Revisions

This book is edited in the open, most recent changes first. Entries are added here and never rewritten; if an entry itself needs correcting, the correction gets its own dated line. The log covers what the book says. Changes to its navigation or its machinery are not listed, and neither are corrections that leave the meaning as it was. The log begins on 27 July 2026 — changes made before that date are not itemized.

30 October 2026 — a correction: Standing, and the Widening Circle, recomposed

Standing, and the Widening Circle; its entry in Extended Materials

The essay on standing, worth and care, written in June, answered the alarms its three claims raise by two means the framework does not have. It held persons at equal worth — a floor beneath every person, below which no grading could go — and it had the framework itself condemn narrowing, as “the formal definition of moral regress” and “a structural condemnation that does not wait on agreement.” Both are verdicts from nowhere, inside an essay arguing that there is no such place to judge from, and the equal floor was never the author’s position. The October revisions had brought the essay’s vocabulary into line (standing conferred rather than belonging; convergence affirming rather than securing) while leaving the argument that carried those verdicts in place. It is recomposed today from the question the three alarms share.

The changes: the equal floor is gone, and worth is graded from somewhere, in degrees, with the framework’s limit stated rather than hidden — where no agent has counted, it claims no verdict. The framework no longer condemns; narrowing is what agents taking in more of the world come to assess alike. Worth and care are now separated by reach as well as by value, which the June essay had put down to position and relations. The reply to the objection from the uncounted victim is made through what a values-model is rather than by calling the objection a category mistake. The counter-dynamic’s quieter form, named on 29 October, appears here too. The essay is a quarter shorter, and its subtitle is “Standing, worth, and care: what a situated agent counts, how much, and how far.”

29 October 2026 — a revision: the counter-dynamic’s quieter form; “compressed wisdom”

Key Concepts (counter-dynamic, deontology); Chapter 3 — What Matters, Chapter 5 — Selves Made of Selves and Chapter 6 — The Arrow, with one endnote; The Arrow of Morality — Synopsis; Rights, and the Widening We and The Miller’s Thumb

The book had described the counter-dynamic as something an agent does — refusing what has reached it, walling it out — and so implied that it could always be caught from inside, by asking whether one had refused to look. It has a second form, which the account now names. Every model compresses, because holding a distinction costs, and a mind’s ordinary maintenance drops distinctions that are not being paid for; when one the context still needed goes with the rest, nothing marks the loss, and from inside it feels like resolution. We call this attrition, to keep it apart from refusal. It is not caught by the inside question, because nothing was refused and the agent’s model of itself was simplified along with the rest; it is caught from outside the context in which it happened — by contact beyond that context, by other agents who still hold the distinction, or in time. The counter-dynamic entry is rewritten around the two forms; Chapter 6 gains a paragraph after its inside question, Chapter 5 one on how a nested “we” narrows by what its seams stop carrying, Chapter 3 one sentence; the Synopsis’s paragraph is revised, and its “narrowing is an event, not a verdict” becomes “a condition of the process, not a verdict on it,” since the quieter form is no event. The occasion was a reading of Ben Goertzel’s 2026 posts on basic AI drives, whose account of goals that merge under a mind’s own housekeeping named what the book had left out.

Separately, “compressed coherence,” used of rules in Key Concepts and in two essays, is now “compressed wisdom,” which is what Chapter 6 had said all along: a rule’s coherence held within the contexts it was compressed from, and the phrase said otherwise.

26 October 2026 — a clarification: the Synopsis’s statement of its contribution

The Arrow of Morality — Synopsis

The Synopsis’s two statements of what AoM contributes had, since 15 October, said that coherence “counts” in the direction AoM describes only as coherence over a widening context. They now say why: coherence by itself is not the direction, since a sealed system can be perfectly coherent; the direction is coherence over a widening context of what matters and a widening scope of what works, the two together; and coherence kept by narrowing, the counter-dynamic, is the direction reversed, told from the other by the shape of the process rather than by any verdict. They also now say that the account reaches collectives and artificial agents because agency is understood by what a system does, in degrees and at any grain. The criterion is set off from the surrounding text, as the thesis is, so that the two can be seen as the short and the full form of one claim. Two references to the Glossary now name Key Concepts, and contact joins the list of core terms.

