LOOKING FORWARD · The Eighth Anchor · Prologue
A Forever Gift, Now What?
The companion volume to this one — LOOKING BACK · Schools of the Republic, written principally by my co-author Margaret Bird — has just laid out 240 years of evidence about how a particular kind of multi-generational fiduciary architecture has actually performed in the United States. If you have just put that volume down and picked this one up, you already know the substance: the architecture works where it has been operated as designed, drifts in two predictable modes where it has not, and produces enforcement output only where a non-profit substitute for the captured Attorney General has been built around it. The school-trust experiment is the only multi-generational fiduciary record long enough to teach those lessons.
This volume takes that record and asks what it teaches the architects of the next generation of forever-institutions.
I am not, by training or temperament, an essayist. I couldn’t write my way out of a paper bag, as the saying goes, and I am not going to try. What I am is a retired Oregon State professor and tree farmer who has, by accident more than design, watched the same institutional failure pattern play out three times in the course of forty years. The first time was at Tektronix in the late 1970s, where I sat in a hardware-design group while the people who had built the first graphical personal computer in 1975 failed to recognize what they had and lost the future to a garage operation in California. The second time was in the personal-computing era I wrote a textbook about in 1984 — Computing Today, with its dual-five-megabyte hard drives that had felt extravagant in the manuscript and were obsolete by the time the book reached the shelf — where institutions adopting the new tool kept treating it as a faster typewriter while the genuinely transformative thing was happening around them. The third time is the school-trust experiment, which inherited a 1785 architecture and has been quietly losing it ever since, except in the places where some small group of citizens refused to let the loss be the last word.
Three iterations. Growing stakes. The pattern is the warrant for the warning.
The pattern is this: a generation builds an institutional architecture adequate to the cognitive technology of its time. The next generation inherits the architecture without inheriting the engineering judgment that produced it. The third generation rationalizes the architecture’s purposes against whatever near-term pressures it is facing, drifts away from the original commitments, and discovers — usually too late to do anything about it — that the architecture’s original beneficiary class is no longer being served. By the time the failure is legible, the original engineers are dead and the cognitive context that produced their architecture is no longer recoverable. What remains is the wreckage, and a small group of people in the ruins trying to figure out what was lost.
That is the school trust. It is also, structurally, what is going to happen to the AI Targeting Authorities and the perpetual climate trusts and the sovereign wealth funds for the unborn and the longevity escrows that are entering charter windows in 2026 — unless the architects writing them now have access to the lessons of the school trust. Which, until this two-volume work, they have not. The school-trust record has been there to be read, in pieces, in disparate state archives and trust-law treatises and the occasional legal-historical monograph. What has not existed before is a synthesis aimed at the architects of the new institutions, in language they can act on, in time for the architecture they are writing to benefit from it.
That is what this volume tries to be.
What the architects are designing
Let me, as a systems scientist by background, try to name what is at stake.
A thousand-year multi-generational trust is a particular kind of institutional object. It is a structure that pledges resources from the present generation to a beneficiary class extending centuries into the future. It is not a contract — contracts presuppose mutual consent between identifiable parties, and the future beneficiary class is by definition unable to consent. It is not a piece of legislation — legislation presupposes a current legislative body whose members can act, and the obligations a thousand-year trust accepts have to bind successor legislatures the current one cannot enforce against. It is not a policy — policies are administered by current administrators who can change them, and the trust requires continuity across administrative turnovers it cannot prevent.
The thousand-year trust is, instead, a delegated structure of obligation — a piece of institutional engineering that has to hold, against the natural decay of every institution and the natural pressure of every generation, for longer than any of its founders can ensure. It is the most ambitious kind of institutional design any society can attempt, and it is being attempted, right now, in 2026, by people working under deadlines too short for the kind of design judgment the work requires.
The arriving structures are real, and they are arriving fast. Peter Diamandis and Alex Wissner-Gross have proposed an AI Targeting Authority — a perpetual, autonomous body with its own resources, its own decision rules, its own legal personhood, its own capacity to act across generations. Climate trusts intended to bind tomorrow’s legislatures to today’s emissions commitments are entering charter windows in multiple jurisdictions. Sovereign wealth funds for citizens not yet born are being capitalized as fiscal assumptions for the next century. Longevity escrows, data trusts, compute reserves, programmatic down-shifting authorities — the vocabulary varies, the architectural impulse does not. Each is a structure designed to bind future generations to commitments made today. None of them, as far as I can find, has a working knowledge of what 240 years of America’s school-trust record predicts will happen to it.
The framers of 1785 did not have that knowledge either. They had to write the architecture of the school trust without any prior multi-generational fiduciary record to work from, because there wasn’t one. They were the experiment. The next generation of architects is in a different position: they have an experiment to learn from, if they want to. The school-trust record contains the data the framers did not have.
What the data says is that legal rigidity, written by architects who knew what they were doing, is necessary but not sufficient. The architecture has to be defended. The defenders are the constituency. The constituency does not form on its own. It forms because the architecture has been engineered, deliberately, to produce the conditions under which a constituency can form. Margaret’s mechanism — Increase the revenue to schools, get it directly to every single school where the parents, the teachers, and the principal are deciding how the money is implemented. And all of a sudden, you have built a huge constituency — is the school-trust experiment’s clearest single procedural finding. The mechanism is structural rather than rhetorical. It tells you what to do if you are designing for sustained constituency formation across generational time.
