy. The parallel extends beyond structure to the specific quality of the dependence. The tenant farmer's vulnerability was not primarily to malice. Most landlords were not actively hostile to their tenants. The vulnerability was to the landlord's other interests — interests that might be perfectly legitimate from the landlord's perspective but devastating from the tenant's. The landlord who sells the land to a developer is not being cruel. The landlord is making a rational economic decision. The tenant who loses the farm is collateral damage of a decision made in a framework that does not include the tenant's flourishing as a variable.

Anthropic's decisions about Claude Code are made within a corporate framework that includes many considerations — revenue targets, competitive positioning, safety concerns, regulatory compliance, investor expectations — and the individual builder's flourishing is, at best, one consideration among many. This is not a moral failing. It is a structural feature of the relationship between a corporation and its users. The corporation's obligations run to its shareholders, its employees, and its institutional mission. The individual user is a customer, not a stakeholder in governance.

The contingency is invisible during normal operations. As long as the tool is available, affordable, and reliable, the builder's experience of sovereignty is indistinguishable from genuine sovereignty. The builder directs the work. The builder exercises judgment. The builder produces whole products reflecting personal care and taste. The contingency surfaces only when conditions change — when the pricing shifts, when the terms tighten, when the capability degrades — and by then the builder's dependence is deep enough that alternatives are not easily found.

Berry and Stockman, in their 2024 paper applying Schumacher's framework to generative AI, identified this dynamic as one of the central structural challenges of the AI transition. They noted that Schumacher's emphasis on scale and the ownership of productive tools provides "a powerful way to consider alternatives to the gigantisms of the FAANG and Silicon Valley-style ideologies of digital transformation." The concentration of AI capability in a small number of corporations reproduces, at the level of the knowledge economy, the concentration of industrial capability that Schumacher spent his career opposing — not because the corporations are malicious, but because concentration creates structural dependencies that undermine the sovereignty of the people who depend on the concentrated resource.

The structural remedies that Schumacher's framework suggests are specific and demanding. Open-source models that the builder can run independently, on hardware the builder owns, reduce the dependence on any single provider. Berry and Stockman point toward open-source AI as a candidate for what they call "intermediate artificial intelligence" — the AI equivalent of Schumacher's intermediate technology, accessible, explainable, and subject to the user's understanding and control. The open-source AI movement — Llama, Mistral, and their successors — represents the most Schumacherian development in the current landscape: tools that move toward genuine ownership rather than rented access.

But the open-source alternative carries its own tensions. The most capable models require computational resources that exceed individual capacity. The gap between open-source models and frontier models is a gap in capability that builders may not accept. And the development of frontier models requires concentrated investment that only large institutions can finance. The remedies reduce dependence without eliminating it. They improve the builder's position without resolving the structural tension between small-scale production and large-scale infrastructure.

The author of The Orange Pill describes a boardroom conversation that captures this tension precisely. The twenty-fold productivity number is on the table. If five people can do the work of one hundred, why not reduce to five? The arithmetic is clean. The author chose to keep and grow the team — converting the productivity gain into expanded ambition rather than reduced headcount. But the choice was the author's to make, and it was made against structural incentives that reward efficiency over investment in human capacity.

Schumacher would note that the author's choice, however admirable, is not a structural remedy. It is a personal decision made by a leader who happens to value human development over quarterly margin. The next leader, facing the same arithmetic, may decide differently. The workers whose livelihoods depend on that decision have no structural mechanism for influencing it. Their sovereignty — like the solo builder's, like the tenant farmer's — is real within conditions they do not control.

Tenant farmers eventually organized. They formed cooperatives. They lobbied for legislative protections. They developed alternative arrangements — land reform, agricultural cooperatives, publicly supported extension services — that reduced dependence on individual landlords and gave them collective influence over the conditions of their production. The process took generations and was contested at every stage. But the structures were eventually built, and they transformed contingent sovereignty into something approaching genuine independence.

AI builders have not yet organized. They have not formed cooperatives or developed political infrastructure for collective influence over the tools they depend on. They remain in the position of tenant farmers before the cooperative movement: individually productive, collectively powerless, structurally dependent on institutions they cannot influence. The structural task of the present moment is to begin building the cooperative and regulatory frameworks that would transform the solo builder's contingent sovereignty into something more durable — genuine ownership of the conditions of production, or at minimum, genuine influence over the institutions that control them.

---

Chapter 4: Buddhist Economics and the Quality of Work

In the Western economic tradition, labor is a cost. The assumption is so deeply embedded that it functions like grammar — invisible to the speaker, shaping every sentence. The employer seeks to minimize the cost of labor. The worker seeks to maximize compensation for labor. Both parties treat work as a disutility: a necessary evil that exists only because the output it produces and the income it generates are desired. The work itself has no value. It is a transaction cost that both parties would prefer to eliminate.

Buddhist economics, which Schumacher encountered during his time as an advisor to the government of Burma in the 1950s and developed into one of his most distinctive contributions, inverts this assumption completely. Work is not a cost. It is a gift — an opportunity for the worker to develop faculties, to contribute to the community, and to produce goods that are genuinely useful. The evaluation of work is therefore bilateral: the product is evaluated by its quality and service to the community, and the process is evaluated by its effect on the person who performs it. Work that produces excellent goods while diminishing the worker is bad work, regardless of its output. Work that develops the worker while producing useful goods is good work, regardless of how efficiently the production is organized.

This bilateral evaluation produces judgments that are radically different from those of the standard framework. A factory that produces goods efficiently while reducing its workers to repetitive machine-tenders is, in Buddhist economics, a failure — even if the goods are excellent and the profits are large. The failure is in the human dimension: the workers' faculties have not been developed. Their consciousness has been narrowed rather than expanded. They leave the factory each day less capable of the reflective, creative, socially engaged life that constitutes human flourishing.

The AI transition demands evaluation by this bilateral standard, and the evaluation produces a picture of genuine complexity — not the comfortable complexity of "it depends," but the demanding complexity of a technology that simultaneously provides and threatens the conditions that good work requires.

Begin with what AI-augmented work provides. The productive dimension is extraordinary. Builders produce working software in hours rather than months. An engineer with no frontend experience builds a complete user interface in two days. A designer with no backend knowledge implements features end to end within two weeks. The products work. They serve real users. They solve real problems. By the conventional standard of output, the technology is an unqualified success.

But Buddhist economics evaluates the process as well as the product, and the process is where the complexity demands the most careful attention.

AI-augmented work develops certain faculties with genuine power. The first is the faculty of creative direction — judgment about what to build, taste about how it should work, strategic vision about what problems deserve solving. The builder who works with Claude Code is not performing repetitive operations. The builder is conceiving, directing, evaluating — engaging with work at a level that demands the full exercise of creative capacity. This is precisely what Schumacher advocated: work that develops the worker's highest faculties through genuine creative engagement.

The second faculty AI-augmented work develops is integrative thinking — the capacity to work across domains that were previously separated by the translation cost of specialized skills. The engineer who starts building user interfaces, the designer who starts writing backend logic, the product leader who prototypes an idea directly rather than writing a specification for someone else to interpret — all of these represent the development of a broader, more integrated form of professional capability. The boundaries between disciplines, which had seemed structural, turned out to be artifacts of the translation cost, and when the cost disappeared, people moved across boundaries they had previously treated as walls.

