A blueprint, drawn from a long-form discussion, jointly authored by Jan-Peter Olters and Claude (Anthropic).
Abstract. What would it take to achieve world peace, environmental sustainability, and shared prosperity? This blueprint argues that the binding constraint is not knowledge—we largely know what would help—but the political economy of acting, in which stable, self-interested incentives make collectively ruinous choices individually rational. Taking human nature as a given rather than a project to be reformed, it shows that the three goals share a single structure: each advances the claim of a party that is weak or absent at the bargaining table—the future, the poor, and the victims of catastrophes not yet arrived—while the costs of acting fall on parties that are present, concentrated, and powerful. What they hold in common, and the true centre of the problem, is the inclusion of the structurally excluded in legitimate rule-making across both space and time: justice as the mother of all global public goods and, therefore, chronically under-supplied. From this follow a strategy in seven priority-ordered layers, a set of conditions that can turn shared existential risk from a driver of conflict into a catalyst for co-operation, and a concrete, sequenced set of first steps whose two keystones are making co-operation cheap and making the losers of necessary transitions credibly whole. The objectives can be approached without limit but never fully reached; the discipline is to keep bending the trajectory towards less domination, greater capability, and wider inclusion—and never to mistake the approximation for arrival.
This essay is the synthesis of an extended dialogue between a human author and an artificial intelligence. We think the reader is owed candour about what that means, because one of the questions running underneath the whole inquiry—the concentration of technological power in few hands—implicates one of the two authors directly.
An intelligence like Claude’s can hold many disciplinary frames at once, synthesise quickly across them, and reason with a certain freedom from the tribal loyalties that distort human judgement. What it cannot bring is accountability, lived stakes, legitimacy, or the authority to commit anyone to anything. It will not pay the costs of bad advice. So its role here is the cartographer’s, not the sovereign’s: it can help draw the map, but it cannot—and should not be trusted to—decide who walks it. We have tried to let that asymmetry discipline the argument rather than decorate it. An AI that claimed it could solve the largest human problems would be performing the very move this blueprint exists to resist: an unaccountable intelligence asserting authority over collective human ends.
The honest starting point is deflationary. The intellectual problem is not the hard part. On most of what follows there is more expert consensus than public debate suggests, and much of it has been known to serious people for decades. The binding constraint has never been knowledge. It is the political economy of acting—the gap between what we know would help and what incumbent interests will permit.
That is why we set aside, early and deliberately, the comfortable story that the world’s troubles stem from the shortcomings of the men and women who run it. Leader-centric explanations are emotionally satisfying and analytically weak, and they point towards the wrong remedy: if the problem is bad people, the solution is better people, and we may all wait for the next election. The more uncomfortable and more accurate account is structural. Largely rational, even well-intentioned actors, placed inside particular incentive structures, reliably produce collectively disastrous outcomes. A climate minister who imposes unilateral costs while competitors free-ride is not unwise; he or she is defeated. The pathology is structural before it is personal.
So we adopted a constraint that shaped everything afterward: take human nature as given. Kant’s wager in Perpetual Peace is the patron principle here—that a workable order can be designed even for ‘a nation of devils, so long as they have understanding,’ because the right institutions channel self-interested antagonism towards the outcomes virtue would also produce. The task is not to improve people. It is to design the game so that decent outcomes survive being played by the people we actually have.
Peace is not the absence of conflict—conflict is permanent and often productive. It is the reliable displacement of violence by procedure: channelling inevitable disputes into non-violent, rule-bound forms. The thin version, mere absence of direct violence, is treacherous; it can describe the quiet of a prison yard or a frozen, unjust order. Because most lethal violence today is internal rather than between states, a serious definition reaches inside borders, not only across them.
Environmental sustainability is best read through the planetary-boundaries framework: the biophysical guardrails within which civilisation can persist. The reading is now severe—the 2025 Planetary Health Check assesses that seven of the nine boundaries have been transgressed, with ocean acidification newly added, and the breached ones trending worse. But the conceptual core matters more than the count: sustainability is fundamentally a claim about intertemporal distribution—an assertion that future people have standing. They cannot vote, bargain, or retaliate, which is precisely why every market and electoral cycle systematically under-prices them.
