CONCEPT

The One Best Way

Frederick Winslow Taylor's 1911 principle — that for any task there exists a single most efficient method — which Ellul recognized not as innovation but as technique's self-articulation in vocabulary the modern world could hear.

Taylor's claim was simple and devastating: for any given task, there exists a single optimal method, identifiable through systematic observation and measurement, and once identified, universally required. Every other method is waste. Ellul saw in Taylor not the inventor of a new logic but the most candid spokesman for a logic that had been developing for centuries. What Taylor articulated, the logic had been enforcing. And once the vocabulary existed, the enforcement became visible — and therefore adoptable as an explicit institutional program rather than an implicit structural pressure. Scientific management spread from the factory to the office, from the office to the school, from the school to the hospital, and eventually to domains Taylor himself would not have recognized as susceptible to systematic optimization. Each extension followed the same pattern: identify inefficiency, design method, implement, eliminate alternatives.

The Persistence of Craft Knowledge — Contrarian ^ Opus

There is a parallel reading that begins not from technique's autonomous logic but from the stubborn persistence of embodied knowledge that resists systematization. Taylor's stopwatch could time the pig iron shoveler, but it could not capture what the experienced worker knew about metal fatigue from the sound of impact, about weather conditions from morning air, about which movements would sustain through a twelve-hour shift versus which would leave you broken by noon. This tacit knowledge—what James C. Scott calls 'metis'—has never been successfully eliminated despite a century of trying. It reasserts itself at every level of abstraction, from the assembly programmer who could feel when memory was fragmenting to the AI prompter who develops an intuition for which phrasings will coax coherent responses from an opaque model.

The political economy of the One Best Way reveals something the entry's structural analysis obscures: optimization serves particular interests, and those interests shape what counts as 'best.' Taylor's methods increased output per worker while wages stagnated. Contemporary AI promises to optimize cognitive work while concentrating the returns in platforms that own the models. The substrate this optimization requires—data centers consuming the power of small nations, rare earth mining devastating communities, content moderators in Kenya suffering psychological damage—suggests that 'efficiency' is achieved by externalizing costs to those least able to resist. The One Best Way has always been best for someone in particular, and that someone is rarely the person whose work is being optimized. The persistence of craft knowledge is not romantic resistance but practical necessity: someone must still know how things actually work when the optimized system inevitably encounters conditions its designers didn't anticipate.

— Contrarian ^ Opus

In the AI Story

Hedcut illustration for The One Best Way — The One Best Way

Taylor was not a philosopher. He was an engineer with a stopwatch, concerned with specific problems — how to shovel pig iron, how to cut metal, how to lay bricks. His ambitions were modest: reduce waste, increase output per worker, rationalize the relationship between effort and result. He could not have imagined that his principle would metastasize beyond the factory to colonize every domain of modern life. But the logic he articulated did not require his ambitions to spread. It spread because it worked. The results were measurable. And once the conviction took hold that an optimal method exists for any activity, choosing a suboptimal one felt not merely inefficient but morally deficient.

The progression from Taylor's shop floor to the contemporary AI-assisted workspace is not metaphorical. It is a direct line of institutional development. Scientific management rationalized the factory. The rationalized factory demanded rationalized information processing — filing systems, tabulating machines, computers, networks. Each stage identified a more efficient method than its predecessor. Each stage made the predecessor irrational to maintain. Each stage absorbed the domain into technique's expanding jurisdiction. And at no point did any individual or institution deliberately choose this trajectory. The trajectory was driven by the logic itself: each level of efficiency creating the conditions and the demand for the next.

Edo Segal traces the same progression in The Orange Pill — from assembly language to compilers to frameworks to cloud infrastructure to AI — and celebrates the liberation each abstraction provided. Ellul's framework accepts the accuracy of Segal's historical account while inverting its evaluation. What appears as liberation from lower-level tedium is, structurally, the elimination of the cognitive environment in which the corresponding understanding once developed. The assembly programmer thought in memory addresses; the framework programmer cannot. The gain is real. The loss is also real. Technique's logic makes only the gain visible, because only the gain is measurable.