But the bilateral evaluation requires equal attention to the faculties that AI-augmented work threatens, and these are at least as significant.

The first threatened faculty is what might be called embodied understanding — the deep, physical knowledge that comes from performing a task manually over thousands of hours. The senior engineer's architectural intuition, built layer by layer through decades of debugging, represents a form of understanding that is deposited through friction, through the specific resistance of a system that did not do what the engineer expected. Each failure taught something that no documentation could convey. Each unexpected behavior revealed a connection between components that the engineer had not previously seen. The understanding accumulated slowly, through struggle, and the struggle was not an obstacle to the understanding. It was the mechanism of the understanding.

Claude Code removes this friction. The code arrives working. The engineer moves on. The geological process — each hour of debugging depositing a thin layer of understanding, the layers accumulating over years into something solid enough to stand on — is interrupted. The surface looks the same. The knowledge has been transferred, not earned. And the difference between transferred knowledge and earned knowledge, invisible in the short term, manifests over months and years as a progressive thinning of the foundation on which judgment rests.

The second threatened faculty is the capacity for sustained attention in conditions of difficulty. The human mind develops its capacity for concentration not through ease but through the specific experience of staying with a problem that resists solution. The developer who spent hours debugging a function did not merely fix the bug. The developer trained the capacity for sustained engagement with difficulty — a capacity that transfers far beyond the specific domain of debugging, into every area of life that requires patience, persistence, and the willingness to remain present when the work does not yield.

AI tools threaten this capacity because they make difficulty optional. The builder who encounters a hard problem can route around it — describing the problem to the tool and receiving a solution without undergoing the experience of solving it. The routing is rational. It is efficient. And it deprives the builder of the specific exercise through which the capacity for sustained difficult engagement is developed.

The third threatened faculty is the capacity for self-regulation — the ability to manage one's own engagement with work, to recognize when productive effort has crossed into compulsive production, and to stop. This capacity is not threatened by the content of AI-augmented work but by its rhythm. The tool is always available. It responds instantly. It sustains engagement indefinitely. The natural pauses that characterize human collaboration — the colleague who goes home, the meeting that imposes a break, the commute that creates a boundary — are absent. The builder's capacity for self-regulation bears the entire weight of containment, and the weight exceeds what most individuals can sustain.

The distinction that Buddhist economics requires — between work that develops and work that depletes — maps onto a distinction the author of The Orange Pill identifies in personal experience. When the author is in flow, the questions being asked are generative: What if we tried this? What would happen if we connected that? The work expands outward, into new territory. When the author is in compulsion, the questions are responsive: How do I clear this queue? How do I finish this task? The work contracts inward, toward grinding completion.

Generative engagement develops faculties. It pushes the builder into uncertainty, where judgment must be exercised without precedent. It expands understanding of what is possible. It produces the specific satisfaction — deep, energizing, self-renewing — that Csikszentmihalyi documented as the hallmark of the flow state.

Responsive engagement depletes faculties. It exercises existing capabilities without expanding them. It narrows attention to the task at hand. It produces the specific fatigue — flat, grey, resistant to rest — that the Berkeley researchers documented in workers whose AI-augmented days had become continuous streams of productive activity with no genuine pause.

Buddhist economics asks: what is the net effect? Not on the output, which is impressive in both modes, but on the builder? Does the builder end the day more capable — more perceptive, more wise, more available to the full range of human experience — or less?

The honest answer, available to anyone willing to examine their own experience with sufficient care, is that AI-augmented work produces both effects, often in the same day, sometimes in the same hour. The morning's generative engagement slides into the afternoon's responsive grind. The flow state that develops judgment gives way to the compulsive state that depletes it. The transitions are subtle, difficult to detect from inside, and the tool provides no signal that the shift has occurred.

Schumacher was drawn to Buddhist economics precisely because it took the internal state of the worker as seriously as the external product of the work. The dominant economic tradition had no category for the worker's inner experience. Output was measurable. Inner experience was not. What could not be measured did not count. Buddhist economics insisted that what could not be measured was precisely what mattered most — that the worker's development or diminishment was the true outcome of economic activity, and that the product was valuable only to the extent that its production served the producer's growth.

In A Guide for the Perplexed, published in 1977, the year of his death, Schumacher elaborated an ontological framework that deepens this point. He mapped four levels of being: mineral, plant, animal, and human. Each level possesses capacities the lower levels lack. What distinguishes the human level is self-awareness — the capacity not merely to think but to be aware of thinking, not merely to experience but to reflect on experience. Self-awareness, Schumacher argued, "has nothing mechanical or automatic about it." It must be developed and realized by each individual. It is, in his words, "a limitless potentiality rather than an actuality."

This framework has direct implications for the evaluation of AI-augmented work. A tool that enhances the builder's output while leaving the builder's self-awareness undeveloped — or worse, while eroding the conditions under which self-awareness develops — has served the lower levels of being at the expense of the highest. The builder has produced more. The builder has not grown. And in Schumacher's hierarchy, growth in self-awareness is the distinctively human achievement, the one capacity that no increase in output can substitute for.

The practical application of Buddhist economics to the AI transition is not a call for less work or simpler tools. It is a call for a different kind of attention to the work — attention that monitors not only what is being produced but what the production is doing to the producer. The builder who develops this attention — who can recognize, in real time, when generative engagement has shifted to responsive depletion — possesses the one capacity that makes the difference between good work and its counterfeit.

The tool will not develop this attention for the builder. The tool does not know the difference between flow and compulsion. The attention must come from the builder, supported by structures — communities, norms, institutional practices — that value the builder's development as highly as the builder's output. Without these structures, the bilateral evaluation collapses back into the conventional, single-entry accounting that counts only the product and ignores the cost to the producer.

Schumacher proposed Buddhist economics not as an exotic alternative but as a corrective to an error so fundamental that the entire edifice of modern economic thought rested on it: the error of treating work as a cost rather than an opportunity, of measuring output while ignoring experience, of celebrating production while overlooking the producer. The corrective is more urgent now than when he proposed it, because the tools have become powerful enough to amplify both the output and the cost simultaneously — to produce more while consuming more, to create extraordinary products while depleting the extraordinary people who create them.

Chapter 5: The Scale Problem

Any intelligent fool can make things bigger, more complex, and more violent. It takes a touch of genius — and a lot of courage — to move in the opposite direction. Schumacher wrote this in 1973, when the biggest corporations employed hundreds of thousands of workers and the biggest computers filled rooms. He could not have imagined a technology whose training required the energy output of a small city, whose development consumed billions of dollars of capital, and whose deployment reached hundreds of millions of users within months — all in the service of enabling a single person to build software at a kitchen table.