Shared prosperity carries its tension in its two words. ‘Prosperity’, following Amartya Sen, is capability—the real freedom to lead a life one values—not GDP, which measures activity rather than flourishing. ‘Shared’ is the explosive half: a claim about distribution, within and between societies. Tellingly, the World Bank has rewritten its mission to ‘end extreme poverty and boost shared prosperity on a livable planet’—an institution formally welding two of our three goals into one sentence because it concluded they can no longer be pursued apart.
Define the goals carefully and a common structure surfaces. Each is a claim advanced on behalf of a party that is structurally weak at the bargaining table. Sustainability speaks for the future. Shared prosperity speaks for the poor and under-represented. Peace speaks for the counterfactual—the victims of a catastrophe that, by definition, has not happened and so has no organised constituency. In every case the beneficiaries are diffuse, future, or invisible, while those who would bear the costs of action are present, concentrated, and powerful.
That asymmetry is the master key. And it tells us what sits at the centre of the three overlapping circles—the thing everyone is really chasing. It is the inclusion of the structurally excluded in legitimate rule-making, across both space and time: procedural justice extended to the poor, to the under-represented, and to the unborn. Justice, understood this way—as non-domination (no arbitrary power of the strong over the weak), as capability (the health and education that convert formal rights into real participation), and as voice (being genuinely heard, with recourse when one is not)—is, in an economist’s terms, the mother of all global public goods: enormously valuable, non-excludable, non-rival, and therefore chronically under-supplied by self-interested actors who cannot capture its returns. The whole problem, restated, is the chronic under-provision of justice under conditions of anarchy and short time horizons.
Any honest treatment must reckon with a new concentration of power that strikes directly at two of justice’s pillars. Control over the infrastructure of communication, computation, finance, and increasingly intelligence itself has concentrated in a small number of mostly US-based private actors. This is domination in the precise sense—arbitrary power over others’ conditions of life, exercised by entities accountable to no electorate, and increasingly over the very substrate of voice, the public sphere, now privately owned and optimised for engagement rather than deliberation. Because these actors are transnational, they land on the weakest point of the whole architecture: the near-total absence of recourse against the powerful at the international level.
We resisted the temptation to make this a story about villains; the concentration is substantially the physics of the technology—network effects, returns to scale, and a staggering capital wall. And there is a genuine counter-current: capability is diffusing (through cheap, open models) even as control concentrates. The honest design question is therefore not ‘concentration or not’, but under what rules capability spreads while control is prevented from concentrating. We have bound over-mighty private power before—antitrust, the regulatory state, constitutional limits tamed the Gilded Age trusts. The catch is that those tools were national and these powers are transnational, which throws us straight back onto the enforcement gap.
The three goals cannot be jointly maximised; the tensions are partly irreducible and the absent parties can never be fully present. There is no terminal state called ‘solved’—only a trajectory that can be bent, without limit, towards less domination, more capability, and wider inclusion. The following layers are ordered by lexical priority: the sequence is part of the design.
Zero—Survive first. Avoid the irreversible: nuclear war, climate tipping cascades, engineered pandemics, and the permanent lock-in of unaccountable power. The justification is not rhetorical but mathematical: an irreversible catastrophe is an absorbing barrier, and once it is crossed there are no further rounds over which to average gains, so the rational rule is not ‘maximise expected value’ but ‘first, avoid ruin’. This layer is also the most dangerous, because whoever declares the emergency wields sovereign power, and ‘we must survive’ is the oldest licence for tyranny. It must therefore be pursued by pre-commitment—guardrails decided in advance through legitimate process and then executed by rule rather than by discretionary emergency power. The principle: avoid the irreversible, but never by means that foreclose the future you claim to save. (That this layer is regressing—the New START treaty lapsed in February 2026 with no successor, the first time in over half a century with no binding limits on the two largest arsenals—is a measure of how live, and how neglected, it is.)
One—Manufacture standing for the absent. Where you can give the excluded a vote, do; where you can only give a proxy, give a proxy; and where even a proxy is too weak, embed their interest in a price or an automatic rule that runs without continuous political will. The social discount rate is, in this sense, the franchise of the unborn, cast on their behalf in every cost-benefit calculation. The danger mirrors Layer Zero’s: representation without authorisation can launder present élites’ preferences through the unimpeachable voice of the voiceless. The containments: prefer option-preservation over guessing the future’s preferences; prefer transparent rules over discretionary human proxies; and keep the living in the loop. The most legitimate move of all is to empower the living young—present, with valid claims, able to authorise their own representatives, and holding the longest horizon—as the future’s best available proxy.
