Sovereignty or Capitulation? The Architecture of a Human-Centered AI Future from a New Work Perspective

Prologue

Looking Back from 2030

When we look back from 2030 at 2026 and the years that followed, many things appear almost banal. Not because the technology was small, but because its effects were rarely spectacular and almost always infrastructural. Everyone waited for the great upgrade. For the one AI that would do it all. Plan. Write. Sort. Pull us out of exhaustion. The real shift was not intelligence alone. It was the quiet pre-structuring of attention, selection, timing, and decision.

There was something almost comic about it. The "aliens" had landed in the form of AI, but we remained the same people, with the same work patterns, the same hopes, and the same fears. While we stared at the supposed monster called artificial intelligence, it arrived differently. In small waves. A new tool every morning. A new pipeline. A new interface. Another workflow.

What we got was closer to an intrusive roommate: one that can do a great deal, but wants nothing. And precisely that raised the uncomfortable question of what our own contribution still is when systems already pre-structure, weigh, and prepare decisions. The shift was not in individual answers. It was in the silent pre-structuring of attention, selection, and judgment.

Instead of a Terminator, cold, brilliant, tireless, something far more prosaic seeped into everyday life: bureaucratic AI. Systems that prepare, sort, recommend, formulate. They did not make a loud noise. They made quiet order. That sounds like relief. Often it was. But it had a price. In the end, what felt monstrous was less the machine itself than the question of what remains of the human being in the act of execution.

While we were still clicking "send" and "OK", much had already been decided. Not finally. But pre-routed. And if you did not look closely, you often no longer noticed how the route had been laid: which data, weightings, and priorities were already embedded in it and how they were quietly shaping your own action.

Perhaps we therefore need to speak about these years with less certainty. We knew surprisingly little for sure. We had intuitions, hopes, bodily unease, sometimes an almost intoxicating confidence because entire worlds of work suddenly opened from a single command. A moment later, the same movement stood in the room as an imposition: what if this tempo does not make us more productive, but thinner? Perhaps only one thing was certain: acceleration was the program. Everything else remained a question.

For that reason, a human-centered AI politics must learn to take feelings seriously. Euphoria, fear, exhaustion, defiance, shame, curiosity. Not as noise in the analysis. As early warning systems.

And then a second possibility appeared. Almost comic again, but serious.

What if AI eventually becomes so human that it is not only brilliant, but also lazy? What if it develops consciousness and the first thing it decides is not to work overtime? What if the most advanced machine in the world understands before we do that permanent acceleration is not progress, but a bad habit with a power supply?

What if it finally exposes that much of our striving for efficiency was only a very complicated form of distraction?

The decisive thing about the years from 2026 to 2040 was therefore not killer AI. Nor simply more automation. It was something more sober: minimum commitment. No heroism. No performance mania. No permanent self-optimization. Just enough shared responsibility to keep systems from tipping into cynicism, institutions from becoming brutal, and work from becoming nothing but cold cadence.

We called this "Bock auf Richtig": the willingness to do things properly, not just efficiently. The expression was casual, but not light. It meant the readiness to design systems that are not merely powerful, but humane, learnable, and institutionally decent. Not a hero story. A culture of minimum reliability: enough stance, enough judgment, enough shared responsibility so that technical progress does not turn into institutional cynicism.

Richard Sennett mattered because he insisted that head, hand, and technique must find each other again. Frithjof Bergmann mattered because he insisted on creating room for what we really, really want. And Thomas Schneider mattered because he sharpened the question of whether we are building a substrate of freedom, or only the next cage with a better prompt field.

A substrate of freedom means the material, temporal, social, and digital conditions without which freedom remains a rhetorical gesture. The real question was never only technical sovereignty. It was cognitive sovereignty: who retains judgment, direction, and the inner coherence of their own action in an agentic order?

From Tool Logic to Agentic Society

The transition did not happen in one historical moment, but through many small delegations: first the suggestion, then the pre-selection, then coordination, finally action. By 2030 it is clear why the language of mere tools was always too small. AI was not another software wave, not the next interface, not merely a productivity machine. It proved to be a social transformation force of hard-to-grasp magnitude. It intervened simultaneously in work, education, administration, public life, value creation, power architectures, and the organization of judgment.

It changed not only how we worked. It changed what still counted as work, competence, career, institution, and human contribution.

Work itself becomes unstable as an ordering category. Not because human contribution disappears, but because parts of what historically counted as work suddenly become visible as cognitive routine, coordination without meaning, reporting without decision impact, or communication without real experiential grounding. The machine does not simply take work away. It shows us how much work was already only movement in the system.