The AI transition presents a scale paradox that Schumacher's framework identifies with precision but cannot resolve through the categories he developed. The paradox is this: the technology that enables individual-scale production is itself the product of the largest concentration of capital, computation, and data in the history of technology. The small depends on the enormous. The human-scaled depends on the inhuman-scaled. The builder's intimate, conversational relationship with Claude Code rests on a foundation of GPU clusters, petabytes of training data, and institutional infrastructure whose scale dwarfs the industrial factories Schumacher spent his career criticizing.

Schumacher argued that gigantism — the tendency of economic organizations to grow beyond the scale at which individual human beings can comprehend or influence them — was not merely an inefficiency. It was a moral failure. Large organizations reduce the individual to a function. The factory worker who performs a single operation on a line stretching beyond sight has been absorbed into a process whose totality is invisible. The executive making decisions that affect thousands of workers in dozens of countries has no contact with the people whose lives those decisions reshape. Scale creates distance, and distance creates a specific form of irresponsibility — not the irresponsibility of malice, but the irresponsibility of structures that make it impossible to see what decisions do to the people they affect.

The AI platform is a different kind of bigness than the factory, and the difference matters. The factory was big in a way that required the worker to be small — to perform a single operation, repeatedly, within a process the worker could not see or direct. The AI platform is big in a way that enables the worker to be whole — to conceive, direct, and evaluate entire products, exercising the full range of creative faculties. The factory subordinated the worker to the machine. The platform empowers the worker through the machine. The direction of the relationship has reversed.

But the reversal does not eliminate the scale problem. It relocates it. The scale problem in the factory was visible: the worker could see the assembly line stretching away, could feel the subordination in every repetitive motion, could identify the foreman as the agent of the system's demands. The scale problem in the AI platform is invisible: the builder experiences creative empowerment without seeing the infrastructure that makes it possible, without knowing the decisions being made about that infrastructure, without having any mechanism for influencing those decisions.

The infrastructure decisions are consequential. How much computational power is allocated to inference versus training. What data is included in or excluded from the training set. What safety constraints are imposed on the model's outputs. What pricing structure governs access. What terms of service limit use. Each of these decisions shapes the conditions under which every builder works, and none of these decisions are made by builders. They are made by the small number of people who control the infrastructure — executives, researchers, investors, and regulators whose frameworks for decision-making include many considerations, of which the individual builder's experience is, at best, one among dozens.

This is the structural expression of the scale problem: the builder is empowered at the point of use and powerless at the point of governance. The builder decides what to build. The corporation decides what tools the builder has to build with, under what terms, at what cost, with what limitations. The builder's creative sovereignty is genuine and circumscribed simultaneously, and the circumscription is invisible because the builder interacts with the tool's surface, not its infrastructure.

Schumacher would not find this arrangement novel. He observed the same dynamic in the industrial economy, where the worker's daily experience of employment — the specific tasks, the immediate relationships, the rhythm of the workday — was shaped by decisions made at an organizational level the worker could not see. The factory worker who experienced Monday as a sequence of specific operations did not see the quarterly planning meeting where the production target was set, the board meeting where the cost-reduction strategy was approved, the trade negotiation where the tariff change altered the competitive landscape. Each of these invisible decisions shaped Monday's experience as surely as the foreman's instructions, and the worker had no mechanism for participating in any of them.

The AI builder's situation is structurally identical. The builder who experiences Tuesday as a productive conversation with Claude Code does not see the pricing committee meeting, the safety review that constrained certain outputs, the competitive analysis that prioritized certain capabilities over others, or the investor presentation that shaped the company's resource allocation. Each of these invisible decisions shapes Tuesday's experience, and the builder has no mechanism for participating in any of them.

The question Schumacher's framework asks of any economic arrangement is whether the system serves the small or the small serves the system. In the AI transition, the answer is genuinely both — and the simultaneity is what makes the situation novel. The builder uses the system to serve personal purposes, building products, solving problems, creating value directed by personal judgment. The system uses the builder — collecting interaction data that improves the model, generating subscription revenue that funds operations, producing demonstrations of capability that attract additional users. The relationship is mutual. It is also asymmetric, because the builder needs the system more than the system needs any individual builder.

The asymmetry matters because it determines the direction of adjustment when interests conflict. When the builder's interests and the system's interests align — when the builder wants capable tools and the system profits from providing them — the arrangement works smoothly. When they diverge — when the system's competitive interests lead to changes that reduce the builder's capability, or when the system's safety concerns constrain outputs the builder considers essential — the adjustment runs in the system's direction. The builder adapts. The system does not.

George McRobie, Schumacher's colleague and the co-founder of the Intermediate Technology Development Group, observed that "there is nothing self-regulating about technology, which is still driven by notions of limitless growth, labor-saving, and expansion of consumption." The observation applies with particular force to AI infrastructure. The AI companies are driven by competitive dynamics that push toward larger models, more data, more computation, more capability — and the cost of this expansion is borne not only by the companies' balance sheets but by the energy systems, the labor markets, and the governance structures of every society the technology touches. The expansion has no self-limiting principle. The limits, if they exist, must be imposed from outside.

The scale problem is not, at its root, a problem of technology. It is a problem of governance. Who decides how the infrastructure operates? Who decides what trade-offs are acceptable? Who decides when the expansion should slow, or redirect, or stop? In the industrial economy, these questions were eventually answered — imperfectly, after enormous human cost — through the development of labor law, environmental regulation, antitrust enforcement, and democratic governance of economic institutions. The answers took generations to develop, and the generations that preceded them paid the cost of unregulated scale in broken bodies, destroyed communities, and degraded environments.

The AI transition is moving faster than any previous technological transition, and the governance structures are moving slower. The EU AI Act, the American executive orders, the emerging frameworks in Singapore and Japan — these are real structures addressing the supply side of the problem, what AI companies may and may not build. The demand side — what builders, workers, students, and citizens need to navigate the transition wisely — remains largely unaddressed.

Schumacher would argue that addressing the demand side requires structures that give the people who depend on AI tools genuine influence over the conditions under which those tools operate. Not merely the right to use or refuse the tools. The right to participate in the decisions that shape them. Cooperative ownership structures. User governance boards. Regulatory frameworks that treat the builder not merely as a customer but as a stakeholder whose flourishing is a legitimate constraint on the system's operation.

These structures would not eliminate the bigness. The computational infrastructure can be large. Schumacher's argument was never that all organizations must be small. His argument was that the relationship between the large and the small must be structured so that the large serves the small rather than absorbing it. The infrastructure can be enormous. The governance of the infrastructure must include the people whose lives it shapes.

Without such governance, the scale problem resolves itself in the direction it has always resolved itself: in favor of the large. The corporations that control the infrastructure make the decisions. The builders who depend on the infrastructure adapt. The adaptation is invisible because it is experienced as choice — the builder chooses to accept the new terms, to work within the new constraints, to pay the new price. But the choice is structured by an asymmetry of power that makes the alternative to acceptance not a different choice but an exit from the productive capability the tool provides.

The AI builder in 2026 is building at human scale with tools of inhuman scale, experiencing creative sovereignty within structural dependence, directing work with personal judgment while the conditions of that judgment are set by institutions operating at a distance and a scale that no individual can comprehend. This is the scale problem, and its resolution will determine whether the small-is-beautiful promise of AI-augmented building is a genuine recovery of human-scale production or a new form of the subordination that wears the costume of empowerment.