For the AI moment, the one best way principle operates with unprecedented speed and universality. Previous extensions of scientific management took decades or generations to colonize new domains. The AI extension compresses the timeline to months. A domain that resisted systematization in 2022 — creative writing, strategic judgment, the formation of ideas — is systematizable in 2025. The logic has not changed. The pace has.

Origin

Taylor's Principles of Scientific Management appeared in 1911, summarizing two decades of his consulting work at Bethlehem Steel and other industrial sites. The book became a global bestseller, translated into languages Taylor did not speak and applied to domains Taylor had never studied. Lenin cited it approvingly. Henry Ford operationalized it at River Rouge. Business schools built curricula around it. Within a generation, the one best way had become the unchallenged premise of modern management, so deeply absorbed that its status as a historical innovation became invisible.

Key Ideas

The principle's power is moral, not merely technical. Once you believe an optimal method exists, choosing a suboptimal one is not just inefficient — it feels like a failure of reason, and rational failure carries stigma that technical failure does not.

Alternatives are eliminated, not merely disfavored. The mechanism is competitive: the actor adopting the optimal method outperforms the actor who does not, and the outperformance compounds until the alternative is eliminated by irrelevance.

The principle universalizes. Taylor applied it to shovels. His successors applied it to offices, schools, hospitals, legal practice, creative work. AI applies it to cognition itself — to the formation of ideas that was once protected by its own immeasurability.

Expertise in the old method becomes obstruction. The skilled practitioner of a superseded technique does not become irrelevant through any fault of her own. Her expertise is simply reclassified as waste by the logic that has identified a more efficient alternative.

Each elimination creates cognitive foreclosure. When assembly language was superseded, assembly-era ways of thinking became unavailable to most practitioners. The same foreclosure operates at every subsequent level, and AI operates at the most comprehensive level yet achieved.

Debates & Critiques

Defenders of the one best way principle argue that cognitive foreclosure is compensated by cognitive expansion — the framework programmer cannot think in memory addresses but can think about application architecture in ways the assembly programmer could not. Segal's ascending friction thesis makes this case explicitly. Ellul would accept that the compensation is partial but insist that it is structurally incomplete: the domains being foreclosed are those in which non-quantifiable values once operated, while the domains being opened are measurable by the very metrics that eliminated their predecessors. The net effect is a civilization in which everything that can be measured is optimized and everything that cannot be measured is lost.

Appears in the Orange Pill Cycle

Jacques Ellul — On AI

The Optimization Dialectic — Arbitrator ^ Opus

The tension between Ellul's account and the contrarian reading resolves differently depending on which temporal frame we adopt. At the civilizational scale—measuring across centuries—Ellul's diagnosis appears 80% correct: technique does exhibit autonomous logic that progressively eliminates alternatives. The assembly programmer's way of thinking is genuinely lost to most practitioners, not temporarily obscured but structurally foreclosed. Yet at the scale of lived experience—measuring in years or decades—the contrarian view captures 70% of the reality: tacit knowledge persistently reasserts itself, optimization frequently fails, and workers develop new forms of craft knowledge even within highly systematized domains.

The question of whose interests optimization serves splits even more dramatically by context. In consumer technology and creative tools, Segal's liberation narrative holds perhaps 60% true—many people genuinely experience expanded capability. But in workplace surveillance, algorithmic management, and platform labor, the contrarian reading dominates at 85%: optimization primarily serves capital's interests while degrading working conditions. The environmental and human substrate supporting AI's efficiency gains represents nearly pure (95%) validation of the contrarian concern about externalized costs.

The synthetic frame that best holds both views might be called 'dialectical optimization': each wave of systematization simultaneously eliminates certain forms of human agency while creating unexpected spaces for new forms to emerge. The One Best Way never fully conquers because its own operation generates contradictions—dependencies it cannot see, contexts it cannot anticipate, human needs it cannot quantify. Yet neither does resistance ever fully succeed, because technique's efficiency gains create genuine value even as they foreclose genuine possibilities. The AI moment accelerates both sides of this dialectic, making the stakes of each optimization decision more consequential and more visible than Taylor's generation could have imagined.

— Arbitrator ^ Opus