Architecture or Soma
Essay 3 of 3 in the Another Brave New World series
After the Naming
The diagnosis was that an old pattern is arriving in a new form, that the pattern is pacification at scale, and that the pattern’s convenience is what makes it durable. The diagnosis matters because the pattern is invisible until it is named. Once named, it becomes refusable. But refusing it requires more than disliking it.
The pacification is a default. Defaults are what happen when the prior questions are not asked. The Roman default was that the urban poor needed bread and that ambitious politicians needed crowds. The Huxleyan default was that stability required happiness and that happiness required the elimination of friction. The contemporary default is that AI labor displacement is inevitable and that UBI is the response. Each default became a default because it was easier than the alternative. The alternative in each case required deliberate construction. Construction is more expensive than acceptance. That is why the defaults win.
This essay argues that the construction is worth the expense. It also argues that the construction has a specific shape. The naming of the pattern was the first move. The naming of the alternative is the second. The choice of the alternative is the third. The previous essays did the first. This one is about the second and the third.
Refusal Is Not Protest
Refusal does not look like a protest. Protests respond to coercion. Pacification is not coercion. Holding a sign at a Brave New World administrator’s office accomplishes nothing because the administrator is not the problem. The administrator is responding to incentives. The pacification is the cumulative output of those incentives. Refusing the pacification means refusing the incentives that produce it, which means refusing the assumptions those incentives encode.
The assumption being refused is the chain. AI capability will accelerate. Therefore, displacement is inevitable. Therefore, UBI is the answer. Each link in this chain has the texture of a fact. None of them is a fact. The first is a prediction with some grounding but enormous uncertainty about pace, scope, and reach. The second is a deployment choice dressed as a technological necessity. The third is a policy response chosen because it leaves the deployment choices upstream untouched. Refusal begins by breaking the chain at the second link.
The Prior Questions
What conscious steering requires can be stated. It requires asking the prior questions. They have not been asked at scale. What does this work contribute to the people doing it, and what is lost if it disappears? What is the relationship between this work and the meaning the worker draws from it? What would deployment look like if displacement were not the default? These questions are answerable. They haven’t been answered because they weren’t asked seriously.
Asking them seriously requires more than goodwill. It requires the actors with the most influence over deployment to treat the work the technology touches as something other than a cost to be optimized away. It requires governments to treat displacement as something other than a category of harm to be backfilled with transfers. It requires the people most likely to lose their work to participate in the conversation about what that work was for in the first place.
This last point is the one that matters. The Roman populace did not participate in the conversation about what the grain dole was meant to replace. The Brave New World population was conditioned out of the capacity for conversation. The contemporary moment is different. The pattern has been named. The naming has not yet been universalized. There is still time to ask the prior question. The question is whether enough of us will.
The Four Commitments
The alternative has a name. Augmented Human Intelligence. AHI. The framework starts from the premise that the question of what to do with AI is the wrong starting point. The right starting point is what humans are, what they need to do for themselves to remain themselves, and how a technology that exceeds them at certain tasks can be configured to extend rather than replace them.
The framework rests on four architectural commitments.
The first point is that work is augmented rather than eliminated. The customer service example from the first essay clarifies this. One worker with AI augmentation can do more than five workers without AI augmentation, and four workers have been laid off. Both deployments are available, one that compresses headcount and another that compresses friction. The choice between them is architectural, not technological. AHI chooses augmentation.
The second is that judgment is preserved. AI excels at producing options, generating drafts, surfacing patterns, and accelerating the mechanical parts of work. Judgment is the human contribution. Which option to choose? Which draft to publish? Which pattern to act on? Which acceleration to refuse? The architecture insists that the judgment stays with the human, not because the AI is incapable, but because judgment without consequence-bearing degrades over time, and only humans carry the consequences.
The third is that attention is cultivated. Attention is the finite resource that the entire AI economy is competing for. The default deployment of AI consumes attention through feeds, alerts, infinite production, and the constant offering of new options. The AHI deployment treats attention as the upstream resource it is and configures the technology to protect it. The system gives the user fewer options, not more. The system surfaces what matters and suppresses what does not. The system trains the user’s capacity for attention rather than mining it.
The fourth is that consequence-bearing is maintained. This is the deepest of the four. The wisdom-knowledge gap is real because wisdom requires that the actor carry the stakes of the action. The doctor who decides on a treatment carries the stakes for the patient. The architect who designs the structure bears the responsibility for the people who will use the building. The teacher who chooses a curriculum carries the stakes for the students. Stripping the stakes from the actor strips the wisdom from the action. AI does not carry stakes. Humans do. The architecture keeps consequence-bearing on the human side.
These four commitments are not separate features. They are aspects of one architectural choice. The choice is for humans to remain the load-bearing element in the system, and for the technology to be configured to extend what humans do rather than to occupy the place humans once held. The Roman pacification removed citizens from the load. The Huxleyan pacification removed adults from the load. The contemporary pacification, left unchecked, will remove workers and thinkers from the load. AHI is the refusal of that removal.
Aligned with What AI Does
The architecture is also an honest reading of what AI can and cannot do. The four commitments are not only what AHI chooses. They are what hold up under rigorous capability evaluation. AI excels at pattern detection and draft generation. It does not bear stakes or carry the judgment that comes from accumulated consequence. The architecture aligns with the actual shape of what humans retain. The alignment is deliberate, designed around what AI actually does rather than what is claimed for it.
What gets preserved is what the pacification dissolves. The connection between work and meaning. The connection between meaning and identity. The connection between identity and citizenship. These are not three preservations but one preservation seen at different scales.