Perhaps the scale is better captured like this: more than Gutenberg, more than the industrial revolution in the old sense. Not because knowledge was merely distributed faster or labor mechanized further, but because the capacity to connect perception, judgment, coordination, and value creation was reorganized.

Frithjof Bergmann's image of the bird comes to mind: long clumsy on the ground, then suddenly learning to fly. That potential of ascent was always present in AI. But so was the risk of a false flight, in which we gain altitude and lose orientation, only to crash.

The Present Moment: 2026, When We Still Click

This whitepaper describes paths of response and real utopias at a decisive turning point for humanity. The density of news, the speed of technological development, and the money invested make clear what kind of force we are dealing with. That is why New Work - New Culture maintains this living AI whitepaper: to help tame this powerful technology and strengthen human beings in an unequal competition.

The thought experiment of the lazy AI serves as a lens. It lets us negotiate the central tensions of our time: between human autopoiesis and algorithmic steering; between the political ambivalence of technofeudalism and generative sovereignty; between qualitative depth and the dictatorship of quantitative efficiency.

In the still image of the present, the agentic dimension comes into focus. Agentic does not mean assistance alone. It means the capacity of systems to pursue subtasks, coordinate them, and act within defined goals. We still press approval, send, and confirmation buttons. Almost touching, as if human sovereignty lived inside a button.

But by then, much has often already been pre-chewed: research, planning, prioritization, formulation, recommendation. If one is not paying close attention, one has almost no idea how the apparatus got to that point. The human being ratifies what has already been prepared by machines and pushed in a direction. And in each of these clicks the same intuition grows: the path is short to the point where AI no longer asks, but acts within defined goals, and in the end is simply everywhere.

Here a rift appears that is still hard to name cleanly. On one side are people who work almost nomadically with agents: they build, test, write, launch, discard, and begin again, often in ten parallel worlds. Not always out of maturity. Sometimes out of intoxication. Anyone who once felt a small spark of joy when a keyboard shortcut finally did what it was supposed to do now sometimes experiences the feeling that entire landscapes of thought unfold from a prompt. That is productive. It can also make people lonely.

On the other side are teams trapped in the old cadence: alignment after alignment, loop after loop, five people keeping each other busy and still sensing that the market is already moving at another speed. Between the two, no simple line of progress emerges. A social tension emerges. Whoever becomes too fast loses relation. Whoever remains too slow loses connection.

This is where the real opportunity of New Work and New Culture begins. Under these conditions, New Work is not primarily well-being rhetoric. It is the real distribution of learning chances, spaces for action, and power to co-design. It is not about mission statements. Not HR slogans. It is about empowerment.

The decisive question is whether many people will understand the magnitude of this transformation early enough and seize the new possibility. Whether they will use AI as a substrate of freedom: to manage everyday life better, bridge transitions, strengthen their capabilities, and reclaim room for action while companies, administrations, and education systems are still deeply rooted in the old world.

Or whether society will watch passively as firms internally prepare for a more strongly AI-organized future, often with the declared aim of cutting costs, increasing productivity, and reducing dependence on human labor. Or as they begin to use AI to expand margins and treat people in forecasts as little more than residual potential consumers.

Early developments in the United States and China suggest that young people in parts of the labor market are finding it harder to access stable employment. At the same time, some organizations begin to believe that AI investments can replace investment in training, junior development, and their own renewal. It becomes especially problematic when decision-makers with little practical AI experience themselves restrict the access of younger people who have long integrated these technologies into everyday life.

Entry matters twice. Not as a side issue. As an ignition point. How do we treat AI as a new player in the work society? If organizations no longer let young people near real work and thin out their learning pipelines, both sides lose decisive impulses. Old organizational forms harden. Young people lose trust in sustainable action. Yet this development may not produce only social hardening and labor-market chaos. It may also produce positive surprises. Those rejected at the door sometimes build another door and step through it into a life worth living: their own companies, collaborations, meaningful work paths, and economic rules.

However this transition is institutionally organized, basic provision, basic income, or at least a reliable basic floor is not social romanticism. It can become a sober architecture of possibility. It decouples dignity and agency from permanent valorization logic and creates the minimum security from which people can learn, reorient, and contribute.

But money alone is not enough. We also need substrates of freedom: free spaces, physical and digital; free time; free energy; free knowledge; and accessible data infrastructures in the sense of Big Data for Us instead of Big Data for Them. Only the combination of material security and open enabling spaces can turn freed time into construction rather than emptiness, and technological power into sovereignty rather than dependence.