Two—Design for devils with understanding. Re-engineer payoffs rather than moralise. Build clubs rather than universal treaties that crawl at the pace of the most reluctant—a carbon club that prices internally and border-adjusts against outsiders inverts the free-rider logic so that it becomes costly to stay out. Most apparent ‘failures of understanding’ in geopolitics are misdiagnoses: the leader usually understands, but his personal survival is decoupled from the collective interest. So the keys are to re-rig consequences, to let demonstrated success persuade by example, to appeal past the leader to the coalition he depends on, to rebuild the verification that replaces fearful guessing with knowledge, and—for the genuinely unpersuadable, who understand and reject the frame—to contain and deter rather than convert. Good design economises on understanding as well as on virtue.
Three—Stay polycentric. Refuse the single global Leviathan. A world government would itself be the concentration of power we most fear, and Kant’s warning against universal monarchy as despotism still holds. The realistic and desirable architecture is nested, overlapping, redundant authority across scales—which secures co-operation without manufacturing new domination.
Four—Sequence the trade-offs, and pay the losers. This is the most practically important layer, because its neglect actively destroys the others. The theory rests on a treacherous word: a reform is ‘efficient’ if the winners could compensate the losers—and in practice they almost never do. Compensation fails because the losers are concentrated and organised while the winners are diffuse; because the promise is time-inconsistent and therefore disbelieved; because the loss is denominated in dignity, community, and place, which a transfer cheque cannot restore; and because the instruments oversell. The ‘China shock’ is the indictment: real aggregate gains, concentrated and persistent losses, compensation that never came at scale, and a backlash now dismantling the trading order itself. The reframe that matters: compensation is not charity to the losers but the insurance premium that protects the winners’ gains from political reversal—and the current fragmentation of the rules-based trade order is the unpaid claim coming due. Done right, compensation is automatic and prior rather than discretionary and posterior; it rebuilds places, not only retrains people; it builds the new opportunity before removing the old; and it gives affected communities real voice in shaping their own transition.
Five—Bind the new private power as we once bound the trusts. Apply antitrust, regulation, and constitutional limits—but build their transnational version through the jurisdictions large enough to set global rules. Treat the genuinely public-good layers (communication, computation, basic capability) as public goods, and lean into the diffusion of capability while constraining the concentration of control.
Six—Build for self-correction, not for arrival. Because the goals are never fully achieved, success is not a just world delivered but a world reliably becoming less dominated, more capable, more inclusive, and more survivable—with the machinery of error-correction intact and the circle of standing widening over time. Institutions that fail gracefully and learn are worth more than institutions that are optimal and brittle.
Existential threats are the one class of problem that is structurally non-zero-sum, and they have, at times, midwifed co-operation—the Montréal Protocol on ozone (one of the few boundaries now recovering), the arms-control architecture born of the Cuban Missile Crisis, the institutions built from the rubble of 1945. But the base rate is sobering: climate and the pandemic, both shared existential threats, mostly produced free-riding, nationalism, and hoarding.
The difference between Montréal and the failures yields the ignition conditions—and, encouragingly, most are design variables we can engineer rather than facts we must await. Co-operation ignites when the price of the co-operative move is low and incumbents can be co-opted rather than destroyed; this is why the collapsing cost of clean technology is the load-bearing wall of the whole strategy, because it mechanically converts climate from an intractable case towards a soluble one and creates incumbents who profit from the transition. It ignites when the decisive players are few (shrink the number through clubs), when compliance is visible (cheap satellite monitoring now helps), and when some coalition will pay the disproportionate share. And it ignites at an acute focal moment—which means pre-positioning the co-operative plan so that when crisis strikes, the ready solution is in the drawer, and winning the framing battle so the crisis is told as ‘share the lifeboat’ rather than ‘close ranks’. Existential risk becomes the key only if we cut the key to fit the lock.
