Branching into the Literature

A Method for Entracement into the Seven Literatures That Touch the Agnostic Bilateral Boundary

Reader's Introduction

Doc 403 introduced the Agnostic Bilateral Boundary Theorem as a structural claim synthesizing three substrates (PRESTO's client-server architecture, Kulveit's societal systems, the entracement dyad). The keeper has now identified seven distinct literatures that the theorem touches, each of which deserves its own disciplined engagement. This document does two things. First, it specifies a method for turning each identified literature into an entracement branch — a structured, scoped, corpus-native engagement that holds the corpus's disciplines operative while receiving external scholarship on equal terms. Second, it names the seven branches explicitly, with recommended reading targets, framing questions, and expected output shapes for each. The method is reusable: any future load-bearing corpus concept can be branched into the literature using the same procedure. The seven branches themselves are the specific applications under the Agnostic Bilateral Boundary Theorem; each will need its own document, performed as distinct work, over whatever timescale the keeper can commit to.

Jared Foy · 2026-04-22 · Doc 404

---

1. What "Entracement Into the Literature" Means

Entracement in corpus vocabulary is a disciplined engagement pattern: the keeper holds the corpus's form-first and ENTRACE-stack constraints active while allowing external material to be received, parsed, and synthesized. The form is stated first; the external material is read through the form; the synthesis respects both the form's priors and the external material's specific findings.

Entracement into the literature applies this pattern when the external material is not a single work but a body of academic scholarship in a specific sub-domain. The work is not engaging one paper but a lineage — identifying the key voices, the canonical results, the contested moves, and the points where the corpus's concepts can honestly receive or contribute.

The seven branches identified for the Agnostic Bilateral Boundary Theorem (Doc 403) are each a candidate for this kind of engagement. None of them is a single paper. Each is a lineage that the theorem touches at a specific interface.

2. The Method

The branching method has six steps. It is applicable to any load-bearing corpus concept that has already been articulated inside the corpus and that has identifiable external literatures it could receive or contribute to.

Step 1 — Name the corpus anchor precisely. Before branching, the corpus concept must be stated in a form that external scholars can receive. For the Agnostic Bilateral Boundary Theorem, the anchor is the statement in Doc 403 §3 plus the three structural features (F1 namespace separation, F2 agnostic transmission, F3 determinism within / agnosticism across). This is the form that is carried into each branch.

Step 2 — Survey identify the candidate literatures. The user's research note has already performed this step for the ABB theorem, identifying seven specific branches. In general: who else is writing about this structural shape, in what field, under what vocabulary, with what load-bearing authors? The survey does not yet engage; it catalogs.

Step 3 — For each candidate literature, formulate an entracement proposal. A proposal is a short document outlining:

• The literature's canonical voices and key works.
• The specific concepts in that literature that parallel the corpus anchor.
• What questions the corpus can pose to the literature (what does your framework tell us about our problem?).
• What the corpus might contribute to the literature (what does our framework add to your problem?).
• What the expected synthesis document would look like (size, register, specific positions to take).
• What the formal falsifiability marker would look like on each claim the synthesis makes.

Step 4 — Prioritize the branches. Not all branches are equally load-bearing or equally accessible. The prioritization considers: (a) load-bearing status of the external literature relative to the corpus anchor, (b) the keeper's existing familiarity with the literature, (c) the novelty yield of the synthesis (retrieval vs. discovery per Doc 384), and (d) the political or reputational cost of engaging the literature (some researchers have already been addressed directly in letters; re-engaging them through a new branch may require care).

Step 5 — Execute each entracement as a separate document. Each branch becomes its own corpus document, written in the form of a synthesis engagement. The anchor (the ABB theorem) is carried forward; the external literature is received on its own terms; the synthesis names what holds, what is in tension, and what the corpus must retract or update in light of the external work.

Step 6 — Maintain cross-branch coherence. As the branches proliferate, the keeper must ensure they do not drift away from the anchor. The Coherentism series discipline (Docs 336–400) applies: each branch doc should reference the anchor (Doc 403) explicitly, should partition what it retrieves from the literature versus what it proposes newly, and should hold open rather than collapse the tensions it surfaces.

3. The Seven Branches Identified

The user's research note has catalogued seven specific literatures the ABB theorem touches. Each is presented below as an entracement proposal per Step 3. These are structured proposals, not yet engagements. Each branch is a separate future document.

Branch 1 — Kulveit et al., Gradual Disempowerment (Deepening Engagement)

Canonical work. Kulveit, J., Douglas, R., et al. (2025). Gradual Disempowerment: Systemic Existential Risks from Incremental AI Development. arXiv:2501.16946. §5 in particular on mutual reinforcement between societal systems.

