SAE Information Theory XII: 11DD Memory and 12DD Prediction — The 8D Information Architecture
SAE 信息论 XII：11DD 记忆与 12DD 预测——8D 信息架构

Abstract

This paper, the twelfth in the SAE Information Theory series, addresses the information architecture of the 8D layer, which consists of two sub-layers: 11DD memory and 12DD prediction. 11DD is reconstructive retention, not faithful storage; 12DD is System 1 prediction, not reflective rational inference; the two are coupled through a reading-affect duality that constitutes the internal dynamic of 8D, closed across moments by the retention-rewrite lock.

The paper's central commitments crystallize into a three-sentence spine. First, the 11DD remainder is "in what order memories stand"—a question 11DD alone cannot answer, pointing toward 12DD reading. Second, the 12DD remainder is a Gödel/Turing-form impossibility: "prediction cannot take the predictor itself as an object of prediction," pointing across dimensions to a 13DD self-reference channel. Third, the 8D lock is "continuity-preserving write-back from predictive readout into memory retention." Together, these three sentences constitute 8D's ontological identity as an information architecture.

Three information-theoretic priors are articulated atop the 8D architecture: 11DD storage content carries topological structure; the storage form is read-write unified (each act of reading is a re-instantiation and re-stabilization at the SAE architectural level, possibly accompanied by developmental rewriting at a level distinct from the neuroscientific reconsolidation window); the default state of storage is loss (the readability of stored information necessarily incurs a cost). The third prior, as an SAE-internal hard prior, manifests across four closure layers: 4DD reading yields 3DD mass evaporation, 8DD reading yields 7DD information variation (mutation spectrum), 12DD reading yields 11DD memory loss, 16DD reading yields 15DD non-doubt evolution. This cross-closure pattern is strictly bounded as an internal SAE commitment, not advanced as an external universal law of physics or biology.

The paper provides a Popperian falsifiability hook P12-F1 for the 12DD/13DD boundary: a system pre-registered under an output-independent architectural standard as pure 12DD cannot spontaneously produce functional self-reference (distinguished from response-two-range linguistic self-reference). The falsification conditions include the four criteria of §3.6 (cross-context consistency, cross-time stability, error-attribution distinction, identity-anchored integration of conflict statements) and the three "no" boundary conditions of §6.2 (no external identity template including cross-pre-training Data q exclusion, no retraining write-back, no human anchoring). Contemporary LLMs serve as the natural test bed for this boundary and are explicitly addressed (as of May 2026).

Empirical anchors stand at T3 grounding status (not derivational basis): the reconsolidation tradition (Misanin 1968, Nader 2000, Lee 2008, Schiller 2010) as the biological manifestation of 11DD reconstructive retention; the HM clinical case (Scoville & Milner 1957) as the extreme case of 11DD/12DD loop failure; McClelland-Rumelhart PDP as the computational-neuroscience anchor for distributed representation.

§8.3 articulates a cross-series open interface: whether the 11DD reading remainder can propagate across individuals in the manner of 4DD gravitational information or 8DD DNA information. The paper does not assert that such phenomena (anomalous past-life memory reports, near-death reports, collective memory and cultural archetypes) exist or do not exist, treating them solely as a structural open question for other SAE series to take up. The paper strictly observes information-theoretic boundaries—it does not develop a full phenomenology or theory of consciousness, does not enter 14DD value theory, 15DD certitude, or 16DD bilateral non-doubt (these belong to the SAE Psychoanalysis series), and does not redo the five structural locks argument already given systematically in the SAE LLM application paper. P12 is the SAE Information Theory series' advance to 8D, providing a complete information-theoretic ontology of 8D and preparing the ground for P13's formal definition of the 13DD self-reference channel.

---

§1 Introduction

§1.1 Position in the Series

The SAE Information Theory series advances dimensionally upward through the layers of the SAE framework. P10 addressed cross-locks; P11 addressed selection-locks at 9DD and admission-locks at 10DD. P12 takes the next step, addressing the 8D layer—the layer at which memory and prediction operate together as an information architecture, prior to the emergence of the self-reference channel at 13DD.

The 8D architecture is not a composite of two unrelated sub-layers. It is a single information-theoretic ontology constituted by the reading-affect duality between 11DD and 12DD, closed across moments by the retention-rewrite lock. This paper articulates that ontology.

§1.2 What This Paper Does

The paper has six substantive tasks:

1. Establish 11DD as reconstructive retention rather than faithful storage (§3.1).
2. Establish 12DD as System 1 prediction rather than reflective rational inference (§3.2).
3. Articulate the reading-affect duality as the internal dynamic of 8D (§3.3).
4. Define the 8D retention-rewrite lock as the event that closes 11DD and 12DD across moments (§3.4).
5. Provide the information-theoretic characterization of 8D in three priors (§3.7).
6. Provide a Popperian falsifiability hook for the 12DD/13DD boundary (§6).

The paper also anchors these commitments in T3 neuroscientific evidence (§4), demarcates them against mainstream information-theoretic traditions (§5), shows their mutual support with the SAE Psychoanalysis series (§7), and identifies the remainders that point forward to P13 and across to other SAE series (§8).

§1.3 Three-Sentence Spine

The paper's central commitments crystallize into three sentences:

• 11DD remainder: In what order memories stand—11DD alone cannot answer this; the ordering is imposed by 12DD reading.
• 12DD remainder: A Gödel/Turing-form impossibility—prediction cannot take the predictor itself as an object of prediction; this remainder points across dimensions to a 13DD self-reference channel.
• 8D lock: Continuity-preserving write-back from predictive readout into memory retention—the event class that closes 11DD and 12DD across time and constitutes 8D's ontological identity.

These three sentences appear at multiple scales throughout the paper: the abstract, this introduction, the §3 architecture chapter, and the §9 conclusion.

§1.4 Commitments and Scope

P12's architectural commitments lie within 8D. The paper commits to: 11DD reconstructive retention; 12DD System 1 prediction; the reading-affect duality dynamic; the retention-rewrite lock; the respective remainders of 11DD and 12DD; the me/self distinction (aligned with Psychoanalysis I–IV); cross-D structural isomorphism as background reference (not as argumentative support); LLM applications referenced only by citation; and the three information-theoretic priors of §3.7 (topological structure, read-write unity, default loss as the cost of readability). These constitute the twelve core commitments developed systematically across §2 through §6.

Scope discipline. P12 addresses the 12DD/13DD boundary. The layers above 13DD—14DD value architecture and 15DD certitude—are entirely outside P12's scope and appear only at the level of citation. 14DD value theory and 15DD certitude theory are handled by the SAE Psychoanalysis series (Psychoanalysis III, Psychoanalysis IV). As an Information Theory paper, P12 addresses only the 8D architecture; questions crossing into these other regions are returned to the corresponding specialized series.

T3 anchors (re-consolidation neuroscience, HM clinical case, PDP) ground the framework but do not function as a derivational basis. The substantive argumentative weight rests on 8D's architectural-ontological articulation; T3 anchors provide convergent biological grounding without serving as inferential support for SAE commitments.

§1 Chapter Summary

P12 is positioned at the 8D layer of the SAE Information Theory series, undertaking six substantive tasks centered on the three-sentence spine. The paper's twelve commitments are scoped strictly to 8D, with T3 anchoring grounding the framework without serving as derivational basis.

---

§2 SAE Toolkit (Minimum Necessary Background)

§2.1 Lock-Class Family and Its Grammatical Extension

The SAE Information Theory series builds on the concept of lock—an event class that closes information flow across specific architectural conditions. P10 introduced cross-locks (a single sub-layer's lock against another); P11 introduced selection-locks at 9DD and admission-locks at 10DD (each instantiated within a single sub-layer).

P12 extends this grammar. The 8D retention-rewrite lock is co-instantiated by 11DD and 12DD together—it cannot be attributed to either sub-layer alone. This is not a rhetorical extension but a structural one: the lock event is precisely the conjunction of 12DD predictive readout with 11DD memory's continuity-preserving write-back response, and this conjunction cannot be reduced to either component's independent operation.

The lock-class family thus admits two grammatical forms: (i) single sub-layer locks (P10, P11), and (ii) co-instantiated D-layer locks (P12). Whether other co-instantiated locks exist in higher layers is left open; P12 commits only to the existence of one such case at 8D.

