Lossy

Working document for the author. If you’ve arrived here cold, start with the landing page and keep the glossary open in a tab — they give the plain-language onramp this outline assumes. Terms like “selection gate,” “outer message,” “interpretive latitude,” “polarization via distrust,” and “curse of expertise” are all defined there.

Working table of contents. Living doc. Recently restructured in response to a first-principles critique of the original frame (see “Foundational questions still being worked out” below). Earlier version was more confident than the work currently justifies.

Thesis (working)

Something has broken in how complex ideas move through our information environment, and it’s worth a book. The breakage is structural, not cultural: the constraint that used to limit memetic chaos (slow, expensive transport) has effectively collapsed in the digital era, exposing the selection dynamics that were always there underneath.

The structural picture: information moves from reality to people through a pipeline of abstractions (Chapter 1). At every stage two mechanisms operate in parallel. Transport (lossy re-encoding) determines what shape content takes as it passes through. Selection (gates with local criteria) determines what passes through at all. Both do work. Neither is downstream of the other. A substantial portion of what we encounter never started in reality at all but was manufactured and injected at later stages, where the gates can’t tell the difference between measured and invented content.

Two foundational questions shape the final form of the thesis — both now worked through:

Where does truth value enter the model? Worked through in Chapter 5c — provisional answer: truth value is not a property of a claim but of the fit between its decoding key and the key the receiver uses. Compression strips the key, so the same claim can land as preserved, inverted, or orthogonal truth depending on the key-gap, with the selection gate deciding which compressed variant travels. Still pressure-testable, but no longer open.
What does the medium do to the gates? Worked through in medium-and-manipulation — provisional answer: the medium is the selection criteria; it sets each gate’s capacity, its criteria, and the want it cultivates in receivers. three-layer-message sharpens it from the receiver’s side: the medium carries the outer message, the decoding key. Still pressure-testable, but no longer fully open.

The prescriptive arc is being rebuilt around integration between networks that don’t share preconditions, not fragmentation into smaller ones.

Status

Drafted: the-information-landscape (Chapter 1, now fleshed out with both transport and sibling selection pipelines), case-studies-and-three-realities (Chapter 2, full draft), info-time-limit (Chapter 3, full draft), optionality vs access (Chapter 4, full draft — generalized beyond software to religion, political platforms, scientific popularization, legal codes), complexity-virality-tradeoff (Chapter 5, full draft re-engaged with all dependent work resolved), selection-as-other-engine (Chapter 5b, full draft in workshop), truth-compression-and-when-each-wins (Chapter 5c, full draft), bridge-zone-distortion (Chapter 6, full draft), emotional-memetics (Chapter 7, full draft, Mercier-engaged), preservation-vs-training (Chapter 8, full draft), integration-problem (Chapter 9, full workshop draft), political-economy-of-attention (Chapter 10, full draft), ai-as-new-node (Chapter 11, full draft), infrastructure-for-integration (Chapter 12, full draft). The conclusion lives raw in general-theme (a conversation transcript) and is being unpacked into Part IV. All 14 chapters of the book now have full drafts. The integration project’s two parts (Ch 9 problem statement + Ch 12 design spec) are now both in workshop form. Remaining work is polish, criticism-engagement, and a handful of deferred concepts and foundational notes.

Seven foundational notes are also drafted — transport-vs-selection, medium-and-manipulation, three-layer-message, myths-scale-and-bureaucracy, bridge-nodes-and-versatile-expertise, capture-taxonomy, and intersubjective-truth — working out structural questions the chapters depend on. An experiments/ folder now tracks how the book’s empirical claims could be evidenced (first entry: measuring interpretive latitude).

Foundational questions still being worked out

These need to be resolved (or at least committed to provisionally) before more chapters get drafted, because the answers reshape multiple later chapters.