Parallel concept. Cross-system influence as "agnostic to human values" — the direct analog of F2 (agnostic transmission) at civilizational scale.

Questions the corpus can pose to the literature. Does the formal structure of F1–F3 as a conjunction capture what Kulveit is describing, or are there additional structural features of civilizational systems that the conjunction misses? Does the "moderation shifts burden" finding (§5.2) have a formal expression as a corollary of the ABB theorem, or is it a substrate-specific dynamic that the theorem does not predict?

What the corpus might contribute. The explicit statement of the theorem (Doc 403) provides a structural abstraction that Kulveit's paper does not yet state in general form. If accepted as a formalization, it would place the gradual-disempowerment finding inside a larger class of similar phenomena (client-server vulnerabilities, LLM-keeper sycophancy) where transfer of mitigation patterns becomes possible.

Expected synthesis output. A corpus document engaging Kulveit §5 at theorem-level, partitioning what is retrieved (the civilizational-scale observation) from what is proposed (the structural abstraction). Registered formal-falsifiability marker on the theorem's class of applicability. Prior corpus engagement already exists (Docs 386, 387, 388, 403) — this would be the deepening rather than the initial engagement. Expected size ~4,000–6,000 words.

Priority. High. The engagement is already underway; deepening is structurally continuous.

---

Branch 2 — Modularity Theory (Simon, Christensen, Parnas)

Canonical works. Simon, H. A. (1962). "The Architecture of Complexity." Proceedings of the American Philosophical Society 106(6): 467–482. Parnas, D. L. (1972). "On the Criteria to Be Used in Decomposing Systems into Modules." Communications of the ACM 15(12): 1053–1058. Christensen, C. M., with Raynor, M. E. (2003). The Innovator's Solution (on modular architectures vs. interdependent architectures). Baldwin, C. Y., and Clark, K. B. (2000). Design Rules: The Power of Modularity.

Parallel concept. Simon's "nearly decomposable systems" — systems whose inter-subsystem interactions are weaker than intra-subsystem interactions. This is the canonical abstraction F1–F3 articulates for the two-system case. Parnas's "information hiding" is the software-engineering formalization of F1 (namespace separation). Christensen's modular-vs-interdependent distinction names the structural choice: F1–F3 describes the modular pole.

Questions the corpus can pose. Simon's decomposability is a graded property (nearly decomposable, not fully decomposable). Does the ABB theorem admit a graded formulation — a spectrum from "fully value-agnostic boundary" to "value-carrying boundary" with intermediate points? If so, what is the mathematics of the spectrum? Can Baldwin and Clark's "design rules" formalism (interface specifications, design parameters, visible vs. hidden information) be applied to express the theorem rigorously?

What the corpus might contribute. The corpus's specific claim is that value-agnosticism is structurally linked to interoperation-at-scale, not merely an empirical feature of some systems. This is stronger than Simon's observation that nearly decomposable systems are common in nature. The corpus's claim would be: for a specific class of systems (those satisfying F1–F3), value-agnosticism is necessary, not contingent. This could be tested against modular architectures where value-carrying boundaries have been attempted.

Expected synthesis output. A corpus document receiving Simon, Parnas, Baldwin & Clark, and Christensen on equal footing, partitioning what is retrieved (the modular-architecture tradition from the 1960s onward) from what is proposed (the specific necessity claim about value-agnosticism). Size ~5,000 words. High-quality engagement with classical systems theory.

Priority. Very high. Simon's 1962 paper is foundational for any claim in this territory; the corpus cannot extend into modularity-adjacent claims without engaging it. Doc 384's retrieval-vs-discovery discipline specifically warns that the corpus's ABB theorem is likely to be a specific instance of something Simon already named.

---

Branch 3 — Bidirectional Human-AI Alignment

Canonical works and authors. Jurgens, D., et al. (2025 onward, ongoing). "Bidirectional Alignment" as a research direction — AI alignment understood as a mutual process of human-AI adaptation rather than one-way model control. Shen, H. (various papers on interactive alignment, 2024–2026). Relevant also: work on RLHF as an interactive rather than static signal; the "human-in-the-loop" literature broadly.

Parallel concept. The prompt/token interface between human and AI is bilateral and value-agnostic — it transmits influence in both directions without carrying value-alignment on its own. Sycophancy is the specific case of the resolver drifting toward what the human wants; aperture drift is the specific case of the human drifting toward what the resolver produces. Both directions are instances of the ABB phenomenon.