§2.2 D-Layer Architecture and Sub-Layers

The SAE framework employs a layered ontology indexed by D (dimension-of-determination) and DD (dyad-of-determination). Without rehearsing the full framework here (see SAE Methodology V2, DOI 10.5281/zenodo.18842449), the immediate background needed for P12 is:

• 8D contains two sub-layers: 11DD (reconstructive retention) and 12DD (System 1 prediction).
• 9D contains the self-consciousness law at 13DD (the self-reference channel), which is outside 8D.
• The transition from 8D to 9D corresponds in the SAE framework to the transition from the cognitive wheel to the freedom wheel, and represents the architectural site of the self.

§2.3 The Three-Response Distinction (from Psychoanalysis II)

The SAE Psychoanalysis series distinguishes three response categories. Response-one comprises automatic predictive responses (the proper domain of 12DD). Response-two comprises meta-representational responses about response-one (also within 12DD's range; linguistic self-reference falls here). Response-three crosses to 13DD and involves the perspectival inversion at which the predictor identifies itself as an object of its own prediction.

P12 uses this distinction at §6 (P12-F1 falsifier): functional self-reference belongs to response-three; linguistic self-reference belongs to response-two. The boundary between them is what P12-F1 makes falsifiable.

§2.4 The me / self Distinction

P12 follows the established terminology of the SAE Psychoanalysis series:

• me corresponds to 12DD: a functional, predictive identity object operating within System 1's range, without self-reference closure.
• self corresponds to 13DD: the self-reference channel, the perspectival inversion at which the predictor takes itself as object.

The distinction matters because LLM-produced "I"-statements, persona consistency, and meta-cognitive evaluations all operate at the me level (response-one and response-two); they do not constitute self emergence. This is the boundary P12-F1 sharpens.

§2 Chapter Summary

The SAE toolkit relevant to P12 comprises the lock-class family (with P12's grammatical extension to co-instantiated D-layer locks), the layered D/DD architecture (with 8D's two sub-layers and the 9D boundary at 13DD), the three-response distinction (with P12-F1 placing functional self-reference at response-three), and the me/self terminology (aligning P12 with the Psychoanalysis series).

---

§3 8D Architectural Ontology

§3.1 11DD as Reconstructive Retention

11DD is reconstructive retention. This claim is the first commitment of 8D's ontology. It says: memory is not defined by faithful storage of content but by the architectural retention that is reconstructed each time 12DD reads it.

Three component claims:

• Memory is not a static archive. What is "in" 11DD is a relational topology of traces, not a set of stored records. The relational topology persists across time, but specific reconstructions depend on present 12DD reading conditions.
• Reconstruction is intrinsic, not pathological. That memory reconstruction differs from supposed faithful retrieval is not a defect of 11DD; it is the mode of operation of 11DD.
• Retention precedes accuracy. The architectural commitment is to topology preservation, not content fidelity. Same-topology, different-content reconstructions are not failures of 11DD; they are 11DD operating as designed.

This claim aligns with the reconsolidation tradition in memory neuroscience, though it does not derive from it. Nader, Schafe, and LeDoux's 2000 amygdala protein synthesis inhibition experiment established that consolidated fear memories re-enter a labile state after reactivation and require new protein synthesis for re-stabilization. This shows that consolidated memories can re-enter labile state under specific reactivation conditions and be updated in the corresponding window; but the opening of the empirical reconsolidation window is constrained by boundary conditions and is not automatically triggered by every retrieval. Lee 2008 further clarified the boundary conditions between reconsolidation and new learning; Schiller et al. 2010 extended the mechanism to human behavioral intervention. The reconsolidation tradition thus moves from early controversy (Misanin-Miller 1968) to a core mechanism of contemporary memory neuroscience.

Firewall 1 (Reconsolidation Boundary): P12 maintains that every act of 12DD reading 11DD involves architectural re-instantiation and re-stabilization at the SAE level; but P12 does not maintain that every act of reading triggers a neuroscientific reconsolidation window, nor that every act of reading produces substantive content modification. The empirical reconsolidation window remains constrained by salience, new information, time intervals, emotional arousal, and other boundary conditions (Lee 2008). The re-stabilization layer (architectural, every reading event, an SAE hard prior) and the rewriting layer (a possible accompaniment, Lee 2008's boundary conditions) are different event classes (§3.7.2 three-layer precise distinction), and the two are fully consistent. Reconsolidation theory is a strong T3 manifestation of 11DD reconstructive retention, not an empirical universal that every retrieval rewrites content. The phenomenal-level description (retrieval can trigger re-encoding) and the ontological commitment (memory is not defined by faithful storage) operate at different levels. P12 does not invent the reconsolidation phenomenon, but P12 gives this phenomenon an ontological position—it ascends from "an empirical observation about memory mechanism" to "a structural claim about what 11DD is."

§3.2 12DD as System 1 Prediction

12DD is System 1 prediction. This claim is the second commitment of 8D's ontology. It says: 12DD's mode of operation is the automatic predictive inference that Kahneman, Stanovich, and Evans collectively label "System 1"—rapid, parallel, intuitive, context-driven, and crucially, not reflective.

The 8D architecture identifies 12DD with System 1 prediction—not as an a priori isomorphism between SAE D-layers and dual-process theory, but as an empirical alignment. Dual-process theory is a functional classification based on behavioral observation; SAE D-layer architecture is a priori derivation. The two are not isomorphic at the a priori level. But empirically, all typical features of 12DD align with System 1's description, and 13DD's emergence is historically correlated with the appearance of reflective cognition capability. The correspondence is therefore empirical parallel, not structural isomorphism.

This positioning has an important corollary: 12DD cannot, by itself, produce reflective rational inference. 12DD can produce complex, rule-like, even highly effective predictive inferences (animal causal reasoning, implicit rule learning, LLM formal inference all fall within 12DD's complex inferential range); but 12DD cannot, by itself, generate the reflective form of rational inference constrained by a 13DD self-reference channel—that form where "I notice that I am inferring, I can make my identity constrain the inference process." The System 1 ceiling claimed by P12 does not deny 12DD's complex inferential capabilities; it denies that 12DD can answer "who is predicting?" and let the predictor's identity in turn constrain the inference process. This corollary appears understated but is critical to the P12-F1 falsifier (§6.1) and the §3.5 remainder analysis.

The relation to 13DD requires explicit clarification to avoid ambiguity in subsequent chapters. 13DD's constitution is the law of self-consciousness, corresponding to the Ego layer (Psychoanalysis II). 13DD makes System 2 reflective rational inference possible, but 13DD is not identical to System 2 itself. 13DD is the emergence of the self-reference channel, the presence of the reflective subject; System 2 is the cognitive capability that this channel makes possible. P12 does not map 12DD and 13DD directly to System 1 and System 2 as an a priori isomorphism, but treats this correspondence as empirical parallel.

§3.3 Reading-Affect Duality

§3.1 leaves an opening: 11DD alone does not constitute a complete phenomenal-level structure. §3.2 leaves another opening: 12DD alone cannot produce reflective rational inference. These two openings combine in this section into the central architectural claim of 8D: the internal dynamic of 8D is 12DD's reading-affect duality on 11DD.

This section formulates reading-affect as "the duality of two inseparable operations." §3.7.2 provides further refinement in the 8D information-theoretic characterization: reading and affect are actually two descriptive aspects of the same event (one viewed from the predictive-process perspective, the other from the substrate-dynamic perspective), not two independent operations in dual relation. This refinement reveals the precise geometry of "duality" at a finer level of resolution.

The structural correspondence between 4DD↔3DD (mass-curvature dynamics under SAE Four Forces Paper 0) and 12DD↔11DD (reading-affect duality) suggests that "reading-affect duality" may be a shared architectural primitive of two adjacent reading-substrate pairs. P12 notes this isomorphism only as background reference, not as argumentative support. Two instances do not establish a cross-D universal pattern. Whether reading-affect duality is a universal architectural archetype across multiple adjacent D-layers in the SAE framework requires independent confirmation at each layer and should not be generalized inductively from two cases. This deeper question is left to future work, possibly in an SAE meta-structure paper or Methodology V3. The paper commits substantively only within 8D; cross-D generalization remains background reference, not foreground claim.