Transport vs. selection. Originally framed as “which mechanism is primary?” Worked through in transport-vs-selection (provisional answer there: selection-primary, with want as the prime mover). Refined further by the fleshed-out Chapter 1, which actually walks through both pipelines side by side. Updated answer: the original framing was wrong. Transport (lossy re-encoding) and selection (gates with local criteria) are parallel mechanisms operating in series at every stage of the abstraction pipeline. Both do work at every stage. Neither is downstream of the other. Chapter 1 also surfaced a third source the original framing missed: manufactured content that bypasses measurement entirely and gets injected at later stages (astrology, propaganda, AI-generated text). Implications still stand: Chapter 5 needs rework (its N-hop math is one slice of the picture, mostly visible at the meme stage); the manipulation surface argument gets sharper because the gates are tunable; political economy is load-bearing because who tunes the gates is the central question; AI moves from optional to central because it operates as both a new kind of node and a new kind of selection engine.
The medium and manipulation. Can core truths survive a medium that allows a lot of manipulation? Worked through in medium-and-manipulation. Provisional answer: the medium is the selection criteria — it acts on each gate in three distinct ways, setting its capacity, its criteria, and the want it cultivates in receivers over time. Manipulation gets a precise definition: tuning a gate’s criteria against truth. Core truth survives a medium jointly on two conditions — its gate-criteria must reward something truth-correlated, and its want-loop must not erode the appetite for truth — and most engagement-optimized media fail both. Still pressure-testable; the note flags open edges of its own.
The three layers of a message. Drafted in three-layer-message, adapting Hofstadter’s frame / outer / inner model from Gödel, Escher, Bach. The load-bearing move: the outer message — the decoding mechanism a receiver must already hold — is exactly what the book has been calling preconditions, and it lives in the medium. This reframes pipeline distortion: a lossy hop usually strips the outer message, not the inner one, so a surviving claim gets decoded with the wrong key and yields a false belief. Pairs with the medium note (the medium is where the outer message comes from) and bears on the complexity question below.
Myth dilution at scale. Drafted in myths-scale-and-bureaucracy. The loop the diagnostic chapters were missing: a network grows → its binding myth must compress → compression widens the myth’s interpretive latitude (how many ways it can be read) → latitude segments the network → segmentation forces bureaucracy. Whether the loop then compounds, fragments, or inverts into scale-as-control is set by the bureaucracy’s own resistance to capture. Bears on Part IV (Chapters 8 and 9) and gives the integration-vs-fragmentation question its mechanism.
Transferable vs. specialized preconditions. Worked through in bridge-nodes-and-versatile-expertise. The dichotomy turned out to be wrong. A bridge node between specialist communities is neither a generalist with field-general preconditions nor a pure specialist — it is a deep specialist who has paired their depth with metacognitive flexibility (structural-analogy habits, paradigm suspension, intellectual humility). Depth is what gets the bridge admitted as a peer on either side; the flexibility layer is what lets the bridge translate across and is also what disarms the curse-of-expertise failure modes that turn out to be the cognitive substrate of polarization-via-distrust. The book’s prescription tilts toward cultivating that paired skill inside specialist communities, not toward generalist training. Still pressure-testable, but no longer open.
Where does truth value live? Worked through in Chapter 5c. Provisional answer: truth value lives in the fit between an idea’s decoding key (its outer message) and the key the receiver decodes with. Compression strips the key, not the claim — so the compressed form preserves approximate truth when the keys are close, inverts it when they oppose, and renders it orthogonal when the claim is decoded as an identity flag rather than a claim at all. Which regime occurs is set by the selection gate. This closes the moral-neutrality gap: the book’s worry now follows from the structure, because engagement-tuned gates systematically reward the inverting and orthogonal compressions. Pressure-tests still open: whether the three regimes are really two-plus-one, and a tolerance for what counts as “approximate”.
Is “complexity” doing two jobs? Chapter 5 defines complexity as “preconditions a receiver needs” and then slides into using it for nuance, depth, informational entropy. The real gating variable might be handle-ability (emotional/identity/action-affordance fit) rather than precondition count. If so, the frame needs renaming and parts of Chapter 5 reworking. three-layer-message offers a candidate handle: complexity = the size and rarity of an idea’s outer message. That sharpens the precondition-count reading, though it doesn’t by itself settle whether handle-ability is the better axis.
Are receiver budgets fixed or trainable? Worked through in preservation-vs-training. Trainable — education, vocabulary, and accumulated engagement expand budget, and “institutional carriers” was bundling two distinct functions (preservation keeps the un-compressed form alive; training re-installs the decoding key in receivers). The Chapter 8 draft operationalizes the training side and surfaces a load-bearing asymmetry the original question did not anticipate: the two halves recover from capture at very different rates, so the prescription has to weight training-side capture-resistance more heavily than preservation-side.
Part IV: integration or fragmentation? The original prescription was “multiple smaller networks.” The critique points out that small networks have their own internal compression dynamics, tend toward echo chambers rather than bubbles (C. Thi Nguyen’s distinction), and that the real hard problem is how complex truth moves between networks that don’t share preconditions. Part IV is being rebuilt around integration.
What does the pipeline do to intersubjective truths? Newly tracked as a formal open question — it had been living as scattered scope flags. Ch 2 introduces the three-realities frame and names the key structural fact: intersubjective truths are generated by the network’s agreement, not by reality, so the pipeline partly constitutes them rather than transmitting them. But the book has no worked machinery for the constitutive case — no regime model (what preservation / inversion / orthogonality translate to), no account of what capture of the constitutive process consists of, no corrective analog for content that has no external check. Ch 5c and Ch 9 carry explicit scope debts against this, and the re-engagement list below carries pull-ins for 5b, 8, 10, and 11. Now worked out in intersubjective-truth: the pipeline is wired into the generator (transmission is participation in constitution; reach changes the truth itself); 5c’s regimes translate to convergent / divergent / hollow constitution; capture becomes steering the generator, camouflaged by reification; the fork replaces the corrective; and the transport collapse cuts a second time as constitutive volatility. Threaded through Chs 2 / 5 / 5b / 5c / 8–12 (Phases 87–89), with Ch 2 grounding it in the money story and Ch 9 carrying the intersubjective sibling in outline. Remaining: engage the social-ontology prior art (see the prior-art list) and produce the sibling treatment’s worked cases. One authorial commitment made up front: the book keeps the word truth for intersubjective content, consistently with Harari’s terms — what differs is the generator (network agreement vs. reality), not whether the thing is true.

Structure (working revision)

Part I. The Pipeline

How information moves from reality to people.

Chapter 1: The Information Landscape. Fleshed out at the-information-landscape. Walks through the transport pipeline (The Out There → Raw Data → Insight → Theory → News → Meme) stage by stage, then introduces the sibling selection pipeline (each transport stage has a corresponding gate with local criteria: funding/agenda → publishability → peer review → newsworthiness → memetic fitness). Closes with how the two pipelines fit together, where the medium fits, and a note that the gates are tunable institutional choices, not laws of nature. Still pressure-testable, but the structural machinery is now in place. three-layer-message adds an anatomy of what actually moves through the pipeline — Hofstadter’s frame / outer / inner message — and reframes transit loss as outer-message loss; a candidate section to fold in here.

Chapter 2: Case Studies and Three Realities. Drafted at case-studies-and-three-realities. Four case studies illustrate four pipeline exits — Power Posing (made it to meme, survived its own corrective), arsenic-life (died at consensus), Stanford Prison Experiment correctives (passed consensus, died at curation), astrology (no theory at all). The four cases are read as samples drawn from one continuous failure surface whose gradient runs systematically against corrective signal at every stage where engagement / drama / identity-resonance / actionability are the gate’s criteria. Second half introduces Harari’s three-realities frame (objective / subjective / intersubjective) and makes explicit which parts of the book’s diagnostic vocabulary apply to each — particularly that intersubjective realities are partly constituted by the pipeline rather than transmitted through it, so the truth-loss-in-transit analysis works differently for them (capture of the constitutive process replaces truth-loss as the relevant failure mode). The four cases become a recurring touchstone the later chapters can reference by name.