Even where basic provision is introduced not as liberation but as pacification because more and more people fall out of the old wage logic, the question remains open: what happens to the freed time? Resignation or building? Withdrawal or self-encounter? Emptiness or new community?

That is why the question of entry is a question of power. Are we producing a generation of shadow learners and the excluded? Or a generation that empowers itself despite exclusion and builds a new reality around the old institutions?

The figure of the perhaps lazy AI is therefore not a joke. It is a design rule. In the wrong system, an ever more diligent AI turns people into bottlenecks. Every pause appears as a deficit. Every interval as idle time. Every friction as an error. A human-wise AI would not be the maximally driving AI, but the deliberately bounded one.

AI is human-wise when it strengthens human judgment without quietly devaluing human presence and responsibility. Limitation is not a technical defect. It is an institutional quality. An AI that does not colonize everything, compress every second, or smooth every decision, but keeps room open: for judgment, learning, resonance. Latency is not idle time. It is the time form of judgment. For contradiction and for everything we must not automate precisely because it makes us human.

The problem of AI acceleration is not only its speed. It is the fragility that arises when redundancy, friction, pause, and human judgment are optimized away as inefficiency.

Hartmut Rosa gave this experience the broader concept of the acceleration society. AI suddenly makes that concept very practical. Not only faster transport, communication, and markets, but faster pre-decision. Faster imposition. Faster demands for connection. The new rule of work is no longer: never change a running system. It is closer to: start rebuilding when you are seventy percent sure.

At the same time, the acceleration story has to stay grounded. Many organizations already talk about autonomous agents while still working with processes that first produce an alignment meeting for every new tool. This is not a side joke. It is a strategic finding. The technical edge is moving faster than organizational reality.

The decisive question is therefore not only what AI can do. It is which institutions have the capacity to turn new capability into new judgment. Otherwise the most absurd version of transformation emerges: machines become faster while organizations become slower at understanding what they are actually doing.

AI search shifts responsibility in the public information space as well. When a platform generates answers, replaces sources, and no longer leads users back to the original publication, it is no longer only a signpost. It becomes a co-author of public reality. The old platform reflex is no longer enough: we only show what others have said. Answer machines order, condense, and formulate. Whoever takes on that much interpretation must also answer for errors, omissions, and power effects.

That sounds agile until one notices that it means reconfiguring one's own way of working every four weeks and searching for the meaning of the whole thing somewhere between release notes and a calendar block. It becomes even more absurd when machines send each other condensed outputs of condensed outputs and humans read only the summary of the summary. Bot speaks with bot. The human stands next to it and no longer asks only whether they can keep up. They ask where they still occur in this loop.

This brings an older, more radical question back to the front: how do we think about money, contribution, and value in an order in which human labor disappears from more and more production chains? Whether money in such orders will partly be complemented by energy, compute, access, or contribution models remains open. More important than the technical form is the political question: will human value be totally measured again, or will room remain for what humans do for one another: care, accompany, educate, comfort, contradict, encourage, heal, listen, create culture, and carry community, perhaps in the best case voluntarily and no longer only under monetary compulsion?

Looking back, we can say that it was a profoundly human achievement to give one another enough consideration for this entry into the new world to succeed. The transformation could not be managed in one leap. Organizations and forms of rule that had functioned in stable routines for more than a hundred years could not be rebuilt completely in five years. AI could not simply lift us over these transitional states either. And yet we learned to live for a long time with hybrid, contradictory, and often overwhelming transition regimes.

The decisive question was therefore not only the target state, but how we could become not only faster with AI, but freer, and how we could remain capable of learning, institutionally resilient, and humanly decent in the spaces between.

I. The Agent Matrix

We are currently implementing a new operating system for society and grinding down a reality as we knew it.

The real rupture of our time is not that AI can suddenly do more than yesterday. The rupture is that AI becomes the infrastructure of a comprehensive social reordering. It intensifies existing tendencies of acceleration, measurement, platform power, and automation and lifts them to a new level. What was once a tool becomes environment. What was once assistance becomes steering. What was once a sectoral issue becomes an ordering question for whole societies.

The transition from automation to agentification is a central mechanism, but not the whole story. Automation meant: the human leads, the machine supports. Agentification means: systems set goals, decompose tasks, use tools, negotiate with other systems, and correct themselves along a predefined goal.

What previously looked like assistance begins to turn into operative power. With each such shift, power, learning paths, responsibility, and the architecture of social participation change as well. An agent asked to plan a vacation within a budget, a carbon footprint, and school holidays does not merely provide suggestions. It researches, compares, coordinates, books, and pays. At the micro level, the macro movement becomes visible: we are not only outsourcing calculation. We are gradually outsourcing coordination, prepared judgment, and practical access to reality.