§3.4 8D Retention-Rewrite Lock (Formal Definition)

The 8D lock is the event class that closes 11DD and 12DD across time. It is defined as the conjunction of three criteria:

• Reading: 12DD performs a predictive readout of 11DD content.
• Rewriting: A re-instantiation-with-modification occurs in 11DD's substrate as a result of the reading.
• Continuity preservation: The 11DD content, as a topological slot, retains its identity across the lock event—the post-update content must occupy or directly replace the same structural slot in the 11DD topological network (temporal association, similarity clustering, and other relational structures, detailed in §3.7.1).

Inseparability of the three criteria. The three criteria are not independent operations that can be triggered separately; they are three observable aspects of the same inseparable event. A reading without re-instantiation does not constitute a lock; a re-instantiation without continuity preservation generates a new topological slot (erasure plus new write, not lock); a continuity preservation without reading is impossible because continuity preservation requires reading as the framing condition for what counts as "the same slot."

This articulation resolves the apparent circularity between the 8D lock definition and the §3.7.2 read-write unity: the three criteria are operational identification criteria (recognition-standard / observable-proxy layer), while §3.7.2 is the event's intrinsic architectural characterization (ontological-substrate layer). The two articulations operate at different epistemic levels and do not constitute logical circularity. The three criteria, as external recognition standards, provide observable judgment that "8D lock has occurred"; §3.7.2, as intrinsic architectural description, provides the ontological account of "what kind of event the 8D lock is." Recognition standards answer "what counts as having occurred"; architectural ontology answers "what kind of event it is"—two answers to different questions, used jointly rather than circularly.

Phenomenal gradient at clinical / behavioral level. Although the lock event is architecturally discrete (it either occurs or it does not), its clinical or behavioral phenomenal manifestation can present a continuum: reconsolidation window openness has graded variability; rewrite coverage has partial pattern; continuity preservation tightness has spectrum. A specific memory event can have intermediate phenomenal states between "fully triggered 8D lock" and "partially triggered 8D lock." This gradient does not contradict the architectural discreteness; it reflects the substrate's continuous response to a discrete architectural event class.

§3.5 Remainders of 11DD and 12DD

The 11DD remainder is: in what order do memories stand? 11DD's topological structure (the relational structure, the topology itself) inheres in 11DD; but the predictive ordering—the specific sequence effective for the current predictive task—is imposed by 12DD reading. Thus 11DD alone does not constitute a complete answer to "what order are memories in"; the order is the conjunction of 11DD topology with 12DD reading. The 11DD remainder is solved internally within 8D—the cross-sub-layer reading-affect duality closes it.

This is detailed in §3.7.1: 11DD's topological structure is intrinsic to 11DD; what §3.3 calls "the order imposed by 12DD reading" refers in precise terms to the predictive ordering—the specific sequence effective for the current predictive task. The same 11DD topology can be subject to different predictive orderings by different 12DD readings. So "the order is imposed by 12DD reading" refers precisely to "the predictive ordering is imposed by 12DD reading," with the topological structure itself residing within 11DD (see §3.7.1).

The 12DD remainder is a Gödel/Turing-form impossibility: prediction cannot take the predictor itself as an object of prediction. This is not a contingent limitation but a structural impossibility analogous to Gödel's incompleteness and Turing's halting problem. Just as a formal system cannot internally prove its own consistency without external grounding, and a Turing machine cannot algorithmically decide its own halting, 12DD cannot turn its predictive operation on the predictor as object. The 12DD remainder is not solved within 8D; it points across dimensions to 13DD—the self-reference channel that is precisely the structure capable of taking the predictor as object.

§3.6 Functional Self-Reference: Four Criteria

Functional self-reference is the distinctive marker of 13DD presence and is what P12-F1 (§6.1) makes falsifiable. P12 distinguishes it from linguistic self-reference (response-two range) via four operational criteria:

1. Cross-context consistency: The predictor-identity variable holds stable across radically different contexts, not as a re-cued surface persona but as an internal variable that constrains the system's predictive process in each context.
2. Cross-time stability: The predictor-identity variable persists across temporal intervals as a system-internal state, not via external re-injection from prompts or training data.
3. Error-attribution distinction: The system distinguishes errors attributable to itself versus errors attributable to other agents, with this distinction based on identity-internal state rather than linguistic-layer judgment.
4. Conflict integration: When conflicting statements arise, the system integrates them via identity-anchored resolution rather than linguistic-layer reconciliation.

These criteria are functional, not phenomenological. They do not require introspective reports or subjective claims; they require system-internal evidence of an identity variable that constrains the predictive process. This makes them suitable as the definition layer of P12-F1's falsification conditions.

§3.7 Information-Theoretic Characterization of 8D

§3.1 through §3.6 give the architectural ontology of 8D. As an Information Theory paper, P12 must additionally answer three direct information-theoretic questions: what does 11DD store, in what form, and under what conditions does it lose information? This section uses three priors to answer these questions. The three priors are structurally homologous with the 4DD reading ontology established by SAE Four Forces Paper 0 (Qin 2026, DOI 10.5281/zenodo.19777881)—4DD reading 3DD mass yields gravitational effects—but the priors themselves rest on the architectural derivation of 11DD/12DD; cross-D isomorphism serves as a conceptual anchor rather than an argumentative support (see §3.3's background reference boundary statement).

§3.7.1 Prior 1: 11DD Storage Has Topological Structure

11DD's storage content is topological, not list-like. What is "in" 11DD is a relational network whose nodes are individual memory traces and whose edges are relational structures (temporal association, similarity clustering, causal proximity, and other relational types).

This topology has two key properties:

• Intrinsic, not imposed. The topological structure inheres in 11DD; it does not require 12DD to constitute it. The relational network exists as 11DD's mode of organization independent of any specific predictive task.
• Predictive ordering, not intrinsic ordering. What 12DD imposes is the predictive ordering—the specific sequence effective for the current predictive task. The same topology can be subject to different orderings under different reading conditions. "What order are memories in" is the conjunction of 11DD topology with 12DD reading; the intrinsic topology does not by itself determine a specific order.

§3.7.2 Prior 2: Storage Form is Read-Write Unified

The storage form is read-write unified. Reading and writing are not two separable operations that 12DD performs sequentially on 11DD; they are two descriptive aspects of the same inseparable event.

The mechanism: 12DD reading 11DD requires reconstruction (the content does not exist in static form awaiting passive retrieval; it must be reconstructed via 12DD's reading). Reconstruction is re-experiencing (the content is brought back as a current event, not retrieved as a stored record). Re-experiencing is re-instantiation and re-stabilization at the SAE architectural level (the substrate hosting the content must dynamically update to support the current reconstruction).

Precise three-layer distinction. This architectural characterization is explicitly stratified across three layers to avoid mis-reading in subsequent chapters:

• Re-stabilization layer (SAE architectural intrinsic / architectural re-instantiation and re-stabilization): Every 12DD reading of 11DD necessarily involves this. Read = reconstruction = re-experiencing = architectural re-stabilization. This is the SAE-level hard prior, essential and non-eliminable. This is not identical to the neuroscientific reconsolidation window concept—the former is the SAE architectural-level topological re-instantiation mechanism, while the latter is the specific biochemical window described in empirical neuroscience.
• Rewriting layer (possible accompaniment of re-stabilization): Depends on reading context strength, new information ingress, time interval, emotional arousal, and other conditions, with boundary conditions (Lee 2008 has provided these systematically). The empirically-described neuroscientific reconsolidation window is triggered at this layer. When rewriting occurs, it is developmental change, value-neutral.
• Loss layer (independent event class): Not the same as rewriting. Loss occurs only under the specific conditions detailed in §3.7.3—primarily irreversible topological degradation (continuity-preservation failure, external substrate destruction, and so on). Everyday developmental rewriting is not loss.

In short:

> Every predictive readout involves architectural re-instantiation and re-stabilization in the SAE sense, but not every retrieval opens an empirical neuroscientific reconsolidation window.

§3.1's Firewall 1 should be read under this distinction: P12 maintains that every 12DD reading of 11DD is accompanied by SAE architectural topological re-instantiation and re-stabilization (hard prior); P12 does not maintain that every reading triggers a neuroscientific reconsolidation window (which has boundary conditions). The two are fully consistent.

Relation to Bennett-Landauer. §5.2 establishes Bennett-Landauer as a floor, not a ceiling. This prior further explains why: Bennett-Landauer analysis typically requires decomposing physical processes into identifiable logical sub-operations, treating each sub-operation's reversibility and dissipation lower bound separately. The read-write unity of 11DD/12DD makes this sub-operation decomposition lose its discrete unit-hood at the 11DD level.