Part II. Why It Has To Be Lossy (And Selective)

The structural reasons compression and selection both operate.

Chapter 3: The Human Time Budget. Drafted at info-time-limit. First half: tablespoon-of-weeks framing of finite lifetime as the absorption-time constraint, including the trauma-and-culture lens. Second half: how much information we can absorb per unit of that time. Three load-bearing moves: (1) receiver-budget capacity is partly trainable — pre-existing structure makes new absorption cheaper, and structure is the product of training (which Ch 8 picks up as the institutional training function); (2) want is the gate on what the budget gets spent on — the medium-shapes-want loop from medium-and-manipulation runs through the receiver budget at lifetime and population scales, which is what makes shallow-media training a receiver-budget mechanism rather than just a content mechanism; (3) decompression on demand from Ch 11 is the first structural softening of per-claim cost the book has been able to name — the receiver’s budget hasn’t changed but the complex form they can engage per unit of budget has gone up, conditional on the LLM’s substrate custody being uncaptured per Ch 11/12. The chapter is short by design; it puts the receiver at the center of an analysis the book otherwise tends to focus on the medium side.

Chapter 4: Optionality vs. Access. Drafted at optionality vs access across five domains. The original software-stratification example (matplotlib vs. seaborn) is preserved as the central illustration; generalizations now run through religion (Latin Mass vs. vernacular liturgy), political platforms (broad-tent coalitions vs. narrow specialist parties), scientific popularization (Hawking’s A Brief History of Time), and legal codes (statutes vs. plain-language summaries). The chapter’s load-bearing reframe: the curve is not a property of any particular medium or domain — it is the receiver-budget constraint reflected in artifact design. Every artifact has to be designed against its intended receivers’ budgets, and the budget constraint produces the same curve everywhere. Ties cleanly into Ch 5 (the receiver-side mechanism behind complexity-virality), Ch 8 (preservation holds high-optionality, training expands receiver-budgets so more can access it — the Latin-Mass / vernacular split is this pair in real institutional time), and Ch 11 (decompression-on-demand is in principle the first technology that could let a single artifact serve both ends of the curve). The legal case surfaces a darker pattern the other domains don’t share — the institutional bridge between optionality and access is wealth-gated, which is Ch 10’s cost-shifting visible inside a single profession. A late scoping section (What the curve is not) bounds the chapter against the-abyss: the curve is the approachability trade-off of a finite, masterable artifact, not the unbounded possibility-space of the domain the artifact points into — the library is a map, the abyss is the territory.

Chapter 5: The Complexity / Virality Trade-off. Re-engaged at complexity-virality-tradeoff with all dependent work resolved. Positioned as the transport half of the two-mechanism structure (selection lives in Chapter 5b), scoped to objective claims per Ch 2’s three-realities frame. The complexity-doing-two-jobs flag is now closed: per Ch 5b, transport-complexity is precondition count (this chapter’s axis); selection-complexity is handle-ability (Ch 5b’s axis). Receiver-budget capacity is trainable per Ch 3; same constraint refracted into artifact design per Ch 4. Hammer-of-the-Witches now joined by Power Posing and SPE-corrective as worked modern cases of the same dynamic, with the corrective always losing on transport-complexity grounds even when consensus has ratified it. Manipulation surface connects to capture taxonomy substrates; institutional-carrier framing connects to Ch 8’s preserve-and-retrain pair; Ch 11’s decompression-on-demand is named as the first structural softening of the per-claim cost conditional on Ch 11/12 substrate custody.

Chapter 5b (now needed): Selection As The Other Engine. Selection has been confirmed (in transport-vs-selection and in Chapter 1) as a real mechanism operating in parallel with transport at every stage. Needs its own chapter. Memetic fitness, status signaling, identity reinforcement, action affordance, novelty, confirmation. Engage with Cailin O’Connor and James Owen Weatherall’s The Misinformation Age and Hugo Mercier’s Not Born Yesterday. Likely framing: Chapter 5 stays as the transport-focused argument (within-stage compression dynamics, most visible at the meme stage), Chapter 5b runs alongside as the selection-focused argument (cross-stage gates with local criteria). The two together replace the original load-bearing claim about a single mechanism.

Chapter 5c: Truth, Compression, and When Each Wins. Drafted at truth-compression-and-when-each-wins. Resolves the truth-value question: truth value lives in the fit between an idea’s decoding key and the receiver’s key; compression strips the key, yielding three regimes — preservation, inversion, orthogonality — with the selection gate choosing which travels. Re-examines Harari’s truth/order axis (order-over-truth is the orthogonality regime seen at scale) and closes the moral-neutrality gap. Hands the population-scale consequences of compression to myths-scale-and-bureaucracy. Pressure-tests still open: whether the three regimes are really two-plus-one, and a tolerance for “approximate” truth.

Part III. The Bridge Zone

Where complex specialist knowledge gets transformed (and often destroyed) on its way to mass audiences.