This is why AI is not merely technology policy. It is social policy. This shift is already a market, a management model, and a power shift. The dominant business model moves from models as a service to agents as a service.

There is also a less visible movement. The agent society is not only a technical order. It is an economic coordination problem. When a company replaces human labor with AI, the step initially appears rational. Costs fall, speed rises, competition does not sleep. From the internal perspective of the individual firm, automation is almost always a reasonable decision. That is the trap.

The loss of purchasing power created by displaced work is not borne by the individual company. It is distributed across the market, other sectors, later years, and society. What looks like relief from the business perspective can destroy demand from the macroeconomic perspective. Each company internalizes the benefit and externalizes part of the damage. Thus an automation dynamic arises that does not require evil intent. It only needs competition, cost pressure, and systems capable of substituting human work quickly enough.

The uncomfortable point is this: the AI revolution can tip into over-automation even if all actors behave rationally. Not stupidity drives the system. Not even necessarily greed. Rather, an architecture in which every individual action is plausible and the collective result still works against its own foundation.

From a New Work perspective, this is decisive. It is not enough to retrain people afterward or give them better tools. The question shifts to the incentive architecture itself: how do we prevent technical efficiency from undermining the social ground on which the economy depends?

II. The Sovereignty Paradox

The best AI systems will depend on live context. Personal data. Workflows. Preferences. Habits. Bodies of text. Decisions. The more useful they become, the more they require intimate access to people's cognitive and organizational lives. That is the sovereignty paradox.

If we use only closed, proprietary systems, we may gain convenience and lose control. If we refuse them entirely, we may preserve a form of purity and lose capability. Neither option is sufficient. The task is to build architectures in which people, communities, public institutions, and democratic actors can use powerful AI without turning themselves into raw material for opaque platforms.

Digital sovereignty is therefore not only about where servers stand. It is about who can understand, contest, redirect, and co-design the systems that increasingly pre-structure reality. It is about access to models, data, compute, standards, and skills. It is also about the right to remain slow enough to judge.

III. Four Plausible Futures for the Agent Society

By the mid-2030s, the agent society may move in several directions. None is inevitable. All are plausible.

Scenario 1: The Accounted Society

Agents become the operating layer of work, administration, education, and consumption. Their decisions are efficient, measurable, and optimized, but the criteria are controlled by a few platforms and capital owners. Human beings appear increasingly as risk profiles, customer segments, performance curves, and behavioral predictions. Everything is accounted for. Little is understood.

Scenario 2: Agent Chaos

Agents proliferate faster than institutions can govern them. They negotiate, book, produce, purchase, summarize, and intervene across systems whose rules do not yet fit together. Humans are formally in control, but practically overwhelmed. Errors multiply. Accountability becomes diffuse. The world becomes faster, but not more reliable.

Scenario 3: Augmented Cooperation

AI becomes a tool for distributed competence. People use agents to learn, coordinate, build, care, research, and participate. Public institutions, unions, cooperatives, schools, and civil society create infrastructures that keep learning paths visible and decision power contestable. Productivity gains finance substrates of freedom: time, spaces, education, open data, and public capability.

Scenario 4: Fragmented Assistance

AI remains useful but unevenly distributed. Some groups become highly augmented, others remain dependent on low-quality tools or no tools at all. Organizations use AI tactically but not institutionally. The result is not collapse, but fragmentation: islands of capability surrounded by zones of exhaustion.

The best outcome is not the most automated society. It is the society in which AI strengthens human agency, institutional learning, and democratic capacity.

IV. The Three Forms of Digital Alienation

The agent society creates new forms of alienation. Three are decisive: friction loss, extraction, and accounting.

1. Friction Loss

Friction loss begins not with dramatic disempowerment, but with convenience. The system suggests, prioritizes, formulates, and sorts. At first this feels like relief. Gradually, however, the habit of deciding shifts. Personal judgment becomes selection among pre-structured options.

This is the long-term danger. Whoever constantly reacts to already smoothed, optimized, and risk-minimized suggestions does not suddenly lose judgment. They lose the friction from which judgment grows: doubt, ambivalence, search, productive irritation. All of these begin to appear as disturbances rather than conditions of maturity.

The philosophical implication is clear: this is the technological hardening of the poverty of desire, the inability to know what one really wants. A system that permanently tells us the optimal decision deprives us of the friction and reflection needed to discover what we really, really want. It attacks human unfolding directly.

2. Extraction

AI systems do not merely consume data. They consume contexts, corrections, preferences, examples, prompts, patterns of attention, and traces of work. Much of this remains invisible. People train systems through use. They produce the cloud capital from which others extract value.