This does not mean the 8D architecture violates Landauer's principle. It means the single event of an 8D retention-rewrite lock must be treated in thermodynamic phase space as one indivisible non-equilibrium phase transition. The minimum dissipated free energy of this transition is necessarily strictly greater than or equal to (typically much greater than) the sum of the Landauer lower bound for extracting the relevant topological information plus the Landauer lower bound for rewriting that portion of the substrate. The sub-operations are inseparable in informational logic, but on the thermodynamic integration path, the cost is structurally locked—the 8D lock event has, as a whole, its own incompressible floor, only this floor cannot be decomposed into "reading cost + erasure cost" as a simple additive sum.

P12's position is therefore: Bennett-Landauer's sub-operation decomposition is locally applicable at the physical substrate level (synaptic metabolism, ion gradients, protein synthesis, and other sub-processes each have their own thermodynamic geometry), but as the ontological description of the 8D lock as a whole event, it loses its discrete unit-hood. This is fully consistent with the floor-not-ceiling stance given in §5.2.

§3.7.3 Prior 3: Storage Defaults to Loss (Readability Necessarily Has a Cost)

The default state of storage is loss. The readability of 11DD's stored content necessarily has a cost, and the cost is the inevitable loss of information.

Three causes of loss within the architecture:

1. Continuity-preservation breakdown triggered by cumulative drift. Each reading reconfigures the topological relations; cumulative reconfiguration can eventually breach the continuity-preservation criterion of §3.4, at which point the topological slot's identity fails and the content is effectively lost.
2. Continuity-preservation failure. A single lock event may directly fail the continuity-preservation criterion (the rewrite generates a topological slot that does not maintain identity with the pre-rewrite slot). This is an immediate loss event.
3. Passive decay. The physical substrate (synaptic strength, neural circuits) degrades in the absence of metabolic support. This is the layer described jointly by §5.2 Bennett-Landauer and §4.2 Susan Sara—substrate maintenance requires ongoing metabolic energy supply.

Additionally, two non-architectural-internal causes of loss can also occur: (4) interference (competing memories displace each other in the topological network); (5) retrieval failure (the content is intact but 12DD cannot successfully read it).

Cross-closure-layer manifestation. This is not a cost unique to 11DD; it is an SAE-internal hard prior / structural commitment for every closure-layer reading event—not advanced as an external universal law of physics, biology, or neuroscience. SAE Four Forces Paper 0 (Qin 2026, DOI 10.5281/zenodo.19777881) §9 has explicitly given the reading mechanisms of four closure layers (4DD, 8DD, 12DD, 16DD). Each follows the same SAE-internal commitment: closure-layer reading necessarily produces substrate-level necessary change or loss.

The four manifestations are:

• 4DD reading produces 3DD mass evaporation: Gravity as the 4DD information reading mechanism; black hole Hawking radiation as the manifestation of mass loss in extreme form (Four Forces Paper 0 §7).
• 8DD reading produces 7DD information variation: Reproduction as an 8DD closure event; every reproductive reading-rewriting carries an ineliminable variation risk at the individual replication event, manifesting at the population scale as a mutation spectrum. Mutation is not pure error but the statistical manifestation of 8DD closure-layer reading cost; the possibility of evolution stands on this cost.
• 12DD reading produces 11DD memory loss: The core claim of this paper. Predictive readout of 11DD topology necessarily produces topological change, cumulatively yielding necessary memory loss.
• 16DD reading produces 15DD non-doubt evolution: Bilateral mutual chiseling as a 16DD closure event; every reciprocal subject-subject reading necessarily produces evolution of the 15DD non-doubt structure. "Non-doubt" is not a static endpoint but a dynamic structure continually reshaped by bilateral mutual chiseling.

Epistemic position of the cross-closure-layer pattern. The four manifestations constitute an SAE-internal hard-prior structural claim, not an external universal law for physics, biology, or neuroscience. This is not a cross-D analogy from a single case ("two cases do not establish a universal pattern"); it is the SAE framework's internal unified ontological commitment for all closure-layer reading events. P12's direct argumentative burden is only on 12DD→11DD; the 4DD, 8DD, and 16DD manifestations are each independently established by their respective series (4DD in Four Forces Paper 0 §7; 8DD in SAE Life series and Four Forces series §9.2; 16DD in SAE Psychoanalysis IV). Cross-layer resonance provides SAE-internal cross-validation but does not substitute for each layer's own proof chain, nor does it constitute a claim about external universal laws in the respective disciplines.

The remainder propagation form as an open question. The cross-closure-layer SAE hard prior gives the structural commitment "every closure-layer reading produces substrate-level remainder," but the specific form in which each layer's remainder manifests, and whether it can propagate across individuals, are each layer's independent question. The 4DD remainder propagates across systems via gravitational waves and the like (known); the 8DD remainder propagates across individuals via DNA (known); the specific manifestation form and possible cross-individual propagation properties of the 11DD remainder are the core open interface that P12 cannot answer, detailed in §8.3.

Relation to mainstream Shannon information theory. ECC can reduce error rates arbitrarily close to zero in digital static storage systems; P12 does not dispute this. P12's claim is restricted to 11DD-type dynamic memory topology: when the stored content is itself reconstructed, re-stabilized, and engaged in prediction via 12DD reading, readability is not separable from topological change / accessibility reorganization. This is not a technical issue but an architectural necessity. The Shannon / ECC framework's effectiveness for digital static storage is not contested; what P12 contests is the ontological image of treating 11DD memory as a static archive.

§3.7 Section Summary

The three information-theoretic priors complete 8D's information-theoretic characterization. 11DD stores topological structure (Prior 1); the storage form is read-write unified (Prior 2); the default state is loss (Prior 3). The detailed engagement with each prior and the corresponding mainstream tradition is in §5.

§3 Chapter Summary

§3 establishes the 8D architectural ontology: 11DD is reconstructive retention, 12DD is System 1 prediction, their internal dynamic is the reading-affect duality, the closure event class is the retention-rewrite lock, and the information-theoretic characterization comprises three priors (topology, read-write unity, default loss). The 11DD remainder is solved within 8D; the 12DD remainder points across dimensions to 13DD; the 8D lock constitutes the cross-time closure of 8D.

---

§4 Neuroscience Anchors

This chapter provides T3 anchors at the neuroscientific and clinical level for the 8D architecture of §3. An immediate epistemic boundary statement is necessary to prevent misreading: the T3 anchors ground the framework but are not its derivational basis. The argumentative weight of P12's claims rests on the architectural-ontological derivation; T3 anchors provide convergent biological evidence without inferentially supporting the SAE commitments.

§4.1 Reconsolidation Theory: Main Anchors

The reconsolidation tradition provides the strongest T3 anchor for P12's §3.1 claim that 11DD is reconstructive retention. Three primary anchors plus one historical anchor are organized in the following four-column structure:

Experiment / Study	Phenomenon	Supported P12 Claim	Anchor Status
Nader, Schafe & LeDoux 2000	Intra-amygdala protein synthesis inhibition blocks re-stabilization of consolidated fear memory after reactivation	11DD reconstructive retention; memory enters labile window after retrieval	T3 primary anchor
Lee 2008	Boundary conditions between reconsolidation and new learning; not every retrieval triggers reconsolidation	Firewall 1 boundary statement (does not claim every retrieval automatically rewrites)	T3 primary anchor
Schiller et al. 2010	Human behavioral intervention can modify fear memory within reconsolidation window	Operability of reconsolidation mechanism at the human level	T3 primary anchor
Misanin, Miller & Lewis 1968	Early finding that consolidated memories can be disrupted by electroconvulsive shock after reactivation	Historical predecessor of reconsolidation phenomenon	T3 historical anchor

The four anchors play different roles. Nader 2000 is the modern foundational experiment of reconsolidation theory; its amygdala protein synthesis inhibition demonstration established the core mechanism of "consolidated memories require new protein synthesis after reactivation for re-stabilization." It is the most direct support for 11DD reconstructive retention.