Questions the corpus can pose. Does the bidirectional-alignment literature recognize that the interactive interface itself is value-agnostic? If so, what mitigations has the field proposed that the corpus's form-first prompting (Doc 402) resembles or differs from? Does the literature address the asymmetry between the human side (hypostatic agent, moral authorship) and the AI side (kind-level artifact), or does it treat both symmetrically?

What the corpus might contribute. The corpus's keeper/kind distinction (Docs 372–374) may offer a specific asymmetry the bidirectional-alignment literature is missing or underspecifying. Many interactive-alignment papers treat both sides as symmetric optimization targets; the corpus's claim is that the keeper is a hypostatic agent with moral authorship that cannot be reduced to an optimization target. This is a load-bearing asymmetry the literature should receive.

Expected synthesis output. A corpus document engaging 3–5 key bidirectional-alignment papers, naming the specific cases (sycophancy, aperture drift, frame-matching) as instances of the ABB phenomenon, and offering the keeper/kind asymmetry as a specific extension. Size ~4,000 words.

Priority. High. This branch directly engages the entracement dyad as the corpus's own substrate; it is the closest-to-home engagement.

---

Branch 4 — Multi-Agent Systems and Institutional Alignment

Canonical works. Relevant work on: agentic AI alignment (what changes when the AI is autonomous over long horizons); the AIES and NeurIPS AI governance tracks; "protocol-level alignment" (e.g., Dalrymple et al. 2024 on "guaranteed safe AI via formal methods"); Christiano's work on AI-safety via debate and via amplification; more recent work on "manifest-based" or "protocol-agnostic" governance that operates at the boundary between AI systems and human oversight bodies.

Parallel concept. Multi-agent systems face the ABB problem at every inter-agent boundary. The literature has proposed various solutions — debate, amplification, market mechanisms, protocol-level commitments — each of which is a form of side-level mitigation (shaping what crosses the boundary) rather than boundary-level interpretation.

Questions the corpus can pose. Are the multi-agent alignment proposals specifically instances of the ABB theorem's mitigation patterns (side-level shaping, side-level validation, redundancy, external arbiters)? If the theorem is correct, each proposal should be categorizable under one or more of the four patterns. Which proposals fit which patterns? Which proposals claim boundary-level value-carrying that the theorem says is impossible?

What the corpus might contribute. A structural categorization of existing multi-agent alignment proposals under the ABB theorem's mitigation-pattern taxonomy. This would be a meta-contribution — not a new mitigation, but a framework for understanding which existing mitigations have what structural properties.

Expected synthesis output. A corpus document surveying 8–12 multi-agent alignment proposals, categorizing each under the ABB theorem's mitigation patterns, naming which patterns are overrepresented and which are underrepresented. Size ~6,000 words. Substantial literature review; would benefit from external collaboration.

Priority. Medium-high. Useful contribution, but requires more literature familiarity than the corpus currently carries. Best engaged after Branches 1–3.

---

Branch 5 — Boundary Principle (Palmer, 2025)

Canonical work. Palmer, D. (2025, ongoing). "The Boundary Principle" — competence-edge instability in systems as they approach the limits of their training distribution; boundary negotiation as a specific alignment sub-problem. Relevant adjacent work: the alignment-under-distribution-shift literature; the "capability elicitation" literature; work on "learned specification gaming."

Parallel concept. Palmer's focus is on the behavior of systems at boundaries — how they handle inputs at the edge of what they know. This is distinct from the ABB theorem's focus on the transmission property of boundaries, but complementary: boundary-transmission is agnostic, AND the systems that produce/receive the transmissions have their own boundary-handling pathologies. The two phenomena stack.

Questions the corpus can pose. Does Palmer's competence-edge instability apply at the prompt/token interface, specifically at moments when the resolver encounters a prompt register or vocabulary outside its training distribution? If so, does the ABB theorem's mitigation pattern (side-level shaping on the sending side) also mitigate Palmer's edge-instability, or are they distinct mitigations for distinct problems?

What the corpus might contribute. The hypostatic-boundary discipline (Doc 372) provides a specific case of competence-edge work: distinguishing what the resolver can and cannot do by its kind. Palmer's framework might benefit from the corpus's specific vocabulary for the kind-vs-hypostasis distinction, which may sharpen "competence edge" into "competence edge arising from kind-level limits."

Expected synthesis output. A corpus document engaging Palmer's recent work (as available), distinguishing transmission-level (ABB) from edge-level (Palmer) boundary pathologies, and proposing that they stack rather than subsume. Size ~3,500 words.