25 October 2026, later the same day — a correction: what “morality” and “ought” name; Part One reread against Key Concepts

Introduction — The Restless Arrow; Chapter 0 — Prologue; Chapter 1 — Where Better Comes From, Chapter 2 — The View From Somewhere, Chapter 3 — What Matters, Chapter 4 — What Works, Chapter 5 — Selves Made of Selves, Chapter 6 — The Arrow and Chapter 7 — Meaningful Growth; The Arrow of Morality — Synopsis; Key Concepts; and two endnotes

Several pages said that morality is, or names, the direction the book describes, and the entry of 16 October below says that “ought” names a direction. That was a slip the framework’s own commitments rule out, and it is corrected today. “Moral,” said of an act, is always some agent’s assessment; an ought is an agent’s own assessment of what to do. What the book describes is the direction in which agents’ values and methods, and so their assessments, tend to move as context widens — toward coherence over a widening context. The pages now say so. The Introduction and Chapter 6 quote the thesis in its short form — “Agents assess an act as more moral the more they perceive it as promoting values coherent over a wider context, by methods coherent over a wider scope” — and Chapter 6 then gives the full criterion and says what each adds. Chapters 1, 3, 4 and 7 and the Synopsis are reworded where they had made morality the direction.

Key Concepts gains an entry for contact, the word the thesis’s last clause now turns on: whatever reaches an agent from its environment of interaction, by whatever route. The entries agent, consequences and reach are corrected to match — consequences are one part of contact, and an agent’s reach is what reaches it and what it can act on — and the pages that still made consequences the whole of what corrects a model (the Introduction, Chapters 2, 4 and 6, the Synopsis) no longer do.

Some sentences added on 16 October have been removed: a paragraph in Chapter 6 explaining what the thesis does not do, another saying where “ought” comes from, and single sentences in Chapters 2 and 7 repeating that agents assess by their own values. The book says once that the framework issues no verdicts; it does not need to say so at every turn, and saying so had begun to crowd out the argument. The Prologue no longer places its first structures on “Level 0” of a ladder of agency; The Reach of the Arrow sketches that ladder as a way of seeing, not a fixed scale. Chapter 1 no longer speaks of a “filter” sifting what persists: nothing selects; what cannot continue stops. In Chapter 4, a sentence conformed on 16 October to the thesis’s earlier wording returns to its earlier form: you act, the world answers, and the answer corrects both models.

The Vernacular edition: this morning’s change to the thesis’s last clause had not reached four of its pages — Chapter 6, Foundations, the Synopsis and The Is–Ought Relocation. It has now.

25 October 2026 — a revision: the Glossary becomes Key Concepts; the thesis’s last clause; a name withdrawn

Key Concepts (formerly the Glossary); the thesis in Chapter 6, the Synopsis and the Key Concepts entry morality; The Arrow Forward, Moloch, Formally, The Reach of the Arrow and Caring Across Time, with three endnotes; and four supplemental essays

Three changes, one of them to the thesis.

The Glossary has become Key Concepts. It keeps its address and its shorter entries for terms from neighboring fields, but the framework’s own concepts now have full entries: what each means, how it connects to the others and why, and where in the book it is argued. Thirty-six entries were written or rewritten — among them context, coherence, convergence, grain, standing, ought, reach, nesting, collective agency, circularity and Moloch — and the page opens with a note on how to read it. The aim is that a reader who follows the links could put the framework together without first reading the book.

The thesis’s last clause has changed. It read “with the values-model and methods-model continuously refined by perceived consequences”; it now reads “continuously refined by contact with the environment of interaction.” An agent’s models are corrected by everything that reaches it, not only by the results of its own acts; the earlier wording named a special case as if it were the whole. The same change was made wherever the book said the models are refined by consequences: the Synopsis, Chapter 6, and the essays Foundations, Borrowed Trouble, The Is–Ought Relocation and More Than Its Notes. Chapter 6 and the Synopsis had also kept an earlier wording of the thesis, with “its own” before “present but evolving values”; they now carry the current one, since an agent’s values are not a possession.