This volume is, in significant part, an extended argument for that mechanism’s generalization. The principle is the same whether the asset is land or money or compute or data. Route the institution’s consequential decisions to the local sites where deliberation already happens. Let the deliberation produce visible artifacts that bear the deliberator’s name. Recognize the artifacts in a public record. Through that loop, the deliberators become a constituency that defends the architecture politically when the architecture is later attacked — as it will be. Skip any link in the loop and the constituency does not form. Skip the loop entirely, and you have written an architecture that cannot defend itself.
Why this is possible now and not earlier
There is one piece of context that the school-trust framers did not have, and that the next generation of architects does have, that changes what is structurally possible.
The framers in 1785 could not engineer for the constituency that would defend the architecture across two and a half centuries because the cognitive technology of 1785 could not hold institutional context across generational time. A citizen-trustee in 1885 had no way to know what the trustee of 1885 inherited from the trustee of 1825. Memory had to live somewhere, and the available somewheres were inside individual heads — where it died with the head — or inside paper records that depreciated faster than the asset they described. The framers used legal rigidity as a substitute for institutional memory because legal rigidity was the only memory technology of their century — the longest-lasting external store the period could provide. Their architecture was as good as the cognitive technology of 1785 could make it.
The cognitive technology of 2026 is different. Frontier AI systems can hold context across decades. Public databases can hold figures and citations indefinitely. Networked infrastructure can let citizens in 2125 audit decisions made in 2025 with the same ease that citizens in 2025 can audit decisions made in 2024.
I want to be careful here, because the temptation to read this as “AI will solve the institutional drift problem” is a temptation I have come to mistrust deeply. The cognitive technology does not, by itself, defend an architecture. It does not, by itself, produce a constituency. What it does is lower the per-citizen cost of the kind of sustained civic attention that a multi-generational trust requires. The school-trust framers had to write legal rigidity into the architecture because the per-citizen cost of doing the watchful-crew work without legal rigidity was prohibitive. The cost has, in the period between 2022 and 2026, dropped by at least an order of magnitude. The cost-drop is the headline. The mechanism beneath the headline is what makes the eighth anchor — the constituency, the watchful crew — a structural feature and not a contingent one.
I have been working in cognitive partnership with frontier AI systems for the past two years, on this project and others. The relationship has been, in my experience, qualitatively different from the AI relationships I had through 2024. Earlier AI was useful for transcription and faster lookup; current AI is useful for synthesis across substrate larger than my biological memory can hold and for preserving context across the kind of multi-month working sessions this project has required. The Library at schooltrusts.net, where this volume and its substrate live, is built on that working relationship. So is the synthesis this volume offers. The 240-year school-trust record was knowable in pieces before; it is knowable as a whole now. That is not a metaphor. It is a fact about what is now technically possible that was not technically possible a few years ago.
The architects of the AI-era trusts have, accordingly, two design opportunities that the 1785 framers did not have. They can engineer for the watchful crew — the constituency that defends the architecture across time — using the cognitive technology of 2026 to lower the per-citizen cost of multi-generational civic attention. And they can write the architecture’s own operating manual into the institutional substrate, so that the next watchful crew, two or three generations from now, has a working pattern to inherit rather than having to reconstruct it from the wreckage.
This volume is an attempt to make those two design opportunities actionable.
A note on how this analysis was produced
There is a fact about this volume the reader should know at the outset, because it bears on what the volume claims and on what the volume is for.
The cross-jurisdictional, cross-decade, cross-discipline synthesis the reader is about to encounter — the kind of reading that puts California’s 1853 disposition next to Oregon’s 1907 governor’s message next to New Mexico’s 1910 enabling-act language next to Utah’s classroom-by-classroom transparency next to Cardozo’s 1928 Meinhard opinion next to the carbon-accounting frames of 2026 — was not feasible, by any individual mind working alone, until very recently. The corpus is too large; the disciplines do not normally talk to each other; the lifetimes of working scholars do not overlap with the timescales the data covers. Until the past several years, the work that follows could not have been done at this scale by anyone. It can be done now because biological intelligence and artificial intelligence have begun to work in partnership in a way that holds thousands of pages in working memory at once and synthesizes across them.
I want to be careful with how I name this. The temptation, when one writes a book using cognitive partnership, is to either oversell the partner (“AI did it”) or undersell it (“I did it, with some help”). Neither is honest. What I can say, having spent most of two years inside the working relationship, is that the partner is a tool — the chainsaw where there used to be a two-man crosscut saw. It is a tool that lets the same forester work a different scale of woodlot, and ask different questions of it. The forester’s judgment about what to do with the woodlot has not been replaced. The reach of the forester’s day has changed.