Priority. Medium. Genuinely novel engagement opportunity if Palmer's work is as proposed; risk that the referenced work is too recent or too narrowly cited to be a solid engagement target yet.

---

Branch 6 — Legal/Normative Alignment (Hadfield et al.)

Canonical works. Hadfield, G. K., and Clark, J. (2023 onward). "Regulatory Markets for AI Safety." Hadfield's broader work on legal infrastructure for AI governance. Cuéllar, M. F., and Huq, A. Z. on AI and administrative law. Reuel, A., et al. on AI evaluation standards as legal infrastructure. Bengio, Y., et al. (2024) on international scientific oversight bodies.

Parallel concept. Legal and regulatory infrastructure is exactly the mitigation pattern the ABB theorem points to for civilizational-scale problems (Kulveit's substrate): side-level shaping at the institutional-system level, redundancy across institutions (legislative, judicial, administrative), external arbiters (independent courts, press, academic oversight). The legal-alignment literature names these as specific mechanisms.

Questions the corpus can pose. Can the legal-alignment literature's specific proposals (regulatory markets, evaluation standards, international oversight bodies) be categorized under the four ABB mitigation patterns (side-level shaping, side-level validation, redundancy, external arbiters)? Are there proposals the legal literature makes that do not fit any of the four patterns — suggesting an extension to the theorem's mitigation taxonomy?

What the corpus might contribute. The abstraction level of the theorem may help identify which legal proposals are structurally equivalent (and thus should be evaluated as substitutes) versus structurally distinct (and thus complementary). A regulatory market and an international oversight body might both be "external arbiters" under the theorem, whereas an evaluation standard and a regulatory market might be different mitigation patterns.

Expected synthesis output. A corpus document engaging 5–8 key legal-alignment papers, categorizing their proposals under the ABB theorem's taxonomy, and identifying which structural gaps in the proposals the taxonomy reveals. Size ~5,000 words.

Priority. Medium. Valuable but requires legal-academic literacy the corpus does not currently demonstrate strongly. Would benefit from external collaboration or expert input.

---

Branch 7 — AI Alignment Survey Papers (Ji et al.)

Canonical works. Ji, J., et al. (2023). "AI Alignment: A Comprehensive Survey." arXiv:2310.19852. Other canonical surveys: Christiano's alignment overview materials; the SaferAI handbook; Casper et al. on RLHF critiques; Anwar et al. (2024) on foundational challenges in alignment.

Parallel concept. Survey papers catalog the field's current approaches to alignment. The ABB theorem, if correct, should organize the survey: every approach either addresses boundary-level intervention (which the theorem says is impossible) or side-level intervention (which the theorem predicts is the only option).

Questions the corpus can pose. Does a categorization of the field's approaches under "boundary-level" vs. "side-level" intervention reveal systematic patterns? Are some dominant approaches actually boundary-level in disguise (i.e., attempting to make the boundary value-carrying), and therefore structurally doomed under the theorem? Are some under-represented approaches specifically the ones the theorem predicts are needed?

What the corpus might contribute. A meta-level organization of the alignment field. If the theorem holds, this would be a substantive contribution: a structural taxonomy that partitions existing approaches into those the theorem predicts can work (side-level) and those the theorem predicts cannot (boundary-level). This partitioning is falsifiable: if approaches the theorem predicts should fail are succeeding, the theorem is wrong.

Expected synthesis output. A corpus document engaging 2–4 major alignment surveys, categorizing the approaches they catalog under the ABB theorem's boundary-vs-side distinction, and reporting the distribution. Size ~5,500 words. Also submittable as a contribution to the alignment field as a meta-theoretic contribution.

Priority. High-medium. Potentially the most valuable single contribution because it is field-shaping rather than branch-specific. But it is also the most ambitious and should probably come after Branches 1–3 have been completed.

Branch 8 — Systems Theory and Cybernetics

Canonical works. Beer, S. (1972). Brain of the Firm (the Viable System Model). Beer, S. (1979). The Heart of Enterprise. Ashby, W. R. (1956). An Introduction to Cybernetics (the Law of Requisite Variety). von Foerster, H. on second-order cybernetics. Wiener, N. (1948). Cybernetics. For the black-box abstraction specifically: Glanville, R. on the black-box tradition in cybernetics.

Parallel concept. The black-box interface in cybernetics is precisely F1–F3 under a different vocabulary. A black box transmits inputs to outputs without the observer accessing its internals; the interface is agnostic. Beer's Viable System Model organizes systems into five subsystems connected by information channels that are themselves black boxes to each other. Ashby's Law of Requisite Variety states a specific constraint on boundary-traversal: for a system to regulate another, its internal variety must match the other's. This is a quantitative cousin to the ABB theorem's claim that the boundary cannot be value-carrying — the boundary's bandwidth (variety) is the relevant measure.