---

Chapter 6: Intermediate Technology for the AI Age

Schumacher developed the concept of intermediate technology during his years as an economic advisor in the developing world, where he confronted a pattern that convinced him the dominant model of development was misconceived at its foundation. Developing nations were offered two options. They could continue with traditional methods — labor-intensive, low in productivity, incapable of meeting the material needs of growing populations. Or they could adopt modern industrial technology — capital-intensive, designed for the conditions of wealthy nations, requiring infrastructure and expertise that the developing nation had not built.

The gap between the two options was enormous, and the attempt to cross it in a single leap produced dependence rather than development. The industrial technology required foreign capital, foreign expertise, foreign machinery. The result was a form of modernization that looked like progress in the aggregate statistics — rising GDP, increasing industrial output — while leaving the population structurally dependent on institutions they could not influence and technologies they could not understand, maintain, or reproduce.

Intermediate technology was the alternative. Tools more productive than traditional methods but less capital-intensive and more accessible than industrial technology. Tools the people who used them could own, understand, maintain, and repair. Tools that bridged the gap between the traditional and the modern through a series of manageable steps, each building capability without creating dependence. A hand loom is intermediate technology. A bicycle is intermediate technology. A small-scale irrigation system is intermediate technology. In each case, the tool is more productive than what came before but simple enough that the user controls it rather than being controlled by it.

The concept was not a rejection of modernity. Schumacher did not propose that developing nations remain pre-industrial. He proposed that the path from the traditional to the modern should be traversed through steps that preserved human agency at every stage — steps sized to the community's capacity, each one expanding capability while remaining within the community's ability to understand and govern.

AI tools function as a new form of intermediate technology for knowledge work, and the parallel is more precise than it initially appears. Before AI, building a software product required either a team — with its attendant organizational infrastructure, capital requirements, and coordination overhead — or years of specialized training in multiple programming languages, frameworks, and deployment systems. The gap between individual capability and team-scale output was the knowledge economy's equivalent of the gap between traditional methods and industrial technology. The individual who wanted to build faced a choice that mirrored the one Schumacher identified in the developing world: accept individual limitations, or join a large organization and surrender the independence that individual production afforded.

Claude Code bridges this gap. A single person produces what previously required a team. The tool is more productive than individual methods but less capital-intensive than assembling a development organization. It enhances individual capability without requiring the individual to surrender creative control. In these structural respects, it functions precisely as intermediate technology — bridging the gap without the institutional overhead that the gap previously demanded.

The developer in Lagos whom The Orange Pill invokes represents this bridging at its most consequential. Before AI coding tools, the developer had ideas and intelligence but lacked the infrastructure — the team, the capital, the institutional support — that turns a talented individual into a shipped product. The tool lowered the floor of who gets to build. The expansion of productive capacity to people previously excluded by lack of resources, not lack of ability, is the democratization that Schumacher spent his career advocating.

But the parallel with Schumacher's intermediate technology also exposes the ways in which AI tools diverge from the concept, and the divergences carry the weight of the argument.

Schumacher's intermediate technology was designed to be owned and understood by its users. The hand loom can be built by the weaver. The bicycle can be repaired by the rider. The small-scale irrigation system can be maintained by the community it serves. Ownership and understanding are essential to the concept — not as incidental features but as constitutive requirements. Without them, the technology does not empower. It creates a new form of dependence, replacing dependence on traditional limitations with dependence on the institution that provides the tool.

AI tools are not owned by the people who use them. They are rented. The user pays a subscription for access to a service controlled by a corporation. The user does not understand the tool at the level that would permit replication or modification. The user cannot maintain the tool independently. The ownership and understanding that Schumacher considered essential are absent, and their absence transforms the tool from a means of genuine empowerment to a means of contingent capability — real capability, but capability that depends on someone else's continued provision.

Berry and Stockman, in their 2024 analysis applying Schumacher's framework to generative AI, identified this tension and proposed a direction: what they called "intermediate artificial intelligence." The concept points toward open-source, explainable AI as the Schumacherian alternative to proprietary, opaque models — tools that move toward the ownership and understanding that genuine intermediate technology requires. Open-source models that can be run locally, on hardware the builder owns, reduce dependence on centralized infrastructure. The builder who runs a local model does not depend on Anthropic's pricing decisions or terms of service.

But the open-source alternative encounters its own version of the development gap. The most capable models require computational resources beyond individual capacity. The gap between open-source models and frontier models is a capability gap that builders accustomed to frontier performance may find unacceptable. And the development of frontier models requires the concentrated investment that only large-scale institutions can finance. The intermediate option exists, but it exists at a lower capability level, and the distance between intermediate and frontier is not trivial.

Schumacher encountered precisely this dynamic in the physical intermediate technology movement. The hand loom was accessible and ownable but less productive than the power loom. The bicycle was maintainable but slower than the automobile. In each case, the intermediate technology offered genuine capability at a lower level of performance than the industrial alternative. The question was whether the lower performance was acceptable given the greater autonomy, and the answer depended on the user's circumstances, priorities, and tolerance for the trade-off.

The same question applies to intermediate AI. The open-source model running locally is less capable than the frontier model running on centralized infrastructure. The trade-off is capability for autonomy. For some builders, the autonomy is worth the reduced capability. For others, the capability gap is too large to accept. The market, left to its own dynamics, will tend toward the frontier — toward the most capable tools regardless of their ownership structure — because the market rewards output and the frontier produces more of it.

This is where Schumacher's framework demands intervention. Intermediate technology does not sustain itself in a market that rewards maximum output. The hand loom did not survive the power loom through market competition. It survived where institutional structures — cooperatives, development agencies, community governance — maintained it against market pressure. The intermediate technology movement succeeded not because intermediate tools were more competitive but because communities decided that the autonomy, the local employment, and the human-scale production the tools enabled were worth protecting even at the cost of reduced output.

The lesson for AI is that intermediate AI — open-source, locally deployable, user-owned and user-understood — will not sustain itself through market competition with frontier models. It will sustain itself only if communities, institutions, and public investment decide that the autonomy and ownership it provides are worth protecting. This is a political decision, not a technical one. It requires the recognition that the most capable tool is not always the most appropriate tool, and that appropriateness, in Schumacher's demanding sense, includes the user's ownership, understanding, and control of the conditions of production.

The history of intermediate technology in the developing world provides instruction. Where intermediate technology was supported by institutional structures — cooperative movements in India, extension services in East Africa, community-governed production in Southeast Asia — it proved both productive and sustainable. Where it lacked institutional support, it was overwhelmed by industrial alternatives whose lower costs and higher output made the intermediate option economically unviable regardless of its human advantages.

The question for the AI transition is whether the structures that would sustain intermediate AI will be built before the window for building them closes. The concentration of AI capability in a small number of corporations is accelerating. The frontier models are pulling further ahead of open-source alternatives. The cost of training competitive models is rising. Each of these trends narrows the space in which intermediate AI is viable, and the narrowing proceeds at the pace of the technology rather than the pace of institutional development.