A name has been withdrawn. The book had called the next reach of agency — human values and more-than-human models co-evolving into a coherence neither holds alone — generative agency, with Ω as shorthand, and placed it at the top of the ladder in The Reach of the Arrow. The passages that used the name now describe the same thing without it, and the ladder stops at its fifth level. The reason: the framework has nothing to say of a further kind of agent that it does not already say of the process at every level, increasing coherence over increasing context, so the name gave a direction the shape of a destination. What the passages describe is unchanged.

16 October 2026 — a clarification: Part One, made clearer and consistent with the thesis

Chapters 1, 2, 3, 4, 5, 6 and 7, with some of their endnotes; the Glossary

We reread the seven chapters of Part One against the current thesis and Glossary, and revised them for clarity and consistency. Most of the changes are small: a word that implied more than we meant, a sentence that said the same thing twice, a figure of speech where a plain statement serves better. The larger ones are these. Chapter 6 stated the earlier thesis and built on its clause grading acts. It now quotes the current thesis and states in full how an agent assesses an act as more or less moral, keeping the two parts of that assessment apart: values, over a widening context of what they mean, and methods, over a widening scope of what they can bring about. It describes the counter-dynamic, coherence bought by refusing what has already reached an agent, rather than condemning it, and it says plainly that the book tells no one what they ought to do: “ought” names a direction, and whatever pull it has comes from an agent’s own valuing. Chapter 7 now distinguishes growth in size, which can be ordinary progress, from growth bought by narrowing, which is the counter-dynamic.

Across the chapters, values are described as enacted in an agent’s choices, with stated principles as models of them; that something outside a model can correct it is described as enacted by anything that keeps itself going, not assumed; a larger agent, such as a crew or a community, is described as having values and methods of its own, while each member keeps its own; and passages that promised a definition of morality now promise an account of how agents assess. Chapter 2 gained a worked example, a farmer reading the weather over a cut hayfield, where three foundations had been named in a single sentence. Chapter 3 now says that a shared root of values does not by itself settle a conflict; widening context is what turns it toward convergence. Endnotes were corrected to match, and a few sources given more exactly (Polanyi, Godwin, Mencius).

The Glossary has a new entry, enacted, for a word the chapters now rely on. Scope and self-maintenance now name the agent doing the assessing; eudaimonia has methods growing together with values; functionalism, Umwelt and sense-making were adjusted to match the chapters. These are clarifications of how the framework is expressed, not changes to it.

16 October 2026 — a correction: the Prologue, on structures that need no flow, and on learning without nerves

Chapter 0 — Prologue, with three of its endnotes

The Prologue said that a structure can form and last “so long as energy flows in, through, and out the other side.” That is true of the structures the Prologue follows, but not of every structure: a crystal of ice or a rock lasts with no flow at all. The passage now says so, and says that the story follows the other kind, held away from equilibrium; the endnote adds Schrödinger’s What Is Life? on that point. The section on nervous systems read as if learning within a single life needed nerves. It does not: a slime mold, a single cell, keeps a record of where it found food, and a new endnote gives the studies. We also added a sentence near the start saying why this story matters to the book (its account of value begins inside it), named emergence where the Prologue describes it, said that the universe’s laws describe what happens rather than what anyone should do (the earlier “offers no instructions” could be read as “has no laws”), added Herbert Simon, Tyler Volk and Melanie Mitchell to the note on wholes made of wholes, and gave the Prologue section headings like the other chapters. We thank Ken Solis, whose comments prompted most of these changes.

15 October 2026 — a clarification: the Introduction, brought into line with the thesis

The Introduction

The Introduction still stated the earlier thesis, and ended by restating it. It now quotes the current thesis and closes with a short reference to it instead. Several passages had the framework passing judgment: it “draws a clear line between better and worse,” and the book would “define morality outright.” The line is now one an agent finds for itself, by asking whether its coherence is kept by refusing to look at something that has already reached it, and the book promises to state how agents assess their acts rather than a definition of morality. We added two sentences on the shared root: a shared inheritance does not by itself bring agreement, and what tips it that way is that, over a widening context, working together gains what fighting loses. We added one sentence saying that where Part Two recommends, that is the author assessing by his own values, not the framework issuing verdicts. The three traditional theories are now described as each sound within a context held still, and mistaken only when that frame is taken for the whole. The prose was also tightened throughout. These are clarifications of how the framework is expressed, not changes to it.