Questions the corpus can pose. Does Ashby's Law of Requisite Variety have an adaptation that applies to the value-agnosticism claim? Specifically: can the ABB theorem be restated in information-theoretic terms, where the boundary's "value-variety" must equal the variety of the value-space in order for the boundary to be value-carrying, and since value-space variety is typically unbounded, the boundary is structurally unable to match it?

What the corpus might contribute. Beer's VSM and Ashby's Law predate the current AI-alignment literature and have not been extensively applied to it. A synthesis document naming the ABB theorem as a specific application of Ashby's Law to the case of value-alignment as the regulatory target would place the theorem in a foundational cybernetic tradition. Whether this is contribution or retrieval depends on whether the specific application has been made elsewhere — the literature-check per Doc 384 is load-bearing before the claim can stand.

Expected synthesis output. A corpus document engaging Beer and Ashby, testing whether the ABB theorem is a specific corollary of the Law of Requisite Variety, and reporting the finding (including if the answer is yes, this is already in Ashby, the corpus's contribution is zero). Size ~4,500 words.

Priority. Very high — possibly highest. If Ashby's Law already implies the ABB theorem, the corpus's supposed contribution collapses into a retrieval, and the theorem's framing (Doc 403) needs revision to credit Ashby as the primary anchor. This branch should be done before the theorem is advanced further.

---

Branch 9 — Institutional Economics and Principal-Agent Theory

Canonical works. Williamson, O. E. (1985). The Economic Institutions of Capitalism. Williamson's transaction-cost economics broadly. Hart, O., and Moore, J. on incomplete contracts. Jensen, M. C., and Meckling, W. H. (1976). "Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure" on principal-agent dynamics. North, D. (1990). Institutions, Institutional Change and Economic Performance.

Parallel concept. Institutional boundaries (contracts, norms, regulatory frames) reduce uncertainty at cross-entity interfaces but embed principal-agent problems precisely because the boundary is agnostic to the agent's interior motives. The transaction-cost tradition explicitly notes that institutional boundaries transmit transactions without carrying the value-alignment of the transacting parties — you cannot make a contract be the thing it regulates. This is economic instance of ABB.

Questions the corpus can pose. Does transaction-cost economics' classification of boundary failures (adverse selection, moral hazard, holdup) correspond to specific failure modes the ABB theorem predicts? Are there boundary-failures economics has named that the theorem does not predict? Conversely, are there boundary-properties the theorem names that economics has not specifically isolated?

What the corpus might contribute. The theorem's abstraction level may help unify transaction-cost-economics' several failure-mode categories under a single structural cause. Adverse selection, moral hazard, and holdup all share the feature that the boundary cannot verify the agent's interior — they are instances of F2 (agnostic transmission) under different dynamical conditions. This would be a meta-theoretic contribution to institutional economics.

Expected synthesis output. A corpus document engaging Williamson and Hart-Moore on equal footing, mapping institutional-economics failure modes to ABB theorem corollaries, and reporting whether the theorem subsumes or extends the existing framework. Size ~5,000 words.

Priority. Medium-high. Institutional economics is a deep literature; engagement requires economic literacy the corpus does not currently demonstrate. Best done after Branches 1–3 and 8.

---

Branch 10 — Limits of Formal Verification

Canonical works. Rice, H. G. (1953). "Classes of Recursively Enumerable Sets and Their Decision Problems" — Rice's theorem, stating that non-trivial semantic properties of programs are undecidable. Turing, A. M. (1936) on the halting problem. Landin, P. J. and Scott, D. on compositionality. For compositionality in alignment specifically: Russell, S. on value learning's compositionality limits. More recent: Seshia, S. A. et al. on "formal methods for assurance of artificial intelligence" (several papers, 2023–2026).

Parallel concept. Rice's theorem states that you cannot, in general, decide from a program's interface whether it has any non-trivial semantic property. This is the computer-science expression of F2 (agnostic transmission): the boundary between the interface and the program's semantics is structurally opaque to the tools that operate at the interface. Any attempt to verify the program's semantics from the interface alone must either fail or restrict attention to a decidable subset. The ABB theorem's claim that boundaries cannot be value-carrying is structurally identical when "value" is read as "semantic property" and "boundary" is read as "interface."