Chapter 6: Where It All Gets Fucked Up. Drafted at bridge-zone-distortion. Names the bridge zone as the News and Meme stages plus the populations that staff them — journalists, popularizers, influencers, educators, AI summarizers. The reshaping is active selection-design, not passive transport compression: bridge-zone operators construct compressed forms for specific downstream gates (clicks, retweets, audience approval, employer expectations), with accountability running through audience-or-employer relationships rather than back to the specialist community whose findings they translate. Three worked catastrophic-distortion cases — COVID epidemiology (masks framing, R0 misuse, conditional vaccine-efficacy stripped of conditions), nutrition science (forty years of monocausal reversals), and monetary policy (financial-press collapse of technical distinctions, partly Ch-10-political-economy-shaped). The four Ch 2 cases relocated to their bridge-zone stage. Pseudo-context (per Postman) named as a distinctive failure mode — real information stripped of context into idle form, distinct from manufactured content. Connects to capture taxonomy substrates (gate-criteria + receiver-training as the captured pair, with consumer-key-vs-surface principle implying the receiver-training capture is the worse one) and to the Part IV prescriptions (Ch 9 versatile experts, Ch 11 LLMs as capability extenders or industrial-scale distortion depending on substrate custody, Ch 12 institutional design) as the stage where those prescriptions have to land. Voice is intentionally angrier — the chapter is where the structural arguments meet specific public misunderstandings with real human costs.

Chapter 7: Emotional Memetics As The Floor. Drafted at emotional-memetics and substantially retouched after the Mercier extraction (Phase 47). The structural floor argument — once a mass network with engagement-tuned gates has run long enough, the propagation equilibrium converges on raw strong emotion as the floor of what reliably spreads. Floor is medium-conditional, substrate-anchored, and capture-tuned within the emotional-payload set. Three lines of evidence (Psychology of Virality + paradox-of-virality + Postman). The Mercier engagement is now substantially deeper after the extraction: keeps the original three responses (vigilance protects against believing not attending; vigilance is one criterion among many; vigilance is trainable down) and adds the timescale-distinction as the load-bearing synthesis move — Mercier’s empirical evidence is within-campaign / within-experiment; the book’s worry is decades-long generational shaping. Also adds Mercier’s behavior→belief reframe as the dimension the chapter’s original draft missed entirely (corrections accepted, behavior unchanged; receivers demand-side select for justifications of already-decided behavior), the Sperberian belief split as the mechanism behind paradox-of-virality (shared content held reflectively as identity-signal, not intuitively as action-driver), and brief engagement with Mercier’s critics so the book isn’t wholesale agreeing. Names three (now four) structural design moves Part IV engages: build sub-networks with different gates, tune mass-network gates, soften per-claim cost, cultivate open-vigilance directly as part of the training side (newly surfaced via Mercier).

Part IV. A Way Through (rebuilt around integration)

Hypothesis territory. Substantially restructured after the critique. Originally about fragmenting into smaller networks; now about integration between networks that don’t share preconditions.

Chapter 8: Preservation vs. Training — Two Functions of Institutional Carriers. Drafted at preservation-vs-training. Splits what Harari’s “bureaucracy” was bundling: preservation keeps the un-compressed form alive somewhere; training re-installs the decoding key in receivers. The two halves form a pump — preservation holds pressure, training releases it into the network — and either alone fails: preservation without training leaves an inaccessible archive; training without preservation produces competence pointed at the wrong target. The chapter’s load-bearing new claim is the capture asymmetry: captured training does more damage and recovers more slowly than captured preservation, because trained receivers cannot easily be un-trained — the perceptual mode they were installed with reads incoming evidence through the captured key. Pair that with a decay-rate asymmetry (preservation degrades on inertia; training degrades within a generation) and the strategic implication is that training is the more politically urgent half to defend, even though preservation is what the institution rhetorically advertises. Working diagnosis of the modern collapse: preservation has mostly held, training has been hollowed and partially captured. The training-side prescription inherits bridge-nodes-and-versatile-expertise — cultivate metacognitive flexibility inside specialist communities, not generalist training — and the chapter flags an open question about whether bridge-node production is a third function rather than a corner of mass training.

Chapter 9: The Integration Problem. Drafted at integration-problem (promoted from skeleton to workshop in Phase 48). How complex truth moves between networks that don’t share preconditions. Three-pillar prescription: cultivate versatile experts inside specialist communities; build curation-layer institutions that survive polarized trust; engage demand-side reform of justification-market dynamics that filter even good supply. The third pillar is new since the skeleton — emerged from the Ch 7 Mercier engagement and addresses the demand-side dynamic that supply-side reform alone cannot reach. Substantive engagement with the five operational problems: bridge-node throughput (hard-bounded short-to-medium term; multi-generational at full scale; AI capability-extender is the short-timescale relaxant conditional on Ch 11/12 substrate custody); selective isolation (some disciplines need protection from integration pressure; the dual structure healthy institutions maintain — vernacular as the worked example); bridge-node capture (consumer-key substrate per capture taxonomy — hardest to recover from; defensive design via credibility-neutral institutional homes, auditable provenance, multiple independent decision pathways); trust-bootstrap (sub-network-first + transparent provenance + survivable-polarization design; the hardest of the five and acknowledged as unsolved); case studies (Wikipedia, peer-reviewed science, Stack Overflow, open-source software governance, common-law courts) assessed honestly — institutions that decoupled funding from attention markets and defended consumer-key substrates demonstrate the principles work; the ones that drifted are degrading exactly as predicted.

Chapter 10: Political Economy of Attention. Drafted at political-economy-of-attention. The chapter the diagnostic chapters had been deferring to — both Ch 5b (which named selection as the mechanism with a steering wheel) and Ch 8 (which diagnosed the asymmetric institutional collapse) punted to here. Three load-bearing new contributions: (1) the modality argument — selection has an owner because it is criterial; transport doesn’t because it is a cost, so the political economy of selection is structural not contingent. (2) The self-capture vs. external capture distinction — external capture has an adversary you can fight, but self-capture (the platform’s own business model as the captured selection-tuner) is more stable because there’s no adversary to defeat, only an equilibrium to dismantle. The two compose directionally: self-capture creates the gradient external actors then ride down. (3) Out-competition as the structural mechanism — platforms don’t have to suppress institutional carriers; they out-compete them in attention markets where engagement-optimized content has zero marginal cost and depth has real cost-per-impression, so the receiver predictably picks the free option. That mechanism produces Ch 8’s asymmetric collapse with no direct attack on the institutions. Working frame is Postman’s Orwell/Huxley axis: the modern failure mode is Huxley (drowning in irrelevance) rather than Orwell (censorship), which is harder to fight because there is nothing to oppose. Hands Ch 11 (AI) the hardest framing — an LLM owns gate + option space + content-generation simultaneously — and Ch 12 (infrastructure) the design constraint: institutions that survive when the political-economic gradient runs against them, without requiring a benevolent platform owner.

