I. The Question

I.A — The trusts we are about to build

In 2026, Peter Diamandis and Alex Wissner-Gross published Solve Everything, a book that proposes a new governance institution they call an AI Targeting Authority — a perpetual, autonomous body empowered to direct compute, capital, and human attention toward humanity’s most pressing problems. The proposal is serious. The authors argue that the AI moment requires institutions designed to outlast the political cycles that would otherwise consume them: structures with their own resources, their own decision rules, their own legal personhood, their own capacity to act across generations.¹

They are not alone, and the architectural impulse they articulate is not idiosyncratic. The same instinct is producing a thicket of new aspirational structures, each in its own vocabulary but all of them solving for the same problem — how to bind the future to a commitment the present can see but the future cannot yet defend. Climate trusts designed to bind tomorrow’s legislatures to today’s emissions commitments are entering charter windows in multiple jurisdictions. Sovereign wealth funds chartered for the benefit of citizens not yet born are being capitalized as fiscal assumptions for the next century. Longevity-research escrows are being committed to fifty- and hundred-year horizons. Data trusts are being designed to hold ownership of consumer datasets across changes in corporate ownership. Compute reserves are being proposed to hold AI infrastructure for public benefit on a perpetual basis. The vocabulary varies. The architectural impulse does not.

These are the trusts we are about to build. The architects are serious people doing serious work, with the best intentions and access to the best legal talent the present moment can buy. None of the structures they are designing has been pressure-tested across generational time. We do not know how any of them will behave in their fortieth year, their seventieth year, their hundred-and-twentieth year — when the people who designed them are gone, the political world that surrounded their founding is unrecognizable, and the beneficiaries the structures were chartered to serve are still politically silent.

We do not know, that is, except for one example.

I.B — The trust we already built

The companion volume — LOOKING BACK · Schools of the Republic — tells the story of the original design at length: how the American school trust was built, what its framers thought they were doing, what legal tools they used, and what tools they did not have. Two and a quarter centuries of operation across fifty states, six chapters of synthesis, and a closing record of what holds and what fails.

The short version, for readers picking up this volume first, is this. In May of 1785, the Confederation Congress put one sentence into a long surveying ordinance. The sentence reserved a parcel — section sixteen — in every six-mile-square township of the federal public domain “for the maintenance of public schools.” The compact form was elaborated in the Northwest Ordinance of 1787, which declared “Religion, morality, and knowledge being necessary to good government and the happiness of mankind, schools and the means of education shall forever be encouraged” and made acceptance of the compact a condition of new states’ admission. The U.S. Supreme Court would later affirm, in Cooper v. Roberts (1855), that admission-act school grants create real, enforceable trust obligations on state public faith.

The framers were operating under conditions that should be familiar to anyone designing the new aspirational structures of 2026. They had catastrophic public debt. They had a barely-functional capital market. They had an extractive geopolitical environment they did not control. They had an asset that was abundant in the moment but visibly finite over time — in their case, land; in the AI architects’ case, attention and compute. And they had a beneficiary class — children — with no political power to defend its share, then or ever. They reached for the institutional form English equity courts had spent two centuries hardening against precisely the failure modes their successors would face: the charitable trust. They locked the asset into the geometry of national settlement. They stated the principle in compact articles declared to be perpetual.

It was, in its design, an impressive piece of work. It was the longest-time-horizon institution that the eighteenth century could engineer with the cognitive technology and the legal vocabulary available, and the framers built it with care. The compact they built has been running continuously for 240 years. It is older than the United States Constitution. It is older than every existing Western democracy. It is older than most of the corporate and philanthropic forms today’s institution-builders take for granted. The data from that experiment is in.

This volume is about what the data says.

I.C — What the experiment returns

The data says, briefly, that aspirational structures decay predictably without an active enforcement constituency, and that the decay is structural. It is not the contingent product of bad actors or unlucky historical moments. It is a consequence of features these institutions share with one another, and that today’s designers are about to share with them whether they intend to or not.

The features are three.

First, hyperbolic discounting — the human cognitive bias that overvalues immediate rewards relative to future ones. In an individual making a retirement decision, the bias is mildly costly. In an institution managing a 250-year trust, it is fitness-destroying. The trust loses ground every time a near-term political pressure meets a long-term fiduciary duty, and it loses ground in a direction the bias guarantees: toward the present, away from the beneficiary. The schoolchildren in the township in 1885, the schoolchildren in the township in 2026, the schoolchildren in the township in 2155 — each successive cohort is the one whose claim, on the bias’s clock, is easiest to discount.