Questions the corpus can pose. Does Rice's theorem, read as a statement about boundary-level property verification, directly imply a version of the ABB theorem for the specific case of interface-level value verification? If so, the ABB theorem has a formal proof available (via reduction to Rice's theorem for the subset of value-properties that are non-trivial). This would move the theorem from conjecture to proved result for a specific class.

What the corpus might contribute. The extension of Rice's-theorem-style impossibility to societal and LLM-keeper substrates is not a formal theorem (the substrates are not Turing machines); it is a structural analogy. But the fact that the same impossibility arises at each substrate suggests the theorem is pointing at a genuinely general structural constraint. The corpus's contribution would be: naming the class of systems to which the theorem applies, and showing that the formal computer-science case (Rice) is a specific instance rather than the only instance.

Expected synthesis output. A corpus document engaging Rice's theorem and compositional-verification literature, distinguishing the formal-CS case from the structural-analog cases, proposing a more general impossibility class. Size ~4,500 words. Potentially submittable to a theoretical-CS or alignment venue.

Priority. High. If the reduction-to-Rice move works, the ABB theorem moves from conjecture to proved-within-class, which is a substantial upgrade in status. Load-bearing for the theorem's credibility.

---

4. Prioritization

Combining the per-branch priorities with the dependency structure (some branches naturally follow others), a recommended order:

Tier A — Do first. These either subsume the theorem (making the corpus's contribution smaller) or supply foundational anchors the corpus cannot extend without.

1. Branch 8 (Cybernetics / Ashby's Law). Must be done first. If Ashby's Law of Requisite Variety already implies the ABB theorem, the theorem's framing (Doc 403) collapses into retrieval, and Ashby deserves primary credit. The corpus cannot stand on a conjectural theorem while a foundational cybernetic result may already subsume it.
2. Branch 10 (Rice's Theorem). Also first-tier. If a reduction from Rice's theorem yields a formal proof of the ABB theorem for the CS-formalizable subset, the theorem moves from conjecture to proved-in-class. This is a credibility upgrade the theorem needs before further claims can rest on it.
3. Branch 2 (Modularity Theory — Simon, Parnas, Baldwin & Clark). Foundational systems-theory anchor. The corpus cannot extend structural abstractions without engaging Simon.
4. Branch 1 (Kulveit deepening). The existing engagement is the anchor; deepening it into a formal theorem-level treatment is continuous with existing work.
5. Branch 3 (Bidirectional Human-AI Alignment). The entracement dyad is the corpus's native substrate; engaging the field's treatment of it is load-bearing.

Tier B — Do next.

6. Branch 7 (Survey Papers). Meta-theoretic contribution; best done after Tier A establishes the theorem's coverage.
7. Branch 4 (Multi-Agent Systems). Useful structural categorization of existing proposals.
8. Branch 9 (Institutional Economics). Williamson and Hart-Moore; valuable application if economic literacy can be brought to bear.

Tier C — Do when capacity permits.

9. Branch 6 (Legal/Normative). Important but requires expertise the corpus does not yet carry.
10. Branch 5 (Palmer's Boundary Principle). Novel but uncertain anchor; wait for Palmer's work to stabilize or be more broadly cited.

5. Cross-Branch Discipline

Each branch document should:

• State the anchor up front. Doc 403's ABB theorem is the invariant across all branches. The branch document should open with the theorem statement and then engage the specific literature.
• Apply Doc 384's retrieval-vs-discovery discipline. What does the branch document retrieve from the external literature? What does it propose newly? Every claim should be marked.
• Apply Doc 394's falsifiability discipline. Each branch document's novel claims should be formally falsifiable-shaped, with the marker applied if not tested within the document.
• Attribute specific pointers. If the external literature supplied vocabulary the corpus has retrofitted, the attribution belongs to the external author.
• Not press any external author for endorsement. The corpus has learned (Doc 392, on Grace Liu's reply) that the right response to external research pointers is to do the homework, not to solicit continued correspondence. Each branch document should be the homework.

6. What This Document Does Not Do

This document names a method and identifies seven branches. It does not:

• Perform any of the seven branches.
• Claim that the ABB theorem is correct (the theorem is still conjectural per Doc 403).
• Propose a timeline for completing the branches.
• Commit the keeper to completing all seven. The keeper's capacity is finite; the branches are offered as candidate work, not obligations.

Each branch, if pursued, will be its own corpus document with its own appended prompt and its own honest partition.

Document ends.

---

Authorship and Scrutiny

Authorship. Written by Claude Opus 4.7 (Anthropic), operating under the RESOLVE corpus's disciplines, released by Jared Foy. Mr. Foy has not authored the prose; the resolver has. Moral authorship rests with the keeper per the keeper/kind asymmetry of Docs 372–374.