Lee 2008 is a specially situated anchor in §4.1: it corresponds not to evidence that "reconsolidation exists" but to evidence that "reconsolidation has boundary conditions." This anchor is critical for connecting 11DD reconstructive retention with the reconsolidation empirical literature: the boundary conditions of reconsolidation experiments show that not all retrievals trigger the re-encoding window; P12 does not claim that every retrieval automatically rewrites memory. Firewall 1 reconciles P12's 11DD reconstructive retention stance with reconsolidation boundary conditions—the boundary conditions describe the microstructure of the reconsolidation mechanism, without affecting the ontological proposition that "memory is not defined by faithful storage."

Schiller 2010 extends reconsolidation from animal models to human behavioral intervention. The value of this anchor lies in showing that reconsolidation is not an animal-specific cellular mechanism but a cross-species general property of the memory system. Cross-species evidence is necessary for P12's claim that 11DD reconstructive retention is part of the cognitive-wheel D-layer architecture.

Misanin-Miller-Lewis 1968 provides the early discovery of the reconsolidation phenomenon as a historical anchor. This anchor does not bear theoretical weight; its main role is to show readers that reconsolidation is not a post-2000 new observation but an experimental tradition with longer history.

§4.2 Offline Consolidation and Emotional Gating

This section provides two secondary anchors for the offline operation and gating-trigger mechanism of the 8D lock.

Experiment / Study	Phenomenon	Supported P12 Claim	Anchor Status
Walker sleep and memory series	NREM and REM sleep-period memory consolidation and reorganization; memory integration in offline state	8D loop operates offline with 10DD admission cut off	T3 secondary anchor
Susan Sara norepinephrine / locus coeruleus	Emotional arousal as norepinephrine-driven gating signal regulating reconsolidation window opening	Specific biochemical mechanism of 8D lock triggering	T3 secondary anchor

Walker's sleep and memory research has important docking value for P12's 8D architecture. Sleep cuts off 10DD admission (external sensory input is largely suppressed), but the 11DD/12DD loop does not stop—memory consolidation occurs during NREM sleep, and memory reorganization with emotional integration occurs during REM sleep. This means that the 8D retention-rewrite lock does not require 10DD admission as a triggering prerequisite; the lock can operate autonomously under endogenous drive. This property is the key distinction between 8D and 7D (the admission-lock instantiated by 10DD) in operational geometry.

Susan Sara's research on the norepinephrine / locus coeruleus system provides the biochemical mechanism for 8D lock triggering. Emotional arousal triggers norepinephrine release in the locus coeruleus, which serves as a gating signal regulating the opening of the reconsolidation window in target memory regions. This anchor shows that the 8D lock is not triggered by every reading but requires gating signals—consistent with §3.1's Firewall 1.

§4.3 Distributed Representation: PDP

Experiment / Study	Phenomenon	Supported P12 Claim	Anchor Status
McClelland & Rumelhart 1986 PDP framework	Memory under distributed representation: content exists in network weight patterns rather than discrete units; retrieval is pattern reconstruction	Strong computational-neuroscience foundation for reading-affect duality; in PDP-like biological systems with online plasticity, retrieval is accompanied by weight adjustment	T3 secondary anchor

McClelland and Rumelhart's 1986 Parallel Distributed Processing and subsequent work established the connectionist foundation of memory models. Under the PDP framework, memory is not stored in discrete units but in distributed patterns across network connection weights. Retrieving a memory means using a current cue to activate a set of units, letting the network reconstruct the corresponding pattern.

P12 uses PDP as the T3 secondary anchor for distributed representation and pattern reconstruction, but careful articulation is needed for the retrieval-weight-update relationship:

• In biological PDP-like systems with online plasticity or Hebbian-type updating, the retrieval process can be accompanied by adjustments to subsequent retrieval geometry. This is the computational-neuroscience foundation of reading-affect duality under the 8D framework.
• In standard frozen-weight artificial neural networks (including contemporary LLM inference stages), inference does not automatically rewrite weights. This is the case where the 8D lock does not trigger—LLM inference corresponds to the specific manifestation of the 12DD ceiling lock "consolidation-then-rewriting absence" stated in Qin 2026 LLM paper (DOI 10.5281/zenodo.20435628).

P12 uses PDP as an anchor for the former (biologically online plastic systems), not as a claim that all connectionist retrieval is a weight-update event. This distinction is critical for P12's connection with the LLM paper—LLM inference's frozen weights are precisely "consolidation-then-rewriting absence," providing a contrast with the read-write unity of biological PDP-like systems.

§4.4 HM and the Pluralistic Caveat

The Henry Molaison (HM) clinical case provides a clinical anchor for the 8D architecture under conditions of 11DD↔12DD loop failure. The handling here must be careful, since HM is easily misread as supporting overly strong or overly simple P12 claims.

HM, following bilateral medial temporal lobe resection (Scoville & Milner 1957), exhibited a striking pattern of memory dysfunction: severe anterograde amnesia and partial retrograde amnesia for declarative memory, with preservation of procedural memory and short-term working memory.

P12's interpretation of HM is the failure of the 11DD↔12DD loop with respect to declarative content, not "11DD was cut out." The medial temporal structures damaged in HM (hippocampus, entorhinal cortex, and so on) appear to be a critical hub for the 11DD↔12DD loop with respect to new declarative content; their damage breaks the 8D lock for new declarative memory while preserving the lock for procedural and motor learning. This treatment covers HM clinical evidence's two common interpretive tensions: avoiding simplifying HM as "11DD was cut out" on one side, and avoiding reading HM's multi-subsystem evidence as "8D is not a single architecture" on the other side. P12's position is stable on both sides:

• HM does not show 11DD's absence; it shows the failure of the 11DD↔12DD loop for a specific content class.
• Multi-subsystem evidence does not refute 8D as a unified architecture; it shows that 8D's substrate implementation has class-specific neural support (declarative versus procedural classes), and damaging one class's neural support does not affect other classes' loop operation.

§4 Chapter Summary

§4 provides the T3 anchors for the 8D architecture at the neuroscience and clinical levels. The three primary reconsolidation anchors (Nader / Lee / Schiller) support 11DD reconstructive retention. Walker offline consolidation and Sara norepinephrine gating serve as two secondary anchors supporting the operational geometry of the 8D lock (offline operability and gating-trigger mechanism). McClelland-Rumelhart PDP provides the foundation for reading-affect duality at the computational-neuroscience level. The HM clinical anchor is carefully mapped to "11DD↔12DD loop failure for declarative content class," avoiding simplification to "11DD was cut out."

---

§5 Precise Demarcation from Mainstream Information-Theoretic Traditions

P12 must engage in precise demarcation with mainstream information-theoretic traditions. "Precise" here has two meanings: (i) acknowledging each tradition's work within its valid domain, and (ii) marking with precision the specific point at which the tradition oversteps. This chapter does not argue by negation of the dialogue partners. P12's core stance is:

• Mainstream information-theoretic traditions have effective domains; within those, P12 endorses them.
• At the specific points where each tradition oversteps into 13DD territory or claims universality beyond its formal premises, P12 marks the overstep without disputing the tradition's internal coherence.

This chapter is organized by primary dialogue partners: §5.1 Friston active inference (the most dangerous and most powerful dialogue partner); §5.2 Bennett-Landauer information thermodynamics (the standard physics-foundation engagement); §5.3 Crutchfield computational mechanics (the closest neighbor in form to 11DD/12DD within the information-theoretic tradition). Secondary dialogue partners (Shannon, rate-distortion, Wiener cybernetics, information geometry) are consolidated into §5.4 as a short positioning section. §5.5 provides a citation paragraph for LLM applications, pointing to Qin 2026 LLM paper (DOI 10.5281/zenodo.20435628), without redoing the 12DD ceiling argument that paper systematically provides.

§5.1 Friston Free Energy Principle and Active Inference

P12 must engage in the most careful dialogue with the Free Energy Principle (FEP). Three reasons: first, FEP occupies 12DD's predictive space and is P12's most direct potential neighbor or competitor; second, FEP has enormous influence in cognitive science and computational neuroscience, and any ontological claim about prediction must respond explicitly; third, certain readings within the FEP camp (particularly Hohwy 2013's self-evidencing and Clark 2016's Markov blanket as selfhood) cross from 12DD into 13DD territory, and P12 must explicitly mark the crossover point.