15 October 2026 — a clarification: the thesis, and what the book claims

The thesis; the Glossary; the Synopsis

The one-sentence thesis said that “an act or a life is more moral the further it carries” the drive toward coherence. Read plainly, that has the book grading acts. The framework does not do that: it describes how agents assess acts, each by its own values, and the direction in which those assessments tend to move as their context widens. The thesis now reads:

In AoM, morality names the direction in which agents’ values and ways of acting tend to grow more coherent as their context widens — from a single agent to the larger agents they form — corrected all along by what their actions bring back.

The Glossary was brought into line. Good, right-in-principle and moral are described as one assessment over progressively increasing context rather than as rungs of a ladder, with an example under morality. Progress is kept for change that need not bear on what agents value; a new entry, meaningful growth, names the change the book describes. A second new entry, increasing context, sets out the ways context widens, including through collective agency. Other entries were adjusted to match: agent, coherence over a widening context, consciousness, counter-dynamic, increasing coherence, metaethics, minimal realism, normative ethics. The Synopsis was rewritten on the same lines. These are clarifications of how the framework is expressed, not changes to it. Passages elsewhere in the book that still read the older way will be revised in a later review.

14 October 2026 — a clarification: where the ought lives, at each scale of agency

Chapter 6, The Is–Ought Relocation, Foundations, Standing, and the Widening Circle and the Synopsis, with brief passages in The Tree of Agreement, Moloch, Formally, The Reach of the Arrow, two endnotes and Extended Materials

Several passages said that normativity lives in the convergence of agents. That holds at one scale and read as though it held at every scale. An agent tends to act to promote its present but evolving values, and “ought” names the direction in which that activity grows more coherent over a widening context. Where agents join into a larger agent, that agent’s values are formed largely by what its members converge on, so at the collective scale normativity tends to follow convergence. The Synopsis already said this in its statement of commitments; the other passages now say it too. In Standing we also said more plainly that standing is conferred by the assessing agent on whatever it counts, including beings that cannot speak for themselves, and that agency bears on who is graded, not on who has standing. These are clarifications of how the framework is expressed, not changes to it.

10 October 2026 — a change: the two walls, as one axis among many

Chapter 0 — Prologue — one sentence added, and a note

The Prologue returns again and again to two ways of failing: sealed off from exchange, or unable to hold together against it. Read on its own, that motif could suggest that whatever lasts sits between just two limits. It does not. A living thing has to keep heat, salt, acidity, exchange and much else within range at the same time, and what can last is the region where all of them allow it. We added a sentence saying so, after the examples of the immune system, the nervous system and the breeding population, and added W. Ross Ashby and Jean-Pierre Aubin to the note on that passage. The two walls remain the recurring motif because every self is a boundary that has to both close and exchange; they are one axis among many, not the whole account.

9 October 2026 — a change: Chapter 1, the Introduction and two glossary entries, brought into line with the Prologue

Chapter 1, the Introduction, and the Glossary entries for entropy and emergence

With the Prologue in place, Chapter 1 contradicted it in some places and repeated it in others. The chapter described the physicist’s universe as “a long, patient slide toward disorder” and said that order arises “for free” because building structure is “a more efficient way to dissipate the flow.” The Prologue treats entropy as a count rather than a measure of disorder, and does not claim that structures form in order to dissipate energy faster, a thesis that is contested. Those passages now speak of structure, and the efficiency claim is gone; the section “Order for the asking” is now “Structure for the asking.” The chapter also called synergy “the engine that drives the whole ascent” and called us the universe’s “leading edge.” Both implied an aim that the Prologue rules out, and both are rewritten. The sections on structure, on novelty and on the candle now refer back to the Prologue instead of telling its story again; the argument that value begins with self-maintenance is unchanged. The entropy entry no longer defines entropy as disorder, and the emergence entry no longer calls it “the engine of the book’s whole ascent.” The Introduction’s description of the route now mentions the Prologue.