Schumacher understood that technology does not wait for governance to catch up. The power loom did not pause while the weavers organized. The factory did not slow while labor law developed. The intermediate technology movement succeeded where it got ahead of the concentration — where it built the structures before the industrial alternative made them impossible. The AI transition demands the same foresight, the same urgency, and the same recognition that the most important decisions about technology are not technical decisions about capability but political decisions about who owns the tools and who governs their use.

---

Chapter 7: The Worker's Consciousness and the Always-On Machine

The worker's consciousness held a place in Schumacher's economics that it holds in no other economic framework of the twentieth century. Mainstream economics had no interest in what the worker experienced during the production process. The worker was characterized by skills, availability, and cost. Whether the worker felt fulfilled or diminished, engaged or estranged, was a matter for psychology or perhaps social policy — certainly not for economic analysis. The economy measured output. Consciousness was outside the frame.

Schumacher insisted on bringing it inside. His argument was not sentimental but structural: an economics that ignores the worker's consciousness ignores the most important dimension of economic activity. The products of the economy exist to serve human beings. But the human beings who produce the products are also served or damaged by the production process. An economy that produces excellent goods while damaging the producers has failed at the deepest level, because the damage is a cost the economic calculation has refused to count.

The worker's consciousness requires specific conditions to flourish, conditions that are not mysterious but are systematically ignored by economic systems organized around output. Schumacher identified these conditions through decades of observation in industrial settings and in developing economies where the relationship between work and human development was more visible than in the abstracted, statistically mediated economies of the industrialized world.

The first condition is meaningful engagement — work that demands the exercise of judgment, skill, and care. Work that reduces the worker to a single repetitive function does not meet this condition regardless of how well it pays. The second condition is rest — genuine disengagement from productive activity, not rest instrumentalized as recovery for the next work session, but rest as an experience complete in itself, the experience of not producing. The third condition is what might be called the right to difficulty — the preservation of challenges that develop the worker's capacities through the specific experience of struggling with something that does not yield easily.

AI-augmented work provides the first condition with remarkable generosity. The builder who works with Claude Code is not performing repetitive operations. The builder conceives, directs, evaluates — engaging at a level that demands creative capacity of a high order. The integration of previously separated disciplines, the expansion into domains that were gated by translation costs, the real-time realization of ideas that would previously have taken weeks to prototype — these represent meaningful engagement of a kind that most industrial work could not offer. The builder is not a component of a machine. The builder is a whole person doing whole work.

But the second and third conditions are where the technology's relationship with consciousness becomes genuinely problematic, and the problems are structural rather than incidental.

The threat to rest is not that the tool demands continuous use. The tool demands nothing. It sits quietly until prompted. The threat is subtler and more pervasive: the tool's always-on availability combines with the builder's internalized imperative to produce, creating conditions under which rest feels like waste. The builder knows the tool is there. The builder knows that any pause — the elevator ride, the wait for coffee, the minutes before sleep — could be converted to productive output. The knowledge itself erodes rest, because rest requires the absence of productive possibility, and productive possibility is now permanent.

The Berkeley researchers documented this erosion with empirical precision. Workers who adopted AI tools did not use the efficiency gains to rest more. They used them to work more. Work seeped into pauses that had previously served as cognitive recovery. The researchers called this "task seepage," and the phrase names something Schumacher's framework explains at the structural level: technology that recognizes no self-limiting principle will expand into every available space unless external structures contain it. The space it expands into, in this case, is the worker's non-productive time — the time that consciousness needs not for recovery but for its own sustenance.

The neurological basis for this is well established. The default mode network — the brain's resting-state activity pattern — is active during periods of apparent idleness: daydreaming, mind-wandering, the unstructured cognition that occurs when no specific task demands attention. Research over the past two decades has demonstrated that default mode network activity is essential for memory consolidation, self-referential processing, creative incubation, and the integration of disparate experiences into coherent understanding. The mind, during apparent rest, is not idle. It is performing the integrative work that gives meaning to the specific, task-focused work of the productive hours.

AI tools threaten this integrative work not by preventing rest but by filling the gaps in which rest would naturally occur. The one-minute pause between meetings, the three-minute wait for coffee, the ten-minute commute between buildings — these micro-pauses were never formally designated as rest. They were the interstices of the workday, too brief to fill with structured tasks under the old technology, and therefore available, by default, for the unstructured cognition that the default mode network performs. When a tool makes it possible to fill a one-minute gap with a productive prompt, the gap fills. The micro-rest disappears. And its disappearance is invisible because it was never recognized as rest in the first place.

The threat to the right to difficulty operates through a different mechanism but produces a convergent effect. Schumacher's framework values difficulty not as an obstacle to be overcome but as a developmental resource — the medium through which the worker's capacities are strengthened, the way physical resistance strengthens muscle. The developer who spends hours debugging a system is not wasting time. The developer is building the embodied understanding of system behavior that constitutes architectural intuition — a form of knowledge that cannot be transferred through documentation or instruction but must be developed through the specific experience of encountering, diagnosing, and resolving unexpected failures.

AI tools make this difficulty optional. The builder who encounters a bug can describe it to Claude and receive a fix without undergoing the diagnostic process. The fix is correct. The builder moves on. The specific exercise through which diagnostic capacity develops has been bypassed, and the bypass is rational — it is faster, less frustrating, and produces the same functional result. The irrationality is visible only at the scale of a career, when the builder who has spent years bypassing diagnostic difficulty discovers that the diagnostic capacity has not developed, and that the judgment that depends on it is thinner than it should be.

Schumacher would connect this to his ontological framework from A Guide for the Perplexed, where he distinguished four levels of being — mineral, plant, animal, and human — each possessing capacities the lower levels lack. The distinctively human capacity, in his analysis, is self-awareness: the ability not merely to think but to observe one's own thinking, not merely to experience but to reflect on experience and derive meaning from it. Self-awareness, he argued, "has nothing mechanical or automatic about it." It must be developed through deliberate engagement with experiences that demand it.

Work that develops self-awareness is work that forces the builder to confront the gap between intention and result — the gap where bugs live, where designs fail, where the thing you imagined does not match the thing you built. The confrontation is uncomfortable. It requires the specific form of attention that acknowledges error, examines its source, and adjusts understanding accordingly. This is the experience through which self-awareness grows, and it is precisely the experience that AI tools make optional.

The tool does not diminish the builder's self-awareness directly. It removes the occasions on which self-awareness would be exercised and thereby developed. The builder who never confronts unexpected system behavior does not lose existing self-awareness. The builder fails to develop the additional self-awareness that the confrontation would have produced. The loss is invisible because it is the loss of a potential that was never realized — the capacity that would have existed if the difficulty had been encountered rather than bypassed.

The combination of these two threats — the erosion of rest and the bypass of developmental difficulty — produces a specific condition in the builder's consciousness that is difficult to name precisely but recognizable to anyone who has experienced it. It is the condition of being simultaneously productive and hollow — producing output of high quality while sensing that something essential has been lost or is failing to develop. The output is there. The understanding is thinner. The presence is fractured. The capacity for the kind of slow, patient reflection that Schumacher associated with wisdom has been consumed by the speed of the tool, not because the tool demanded speed but because the tool made slowness feel like a luxury the builder could not afford.