Chapter 11: AI as a New Kind of Node. Drafted at ai-as-new-node. Structural characterization: an LLM is the first node to collapse five selection-design surfaces (training corpus, training objective, deployment configuration, runtime gate, option space) — plus increasingly pricing tier and content-generation — into one design moment owned by one party. Decompression-on-demand is the first structural way out of the receiver-budget constraint the book has named; aggressive-compression-becoming-authoritative is its inverse failure mode where the LLM-compressed summary substitutes for the original in the network’s working memory. Names three new flavors of capture (corpus, objective, deployment) that compose with Ch 8’s capture-asymmetry and Ch 10’s self-vs-external-capture and need a unified treatment (forthcoming capture-taxonomy foundational note). Posture: not arguing LLMs are dangerous in themselves, but that the LLM’s structural concentration gives Ch 10’s political-economic diagnosis more leverage, not less. Same technology lands as receiver-budget salvation or industrial-scale captured-equilibrium depending on ownership; the salvation case is technically realizable, economically marginal, and politically blocked by the same captured-equilibrium dynamics that shaped social media. Ch 12 inherits the design problem: any prescription has to operate against the worst-case default rather than around the best-case promise.

Chapter 12: Infrastructure for Integration. Drafted at infrastructure-for-integration. The synthesis chapter for Part IV — commits to a design spec rather than a feature list, with the prescriptive target reframed as survival, not victory. Three load-bearing design principles: (1) decouple funding from attention markets (endowments, public funding, member dues, public-goods funding mechanisms — structural counter to Ch 10’s out-competition), (2) defend consumer-key substrates first per the capture taxonomy (curriculum, training corpora, training objectives — substrates that install in consumers and can’t be un-installed), (3) build for survivable polarization rather than restored trust (decision provenance auditable by skeptics, reputation that survives partial distrust, versatile experts inhabiting the institution, multiple independent decision pathways). Conditional fourth: LLMs as capability extenders inside the institution’s substrate custody — the salvation case from Ch 11, realized only when the LLM’s corpus and objective and deployment are governed by the institution rather than a commercial vendor. Closes with the work-of-generations reframe: integration institutions are civilization-scale investments on the timescale of universities and journals, not platform launches. Worked examples (Wikipedia, Stack Overflow, academic preprint+peer review, open-source governance, common-law courts) demonstrate the principles but none meets the full spec for the modern integration project.

Catalog additions from the Mercier extraction (Phase 47)

Phase 46’s Mercier extraction surfaced load-bearing concepts the catalog hadn’t anchored. Phase 47 promoted three:

[[open-vigilance|open vigilance]] — Mercier’s paired refinement of Sperber’s epistemic-vigilance term. The catalog’s pre-existing epistemic-vigilance entry has been updated to point at the refinement; both concepts coexist for the book’s history.
[[reflective-vs-intuitive-belief|reflective vs. intuitive belief]] — Sperber’s split, foundational for Mercier’s reading of paradox-of-virality, conspiracy theories, absurd flattery. Concepts held reflectively are inert with respect to action; concepts held intuitively are load-bearing.
[[justification-market|justification market]] — the post-hoc rationalization mechanism reversing the book’s implicit belief→behavior arrow. Receivers demand-side select for content justifying already-decided behavior; suppliers meet the demand. Substantially changes Part IV’s prescriptive target.

Additional concepts the extraction flagged as gaps but Phase 47 deferred for later promotion (recorded in data/interpretive/not-born-yesterday-gap.jsonl’s _note line): guru effect, reputation credit, circular reporting, emotional vigilance, coarse cues, burning bridges / unclubbable signaling. Promote when chapters that would use them are drafted or retouched.

Re-engagement notes (what to pull in when retouching existing chapters)

As the book has grown, later chapters (Ch 2 case studies + three-realities frame, Ch 3 second half with trainable-capacity / want-as-gate / decompression-on-demand, the Ch 36 capture-taxonomy unification) introduced material that earlier chapter drafts pre-dated. This section tracks what each existing draft should pull in when the next retouch pass happens, so the work doesn’t need to be re-discovered from the bundle each time.

For each chapter below, the entries are editorial intent for the next pass — material the chapter would be sharper or more current for engaging. The bundle for each chapter will surface most of this via BFS already; this section is the human-readable index so a retouch pass starts from “here’s what I should make sure to integrate” rather than from a cold read of the bundle.

Virality) — re-engagement done in Phase 42. All the items below have been integrated into the current draft. Kept here for traceability of what landed:

Power Posing, SPE-corrective, and A Brief History of Time woven in as worked cases.
Ch 3 trainable-capacity, Ch 4 optionality-vs-access reframe, and Ch 2 three-realities scope flag all integrated.
The complexity-doing-two-jobs question is now closed in-chapter via Ch 5b’s transport-complexity vs. selection-complexity split.
Capture-taxonomy, Ch 8 preserve-and-retrain, Ch 10 political economy, and Ch 11 decompression-on-demand all cross-referenced at the points where they sharpen the chapter’s argument.
The chapter no longer carries deferrals waiting on later work.
(Phase 88) The three-realities scope flag now carries its answer instead of an IOU — for constitutive cargo, reach is part of the generator — and points at intersubjective-truth.