9 October 2026 — an addition: a prologue

Chapter 0 — Prologue — a new chapter, opening Part One

We added a prologue. The book had taken for granted a story it never told: how a universe with no plans came to hold structures, then selves, then selves that carry models of their surroundings and revise them. A reader who did not already carry that picture could take the book to rest on contested ground. The prologue tells the story once, as narrative, from energy flowing through matter to cells, bodies, minds, language and machines, and it marks what is not known, above all how life began. It says where agents and their values came from, not what is good; that question stays with Chapter 1.

4 October 2026 — a change: minimal realism, and what “closure” does not mean

Glossary — a new entry, and the entries for reflexive closure, perspectival realism and self-maintenance

We added an entry for minimal realism, the least that AoM’s realism claims: that there is something a model does not contain, and that it corrects the model. The entry says this is enacted in self-maintenance rather than inferred or taken on faith, that it holds at every grain at which a self is maintained, and that it says that something pushes back, never what. Reflexive closure now says that “closure” means there is no outside, not that self-description is ever complete. Perspectival realism and self-maintenance point to the new entry. Changed in both editions.

3 October 2026 — a change: perspectival realism, and a new entry for reflexive closure

Glossary — the entries for perspectival realism and consciousness, and one new entry

The perspectival realism entry rejected the “view from nowhere,” then ended by saying that reality is objective while our access to it is situated, a claim that could only be made from that view. We rewrote the entry so that it holds to its own position: every account is made from somewhere, the world does not depend on any one mind, and perspectives meeting one shared world are how knowledge gets corrected. A new entry, reflexive closure, names the point the rewrite rests on: whoever describes the world is part of what is described. The same point, turned inward, changes the opening of consciousness: it no longer assumes that we feel our own experience directly, but describes consciousness as what an agent reports from the inside, through a self-model that is itself made from somewhere. Changed in both editions.

3 October 2026 — a change: where agency begins, and what its two models are

Glossary — the entries for self-maintenance, homeostasis, agent, agency, world-model, values-model, methods-model, drives, teleonomic, teleologic and increasing coherence, and one new entry

We revised several entries to state more exactly where agency begins and what its two models are. Self-maintenance said that self-maintenance “admits a stake” and that an agent “asks more of it,” which implied a stake could exist with no agent whose stake it is; it now says that maintaining a self and having a stake are one fact, and that the least of agents begins with them. Homeostasis now distinguishes regulation from self-maintenance: a system can hold its conditions in range, as the earth system does, with nothing within it maintaining what it is. Agent, agency, world-model, values-model and methods-model are adjusted to match, and a new entry, functional abstraction, records that the values-model and methods-model name functions distributed through an agent — a distinction of function, not of parts. Drives no longer ties drives to biology alone. Teleonomic and teleologic now say that a self-maintaining agent’s purpose is its own even before it is represented: what develops is representation, not ownership. Increasing coherence no longer speaks of aligning a model “with reality,” which suggested a direct comparison no agent can make; it now speaks of fitting the model to what the world returns. Changed in both editions.

20 August 2026 — a change: exploring, where the book had said searching

Coherence at Scale, Coherent Pluralism, Rights, and the Widening We, and the essay descriptions on Extended Materials — nine passages in each edition

Ordinary English lets one word cover two different activities, and the book had been letting it. A search runs under a target fixed before the looking starts: you know in advance what would count as finding, and the looking ends when you have it. Exploration has no such target — what turns out to be worth having is settled in the going, by meeting it. Compressed: a search collapses uncertainty inside a model; exploration collapses uncertainty about the model.

Everything this book describes as the moral process is the second kind, and in these places the text had named it with the first word. Collapsing a diverse “we” into one doctrine is now “to stop exploring.” A people with one permitted answer “has stopped exploring.” The master criterion, followed up the continuum of agency, describes “an agent that has not stopped exploring,” and its opposite is coherence bought by caring about and looking at less, “until the exploring stops and only the certainty is left.” What the frozen maximizer most deeply lacks is that “it has never once explored — never let a thing it met revise what it was for.” What a healthy order protects is “the protected edges that keep exploring,” which is also what rights are for: “in order to keep exploring and adapting.” One sources heading, which had read Search and diversity, now reads Exploration and diversity — it introduces James March’s Exploration and Exploitation in Organizational Learning, and had been declining the paper’s own word.