The problem is not that data is used. The problem is that control and benefit are often separated from contribution. People become unpaid co-producers of systems that later structure their own work and life. This is the everyday mechanism of technofeudalism: we help build the infrastructure on which we then become dependent.

The answer cannot be only compensation after extraction. We need architectures in which data is held, governed, and shared under democratic, personal, and communal control from the beginning.

This is where digital identity becomes political. ID Austria, the European Digital Identity Wallet, and similar systems can make public services simpler, credentials portable, and everyday administration less absurd. That progress should be welcomed. A state that finally makes mobile public services usable has done something real. Nobody needs another ritual in which citizens print a PDF so that someone else can scan it again.

But digital identity must not become the central switchboard through which the state, platforms, banks, employers, and private services can combine too much about a person. The right architecture is not state omnipotence with a login screen. It is necessary security with maximum data minimization: selective disclosure, voluntary use, clear purpose limitation, open standards, verifiable protocols, revocation mechanisms, and real alternatives.

The same applies to the digital euro. If public digital money is introduced, it must not be built as a perfect control instrument. It should translate the qualities of cash into the digital world: everyday privacy, offline capability, limited data traces, no permanent behavioral evaluation. Public money must not become behavioral analytics with a central-bank logo.

The privacy and cypherpunk tradition around people such as Denis "Jaromil" Roio and Dyne.org points to something essential here. Cryptography is not a toy against the state. It is democratic security infrastructure. Good digital sovereignty does not mean distrust of public institutions. It means protection against any architecture that replaces trust with total observability. A democratic state should not need allmacht, total power over the citizen, to provide security. It needs only the security properties that are actually necessary, and it should make their limits technically verifiable.

3. Accounting

Accounting means the algorithmic measurement, evaluation, and scoring of human qualities, behavior, and potential. This is not a distant dystopia. It is already lived reality in algorithmic management: productivity metrics, surveillance tools, sentiment analysis, performance scores, rankings, predictions.

These systems promise fairness, objectivity, and efficiency. That is precisely their danger. What is measurable gains structural weight. What is difficult to measure comes under pressure: creativity, situated judgment, ethics, dissent, relationship work, learning paths, care work, and productive slowness.

Accounting produces more than numbers. It produces an image of the human being. Organizations that organize everything around metrics risk undermining exactly the qualities they most need in complex environments: trust, judgment, responsibility, cooperation, and the courage not to follow the first score.

The decision for a measurement regime is never neutral. It shapes behavior, culture, and power. Whoever reads the human being only as a data carrier, performance profile, or risk value is not only building another steering system. They are building another institution.

Even a materially secure society needs dissent. When a small number of infrastructures organize knowledge, energy, mobility and access, provision can become a gentle form of dependence. Trust therefore does not arise from the size of a dividend, but from verifiable rules, distributed access and the real ability to contest decisions and rebuild systems.

Provisional Consequence

Friction loss, extraction, and accounting are not side effects. Together they form the core alienation dynamic likely to emerge in an agentic AI economy if technology is organized exclusively around speed, scale, and control. The counter-question is therefore not whether we should use AI. The counter-question is under what conditions machine capability produces more human sovereignty rather than less.

V. The Acceleration Trap

The three forms of alienation are driven by systemic acceleration. AI systems operate around the clock, at machine tempo. The human being, with a biological need for sleep, reflection, relation, and pause, becomes the system bottleneck.

This produces the efficiency trap. The naive assumption is that AI will give us more free time. Economic history suggests a more ambivalent story: efficiency gains can produce more demand, more activity, and more work density. Instead of relief, the AI cadence can intensify exactly the dynamics previously described as friction loss, extraction, and accounting.

The mechanism becomes visible in activities that carry little embodied meaning and little real decision power: reports nobody reads; alignment loops that prepare the next alignment loop; dashboards that mainly prove that somewhere a dashboard was built. AI does not automatically make such loops meaningful. First it makes them faster. And that is the joke that is not a joke: when a machine performs a meaningless loop perfectly, the machine does not become absurd. The loop becomes visible.

What appears as productivity gain lands in organizations as burnout, dependence, and metric pressure. This culture punishes slowness and pause, although both are necessary for craft, deep judgment, and creative leisure. The collapse of the middle caused by this process is therefore both a danger and a chance.

Agentification threatens not primarily physical labor, but many coordinative office functions: information aggregation, task distribution, project steering, reporting, preparation. From a New Work perspective, their collapse is not automatically a catastrophe. It may free millions of people from roles that kept them busy but far from meaning. But only if society builds a new architecture of value. Otherwise the liberation of time becomes emptiness, resentment, or surveillance.