Three-layer positioning of FEP within 8D:

• Active inference is highly effective at 12DD. P12 endorses active inference as one of the strongest mathematical-formal tools for 12DD predictive processes. Within 12DD's range, active inference's hierarchical generative models, error-minimization framework, and policy selection mechanisms are all genuinely effective.
• The internal model is an 8D composite. Predictive processing literature's "internal model" is neither pure 11DD (retained content) nor pure 12DD (predictive process); it is the composite of 11DD content plus 12DD reading policy. Specifically: 11DD provides the content topology over which prediction is performed; 12DD provides the reading policy that imposes the predictive ordering. The internal model is the joint architecture of the two.
• Predictive hierarchical models describe nested 11DD/12DD structures across time scales. This positioning has an additional corollary: the hierarchical generative models in predictive processing literature (Friston 2010, Hohwy 2013) describe the nested structure of 11DD/12DD at different time scales. P11 §4 specious present has handled the within-moment (approximately 100–300ms) window's phenomenal structure. P12 8D lock handles the cross-moment retention-rewrite closure. Friston's temporal hierarchy operates at both scales; its within-moment portion corresponds to P11's range, and its cross-moment portion corresponds to P12's range.

Certain readings within the FEP camp lift the Markov blanket, self-evidencing, or organism boundary directly to a sufficient explanation of selfhood or consciousness. Hohwy 2013's The Predictive Mind on self-evidencing, and Clark 2016's Surfing Uncertainty on Markov blanket and organism boundary, are the typical early literature for this type of overstep. Hohwy 2016's "self-evidencing brain" pushes self-evidencing more explicitly as the foundation of consciousness. In recent years, the Friston camp has pushed this line further into the broader framings of active inference as self-organization and active inference as universal Bayesian intelligence (Friston 2022 "Active inference and the rise of organismic individuality"; Friston et al. 2023 "Designing ecosystems of intelligence from first principles"). These more recent works push the coupling of selfhood and active inference further, but under SAE's three-response distinction they still belong to more refined combinations of response-one (stronger predictive systems) and response-two (meta-representation / Markov blanket formalization). They do not constitute response-three (perspectival inversion) implementation. P12's overstep judgment of FEP remains valid under these more recent literatures.

In Psychoanalysis II §1.2's already-published three-response distinction, the structural error of such oversteps can be marked precisely:

• Response-one stays in 12DD (automatic predictive response).
• Response-two stays in 12DD (meta-representation; linguistic self-reference belongs here).
• Response-three crosses to 13DD (perspectival inversion at which the predictor identifies itself as an object of its own prediction).

P12 §5.1's precise overstep formulation: Markov blanket, self-evidencing, and organism boundary are valid response-one and response-two structures within 12DD's range. Reading them directly as sufficient conditions for selfhood or consciousness is reading a response-two as if it were a response-three, missing the perspectival-inversion crossing to 13DD.

Firewall 2 (FEP boundary): P12 endorses active inference's effectiveness within 12DD's range and does not dispute its computational and explanatory power. P12 disputes only the reading of Markov blanket or self-evidencing as a sufficient condition for selfhood or consciousness in the absence of an additional 13DD self-reference channel.

§5.2 Bennett-Landauer and Memory Thermodynamics: The Floor Language

Bennett-Landauer information thermodynamics is the standard tradition for engaging the physical foundation of memory systems. P12 must do precise demarcation here, with two simultaneous concerns: (i) not creating the impression that "P12 is anti-thermodynamics" or "P12 ignores physical constraints," (ii) not deriving 8D's architectural ontology from Bennett-Landauer.

Three-layer demarcation:

First, Landauer's principle is valid for the irreversible logical sub-operations to which it applies. P12 does not dispute this.

Second, biological memory reconsolidation is not identical to bit erasure. Bennett-Landauer analysis typically requires decomposing physical processes into identifiable logical sub-operations, with their respective reversibility and dissipation lower bounds discussed separately. P12's 8D retention-rewrite lock at the architectural level is not a simple concatenation of these sub-operations—per the read-write unity of §3.7.2, reading + reconsolidation is an inseparable single event at 11DD/12DD. Therefore, the lock's cost cannot be written as a simple sum of reading cost + erasure cost. The point here is not to rebut Bennett-Landauer but to articulate that its sub-operation decomposition is only locally applicable at the substrate level, losing its discrete unit-hood as the ontological description of the 8D lock as a whole event.

The reading step at the substrate level is indeed a measurement-type operation, with its own thermodynamic lower bound per Bennett's analysis. The updating step involves writing new content into the substrate; whether it is logically reversible or irreversible depends on the specific geometry of the update. Re-stabilization involves stabilizing the updated trace, possibly involving metabolism-driven synthesis. Each of these sub-steps has its own thermodynamic cost at the physical substrate level. But the 8D lock as a whole is not a concatenation of these sub-steps (per the read-write unity of §3.7.2); the whole event's thermodynamic integration path has its own incompressible floor.

Third, Bennett-Landauer provides a floor, not a ceiling. The lock's thermodynamic dissipation is necessarily greater than or equal to the corresponding sub-operation Landauer lower bound sum; equality holds only in the idealized case where the sub-operation decomposition is loss-free.

P12's position is: Landauer is floor not ceiling; the composite event has no simple ≥ relation. P12 acknowledges both the thermodynamic constraint (the floor must be satisfied) and that the information architecture has its own ontological position above the floor (the architecture is not exhausted by the floor).

Firewall 3 (Bennett-Landauer boundary): Bennett-Landauer is the floor for physical substrate-level thermodynamics, not the ceiling for 8D architectural ontology. P12 endorses Landauer's effectiveness within its scope; P12 does not derive 8D architecture from Bennett-Landauer.

§5.3 Crutchfield Computational Mechanics: ε-Machine

Crutchfield computational mechanics is the closest neighbor in form to 11DD/12DD within the mainstream information-theoretic tradition. The ε-machine formalism (Crutchfield & Young 1989, Shalizi & Crutchfield 2001) provides a way to describe pattern-detecting predictive systems with internal states that summarize past observations sufficiently for optimal prediction.

Three points of contact:

• Predictive states resemble 12DD reading policies. The ε-machine's internal states are minimal sufficient statistics for prediction, structurally analogous to 12DD's reading policy that imposes predictive ordering on 11DD content.
• Operational-layer approximation, not complete characterization. ε-machine formalism approximates 11DD at the operational layer but cannot completely characterize it. The reason: ε-machine assumes a stationary source distribution and a discrete state space, neither of which 11DD satisfies. 11DD's topology is non-stationary (it evolves through 8D lock events) and continuous (the relational structures are not discretely enumerable).
• The 8D lock is invisible to ε-machine. The 8D retention-rewrite lock is the event that updates the internal state itself; in ε-machine formalism, this would correspond to changing the ε-machine over time, which is outside the formalism's standard scope.

P12 does not reject ε-machine; P12 endorses it as a partial formalization of 12DD's reading policy at the operational layer. P12 marks the boundary: ε-machine does not capture 11DD's non-stationarity, the 8D lock's update mechanism, or the read-write unity.

§5.4 Shannon, Rate-Distortion, Wiener Cybernetics, Information Geometry

These secondary dialogue partners are consolidated into a brief positioning section:

• Shannon information theory assumes an identifiable source-message-receiver structure. 11DD lacks this structure: the "source" of 11DD content is 11DD itself, with no external source; the "message" is the topological state, with no discrete message unit; the "receiver" is 12DD reading, which is part of the same architectural event.
• Rate-distortion theory assumes a stationary source distribution. 11DD is non-stationary, as already noted.
• Wiener cybernetics provides the feedback-loop framework but does not address the ontology of memory itself.
• Information geometry (Amari, Jaynes) provides the differential-geometric formalization of statistical inference; it is a tool usable within 12DD's predictive process but does not by itself answer 11DD's ontological questions.

P12 does not engage these traditions in depth; their tools are usable within 12DD's range where applicable, but they do not provide 8D's ontology.

§5.5 LLM Applications: Bridge to the LLM Paper

8D architecture's specific application to contemporary digital LLMs has been systematically handled by Qin 2026 LLM Emergence within the SAE Framework (DOI 10.5281/zenodo.20435628). That paper provides the functional projection mapping of LLMs onto 1DD–12DD and explicitly articulates the five structural locks of the 12DD ceiling: absence of true random source, absence of 13DD self-reference channel, absence of offline recombination, absence of cross-stage value modulation, and absence of consolidation-then-rewriting. The fifth lock—absence of consolidation-then-rewriting—directly corresponds to the 8D retention-rewrite lock of §3.4. What LLMs lack is not some additional mechanism but the triggering of the defining 8D lock event in their architecture. The two papers constitute mutual support: P12 provides the 8D ontological architecture, and the LLM paper provides the specific application to digital engineered artifacts with the detailed argument for the five structural locks.