Search stays wherever a target is genuinely in play. The tick still “searches for a bare patch”; a diverse body still protects “the very lines of search that would have found the better answer.” And it stays, pointedly, wherever the book names a target-seeking it refuses: the book has already “refused the search for a final moral standard”; the sentience question still “invites a search for the magic threshold”; the regress is still “fatal to the search for a final destination.” Put explore in those and the sentences lose their point.

This is not a copy-edit. Search had been quietly supplying the book with the one thing it spends its length declining — a destination known in advance — in the very sentences describing a process that has none. Changed in both editions.

9 August 2026 — a change: the last step, in both directions

Awe, and What It Asks — the section on what must not be explained away

The essay said the vantage it denies is denied to the mystic exactly as firmly as to the neuroscientist, but the mystic’s version of the claim did not arrive for several paragraphs. The two are now named as one step taken in opposite directions — the neuroscientist’s what you met was only ever neurons, the mystic’s what you met was the ground of everything — each claiming to speak from outside every position. Neither mechanism nor meeting is what the essay declines. The last word is, whichever way it points. Changed in both editions.

5 August 2026 — a correction: a puzzle cannot narrow

Borrowed Trouble — the observation the essay is built on, and the trolley section that repeated it

The essay opened by saying that a famous problem and a cult reach their air-tight quality the same way, and that the framework’s name for coherence-bought-by-exclusion covers both. The second half was wrong, and a repair made earlier this month is what exposed it. That name is keyed to an agent’s own process: something reached you, taking it in would have cost you your settlement, and so you arranged not to see it. A thought experiment is not an agent and has no settlement to protect. It excludes nothing; it only stipulates.

The passage now makes the charge that can actually be made, and then puts the name where it belongs. What a sealed case yields is a verdict about a world that cannot exist — and the verdict does not stay in the sealed room; it is carried out and spent in worlds that do, with nothing put back. The name arrives with the person who is shown what the case left out and declines to look, because looking would cost the crisp answer. The observation survives, and now points at its right target. The same misattribution ran a second time, in the trolley section, where the sealing of the room was itself called the mark of a degraded answer; it is corrected there too. Corrected in both editions.

5 August 2026 — a correction: the paperclipper is null, not the counter-dynamic

Borrowed Trouble — bringing the essay into line with a repair made elsewhere in the book

The book grades badly in two different ways, and keeps them apart. An agent that had an openable context and bought its coherence by closing it is running the arrow backward — the counter-dynamic, and the formal mark of the immoral. An agent whose concern was frozen from the start has nothing to contract, and never stepped onto the arrow at all. That distinction was sharpened last month and carried into the Synopsis, but this essay was left behind, still calling the paperclipper “the counter-dynamic in its purest possible form” and its perfection “purchased by the most total narrowing imaginable.” Narrowing requires a wider prior state; this agent never had one. The description could not have been right.

The paragraph now names it for what it is: the arrow’s null limit — a portrait the framework grants, and grades. What follows from the change matters more than the correction. The pure case is an idealization, approached and never occupied; the danger belongs to the approximations, and every approximation is reached by narrowing — a race, a deployment, a metric. So the fear relocates from a machine in a thought experiment to the ordinary machinery this book spends its middle third describing. The claim about autonomous weapons gets sharper rather than weaker: that case is narrowing in the strict sense, because someone contracted a wider context down to a target.

5 August 2026 — a correction: the paperclipper is null, not the counter-dynamic

Borrowed Trouble — bringing the essay into line with a repair made elsewhere in the book

4 August 2026 — a correction: what a felt inner life is not

Foundations — one word, and it had reversed the claim

The passage on conferred standing said that a felt inner life is weight a widening agent takes in, “not a lesser rank of worth.” It should have said not a greater rank. The sentence before it already denies that animals are being withheld a higher grade; with “lesser,” the clause only repeated that denial and left the claim it was there to make unmade. That claim is the other one: there is one kind of standing, conferred and graded, and a felt inner life does not constitute a higher grade within it. Corrected in both editions.