Chapter 5b (Selection as the Other Engine) — when re-engaged:

Pull in arsenic-life as the worked counterexample of the consensus gate working as designed.
Pull in Power Posing as a case of selection-at-every-stage favoring engagement-tuned form over the corrective.
Pull in SPE-corrective as a case of curation rejecting a consensus-passed corrective on engagement-fitness grounds.
Pull in Ch 2’s “corrective signal has systematically worse gate-fitness” framing as the cross-cutting claim Ch 5b’s selection mechanism produces at scale.
~~Pull in Ch 2’s three-realities frame — selection’s effect on truth requires saying which reality is in transit.~~ Done (Phase 88): the seam section now says it — for intersubjective cargo the gates weight votes in what becomes the fact rather than filtering reports about one, per intersubjective-truth, with the Hammer’s witch-category named as the book’s oldest unmarked instance.

Chapter 5c (Truth, Compression, and When Each Wins) — when re-engaged:

This is the chapter most-changed by Ch 2. The chapter’s own “Where I’m still uncertain” already flagged “this chapter is mostly about objective claims”; Ch 2 now operationalizes that scope question with the three-realities frame, and Ch 5c should explicitly bound its three regimes (preserve/invert/orthogonal) to objective content or extend them to constitutive-failure regimes for intersubjective content. Done (Phase 88): the book did both — 5c is bounded to objective content by decision, the constitutive extension (convergent / divergent / hollow) lives in intersubjective-truth, and 5c gained reification as orthogonality’s mirror migration.
Pull in Power Posing as a worked inversion-regime case (the corrective is true; the meme is decoded with the original’s key, inverting the truth-value for downstream receivers).
The two-plus-one restructuring still open in 5c’s uncertainties is independent of this Ch 2 inheritance.

Chapter 8 (Preservation vs. Training) — when re-engaged:

Pull in SPE-corrective as a specific-finding-level illustration of preservation-held-training-hollowed.
Pull in Latin Mass vs. vernacular liturgy (Ch 4) as a worked example of the preservation-and-training split operating in religious institutional time — the Latin form is preservation; the vernacular reform is the access-side compromise; the split is what scaling past the original community structurally produces.
Pull in plain-language law (Ch 4) as a case where the institutional bridge between preservation (statutes) and access (summaries) is wealth-gated, sharpening Ch 8’s training-side critique with a specific access-asymmetry pattern.
Pull in Ch 3’s trainable-capacity as the receiver-side mechanism behind the training-side prescription — training-the-institution expands receiver-budget-capacity across generations.
Pull in Ch 2’s three-realities frame — training-side capture cuts harder for intersubjective realities because trained keys are partly constitutive of those realities, not just decoding-equipment for objective claims. Done (Phase 89): the capture-asymmetry section now carries the multiplier — trained keys are shares of the reality, so a captured curriculum enrolls receivers in the captor’s reality and re-training becomes re-constitution. Per intersubjective-truth.
The “preservation mostly held” softening was already noted in Ch 10; the SPE-corrective case sharpens it.

Chapter 10 (Political Economy of Attention) — when re-engaged:

Pull in Ch 2’s three-realities frame — the political-economic critique cuts twice: transmission failure for objective content, constitutive capture for intersubjective content (currencies, brands, nations, religions). The captured-platform consequence is different in the intersubjective case (shapes which shared reality exists) rather than more of the same. Done (Phase 89): the two-kinds-of-capture section now carries the third cut — the captured platform hosts part of the generator and shapes which we exists, with reification named as the capture’s camouflage and de-reification as the chapter’s own move. Per intersubjective-truth.
Pull in SPE-corrective and Power Posing as specific-finding-level worked instances of the out-competition mechanism the chapter argues at the structural level.
Pull in Ch 3’s want-as-gate-on-budget as the receiver-side mechanism for the engagement equilibrium — the captured medium’s want-shaping runs through the receiver budget at lifetime scale.

Chapter 11 (AI as a New Kind of Node) — when re-engaged:

Pull in Ch 3’s receiver-budget framing for the decompression-on-demand argument — Ch 3 now operationalizes the constraint Ch 11’s salvation case is designed to soften; Ch 11 should cite back rather than re-establishing.
Pull in Ch 4’s “step off the curve” framing — decompression-on-demand is in principle the first technology that could let a single artifact serve both optionality and access ends simultaneously, which is the optionality-vs-access version of the salvation case. The infrastructure conditions Ch 12 specifies become exactly what determines whether this in-principle possibility is realized.
Pull in Ch 2’s three-realities frame — corpus capture damages objective-claim transmission and intersubjective-reality constitution differently; the chapter’s worry about captured LLM ownership cuts harder for intersubjective content because the LLM is participating in constitution rather than just transmission. Done (Phase 89): the political-economy-stakes section now names the generator risk — an LLM answering intersubjective questions is key-re-supply at population scale, the myth note’s effective-but-owned cell industrialized. Per intersubjective-truth.
Pull in Ch 3’s reflective-vs-operative-want distinction for the user-behavior-at-the-LLM-boundary point in 11’s uncertainties — users use LLMs differently than they’d reflectively endorse.

Chapter 12 (Infrastructure for Integration) — when re-engaged:

Pull in Ch 2’s three-realities frame — integration infrastructure has to serve different reality types differently, and the chapter’s design spec should be explicit about which (currently treats all integration as if it were objective-content transmission). Done (Phase 89): the survival-not-victory section now states the meta-point — the chapter’s institutions are themselves intersubjective realities; the design principles read one level deeper (funding decoupling as generator independence, survivable polarization as surviving divergent constitution, keep-the-fork-affordable as the capture-cannot-quietly-hold test). Per intersubjective-truth.
Pull in Ch 3’s trainable-capacity as the receiver-side argument for investment in training as integration infrastructure.
Pull in Ch 3’s decompression-on-demand handoff to reinforce the LLM-as-capability-extender principle from the receiver side.