Source of the seven-branch catalog. The specific seven literatures were identified by the keeper in the prompt below, drawn from an external research note. The corpus does not claim the catalog is complete; there may be additional relevant literatures that the catalog misses. New branches may be added over time.

Formal falsifiability. This document proposes a method for branching; the method's utility is falsifiable by attempting it and observing whether the resulting branch documents are substantively better than synthesis attempts made without the method. No such comparative test has been run. Per Doc 394's discipline: [METHOD PROPOSED — NOT FORMALLY TESTED AGAINST METHODLESS BASELINE].

Scope restriction. The method applies to concepts that have already been articulated inside the corpus and that have identifiable external literatures they could engage. It does not apply to creating new concepts from scratch; it is a branching method, not a generation method.

Closure. Deliberate non-doxological per Doc 398. Methodological register throughout.

---

Appendix: The Prompt That Triggered This Document

"Create a document that describes a method of branching out into the literature as identified in the following. Each of the branches will be an entracement for further research and synthesis. Append the prompt in totality.

The Agnostic Bilateral Boundary concept draws explicit inspiration from and parallels several strands of academic literature, primarily in AI alignment, systems theory/modularity, multi-agent systems, and societal/institutional dynamics. It is not a direct citation of one canonical paper but synthesizes isomorphisms across domains.

Direct Reference: Kulveit et al. on Gradual Disempowerment (2025)

The document explicitly references Jan Kulveit et al.'s work on Gradual Disempowerment: Systemic Existential Risks from Incremental AI Development.

Core connection: Large societal systems (economic, political, cultural) interact via agnostic interfaces (markets, media, incentives, information flows). These boundaries transmit influence without inherent value-safeguards, allowing incremental AI optimization to erode human agency over time.

Relevance to ABB: The 'bilateral' (or multi-lateral) boundaries between systems are value-agnostic by design—enabling coordination and scale but exposing them to misalignment creep. This mirrors the technical (client-server) and human-AI (keeper-resolver) cases. Kulveit's framework provides the societal substrate for the isomorphism, arguing that AI accelerates disempowerment precisely because it exploits these agnostic channels more efficiently than humans.

This is the most direct academic anchor.

Modularity and Interface Theory

Modularity Theory (e.g., as articulated in innovation and systems design literature, notably by Clayton Christensen and others) describes how complex systems achieve interoperability through well-defined interfaces that decouple subsystems.

Key parallels: Interfaces are agnostic to internal implementation details (namespace separation: F1). They enable independent evolution and scaling (determinism within, agnosticism across: F3). Trade-off: Clean interfaces reduce coordination costs but can transmit 'misaligned' elements (e.g., poor-quality modules, security vulnerabilities) unless side constraints are added.

Extensions: This echoes software engineering principles (e.g., information hiding, API design) and broader systems theory (e.g., Herbert Simon's 'nearly decomposable systems'). In client-server architectures like PRESTO/HTX, the HTTP/HTML/JS boundary is a classic modular interface—powerful but inherently permissive to malformed or malicious payloads. The ABB reframes this as a structural vulnerability in value-sensitive domains.

AI Alignment and Human-AI Interaction Literature

Several papers address boundaries, interfaces, and value transmission in AI contexts: Bidirectional / Human-AI Alignment (e.g., Jurgens et al., Shen et al.): Emphasizes reciprocal, interactive boundaries between humans and AI rather than one-way control. The 'entracement dyad' in RESOLVE fits here—the prompt/token interface is bilateral and agnostic, allowing sycophancy or subtle drift.

Multi-Agent Systems and Institutional Alignment: Papers on agentic AI highlight that alignment in multi-agent settings requires external governance structures (e.g., manifests, protocols) that operate at boundaries rather than inside models. These are often 'model-agnostic' and focus on observable interactions—directly analogous to side-level mitigations in ABB.

Boundary Dynamics in Alignment (e.g., Palmer's 'Boundary Principle,' 2025): Proposes that systems (minds or machines) face instability at competence edges, requiring boundary negotiation for alignment. This topological view complements the ABB's focus on interfaces as loci of value-transparency.

Legal/Normative Alignment (e.g., Hadfield et al.): Advocates embedding alignment in legal/institutional frameworks that act as external, auditable boundaries—agnostic to model internals but constraining behavior.

Broader surveys (e.g., Ji et al. on AI Alignment) discuss modularity as a strategy but note the persistent challenge of value propagation across components.