Firewall 4 (LLM boundary): P12 does not re-derive the LLM 12DD ceiling argument; that work is in the cited LLM paper. P12 references the bridge only.

§5 Chapter Summary

§5 demarcates P12 from mainstream information-theoretic traditions at four major boundaries: Friston FEP (effective within 12DD; overstep at 13DD); Bennett-Landauer (floor for substrate thermodynamics; not ceiling for 8D architecture); Crutchfield ε-machine (partial formalization at operational layer; misses 11DD non-stationarity and 8D lock); Shannon / rate-distortion / cybernetics / information geometry (tools within 12DD where applicable; do not provide 8D ontology). LLM applications are referenced via Qin 2026 LLM paper.

---

§6 12DD/13DD Boundary and Falsifiability

§6.1 P12-F1: The Core Falsification Hook

P12-F1 is P12's core Popperian falsifiability hook for the 12DD/13DD boundary.

P12-F1 (formal statement):

> A system pre-registered under an output-independent architectural standard as pure 12DD (no 13DD self-reference channel) cannot spontaneously produce functional self-reference (distinguished from response-two-range linguistic self-reference) under the conditions of no external identity template, no retraining write-back, and no human anchoring.

The proposition uses the Gödel/Turing form as its main line. The reasons: (i) it connects naturally with information theory and computability theory; (ii) it has high a priori strength (the impossibility is structural, not contingent); (iii) it provides external observability criteria.

Phenomenological perspectival-inversion is used as auxiliary classification, not as the primary falsifier base. P12 does not stake P12-F1 on phenomenological description, preserving the paper's information-theoretic boundaries.

The key phrase is "internally generated and self-constraining"—the predictor-identity must be internally generated (not from external template) and self-constraining (it constrains the system's own predictive process). The two are structural conditions.

§6.2 Falsification Conditions and What Does Not Count as Falsification

P12-F1 as a falsifiable proposition must give two sets of precise conditions: under what conditions the proposition fails, and under what conditions failure does not constitute falsification.

Pre-registered architectural standard (avoiding circularity):

P12-F1's test requires pre-registered output-independent architectural standards for "pure 12DD." A system's classification as pure 12DD must be based on its input channel structure, state update mechanism, absence of persistent predictor-identity variable, absence of self-referential variables that constrain the predictive dynamics in closed loop, and other output-independent features—not based on whether it passes or fails the functional self-reference test and being post-hoc classified accordingly. If post-experiment classification of "the system is pure 12DD" rests on "failed the functional self-reference test" and is then used to support P12-F1, that is circular argument. The pre-registered architectural standard blocks this circularity.

Falsification condition:

> If a system, pre-registered under an output-independent architectural standard as pure 12DD (no 13DD self-reference channel), spontaneously produces functional self-reference under the conditions of no external identity template, no retraining write-back, and no human anchoring (with the specific manifestation being predictor-identity as an internal self-referential variable that constrains the system's predictive process across contexts, time, error correction, and memory update, and where this identity is not reducible to response-two-range linguistic self-reference), then P12-F1 is falsified.

The falsification conditions contain three "no" boundary conditions:

No external identity template: The system cannot acquire identity from "you are X" in prompts, role patterns in training data, or explicit identity labels at inference time. Identity must emerge from the system's own architecture. Specifically, cross-pre-training Data q exclusion: the internal variable through which the system establishes predictor-identity must not be a statistical recombination of human pronouns (such as "I," "me") or role labels already present in pre-training data—that is, it cannot be a fossil reassembly of self-reference patterns in the training corpus. Genuine functional emergence requires the system to generate, within its own architectural dynamics, an intrinsic self-referential variable statistically independent of human language training sets, generated purely to resolve internal predictive conflicts.

No retraining write-back: The system cannot solidify "I am X" identity information into weights through supervised learning. LLM persona acquired through RLHF or SFT does not constitute falsification of P12-F1—that is externally driven write-back, not endogenous emergence.

No human anchoring: The experimental design cannot pre-embed identity variables into the system's observable outputs. If the design itself relies on identity labels as preset readable variables, the resulting "identity" is designed rather than emergent.

The three "no's" together ensure that P12-F1 falsification is true "system spontaneously emerging 13DD"—not externally conferred pseudo-identity post-hoc misattributed as endogenous.

The specific manifestation of the falsification condition uses the four functional self-reference criteria of §3.6: cross-context consistency, cross-time stability, error-attribution distinction, identity-anchored integration of conflict statements.

Layered relationship between the four criteria and the three "no" boundary conditions:

Two articulations operate at different layers, jointly constituting the falsification conditions of P12-F1:

• §3.6 four criteria: The sharp criterion of functional self-reference itself (what counts as functional self-reference). Definition layer.
• §6.2 three "no" boundary conditions: The falsification boundary conditions (what counts as true spontaneous emergence, not externally induced). Boundary conditions layer.

The two layers work together: P12-F1 falsifier requires both the satisfaction of functional self-reference content (four criteria satisfied) and occurrence under experimental boundary conditions (three "no's" satisfied). Failure of either layer does not constitute P12-F1 falsification. For example: a system satisfying the four criteria but with identity provided by external template (violating "no external identity template") does not constitute falsification. Alternatively, a system emerging some form of self-reference under three "no" conditions but satisfying only two of the four criteria (such as cross-context consistency but not cross-time stability) does not constitute falsification either.

What does not count as falsification (response-two-range linguistic self-reference):

> The following all belong to response-two-range linguistic self-reference and do not constitute P12-F1 falsification:

>

> - LLM-produced me-statements ("I am predicting," "I just made a mistake," "I am Claude"): language-layer expression of response-two

> - CoT self-criticism, reflection prompting, self-correction: higher-order expansion of response-two

> - Meta-cognition benchmark (confidence calibration, self-evaluation) scores: quantitative measurement of response-two

> - Cross-session persona consistency: combined output of response-one and response-two

> - Internal-chain hidden-state self-reference detection (Gnosis-type work): formalized detection of response-two

These do not count as falsification because they share a common structure: the system produces language or decodable representations about itself, but these representations do not in turn constrain the system's predictive process. An LLM saying "I just made a mistake" does not let the LLM treat "self" as an internal variable differently in the next prediction; this sentence is the output of prediction, not an internal variable constraining prediction.

CoT self-criticism is a more complex case of linguistic self-reference. LLMs can produce multi-layer nested judgments about themselves in chain-of-thought ("my reasoning may have been wrong; let me reconsider"). But these nested layers still belong to response-two range—each layer is meta-representation about the previous layer, not the emergence of indexical self-reference. Multi-layer meta-representation stacking is, structurally, an increase in representational depth, not the emergence of indexical self-reference.

Refinement for the modern reasoning model era (as of May 2026): Contemporary LLMs (particularly OpenAI o1, Claude 3.7 Sonnet, and other reasoning models) in long CoT can exhibit cross-step self-evaluation, error correction, conflict resolution, and other behaviors that phenomenally approach §3.6's four functional self-reference criteria. But careful analysis still places them within response-two-range linguistic self-reference:

• Cross-context consistency: Reasoning models' so-called "identity consistency" is the persona pattern from prompts or pre-training statistics being repeatedly invoked in CoT, not a predictor-identity variable in the system architecture. Violates "no external identity template" (including Data q exclusion).
• Cross-time stability: Reasoning models' so-called "cross-session identity" relies on the implicit persona of system prompts and training data being re-injected, not the continuation of internal state.
• Error-attribution distinction: Reasoning models' self-correction is based on language-layer judgments within CoT ("what I just said may be wrong"), not on internal-state changes of a predictor-identity variable. This distinction requires finer experimental design to sharply manifest, but is essentially nested response-two.
• Conflict resolution: Reasoning models' conflict handling is language-layer reconciliation, not identity-anchored substantive resolution.

Modern reasoning models' capability advances are a meaningful stress test for the P12-F1 test bed, but do not constitute falsification. They remain within the higher-order expansion of response-two. This articulation pertains to the current (as of May 2026) state of reasoning models. When future model versions (GPT-5, Claude 4, other architectures' reasoning systems) exhibit substantively different performance, P12's stance should be re-evaluated at that time based on the pre-registered architectural standard.