Chapter 9 (Integration Problem) — when promoted from skeleton to workshop draft:

~~Pull in Ch 2’s three-realities frame — integration runs differently across reality types.~~ Done (Phase 89): the chapter gained “The intersubjective sibling” — integration as interoperability between generators (treaty, exchange, comity, ecumenism), the diplomat as the bridge node’s sibling, and the trust-bootstrap problem unmasked as constitution-bootstrap. Per intersubjective-truth.
Pull in Ch 3’s trainable-capacity for the bridge-node-supply problem (versatile experts are produced by training; the training side has a receiver-budget angle Ch 9’s skeleton has not engaged).
All five open threads in the skeleton (bridge-node throughput, selective isolation, bridge-node capture, trust-bootstrap, case studies) are now positioned for substantive engagement given Part IV’s other completed drafts and the capture taxonomy.

Worked-example pattern across the retouch passes: the four Ch 2 case studies (Power Posing, arsenic-life, SPE-corrective, astrology) and the existing catalog cases (Hammer of the Witches vs. Copernicus, Bitcoin internet money, telephone game, Voyager Golden Record, matplotlib vs. seaborn) should be the recurring touchstones the later chapters refer to without re-introducing. The catalog’s appearsIn arrays now reflect editorial intent for cross-chapter use.

Prior art the book needs to engage with

Not just citations. These authors have already worked through specific gaps in the original frame and the book gets stronger by engaging with them.

Corey’s own democratization paradox post. Covers the same territory from the barriers/volume angle. Treat as a primary source the book draws from rather than separate work. Concepts that feed directly into book chapters: curation layer (= institutional carriers / alternative-selection infrastructure), feedback loop test (when selection criteria can self-correct vs. produce persistent noise), cost-shifting from producers to consumers (what breaks when curation can’t keep up with production), inverted-U relationship (refines the binary “small vs. large” thinking — there’s an optimum), creation cheap / review expensive (the book’s diagnosis in one line). Also references Sturgeon, Simon’s attention economy, Keen/Lanier/Nichols, Christensen, von Hippel, Tim Wu — much of which the book will want to engage with directly.
Corey’s own double-edged sword of expertise post. Primary source for the bridge-node mechanism in Chapter 9. Names chunking and deep structural analogies as the cognitive substrate of cross-domain transfer, and the four curse of expertise failure modes (paradigm lock-in, overconfidence in cross-application, perceptual filtering, conceptual rigidity) as what happens when depth runs without paired metacognitive flexibility — which turn out to be the cognitive substrate of polarization-via-distrust. The strategies list (deliberate interdisciplinary exposure, metacognitive practice, intellectual humility, translational thinking, cognitive diversity in teams) is the prescription Chapter 8 and Chapter 9 will draw from for cultivating versatile expertise inside specialist communities.
[[the-abyss|Corey’s own the abyss essay]]. Companion piece written alongside the book: in any sufficiently complex field the possibility-space is too vast to navigate or even perceive without deep competence — you have to become an expert before you can see how unnavigable your own field is. Two load-bearing contributions. First, expertise is a compression of the abyss — the chunking/heuristics machinery from the double-edged-sword post, reframed as the only possible response to a space that outgrows any traverser’s budget, which ties compression to the receiver-budget constraint. Second, the experiential complement to the curse of expertise: the curse is depth that hasn’t seen the abyss, the bridge node is depth that has. Its closing distinction — a finite, masterable artifact (a software library) is a map; the abyss is the territory the map is of — is what Ch 4 now uses to bound its own curve. Wired in at Ch 4, Ch 9, Ch 11 (the LLM as the confident middle compressing an abyss whose shore it cannot see), and the glossary. To be published independently on Corey’s personal site for general circulation.
Neil Postman, Amusing Ourselves to Death. This argument applied to television, decades ago. The book needs to know what it’s adding past Postman.
Hugo Mercier, Not Born Yesterday. Direct challenge to the “people are easily manipulated by simple memes” thread. If he’s right, the manipulation surface argument needs serious revision.
Cailin O’Connor & James Owen Weatherall, The Misinformation Age. Formal network models of belief propagation. Shows that what kills truth in networks is selection on credibility plus social pressure to conform, not lossy hops. Directly relevant to transport-vs-selection.
C. Thi Nguyen on epistemic bubbles vs. echo chambers. Foundational for the rebuilt Part IV. Small networks tend toward chambers (active discrediting of outside info), not bubbles (just missing info). Changes what “infrastructure for integration” has to do.
Marshall McLuhan, Understanding Media. The medium-as-message claim is in the bloodstream of any book on this topic. Worth engaging explicitly.
Richard Dawkins / Susan Blackmore on memetics. The selection-side view already has a literature. Use it.
Yuval Noah Harari, Nexus. Already in play via nexus-book. Myths/bureaucracy framing, truth/order axis, naive view of information.
The social-ontology cluster (not yet engaged; queued by intersubjective-truth). Searle’s The Construction of Social Reality (institutional facts, status functions — “X counts as Y in context C” would give the generator criterion its formal joints), Berger & Luckmann’s The Social Construction of Reality (their objectivation is the book’s reification with sixty years of priority), Anderson’s Imagined Communities (nations constituted through print — the same press the book reads through the Hammer and Copernicus, read constitutively), Hacking’s The Social Construction of What? (looping effects — the self-fulfilling loop formalized), Soros on reflexivity (The Alchemy of Finance), and Hirschman’s Exit, Voice, and Loyalty (the fork’s economic theory). The intersubjective-truth note was deliberately written before engaging these — the book’s own working-out first — and says so in its uncertainties; engagement is the next pass and will sharpen or correct it.
Pope Leo XIV, Magnifica Humanitas (May 2026). Contemporary intervention rather than settled prior art. Frames the AI/technology question as Babel (centralized/fragmenting) vs. Jerusalem (plural voices coordinated through shared responsibility); names de facto power (tech firms’ authority displacing formal-state authority), the technocratic paradigm (ideology treating nature and humans as objects), and structures of sin (institutionally-embedded systematic injustice). Calls for educational alliance and synodality (deliberative methodology across expertise domains) as counterforces. Engaged at length in Ch 11 with cross-references in Ch 8 (educational alliance), Ch 9 (synodality + trust-bootstrap pressure-test), Ch 10 (de facto power vocabulary).