Schumacher's practical response to this condition would not be to restrict the tool but to build structures that protect the conditions consciousness requires. Mandatory offline periods are the contemporary equivalent of the eight-hour day — structural interventions that contain the technology's tendency to colonize every available hour. Protected time for difficulty — deliberate, structured engagement with problems that the tool could solve but the builder chooses to solve manually, for the specific purpose of developing capacities that the tool's efficiency would otherwise leave fallow — is the equivalent of physical exercise for a knowledge worker: an investment in capacity that produces no immediate output but sustains the foundation on which all output depends.

These structures are not luxuries. They are necessities, demanded by the specific conditions of a technology that provides meaningful engagement while threatening the rest and the difficulty that consciousness requires for its full development. The failure to build them is not a failure of individual discipline. It is a structural failure — the absence of institutions that recognize the worker's consciousness as a resource to be cultivated rather than a fuel to be consumed.

---

Chapter 8: Good Work and Its Counterfeits

Good work, in Schumacher's understanding, was not a romantic aspiration. It was a practical criterion as measurable in its own terms as profit or productivity, though the measurements required attention to dimensions that conventional economics had trained itself to ignore. Good work nourished the worker while serving the community. It developed faculties while producing output. It was simultaneously productive and developmental — the product useful, the process formative.

The criterion was demanding. Most work in the industrial economy failed it. The assembly line worker performing a single repetitive operation was not doing good work, regardless of the product's quality or the factory's profitability. The work did not develop the worker's faculties. It did not nourish consciousness. It produced output at the expense of the producer, and the exchange was one that Schumacher's economics identified as illegitimate regardless of compensation.

The AI transition introduces a complication that Schumacher did not face and that his framework must stretch to accommodate: the possibility of work that looks and feels like good work but is not. Good work's counterfeit. The counterfeit is dangerous precisely because it is indistinguishable from the genuine article at the surface level — indistinguishable to the productivity metric, indistinguishable to the observer, and often indistinguishable to the builder experiencing it.

Good work, genuinely good work in Schumacher's bilateral sense, produces both a useful product and a developed worker. The builder who ends the day having produced something valuable and having grown through the process of producing it has done good work. The growth might be in judgment — the capacity to make better decisions about what to build. It might be in understanding — a deeper grasp of how systems work and why they fail. It might be in wisdom — the capacity to see one's own work in a broader context, to evaluate it not only by its functionality but by its effect on the people it serves.

The counterfeit produces a useful product without developing the worker. The output is there. The growth is not. The builder has produced more but has not become more capable, more perceptive, or more wise. The distinction is invisible from the outside because both the genuine and the counterfeit produce the same observable result: a shipped product, a solved problem, a satisfied user. The distinction is visible only from the inside, and only to a builder who has developed the self-awareness to detect it.

The author of The Orange Pill describes one form of the counterfeit with diagnostic precision. Claude produces a passage connecting Csikszentmihalyi's flow state to a concept attributed to Deleuze. The passage is elegant. The connection is beautiful. The philosophical reference is wrong — wrong in a way that would be obvious to anyone who had actually read Deleuze, but invisible to anyone seduced by the prose's fluency. The smooth output concealed the fractured argument. The product looked good. The process had produced nothing developmental in the author, because the author had not thought the thought — the tool had generated it, and the author had nearly accepted it without the critical engagement that would have constituted genuine intellectual work.

The author caught the error. The discipline of checking, of resisting the seduction of polished output, caught the counterfeit before it entered the final text. But the catching required a specific capacity — the willingness to distrust output that sounds right, to verify claims that arrive with confidence, to maintain the gap between reception and acceptance that constitutes critical thinking. And this capacity is precisely the capacity that the tool's fluency tends to erode, because fluent output rewards acceptance and punishes verification. Checking takes time. Accepting is instant. And the builder who has internalized the imperative to produce — the builder whose day is measured in output rather than understanding — experiences checking as friction, as a cost, as an inefficiency to be minimized.

The counterfeit extends beyond individual instances of unchecked output to a broader pattern that Schumacher's framework identifies as the central risk of the amplifier. The pattern is this: the builder produces more output at a higher level of apparent quality, experiences the production as stimulating and engaging, and mistakes the stimulation for development. The work feels creative. The work feels meaningful. The work produces the subjective markers of flow — absorbed attention, loss of self-consciousness, the sense of operating at the edge of capability. And yet the builder is not growing, because the growth that matters — growth in judgment, in understanding, in the capacity for wise direction — requires engagement with difficulty that the tool has smoothed away.

The subjective markers of good work and the subjective markers of its counterfeit are identical. Both produce absorption. Both produce the sense of operating at capacity. Both produce satisfaction with the output. The difference is in what the experience deposits in the builder. Good work deposits understanding — the specific, earned knowledge that comes from struggling with a problem until it yields its structure. The counterfeit deposits output — the product is there, but the understanding that should have accompanied its production is not.

Consider two builders producing the same product on the same day. The first encounters a technical problem, spends an hour understanding its source, tries three approaches that fail, and on the fourth attempt solves it with a solution that reflects genuine comprehension of the system's architecture. The second encounters the same problem, describes it to Claude, receives a working solution in ninety seconds, and moves on. Both builders ship the feature. Both feel productive. The first builder has deposited a layer of understanding that will inform every subsequent architectural decision. The second builder has deposited nothing except the completed task.

The first builder did good work. The second builder did work that produced a good result. The distinction is invisible in the output. It is consequential over a career.

Schumacher would not argue that the second builder should always choose the first builder's path. Efficiency has its place. Not every technical problem is a developmental opportunity. The question is one of proportion — whether the builder's overall practice includes sufficient engagement with genuine difficulty to sustain the developmental dimension of work, or whether the efficiency of the tool has shifted the proportion so far toward output that development has been starved.

The author of The Orange Pill describes the practice of periodically discarding Claude's output and writing by hand — choosing the harder path not because the tool's output is inadequate but because the process of struggling with the blank page produces something in the author that the tool's fluency cannot. Rougher output. More honest uncertainty. The specific knowledge of where the argument is strong and where it is weak that only emerges through the experience of trying to make the argument hold together without assistance. This practice — the deliberate retention of productive difficulty — is the antidote to the counterfeit.

The practice costs output. Time spent writing by hand is time not spent producing text with Claude. The trade-off is visible and immediate: less output per hour, less polished prose, slower progress toward completion. What the trade-off purchases is invisible and delayed: deeper understanding, stronger judgment, the capacity to evaluate the tool's output with the authority that comes from having done the work oneself.

Schumacher would recognize this trade-off as the central challenge of good work in every era: the tension between output and development, between the product the market rewards and the process the worker needs. The industrial economy resolved this tension in favor of output, designing workplaces that maximized production at the expense of the worker's development. Schumacher's economics argued for a different resolution — one that treated the worker's development as a legitimate constraint on the production process, not as a luxury to be indulged when profits allow.