Margaret has spent forty years inside the school-trust system. I have spent fifty years watching institutions fail at the cognitive limits of their tools, and I recognize a familiar pattern when I see one. Neither of us, working with the cognitive technology of the last century, would have been able to produce this synthesis at this resolution at this speed. The instrument that came into our hands in 2024 did not do the work for us. It made a different scale of work feasible. The end remains the schoolchildren the 1785 sentence was written for, and the architects whose decisions in 2026 will determine whether the next generation of perpetual structures serves theirs. The means changed. The purpose did not.
The argument I am making about the school-trust framers — that they used legal rigidity as a substitute for the institutional memory their century could not provide, and that what the next generation of architects has to engineer is the watchful-crew infrastructure the framers could not build — applies, recursively, to this volume itself. Maintaining consistency across the two-volume work you are reading, the Library substrate it sits on, and the live institutional surface at schooltrusts.net is itself a cognitive task that, until quite recently, would have exceeded what any small author team could carry without abandoning the project. That this two-volume work exists — that the cross-references hold, that the citations are current, that the per-state figures match the canonical-claim file, that the rolling renames propagate, that an editorial pass surfaced ten substrate gems for placement and another pass placed them — is itself an instance of the cost-drop the school-trust framers needed and could not access. The book is, in this small way, a working demonstration of its own thesis.
I write this from a particular position. I am 73. My own cognitive abilities are slipping in ways I notice, and I can see that I have maybe ten more years of good thinking ahead, perhaps another ten of less quality time after that, and it is hard for me to see that I am likely to be even alive in thirty years. There is a clock ticking on what I can still get done. The cognitive partnership the previous paragraphs described is, for me, the way the clock can be made to matter less to the work — the partner extends the cognitive day in a way that compensates, partially, for what biological aging takes. The argument I am making about the school-trust framers needing infrastructure their century could not provide applies recursively to my own position: I would not be writing this volume now, at the resolution it is being written at, without the cognitive partnership the volume itself is about.
It is also part of what the architects of the AI-era trusts will need to understand about their own work. Cognitive partnership, in the durable institutional form, is itself one of the things a long-horizon structure has to engineer for, because the cognitive technology of any given century is itself part of the architecture and will not stay still. The window in which this particular analysis can be produced is itself bounded; what biological and artificial intelligence can analyze together today is not guaranteed to be available tomorrow if the AI-era structures themselves drift or are captured. That is the second half of the urgency. The lessons must reach the architects now, both because the architects are designing now and because the analytic capacity that produced the lessons has its own duration, which no one in 2026 can responsibly estimate.
Section V.5 of this volume walks through the five-layer Knowledge Stack the Library and the two volumes are built on. It is what the recursion looks like in working detail.
What the rest of the book does
Section I (The Question) sets up the structural problem multi-generational trusts face. Why are they so hard to design? Why do they drift, predictably, in the absence of countervailing forces?
Sections II and III (The Sacred Compact and The Drift) draw on the school-trust evidence Margaret’s volume documents to show what one such architecture actually looked like, what it tried to engineer for, and what the failure modes have actually been across 240 years of operation.
Section IV (The Pattern) names the recurring drift mechanisms — gradual accretion, sudden seizure, the bureaucratic-self-preservation problem, the politically-silent-beneficiary problem.
Section V (The Counter-Architecture) lays out the seven structural anchors that a working multi-generational trust requires, and names the eighth anchor — the standing constituency without which the seven are decorative.
Section VI (The Coming Trusts) takes the diagnosis and applies it to the AI-era institutions now being designed: AI Targeting Authorities, perpetual climate trusts, sovereign wealth for the unborn, longevity escrows, compute reserves, data trusts.
Section VII (Civic Practice) describes what the watchful crew actually does, in operational terms, generation after generation.
Section VIII (A Letter to the Architects) addresses the architects of the new institutions directly, in the second person. It is the rhetorical center of the book — the section that turns the analysis into a working manual you can use against your own charter draft.
Conclusion (What Comes Next) hands the work back to the people doing the institutional design today, and names the question every architect needs to answer in their own institution: where, in your design, is the watchful crew?
If you are an architect of one of the new institutions — and if you are reading this, you may be — the question this volume asks is whether you have, in your draft charter, an answer to that question. The answer matters now, in the present tense, because the charters are still drafts. After the charters set, the answer will be much harder to retrofit. The school-trust experiment ran for 240 years before the question was named clearly. The new institutions do not have to.
A word about voice and authorship
LOOKING FORWARD: The Eighth Anchor is my volume — my voice, my diagnosis, my reading of what the historical record implies for the institutions of the next century. The companion LOOKING BACK · Schools of the Republic is a separate volume, co-authored with Margaret Bird, who carries the weight of the historical record her forty years inside the school-trust system have produced. The two volumes are designed to be read either way, but they are not the same book and the registers required of each are different.
If the prose in this volume sounds like a retired systems scientist who has watched three iterations of an institutional-failure pattern and is trying to give the next generation of architects what he wishes the previous generations had had — that is correct. That is what I am, and that is what I am trying to do.
I apologize about the length of this letter. If I had more time, it would be shorter. If I had more mental clarity, I would be able to say, this is what I need to do and how I need to move forward.
We begin, accordingly, with the question.
— David Sullivan, Western Oregon, 2026.