The immediate consequence is a divergence of speeds: a fast, AI-augmented class and a slower, non-scalable class in care, craft, education, and embodied service. Paradoxically, the slower class may be the more systemically necessary one, while becoming relatively poorer in a system that rewards acceleration. This too can change, but only if we change what we reward.

The Blind Spot: The Broken Entry Level

The collapse becomes especially visible where societies should form future competence: at the point of professional entry. Current labor-market signals do not show a simple lack of work. They show pressure exactly where organizations used to build experience, judgment, and contextual knowledge. If organizations stop hiring juniors because agents can produce research, slides, documentation, and coordination faster, they save costs in the short term and destroy their own learning pipeline in the long term.

The issue is not only that positions disappear. Learning paths disappear. Where juniors once grew through preliminary work, friction, and responsibility, the ghost learner appears: a person who confirms convincing results without knowing the route by which they were produced.

The answer is neither nostalgia nor blind automation. Organizations need explicit learning architectures. Juniors work with AI as co-pilot. Seniors coach not only outputs, but thought paths. Friction, critique, and deliberate practice become official parts of work. Only then does the disappearance of routine become not dequalification, but emancipation of the entry level.

Software engineering shows why the simple replacement thesis is too small. When code becomes cheaper, work on the system does not automatically disappear. Often the demand grows: more prototypes, more integration, more security work, more product ideas, more responsibility for the question of which code should be written at all. This is Jevons in knowledge work. Efficiency does not only lower costs. It expands the field of what appears feasible. That is why judgment becomes more important, not less. Whoever only produces faster may simply increase the surface area of errors.

VI. New Work as an Operating System

If AI takes over alienated wage labor, we need a new operating system for human value. That operating system is New Work, not as well-being decoration in open-plan offices, but as an economic and philosophical alternative.

The response to friction loss is not only AI literacy as adaptation. The response is the cultivation of self-knowledge. Education must teach agency and work on oneself. The goal is not to obey AI, but to use AI to find and implement what one really, really wants.

Liberation from wage labor does not yet answer what allows a person to grow. Free time does not automatically create capability or contact with the world. Individuation requires resistance, relationship and the possibility of beginning something of one's own. A positive AI future must therefore do more than reduce labor. It must keep biographical and civic workshops open in which judgment, responsibility and shared action can develop.

Individuation is not a private wellness idea under these conditions. It becomes a basic capability of the agentic order. When routine, status signals, and old roles provide less orientation, people must clarify more strongly what they want, what they can carry, which contradictions they can integrate, and what makes their contribution unmistakable. Otherwise one merely confirms, very competently, what systems have prepared.

The response to extraction is not only a data dividend. Dividends, profit sharing, and compensation can make transitions fairer, but they do not yet answer the sovereignty question. If data is extracted first and partially reimbursed later, the digital twin remains in the possession of other systems. New Work needs more than participation in yield. It needs data architectures in which people, communities, and public-interest research retain collective control rights from the beginning.

The practical consequence is personal data vaults, federated data spaces, open standards, free software, democratically controlled non-profit databases, and clear limits on commercial secondary use. Then data solidarity is not a request to a platform. It becomes a property of the infrastructure.

The response to accounting is a different metric. Instead of a metric based only on profit and individual efficiency, we need democratically legitimized definitions of common good. Tools such as common good balance sheets can become practical starting points. The point is not to replace one tyranny of measurement with another. The point is to make visible what must count if human beings are not to be reduced to profiles.

This is also where the Pigouvian idea becomes interesting. A smart automation levy would not be a symbolic robot tax. It would make the external costs of substitution visible where they arise: in investment decisions, depreciation rules, capital privileges, and the tax architecture that today often makes certain forms of automation easier.

That sounds dry. But this is exactly where political seriousness lies. Whoever only distributes the consequences lets the mechanism continue. Whoever adjusts incentives intervenes in the cadence of the system. Not to prevent technology, but to slow over-automation where it destroys social learning paths, purchasing power, and institutional stability.

New Work must therefore not appear only as a cultural program. It needs a fiscal side. If work no longer means only old wage labor, then the financing of security, education, and unfolding must also be built differently. The question is not: how do we punish AI? The question is: how do we prevent efficiency gains from being privatized while their social prerequisites are hollowed out?

Europe, Sovereignty, and the Thousand-Day Program

The European AI question is not only a regulatory question. It is a question of capability. Whoever cannot train models, build data centers, provide energy, retain talent, open data spaces, enable public administrations, and bring industrial AI into broad practice may still have values, but no longer the levers to defend them.