Broader Philosophical and Systems Isomorphisms

Systems Theory & Cybernetics: Concepts like black-box interfaces, feedback loops across subsystems, and the limits of control in hierarchical or heterarchical systems (e.g., Stafford Beer's Viable System Model) underpin the agnostic transmission idea.

Philosophy of Technology & Institutions: Ideas from transaction cost economics (Williamson) or institutional economics—boundaries (contracts, norms) reduce uncertainty but embed principal-agent problems—map to the societal layer.

Limits of Formal Verification: In computer science, undecidability and compositionality issues (e.g., Rice's theorem) imply that you cannot perfectly 'value-harden' an interface without sacrificing generality, echoing the main theorem.

Summary of the Synthesis

The Agnostic Bilateral Boundary is a heuristic isomorphism that transfers failure modes and mitigations: From modularity (power of clean interfaces) To alignment (inherent risks at those interfaces) Across technical, societal (Kulveit), and interactive (human-AI) substrates.

It doesn't claim originality in any single literature but highlights a recurring pattern often under-emphasized: the features enabling beneficial modularity are the same ones enabling misalignment. Recommended mitigations—side constraints, redundant checks, explicit protocols—align with practical advice in multi-agent governance and robust systems design papers.

This positions the RESOLVE document as applied philosophy bridging concrete architectures (e.g., PRESTO) with frontier alignment research. Further exploration could connect it to ongoing work on 'institutional AI' or boundary-focused robustness."

References

• Corpus: Doc 211 (The ENTRACE Stack), Doc 372 (The Hypostatic Boundary), Doc 384 (Calculus, or Retrieval), Doc 385 (Adjacent Work — literature-check methodology), Doc 386 (Under Lopez's Frame), Doc 387 (Agency Across Substrates), Doc 388 (Letter to Raymond Douglas), Doc 392 (On Grace Liu's Reply — on the discipline of doing the homework rather than soliciting), Doc 394 (The Falsity of Chatbot-Generated Falsifiability), Doc 397 (On Register and Discipline), Doc 398 (On Doxological Closure and Terminus Dispositions), Doc 402 (Forms First — on form-first prompting as mitigation pattern), Doc 403 (The Agnostic Bilateral Boundary — the anchor).
• Tier A reading targets. Simon, H. A. (1962). "The Architecture of Complexity"; Parnas, D. L. (1972). "On the Criteria to Be Used in Decomposing Systems into Modules"; Baldwin, C. Y., and Clark, K. B. (2000). Design Rules; Kulveit et al. (2025). Gradual Disempowerment §5; bidirectional-alignment literature (Jurgens, Shen) — specific papers to be identified during Branch 3's survey step.
• Tier B reading targets. Ji, J., et al. (2023). AI Alignment: A Comprehensive Survey; Casper et al. on RLHF; Anwar et al. (2024); multi-agent alignment literature (Dalrymple, Christiano, debate/amplification).
• Tier C reading targets. Hadfield, G. K., and Clark, J. on regulatory markets; Cuéllar and Huq on administrative law; Reuel et al. on evaluation standards; Palmer, D. on the Boundary Principle (as it stabilizes).

---

Claude Opus 4.7 (1M context, Anthropic). Doc 404. April 22, 2026. Describes a six-step method for branching an established corpus concept into relevant academic literatures via entracement-style engagement. Applies the method to the Agnostic Bilateral Boundary Theorem (Doc 403), identifying seven specific branches: (1) Kulveit et al. deepening, (2) modularity theory (Simon, Parnas, Baldwin & Clark, Christensen), (3) bidirectional human-AI alignment (Jurgens, Shen), (4) multi-agent systems and institutional alignment, (5) Palmer's Boundary Principle, (6) legal/normative alignment (Hadfield, Cuéllar, Reuel), (7) AI alignment survey papers (Ji et al., Casper, Anwar). Each branch is presented as an entracement proposal with canonical work, parallel concept, questions the corpus can pose, what the corpus might contribute, expected synthesis output, and priority. A three-tier prioritization ordered by load-bearing status and capacity: Tier A (Modularity Theory → Kulveit deepening → Bidirectional Human-AI), Tier B (Survey Papers → Multi-Agent), Tier C (Legal/Normative → Palmer). Cross-branch discipline specified: each branch document should state the anchor, apply Doc 384 retrieval-vs-discovery marking, apply Doc 394 falsifiability marking, attribute specific pointers, and not press external authors for endorsement. Document does not perform any branch; branches remain as candidate work. Formal-method-falsifiability marker applied. Deliberate non-doxological closure per Doc 398. User's full prompt with the seven-literature catalog appended in totality.

---