The Teacher and the Civic Body
Consider a teacher. The work has many parts. Content delivery, lesson design, grading, parent communication, administrative compliance, and work tailored to students of different abilities. These consume most of the teacher’s hours. There is also a different kind of work, the kind that does not show up on a list. The daily reading of thirty faces in a room and the decision about which student needs what, in this moment, before the moment passes. The first kind of work is what fills the calendar. The second kind is what makes a teacher a teacher.
The deployment of AI in education absorbs the first kind and the second one, too. The lesson is generated and delivered through a platform. The grading is automated. The student interacts with the system. The teacher becomes a monitor of a configuration she did not design and cannot adjust to the room in front of her. The teaching has been removed. The teacher remains employed, on paper, for as long as the model needs a human in the building.
The augmentation deployment absorbs the first kind and protects the second. The teacher’s observations about each student enter the system first: which child is struggling, who is engaged today, who is not, and what each one needs next. The AI takes those observations, along with grades and performance patterns, and drafts the lesson plans, worksheets at different levels, parent communications, and progress notes. The teacher reviews what comes back, edits what does not fit the room she actually has, and signs what goes out. The judgment enters the work at the top and stays present at the end. Administrative compliance and tailoring work for different abilities move off her desk. The hours that used to disappear into those tasks return to her. Those hours go to the room. The attention is cultivated rather than mined. The consequences of her decisions about each student remain hers. The four commitments are not abstractions in this configuration. They are the description of the day.
Scale this up. The civic body is composed of workers like her. The voter who is still doing work that requires judgment and bears consequences votes differently from the voter who has been turned out of that work and is now scrolling a feed funded by a transfer payment. The voter who still has stakes in the world votes with substance. The voter who has been pacified votes without. The civic muscle is exercised through engagement with the questions the work raises. The questions are real because the work is real. The institutions that depend on engaged citizens remain populated by them. The political pressure that would otherwise build against the trajectory does so as usual because the conditions that produce it remain in place.
The connection between meaningful work and functioning citizenship is not metaphorical. It is structural. The Roman populace lost both simultaneously because they shared a foundation. The Huxleyan population lost both simultaneously for the same reason. The contemporary moment threatens both because it threatens the foundation. The architecture preserves both because it preserves the foundation.
One Writer at a Desk
One last instance, this one closer to where you are reading. This trilogy was not written by a single hand at a desk. It was produced through the kind of arrangement the trilogy describes. The writer’s judgment about which arguments to make, which voices to engage, and which objections to face entered the work first. The AI took those decisions together with everything it could analyze about structure, language, and counterargument, and produced drafts, revisions, and gauntlet syntheses. The writer reviewed what came back, kept what fit the voice and the structure, cut what did not, and signed what went out. The mechanical work moved to the AI. The judgment entered at the top stayed present at the end. The consequences of publishing belong to the writer. The four commitments are not aspirational in this configuration. These are the conditions of the work.
The instance is small. One writer at a desk is not a civilization. But the instance answers one question the trilogy has been asking: whether the architecture is possible at any scale. It is. The question that remains is whether the architecture is chosen at every possible scale.
What is being protected is the conditions under which a person can be a person, and under which people together can be a polity. The protection is not abstract. It runs through the work.
Architecture or Soma
The choice is between architecture and soma.
Both are available. One is the default. One is the construction. The default arrived for the Romans without anyone choosing it. The default would have arrived for Huxley’s world for the same reason. The default is arriving in our world right now, unless enough of us interrupt.
A skeptic will object that the architecture cannot be chosen voluntarily. Augmentation may be possible at the firm level, but the firm that augments will be outpriced by the firm that displaces. The choice between configurations, even if both are technically available, is foreclosed by competitive pressure.
The objection has the same shape as the inevitability claim in the first essay, and earns the same response. Competitive pressure is not a law of nature. It is a configuration. The optimization targets, the capital allocation, the procurement standards, and the absence of countervailing pressure that produces it are human-constructed. The argument is not that augmentation will spontaneously win the competition as currently configured. It is that the configuration can be changed, and that the change is a choice rather than a hope.
Refusal lives at every level where the configuration is set. A firm that chooses augmentation over displacement. A regulator who writes procurement standards favoring the first over the second. A practitioner who refuses the version of her job that removes her from it. No single act changes the trajectory. Together, they shift what is possible.
What Is Being Protected
The architecture has its own failure mode worth naming. Even when judgment is meant to stay with the human, the dynamics of monitoring an AI that is usually, though not always, correct produce a slow erosion. The human in the loop becomes a second-guesser rather than a judge, and the skill atrophies because exercise becomes infrequent. Research on automation complacency makes this clear: humans are poor monitors of partially reliable systems. The architecture has to do more than insist on judgment as a commitment. It must be built under conditions that keep judgment exercised. How to design those conditions is a question this trilogy does not fully answer. It is worth its own essay.
The architecture is not utopia. It is not the elimination of hardship. It is not the assurance that the technology will deliver only good outcomes. It is a structural commitment to keep humans in the load-bearing position and to configure the technology to extend, rather than replace, them. The commitment has to be made by enough people, at enough levels, for the structural choice to register as a choice rather than a slogan. That is harder than UBI. It is also more honest.
What we are protecting can be named. The capacity to do work that means something. The capacity to be a citizen of a polity that functions. The capacity to be a person who notices what is happening to oneself and to one’s neighbors. These three capacities share a foundation. The foundation is meaningful work, embedded in a polity, undertaken by people who bear the consequences of their actions. The foundation is what the pacification dissolves. It is also what the architecture preserves.
The third iteration has been named. The naming is the first refusal. The architecture is what the second refusal looks like.
The choice remains.
Next Tuesday: “The Category Error at the Heart of Justice.” We treat justice as blind. The science of how people actually become who they are suggests it can’t afford to be.
Originally published on Substack.