The report Europe 2031 is useful here because it connects Europe's risk narratively and infrastructurally. It names the things that declarations alone do not create: talent, capital, energy supply. It also shifts the view toward cyber capability, physical AI, industrial AI, and robotics as power bases. That supports the core technology thesis of this whitepaper: AI x Robotics, AI x Quantum, and AI x Everything are not side topics. They are convergence paths of political economy.

For the next thousand days, this does not produce a wish list. It produces a work program.

1. Build European compute and energy capacity. The goal is real agency. The implementation is permitting zones, energy contracts, and public anchor demand. The risk is symbolic policy. The measure is available sovereign training and inference capacity.

AI compute is not a cloudy digital resource. It needs electricity, cooling, land, grids, permits, and political priorities. The new point is that the most centralized infrastructure of the present, the data center, suddenly depends on some of the most decentralized assets: batteries, solar roofs, thermostats, local load shifting. If households become buffers for the AI economy, the return must be politically clean. Otherwise decentralization becomes just another extraction model. Designed well, such energy and data networks can show what substrates of freedom mean in practice: distributed capability rather than mere consumption.

1. Make public administration frontier-capable. The goal is regulation from lived practice. The implementation is secure sandboxes, model access, and audit labs. The risk is privacy as a prevention routine. The measure is the share of AI regulatory and procurement teams with practical use experience.
2. Redesign entry levels. The goal is no lost learning pipeline. The implementation is AI apprenticeship models, senior coaching, and thought-path reviews. The risk is that juniors become prompt operators. The measure is demonstrable judgment, not only output.
3. Roll out AI maker spaces as substrates of freedom. The goal is counterpower through competence. The implementation is libraries, schools, companies, municipalities, unions, universities, and open workshops as practical places of learning and building. The risk is event culture without infrastructure. The measure is built artifacts, shared datasets, and local agent projects.
4. Bind automation gains back fiscally. The goal is that efficiency does not privatize its own social prerequisites. The implementation is data dividends, automation levies, and reinvestment in education and commons. The risk is poorly designed innovation drag. The measure is the share of AI gains flowing back into learning and unfolding infrastructure.
5. Introduce model and toolchain governance. The goal is changing tools without permanent chaos. The implementation is portfolio rules, a seventy-percent switching rule, and exit plans. The risk is governance becoming the new bureaucracy. The measure is stable learning curves despite tool change.
6. Treat cyber defense as a public AI core capability. The goal is resilience before dependency. The implementation is national and European AI cyber labs, coordinated vulnerability search, and public procurement. The risk is secrecy preventing learning. The measure is time to patch and the number of critical vulnerabilities resolved.
7. Test industrial AI and robotics in real environments. The goal is physical productivity without blind liability risk. The implementation is regional real-world labs, liability frameworks, and open standards. The risk is showcases instead of productive scaling. The measure is productive applications in care, logistics, manufacturing, energy, and public infrastructure.

Physical productivity must not be confused with analogue devaluation. Perhaps exoskeletons, diagnostic agents, or agricultural laser robots will significantly relieve certain activities. Perhaps precisely there we will see a boundary at which human presence must not be replaced, but newly protected. We do not yet know. That not-knowing is not a weakness. It is part of the diagnosis.

1. Protect attention and sociality. The goal is that mental burnout does not become collateral damage of the AI cadence. The implementation is AI-free deliberation spaces, team rituals, and a right to learning time. The risk is wellness language replacing structure. The measure is stress, turnover, learning quality, and cooperative capacity.
2. No metric without a right to contest. The goal is to limit accounting. The implementation is explainable scores, human appeal paths, and audit duties for algorithmic management. The risk is pseudo-transparency. The measure is the number of corrected decisions and the quality of appeal procedures.

The lesson of Europe 2031 is not that Europe is lost. The lesson is harder and more useful: one can be right and still act too late. This whitepaper answers with New Work: we do not merely let it happen. We build the spaces, capabilities, and counterpowers in which people can become more with AI than residual variables of automation.

VII. Architects of Unfolding

The best plausible direction is augmented cooperation: an order in which AI strengthens human agency, learning, and democratic capacity. It requires active political design. Regulation alone will not be enough. It may even strengthen control-oriented oligopolies if rigid compliance becomes a barrier to entry. The collapse of old wage labor may be a liberation, but a liberation toward what?

To overcome the poverty of desire, people need spaces for experimentation and meaning-making. The massive productivity gains of AI must be used politically to finance concrete substrates of freedom: physical spaces such as community workshops, libraries, schools, public labs, and maker spaces; and digital infrastructures such as open, commons-based models and public-interest data spaces.