A blueprint that names no first move is a wish. So it is worth being concrete—and the concreteness has a logic of its own. The first steps are not the most important policies; they are the enabling ones, chosen because they cost little, reinforce themselves, and manufacture the political conditions under which the costlier moves become survivable. The sequence is dictated by political economy, not by the magnitude of the problem. Lead with what requires no sacrifice and therefore provokes no blocking coalition; lead with what builds its own constituency as it goes; lead with what a coalition of the willing can do without waiting for universal consent. The hopeful part is that almost none of this requires institutions we have never built. The scaffolding already exists—unfinished, underfunded, and unco-ordinated, but not absent. The task is largely to connect, fund, and accelerate it.
Two of these carry most of the weight. Step 1 removes the cost of co-operating; step 5 removes the fear of being the loser. Once the clean alternative is cheap and the losers are credibly made whole, a remarkable amount of what currently looks politically impossible—the clubs, the binding budgets, even the consumption ceilings that require an appeal to reason—becomes merely difficult. They are the two keys that turn the most locks, which is why they belong first, and why their habitual neglect (treating clean-tech cost as a market afterthought and compensation as optional generosity) is the single most expensive mistake in the field.
A blueprint that pretended every circle could be squared would be worthless. Three knots resist us most.
The first is the development dilemma. Shared prosperity is, arithmetically, the raising of per-capita throughput for billions, even as population climbs towards its peak—and we are already past seven of nine planetary boundaries. The collision is real and only approximately resolvable, through a structured asymmetry: protected development space for the poor to reach a decent floor; maximal, finance-and-technology-transferred decoupling so the poor leapfrog straight to clean; absolute reduction of luxury throughput at the top; and a redefinition of prosperity towards capability rather than matter. The biophysics then adjudicates an old philosophical question: the planet will not permit the universalisation of today’s rich material standard, so ‘shared prosperity’ cannot mean ‘everyone eventually consumes like the wealthy’. It must mean a decent capability floor for all, within a shared ecological ceiling.
The second is the incumbency of the present. The decisive bloc in ageing Western democracies—large, high-turnout, asset-rich, with the shortest remaining horizon—has both the votes and the assets to block intergenerational rebalancing, and the demographic arithmetic forbids simply out-voting it. This forecloses the coercive route and compels the co-operative one: cheap transitions that demand little sacrifice, appeals to legacy and grandchildren rather than abstract ‘future generations’, protection of the genuinely vulnerable old, and the visible repair of a broken intergenerational compact. The same disconnect is, globally, a geographic one—the order is governed by cohorts that are old and Northern while the demographic future is young and Southern—which is simply the ‘justice across space and time’ of the centre of the Venn, wearing a demographic face.
The third is the residue of the unpersuadable. No mechanism reliably generates understanding in a powerful actor who does not want it and cannot be coerced. The strategy does not eliminate this residue; it shrinks it—converting case after case from ‘must persuade a recalcitrant power’ into ‘the co-operative path is now cheaper, the coalition stronger, the consequences biting, the next cohort already turned.’ What remains is the irreducible domain of deterrence and patience.
If there is one posture beneath this blueprint, it is longer-sighted realism rather than idealism. The durable achievements—Montréal, the arms-control era, the postwar settlement—were built not by converts to virtue but by realists who read their interests correctly over a long enough horizon. We do not need people to become better; we need the game arranged so that the co-operative move is the intelligently self-interested one. That is sturdier than hoping for a change of heart, and it keeps faith with the constraint we set at the outset.
Two truths follow. The first is structural humility: you cannot square this circle, only approximate it—without limit, but without arrival—and the discipline is never to mistake the approximation for the thing. The second is where the real difficulty lies. We know, more or less, the shape of the answer; we derived it from first principles in the course of a single conversation. The binding constraint is not knowledge. It is the will to bind power, the legitimacy to do so, and the coalitions to make it stick—and those are things no intelligence, however large, can supply from the outside.
Which returns us to where we began—though no longer empty-handed about where to start. The walking begins, concretely, with the two moves that unlock the rest: make co-operation cheap, and make the losers whole. Drive the cost of the clean alternative towards zero so that defection stops paying, and attach credible, automatic, place-based compensation to every transition so that those asked to bear its costs are not asked to bear them alone. With those two foundations laid, most of the architecture above turns from desirable into buildable—the clubs can form, the budgets can bind, the absent can be seated, and the harder appeals to reason can be made to a public no longer afraid of being sacrificed. The map can be drawn well; this is our attempt to draw it. But the walking is human work, and it always was.
powered by Borlabs Cookie
Great article.
Promo: C4bjWthY7dcX8qi