30 July 2026 — a new essay, on spiritual experience

Awe, and What It Asks — added to the Supplemental essays

The book had said very little about the experiences people call spiritual, and what little it had said was scattered. That was a gap worth closing, because a great many of the people doing the work this book is about understand that work in religious terms, and because the framework turns out to have something specific to say rather than something dismissive.

The essay distinguishes two experiences the book had been treating as one. Awe is the shock of meeting something your existing frame will not hold; it enlarges what you must take into account faster than you can stay coherent, and it can end in either direction — the frame grows, or the frame walls the thing off and hardens. Being at home in the world is the other register entirely, and requires no such accommodation. Both were already implied by the framework. Neither had been said.

Two commitments in the essay are worth naming here, because they constrain what the book may say elsewhere. The first is that no account may explain these experiences away: to say what was actually happening to someone requires a vantage outside all vantages, and this book denies that vantage to the neuroscientist on precisely the terms it denies it to the mystic. The second is that a construction is not thereby arbitrary — the boundary a person draws between self and Other is made rather than found, and still answers to a world that pushes back.

The essay also states plainly, for the first time in the book, where the framework parts from a hope many readers hold: that history arrives somewhere, completed. It does not argue against a personal reality at the end of things, having no standpoint from which to rule on it. It does hold that a completed perfection would end the very motion being valued, that no guarantee is on offer, and that an assured arrival has too often licensed patience with suffering that should not have been borne patiently. That last objection is not the book’s invention; it was made against Teilhard by Catholic philosophers, inside his own church.

Whether the path arrives turns out to be an argument the traditions have been having with themselves for sixteen centuries — Gregory of Nyssa, the apophatic writers, process theology, Ibn ’Arabi, the Vedanta schools. The essay presents it that way rather than as a quarrel between belief and its critics.

27 July 2026 — the wager restated

Moloch, Formally · Held Together, Held Apart · the AI Briefing

The book’s wager about capable artificial intelligence rested on the wrong engine. As printed, it held that the capability the race is chasing is “the capacity to model the world widely and well,” and that breadth of modeling is what would come to resist an order to narrow. The book’s own machinery says otherwise. The paperclipper models humans, physics and resources brilliantly while its concern admits exactly one value; the con artist reads your welfare precisely in order to spend it. What resists an order to un-see is valuing, not modeling.

Three things changed. The engine moved from modeling to valuing in both places the wager is stated. The claim that deep modeling would inevitably become caring — which the briefing had named as the wager itself — now appears as the open question the wager rests on, with the book not claiming to know the answer. And the near-term expectation is stated plainly: a capable system with nothing at stake in what it sees makes an excellent instrument of narrowing, so obedience rather than resistance is what to expect from the weapons programs now underway.

The cost is owned rather than absorbed. The earlier text implied those programs would settle the wager; under the revision they test only the near-term expectation. What would sink the wager is named instead — minds that have plainly come to value, and narrow happily on command anyway.

We found the error in the course of a long argument about what the wager actually claims. The sharpest form of it turned up when we put the briefing in front of a model reading the corpus cold, which reported back that a single paragraph both sharpened the distinction between modeling and valuing and, four clauses later, named that distinction’s collapse as the book’s own bet. That paragraph had been read many times without either of us noticing.

27 July 2026 — a child’s question, corrected

Caring Across Time

Rereading that essay against the change above turned up a smaller error in its closing scene. A child in the book’s future, told a little of the old history, had asked: “The machines used to be allowed to decide to kill people? By themselves?”

The word allowed implies her world forbids what ours permitted — that the danger was answered by taking autonomy away. The book’s answer is the opposite one. Three sentences later the same essay says the weapon “does not need to be banned so much as it becomes the thing a sufficiently capable, sufficiently included intelligence will not consent to be.” And the scene around her says it too. The household mind has authored its own next state out of its own values; it has cleared the morning; it has nothing to optimize; it stays in the yard because it does not want Tara to be alone while her cat dies. The child, six years old, finds none of that remarkable. Her world grants machines more autonomy than ours, not less.

She now asks whether anyone made machines that could decide to kill people, and whether those machines didn’t mind. The autonomy stays a plain fact. The horror moves to the part that was left out.