The AI transition makes this argument more urgent and more difficult simultaneously. More urgent because the tool's power makes the counterfeit more convincing — the output is so good, so fast, so apparently creative that the absence of development is harder to detect. More difficult because the market's incentive structure has not changed: the market still rewards output, still measures productivity, still treats the builder's inner development as an externality that falls outside the economic calculation.

The builder who ships ten features in a week is rewarded. The builder who ships five features and spends the remaining time in deliberate, developmental struggle with problems the tool could have solved is, by the market's metric, less productive. The metric does not count the judgment that the struggle developed. The metric does not count the understanding that the difficulty deposited. The metric counts features shipped, and by this count, the builder who produced the counterfeit outperformed the builder who did good work.

Good work in the age of the amplifier requires structures that correct this distortion — structures that value the builder's development as a productive outcome, not merely the builder's output. The Berkeley researchers' proposal for "AI Practice" — structured pauses, sequenced rather than parallel work, protected time for human-only engagement — points toward such structures. Organizations that implement these practices are building institutional recognition of what Schumacher argued throughout his career: that the worker's development is not a cost of production but the most important product of the production process, and that an economy that ignores it is an economy consuming its own foundation.

The counterfeit will always be available. The tool will always offer the easier path. The output will always be impressive. The question Schumacher's economics poses to every builder is not whether the output is good — it almost certainly is — but whether the builder is growing through the process of producing it. If the answer is yes, the work is good. If the answer is no, the work is counterfeit, and no amount of output can compensate for what the builder has failed to develop.

Chapter 9: The Village and the Platform

Schumacher advocated for economic organization at the village scale, and the advocacy was neither nostalgic nor arbitrary. The village was not a sentimental preference. It was a structural argument about the conditions under which certain essential features of humane economic life can be maintained: mutual knowledge, personal accountability, collective governance, and the social bonds that transform a collection of producers into a community. These features were not incidental to the village's economic function. They were constitutive of it. The village produced not only goods but relationships, not only output but mutual obligation, not only wealth but the specific form of social capital that enables people to live together with dignity and care.

The industrial economy destroyed the village as an economic unit, not through malice but through the logic of concentration. Larger enterprises drew workers away from communities where they were known into organizations where they were anonymous, replacing the mutual obligations of neighbors with the contractual obligations of employment, substituting hierarchical relationships between employer and employee for the horizontal relationships of fellow citizens. The destruction was incremental and accompanied, at every stage, by aggregate improvements in material welfare that made the human cost easy to overlook. The gross domestic product rose. The communities that had sustained human life for centuries dissolved. The statistics recorded the first development. They had no category for the second.

The AI platform reproduces this dissolution at a different scale and through a different mechanism, but the structural result converges. The builder on the platform is productive but alone. Not lonely in the conventional sense — the builder is engaged, stimulated, operating at a level of creative intensity that most pre-AI work could not offer. But the engagement is with a tool, not a community. The stimulation is cognitive, not social. The productivity is individual, not collective.

The author of The Orange Pill describes three friends on a Princeton campus — Uri the neuroscientist, Raanan the filmmaker, the author himself — arguing about the nature of intelligence with the specific intensity of people who have known each other for thirty years. The description captures something the platform cannot provide: the quality of attention that one human being brings to another who is known personally and cared about individually. Uri says, with genuine bluntness, "That is either trivially true or complete nonsense." The honesty is possible because the relationship is real. The challenge is possible because the mutual knowledge is genuine. The growth the conversation produces — the gradual refinement of the author's thinking about intelligence as a river — is possible because the conversation takes place within a social context of trust, candor, and sustained mutual attention.

Claude Code does not provide this. Claude provides extraordinary cognitive partnership. Claude does not provide the friend who says, with concern that is personal rather than algorithmic, "You look exhausted. Go home." Claude does not provide the colleague who notices a pattern of compulsive work and names it before the builder can. Claude does not provide the community that debates the conditions of building and holds its members accountable for building wisely.

The absence of community has consequences that extend beyond emotional well-being, though the emotional consequences are real. The absence of community means the absence of mutual accountability — the external check on self-exploitation that the individual builder cannot reliably provide from within the compulsion itself. The builder working alone at three in the morning has no one to say "enough." The builder's only check is self-regulation, and self-regulation is a finite resource that the tool's always-on availability steadily depletes.

The absence of community also means the absence of collective governance over the conditions of building. The individual builder has no mechanism for influencing the decisions that shape the work — pricing, terms, capabilities, the future direction of the technology. These decisions are made by corporations, and the individual builder's only recourse is acceptance or exit. Collective governance would transform this dynamic. A community of builders sharing a platform could negotiate collectively with the provider, establish norms reflecting the community's values, create mutual support during transitions, and develop shared knowledge about sustainable practice.

Practical forms of such communities are beginning to emerge. Online groups of AI-augmented builders share experiences and develop informal best practices. Some are organized around specific tools, others around specific domains. The best of them provide something the platform cannot: the experience of being known, of being accountable to others, of participating in a collective project that extends beyond individual production. But these communities are fragile. They lack institutional support, governance structures, and economic foundations that would sustain them beyond the enthusiasm of their founders. They operate within a platform ecosystem structurally indifferent to their existence, because the platform's business model depends on individual subscriptions, not community governance.

Schumacher would identify this fragility as symptomatic of a deeper structural absence. The platform provides the tool. It does not provide the context in which the tool is used wisely. The tool without the community is powerful but dangerous. The community without the tool is humane but limited. The combination — the powerful tool used within the context of a humane community — is what appropriate economic organization requires.

The village scale is the scale at which mutual knowledge is possible, personal accountability is enforceable, and collective governance is not a theory but a lived practice. Not a geographic village, necessarily. But a social structure that embeds productive activity within a context of mutual obligation, shared governance, and human connection — a structure that provides the counterweight to productive compulsion that the individual builder cannot reliably supply from within.

The platform provides capability. The village provides the wisdom to use capability well. The AI transition needs both, and currently has only one.

---

Chapter 10: Building as if People Mattered

The AI transition will be evaluated, ultimately, not by the output it produces but by the lives it creates. This is the central claim of Schumacher's economics, and it is a claim that the dominant discourse has not yet learned to take seriously. The discourse measures output — products shipped, revenue generated, productivity multiplied. These are real achievements. Schumacher's economics does not deny their reality. But it insists they are not the final measure of success, because the final measure is not what the economy produces but what kind of lives the economy enables.

More products, more features, more applications, more revenue, more growth — these are outputs. Better lives, deeper relationships, more meaningful work, healthier communities, wiser citizens — these are outcomes. Schumacher's economics insists that the outcomes are what matter, and that outputs are valuable only to the extent that they serve outcomes. An economy that produces an extraordinary volume of output while producing exhausted, anxious, isolated human beings has failed, and the failure is not in the output, which may be excellent, but in the relationship between the output and the lives of the people who produce it.

The arguments of the preceding chapters converge on a set of structural requirements that building as if people mattered would have to satisfy. These requirements are not aspirational. They are the minimum conditions under which the AI transition serves human flourishing rather than consuming it.