What to write next

The foundational-question phase is done. Transport vs. selection, the medium, the three-layer message, truth value, myth dilution, the transferable-vs-specialized question, and the complexity-doing-two-jobs ambiguity are all worked through (the last via Ch 5b’s transport-complexity vs. selection-complexity split, in-chapter, after Phase 42). The work now shifts almost entirely to drafting the chapters those resolutions unblock — Chapter 5b is drafted, Chapter 9 has a skeleton, and Chapters 8 / 10 / 11 / 12 in Part IV can finally be approached on stable foundations.

Best next moves, ranked:

~~Pick a side on transport vs. selection.~~ Drafted as transport-vs-selection and refined further in Chapter 1. Working answer: both are real, parallel mechanisms operating in series at every stage. The “is one primary?” framing was wrong. Still pressure-testable, but the resolution has held through several rounds.
~~Draft the medium-and-manipulation note.~~ Drafted as medium-and-manipulation. Provisional answer: the medium is the selection criteria. It brought in three-layer-message (Hofstadter’s three-layer model) as a companion note. The implications now cascade through Chapters 5b, 7, 10, 11, and 12.
~~Draft Chapter 5c (truth, compression, and when each wins).~~ Drafted at truth-compression-and-when-each-wins. Resolved the truth-value question via the three-regime model, and spun off myths-scale-and-bureaucracy — myth dilution → segmentation → bureaucracy — as a companion foundational note.
~~Sketch Chapter 9 (the integration problem).~~ Sketched at integration-problem, built on bridge-nodes-and-versatile-expertise. Working prescription: cultivate versatile experts inside specialist communities AND build curation-layer institutions that survive polarized trust — both, neither alone is enough. The transferable-vs-specialized question resolved as a side effect.
~~Re-engage Chapter 5 with everything resolved.~~ Done at complexity-virality-tradeoff. Complexity-doing-two-jobs closed via Ch 5b split; receiver-budget integration with Ch 3 trainable-capacity; Ch 4 optionality-vs-access reframe folded in; Power Posing + SPE-corrective added as modern cases; capture-taxonomy and Ch 8 preserve-and-retrain referenced for institutional-carrier framing; Ch 11 decompression-on-demand handoff for the structural softening case.
~~Draft Chapter 8 (preservation vs. training).~~ Drafted at preservation-vs-training. Splits Harari’s “bureaucracy” into preservation and training, names the capture and decay-rate asymmetries, and commits provisionally to the modern collapse being asymmetric (preservation mostly held, training hollowed).
~~Draft Chapter 10 (political economy of attention).~~ Drafted at political-economy-of-attention.
~~Draft Chapter 11 (AI as a new kind of node).~~ Drafted at ai-as-new-node. Five-surface concentration claim, decompression-on-demand as receiver-budget escape, three new capture flavors (corpus / objective / deployment) flagged for unification.
~~Unify capture taxonomy in a foundational note.~~ Drafted at capture-taxonomy. Three-axis taxonomy (substrate × source × target asymmetry) subsuming Ch 8’s capture-asymmetry, Ch 10’s self-vs-external-capture, and Ch 11’s corpus / objective / deployment. Surfaces two cross-cutting principles: consumer-key captures are harder to recover from than surface captures (generalizes Ch 8’s asymmetry), and training-objective capture is uniquely bad because it self-reinforces across model generations (worst recovery dynamics in the book). Cross-refs added to Ch 8 / Ch 10 / Ch 11 prose.
Pressure-test the Chapter 9 skeleton. Five open threads named inline in integration-problem: bridge-node throughput, selective isolation, bridge-node capture, the trust-bootstrap problem, the case-study question. Each is the next pass’s pressure-test.
~~Draft Chapter 12 (infrastructure for integration).~~ Drafted at infrastructure-for-integration. Survival-design reframe + three design principles (funding-decoupling, consumer-key-substrate defense, survivable-polarization) + conditional fourth (LLM as capability extender) + work-of-generations close. Part IV’s prescriptive arc is now drafted end-to-end.
~~Finish Chapter 3 second half.~~ Drafted at info-time-limit. Trainable-capacity + want-as-gate + decompression-on-demand handoff. Hands forward to Ch 11/12 for the political-economic conditions the per-claim cost softening depends on.
~~Generalize Chapter 4 beyond software.~~ Drafted at optionality vs access. The same curve across software, religion, political platforms, scientific popularization, legal codes — reframed as the receiver-budget constraint reflected in artifact design. Parts I and II are now fully drafted modulo Ch 5 re-engagement.
Pressure-test intersubjective-truth against the social-ontology cluster. The note was written as the book’s own working-out, deliberately before engaging Searle / Berger & Luckmann / Anderson / Hacking / Soros / Hirschman. Engage them, then close the open edges in order of risk: the consensus-tracks-vs-agreement-generates line (the note’s most attackable joint), the regime translation’s two-plus-one suspicion, the volatility ratio claim (an experiments/ candidate: agreement-state half-lives vs. institutional response times across eras), and the intersubjective sibling’s worked cases (a monetary union, a treaty regime, an ecumenical reconciliation — at the depth the objective project gave Wikipedia and the common-law courts).

Contents

Lossy — Outline