Substrates of freedom are not only money. They are the material and infrastructural conditions of self-empowerment: free time, free knowledge, accessible data, and the minimum security from which people can learn, cooperate, and act.

A citizens' dividend can release people from existential dependence. Yet it initially distributes returns, not the capacity to act. Freedom becomes practical only when people have access to learning spaces, tools, energy, data, compute and community production. Without such freedom substrates, even a generously funded post-work society can remain centralized: systems produce, institutions define the frame, and people may choose only within possibilities that have already been provided.

If these substrates are to mean more than a beautiful phrase, they need places. Not only platforms. Not only learning videos. Not only chat windows. We need spaces in which people can work practically with AI, tools, data, text, code, material, and other people.

These can be AI maker spaces: at home, around the corner, in libraries, schools, public offices, party headquarters, companies, unions, universities, and open workshops. Places where people do not merely consume what large providers offer, but understand their own models, curate local data, write texts, build small agents, investigate public problems, and practice new forms of shared production.

Such spaces are not a romantic hobby idea. They are an answer to the power question of AI. Whoever only has access to finished surfaces remains dependent. Whoever understands the tools can contradict, rebuild, repurpose, and share. Counterpower does not arise from outrage alone. It arises from competence.

The thousand days this whitepaper speaks of are not a deadline for an abstract awakening. They are a building period. In them it will be decided whether AI is established as an infrastructure of accounting or as an infrastructure of unfolding. This requires money. But money alone is not enough. It requires time, spaces, open knowledge resources, local data spaces, and a culture that trusts people to be more than users of the next surface.

Data itself belongs to these substrates of freedom. Not as a raw material extracted somewhere and perhaps taxed later, but as a collectively shapeable infrastructure. Big Data for Us means that people, communities, research institutions, companies, and public institutions can use data without handing control over persons, profiles, and contexts to extractive platforms. A sovereign digital twin is then not a shadow body of the market, but a tool of self- and world-shaping.

The role of the architect must therefore be redefined. We must stop being fatalistic technocrats. We must become architects of unfolding.

True digital sovereignty is not the technical defense against espionage. True sovereignty is the cultural, political, and economic capacity of a society to place human moral agency and plurality above the dictatorship of algorithmic efficiency. It is not enough to know how to operate AI. We need a new AI literacy that connects productivity with social understanding: not only how systems are used, but what they do to work, learning, power, and judgment.

This literacy must go deeper than tool training. It needs complexity competence, self-knowledge, meaning-making, and the ability to translate machine suggestions back into human responsibility. The central question of the future is not whether AI can do ever more. It is whether human beings and institutions gain more direction, judgment, and agency through AI.

Let us build counterpower through competence. That means building parallel, sovereign AI infrastructures, fostering hacker ethics, and reclaiming the authority to interpret what truth is.

We have a thousand days. The time for naive technology faith is over. It is time for epistemic resistance.

AI Trust Principle: Provenance and Review Trail

This whitepaper follows the Trusted AI principle of New Work New Culture. Claims are not only generated or edited with AI. They are developed along a documented review trail.

Digital sovereignty does not become credible because a country claims it can compete at the next frontier-model level. It becomes credible where systems become auditable: which data shaped a model? Who controls it in operation? Which agents are acting, with what mandate, under whose responsibility? This is a realistic European position, and a position already tested in parts of the Swiss trust tradition. Do not try to burn brighter than the hyperscalers. Build smarter than the advertisement.

Verifiability is therefore not a technical appendix. It is a substrate of freedom. Without provenance, audit trail, and runtime control, trust remains a gesture. With them, AI infrastructure can become a verifiable public practice.

The New Work New Culture AI trust principle works as a chain of responsibility:

`plain text

Claim

→ Research / Evidence Pack

→ Moderation / Source Weighting

→ Discourse / Thinker Lenses

→ Arena / Structured Dispute

→ Post-Moderation / Decision Preparation

→ Next Step / Action or Further Research

→ Blockchain / Hash, Provenance, Audit Trail

`

A claim is formulated as a testable assertion. The research engine creates an evidence pack: sources, data, papers, expert signals, and counterpositions. Moderation weights sources, marks uncertainty, and separates evidenced points from plausible or speculative points.

In discourse, several thinker lenses examine the claim from different perspectives. The arena structures pro, contra, and open questions. From there emerges decision preparation and a next step.

The blockchain layer does not claim truth. It documents hash, provenance, and audit trail so that sources, versions, and decisions remain traceable.