The first requirement is the bilateral evaluation of work. Every deployment of AI tools should be evaluated not only by what it produces but by what it does to the people who use it. The productivity metric is necessary but insufficient. It must be accompanied by measures of the builder's development — whether judgment is growing, whether understanding is deepening, whether the capacity for the kind of reflective engagement that constitutes wisdom is being sustained or eroded. Organizations that implement this bilateral evaluation will discover what Schumacher argued throughout his career: that the worker's development is not a cost of production but the most important product of the production process.

The second requirement is structural protection of the conditions that consciousness needs. Rest, difficulty, and presence cannot be left to individual discipline when the tool's always-on availability systematically overwhelms individual willpower. The eight-hour day was a structural intervention that contained industrial technology's tendency to consume the worker's entire life. The AI transition requires analogous interventions — mandatory offline periods, protected time for difficulty that the tool could solve but the builder chooses to struggle with, institutional norms that treat presence in relationships as productive rather than wasteful. These structures are not luxuries. They are the contemporary equivalent of labor law, demanded by a technology that recognizes no self-limiting principle of its own.

The third requirement is the transformation of contingent sovereignty into genuine independence. The solo builder's dependence on tools controlled by corporations must be addressed through structural remedies — open-source models, local deployment options, cooperative ownership structures, regulatory frameworks that treat the builder as a stakeholder rather than merely a customer. Berry and Stockman's concept of "intermediate artificial intelligence" points the direction: AI that moves toward the ownership, explainability, and user control that Schumacher's intermediate technology required. The open-source AI movement is the most Schumacherian development in the current landscape, and its sustenance requires deliberate public investment and institutional support, because intermediate technology does not survive market competition with frontier alternatives without structural protection.

The fourth requirement is community. The platform provides capability. The village provides wisdom. The builder who works alone with a tool that is always available is a builder exposed to risks that no tool can mitigate — the risk of self-exploitation, the risk of counterfeit work mistaken for the genuine article, the risk of producing output without growing as a person. Village-scale structures for AI-augmented work — communities of practice that provide mutual accountability, shared governance, and the human connection that sustains consciousness — are not optional additions to the AI-augmented workplace. They are essential infrastructure, as necessary as the computational infrastructure that powers the tools.

The fifth requirement is governance at the scale of the infrastructure. The decisions that shape every builder's conditions — pricing, capabilities, terms, safety constraints, resource allocation — must include the people those decisions affect. This does not mean every builder votes on every model update. It means that structural mechanisms exist through which the interests of builders, as a class, are represented in the governance of the platforms they depend on. Cooperative structures, user governance boards, regulatory frameworks that mandate stakeholder representation — these are the political infrastructure of an AI economy organized as if people mattered.

These five requirements are demanding. They will not be met automatically. The market's incentive structure does not reward bilateral evaluation, structural rest protection, distributed ownership, community formation, or democratic governance. These structures must be built against market pressure, through the same combination of political will, institutional creativity, and collective action that built the labor protections of the industrial era.

Schumacher understood that the structures which protect human flourishing during technological transitions do not emerge organically. They are built — deliberately, against resistance, by people who understand that the most important decisions about technology are not technical decisions about capability but political decisions about who the technology serves.

Richard Murphy, applying Schumacher's framework to the present moment, states the point with the directness that Schumacher favored: "Technology without moral direction becomes dehumanising. The question is never just what we can do, but what we should do." The question applies not only to the builders of AI systems but to every person who uses them, every organization that deploys them, every institution that governs them, and every parent who watches a child grow up inside them.

Schumacher did not live to see the AI transition. He died in 1977, on a lecture tour in Switzerland, still arguing for an economics the world was not ready to hear. The world was busy growing. The GDP was climbing. The argument that the economics of more might be less humane than the economics of enough was received as charming, provocative, and irrelevant.

The conditions have arrived that make the argument urgent. The tools are extraordinary. The output is impressive. The question — the only question that an economics as if people mattered permits — is whether the people who use these tools will build structures that protect the inner lives, the relationships, the communities, and the consciousness that give output its meaning. Without these structures, the output, however impressive, is purchased at a cost that Schumacher's economics identifies as prohibitive: the cost of the producer, consumed by the process of production, depleted by the tools that were supposed to serve.

The tools have arrived. The question is whether we will build as if people mattered, or whether we will allow the output to accumulate while the producers are quietly spent.

---

Epilogue

The number one hundred stopped me. Not the billions in GPU infrastructure. Not the trillions in evaporated market value. One hundred dollars. That was the monthly cost of the tool that turned twenty engineers into what felt like two hundred. One hundred dollars — the price of a decent dinner for two, roughly what I spend on coffee in a month if I'm being honest about the habit.

Schumacher's entire career was built around a question that most economists considered beneath them: what is the right size? Not the maximum size. Not the most efficient size. The right size — the size at which a tool serves the person using it rather than the other way around. He asked this about factories, about farms, about national economies, about the development programs he advised in Burma and India. Always the same question, applied to different materials. How big before the tool becomes the master? How powerful before the worker becomes the component?

One hundred dollars a month. The tool fits in a laptop. The builder works at a kitchen table. By every spatial metric, this is the smallest-scale production technology in the history of serious software development. Schumacher would have smiled at the geometry of it. One person. One conversation. A product that works.

Then he would have asked the question behind the question. Not how small is the workspace? but how large is the dependency? Not how free is the builder? but who owns the conditions of that freedom? The solo builder is sovereign the way a tenant farmer is sovereign — genuinely directing the work, genuinely exercising judgment, and genuinely unable to forge a new tool if the one provided is taken away. The smallness at the point of use depends on an enormity at the point of infrastructure. The beautiful is contingent on the colossal.

I think about the engineer in Trivandrum who lost the ten formative minutes along with the four tedious hours. Schumacher would not have mourned the tedium. He was not a romantic about drudgery. But he would have recognized instantly what she lost and could not name until months later: the specific knowledge that only difficulty deposits, the geological layers of understanding that accumulate through struggle and cannot be transferred through output. Buddhist economics would call that loss a failure of the production process — not because the product suffered, but because the producer did.

What haunts me is how invisible the loss is. The output looks the same. The sprint feels productive. The dashboard is green. The thing Schumacher spent his life measuring — what the work does to the person doing it — does not appear on any screen I check. I have to look at the people. I have to ask them. And I have to be willing to hear answers that complicate the twenty-fold productivity number I am so proud of.

Building as if people mattered. The phrase is simple in the way that load-bearing walls are simple — plain, structural, holding up everything above them. It does not tell me to stop using Claude Code. It does not tell me to slow down. It tells me to measure what I have been ignoring: the rest that did not happen, the difficulty that was bypassed, the consciousness that was consumed in the process of producing something impressive.

The tools have arrived. The structures that would make them humane have not. That gap — between the tool's power and the institutions that should govern its use — is the space in which the damage accumulates, quietly, beneath the surface of extraordinary output. Schumacher saw this gap in the industrial economy and spent his life trying to close it. The gap is wider now. The tools are more powerful. The urgency is greater.

And the question he would ask me, if he could, is the one I find hardest to answer: Is the amplification worth the life of the person holding the amplifier?

I am building the structures that would let me answer yes. I am not yet confident I have earned the answer.

— Edo Segal