Second, eusocial pathology in the bureaucracies guarding the trust. Eusociality is the pattern social insects share with mature human bureaucracies: the survival of the hive begins to take precedence over the survival of the species. In a long-running institution, it produces what political scientists have variously called regulatory capture, the iron law of oligarchy, and institutional parasitism. The State Land Board comes to view the trust assets as resources for its own homeostasis. The agency staff comes to view the beneficiary as an inconvenience to be managed rather than a principal to whom a duty is owed. The trust mission becomes the cover under which the agency’s own continuity is funded.

Third, the beneficiary class whose political silence the framers rarely engineered for. Schoolchildren cannot vote. Future generations cannot lobby. Climate beneficiaries are not yet born. Recipients of an AI Targeting Authority’s grants have no idea their interests are being traded away. The institutions are designed for the politically silent, and the political silence becomes the very vector through which the trust is captured. A perfect-host beneficiary, as parasitologists use the term, is one whose welfare the host institution can sacrifice without political cost. Children, future generations, the unborn, and the beneficiaries of trusts whose corpus is held in geographically distant or technically opaque assets all qualify.

Two failure modes compound from these three features, and naming them precisely is the prerequisite for any architecture meant to hold against them. The first is gradual drift — slow rotation of the trust away from its purpose, by accretion of small diversions, redefined beneficiary classes, modernized mandates, and corpus translations that each look defensible in isolation. The second is sudden directed hijack — a single political moment in which the trust’s resources are seized, its mandate rewritten, or its beneficiary class redefined wholesale, often under cover of crisis or fiscal emergency. The drift mode runs continuously and is harder to see; the hijack mode runs in episodes and is harder to oppose. The school-trust record contains both, and both are predicted by the same three features. The architecture this volume proposes must hold against both.²

The three features compound into the two failure modes. Hyperbolic discounting creates the time-preference asymmetry. Eusocial pathology creates the institutional self-preference. The perfect-host beneficiary creates the absence of pushback. The trust drifts; the bureaucracy reframes the drift as prudent management; the beneficiaries do not show up to argue. And periodically, when the drift has done enough work to make a directed seizure look like ordinary policy, the hijack arrives — a single legislative session, a single Land Board vote, a single emergency appropriation in which the corpus changes hands or the mandate changes meaning.

The data also says, importantly, that recoveries are possible. The companion volume’s Conclusion documents three findings about that record: the architecture works where it has been operated as designed (Utah’s contemporary record being the working example); the architecture has drifted in the two predictable modes named above; and every successful enforcement case in the modern record shares one institutional feature — a non-profit plaintiff at the center of the plaintiff team. The recoveries share structural features the failures do not. Those features are not philosophical. They are operational. They can be specified. They can be installed in advance. The claim of this volume is that they should be — and that the next generation of aspirational structures can be built with the operational features of the recoveries already in place, before the drift begins, rather than retrofitted after the drift has already done its work.

I.D — The argument of this volume

Section II names what the framers built in 1785 — the act of writing the trust into law — and what design choices that act made and did not make. Section III is the diagnostic of what went wrong, told as a brief national pattern (the companion volume carries the case-by-case record).

Section IV moves from descriptive pattern to causal model: it makes explicit the three features named above, traces their mechanisms, and shows why the failures occurred when and where they did, and why the recoveries occurred when and where they did.

Section V proposes a counter-architecture: seven structural defenses (codified fiduciary duties, asset restoration, individual accountability, direct distribution, independent beneficiary advocacy, specialized legal defense, mandatory fiduciary education), plus a standing constituency — an active community across generational time, capable of enforcing the formal architecture when the institutional immune system fails. Section V.5 describes the five-layer Knowledge Stack the Library and this volume are built on, as one worked example of the cognitive-infrastructure successor architecture the AI-era trusts will need.

Section VI maps the counter-architecture onto the trusts of the AI age, the climate age, and the longevity age. It shows that the same pathologies operate in the new domains and the same defenses apply, and it draws specific design implications for the AI Targeting Authority, the perpetual climate trust, and the longevity escrow.

Section VII names what civic practice looks like across generational time — the operational requirements of the standing constituency the architecture requires.

Section VIII is A Letter to the Architects — the volume’s rhetorical center, addressed in the second person to the people drafting charters now.

The argument runs across two registers at once. As a piece of historical reflection, it is a story about how a continent and a civilization built a sacred compact for its children and let it slip — and what the slipping teaches. As a piece of governance design, it is a working manual for the people now building structures whose beneficiaries will not be born for another fifty or hundred years. The two registers are not separable. The framers who wrote that one sentence in May 1785 knew they were trying something; they did not know how long the experiment would run, and they could not have predicted what its results would be. The architects of the AI Targeting Authority and the perpetual climate trust and the longevity escrow do know — at least, they will, after this volume.

What that experiment returns is what the rest of this volume is about.

Footnotes / source notes