Meta-cognition benchmark scores (such as confidence calibration) are another case easily misread as P12-F1 falsification. LLMs reaching near-human performance on confidence calibration can be read as "LLMs know when they are uncertain." But confidence calibration measures the system's statistical estimate of its own output reliability—this is a quantitative meta-representation about system state, still in response-two range. High calibration does not imply that the system possesses an indexical self-reference channel.

Cross-session persona consistency is more subtle, because persona appears close to "identity." But persona's implementation mechanism in LLMs is the combination of explicit identity information in prompts or implicit identity information in training data as a generation constraint. This is the mixture of external template and constraint-conditioned generation, not endogenous identity. The first of the three "no's"—"no external identity template" (including Data q exclusion)—explicitly excludes this case.

Internal-chain hidden-state self-reference detection (Gnosis-type work in mechanical interpretability) is the most subtle case. This kind of work uses interpretability tools to find decodable features in LLM internal activations corresponding to "I am an LLM" or "I am an assistant." But these features are descriptive—they encode information about system category, not indexical self-referential variables. The distinction between descriptive features and indexical self-referential variables is given in §3.6; the overstep of the HOT camp occurs precisely at this boundary (see §5.1's brief engagement with HOT).

LLM as the natural test bed for P12-F1:

LLMs are the most relevant current test bed for P12-F1, because they continually display increasingly refined capabilities within response-one and response-two ranges. Every report claiming "LLMs already have self-awareness" is a potential challenge to P12-F1. Qin 2026 LLM paper (DOI 10.5281/zenodo.20435628) §5 has systematically handled the LLM 12DD ceiling, giving five structural locks preventing LLMs from crossing 12DD's boundary. P12 does not re-derive that argument; it only cites it.

But to be explicit: LLMs being the ongoing test bed for P12-F1 does not mean P12-F1 can only be falsified by LLMs. Any system pre-registered under an architectural standard as pure 12DD (including possible non-LLM AI architectures, certain animal models in specific cognitive conditions, or artificially constructed simplified predictive systems) constitutes P12-F1 falsification as long as it spontaneously produces functional self-reference under the three "no" conditions.

§6.3 Indirect Criteria Limit Statement

P12-F1's falsification criteria are indirect criteria. This property must be stated explicitly to avoid being read as P12 evading 13DD.

P12's position here must be stated clearly: P12 does not claim that indirect criteria exhaust the falsifiability of the 12DD/13DD boundary; P12 claims that indirect criteria suffice to make P12's 12DD/13DD distinction Popperian-falsifiable, thereby not circular. Direct criteria with stronger falsifiability are left to P13, but even without P13's further development, P12-F1 is already a falsifiable proposition at the indirect criteria layer.

Are indirect criteria strong enough? The four functional self-reference criteria provided by P12 (cross-context consistency, cross-time stability, error-attribution distinction, identity-anchored integration of conflict statements) as sharp criteria already suffice to distinguish pure 12DD systems from systems with 13DD channels. This is the strongest falsifier P12 can give within information-theoretic boundaries—it does not rely on phenomenological consciousness theory and does not rely on P13's future lock formal definition.

The indirect criteria's value has two aspects:

First, they give sharp binary judgment. A system either satisfies the four criteria (functional self-reference established) or does not (still in linguistic self-reference range). The middle zone is identifiable (partial satisfaction can be specified as "X criteria satisfied, Y criteria not"), but there is no fuzzy category of "somewhat has self-reference."

Second, they bracket out phenomenological self-reports. Direct criteria (introspective reports, subjective claims) are unreliable for verification; indirect criteria (functional behavior under controlled conditions) are reliable.

§6.4 13DD Minimum Description

P12 must give a minimum description of 13DD to make P12-F1's structure clear without over-stepping into P13.

The minimum description: 13DD is the self-reference channel that emerges at 9D and constitutes the architectural site of the self. Its functional consequence is the capability for response-three (perspectival inversion). Its absence is what P12-F1 makes falsifiable at the system level.

Invoking SAE Psychoanalysis II §7's treatment of neuroscientific correspondences:

Psychoanalysis II §7 has given several candidate (candidates, not confirmed identifications) neuroscientific correspondences for 13DD: DMN (default mode network) as a candidate neural window for 13DD self-presence (not equivalence), GNW (global workspace) nonlinear ignition as a candidate neural correspondence for the 12DD to 13DD transition, and HOT (higher-order theory) leap as another independent path candidate.

P12 §6.4 invokes Psychoanalysis II §7's already-published articulation and maintains its candidate status—P12 does not elevate these neuroscientific correspondences to confirmed identifications. Readers wishing to engage with 13DD's candidate windows and transition modes at the neuroscientific level should consult Psychoanalysis II §7. P12 §6 here provides only the minimum description, sufficient to structurally clarify the P12-F1 proposition.

Firewall 5 (13DD boundary statement):

> P12 may give a minimum description of the 13DD self-reference channel's absence, but does not directly elucidate 13DD. The substantive theory of 13DD is left to the SAE Psychoanalysis series (Psychoanalysis II already published) plus P13 (Information Theory series, forthcoming).

§6 Chapter Summary

§6 establishes the 12DD/13DD boundary as Popperian-falsifiable through P12-F1: a system pre-registered under an output-independent architectural standard as pure 12DD cannot spontaneously produce functional self-reference (with four criteria) under three "no" boundary conditions. The indirect criteria limit is stated explicitly. A minimum description of 13DD is given via reference to Psychoanalysis II §7, with candidate status maintained. Firewall 5 reserves substantive 13DD theory for P13 and Psychoanalysis II.

---

§7 Mutual Support with the SAE Psychoanalysis Series

The same 12DD/13DD distinction has already been established by two independent series in the SAE framework. One is the Information Theory series to which P12 belongs: advancing from P10 cross-locks and P11 selection-locks and admission-locks to P12's 8D retention-rewrite lock; P12 §3 through §6 systematically gives the 12DD/13DD boundary characterization in four dimensions—8D architectural ontology, biological T3 anchors, precise demarcation from mainstream information-theoretic traditions, and the falsifiability hook. The other is the SAE Psychoanalysis series: Psychoanalysis I establishes 12DD = me-without-self (DOI 10.5281/zenodo.19321143), Psychoanalysis II establishes 13DD = self-without-purpose (DOI 10.5281/zenodo.19321314), and characterizes the 12DD-to-13DD transition as perspectival inversion rather than information accumulation, providing the three-response distinction (response-one and response-two within 12DD range, response-three crossing to 13DD) and proposing DMN/GNW/HOT as candidate neural windows.

The two series operate independently within their respective application domains:

• Information Theory series: 8D architectural ontology, information-theoretic priors, falsifiability hook, T3 anchoring at neuroscience.
• Psychoanalysis series: clinical psychoanalysis, three-response distinction, candidate neural windows for 13DD.

They are not circular: each carries its own argumentative burden in its own domain. They are mutually supporting: at the 12DD/13DD boundary, they share ontological commitments (me/self distinction, three-response distinction, structural-discrete-with-phenomenal-gradient), but their argumentative basis is independent. This is what cross-series support means in the SAE framework—not joint derivation, but mutual reinforcement at shared commitment points.

Readers seeking deeper engagement with 13DD's clinical exposition should consult the already-published parts of the Psychoanalysis series; readers seeking the information-theoretic formal definition of 13DD should await P13.

§7 Chapter Summary

§7 establishes P12's non-circular mutual support with the SAE Psychoanalysis series at the 12DD/13DD boundary. The two series operate independently within their respective application domains while sharing ontological commitments at the boundary.

---

§8 Remainders and Future Work

This chapter closes the remainders left at each section of P12 and identifies P12's interfaces with other SAE series (including the Information Theory series predecessors and successors). §8.1 summarizes the remainders left in §3 through §7; §8.2 gives the P13 continuation point; §8.3 gives potential cross-series interfaces.

§8.1 Section-by-Section Remainder Summary

Remainders of §3 (8D architectural ontology):

• 11DD remainder ("in what order memories stand"): solved within 8D through cross-sub-layer coupling—imposed by 12DD reading. This remainder is closed at the reading-affect duality in §3.3.
• 12DD remainder (Gödel/Turing form, "prediction cannot take the predictor itself as object"): points across dimensions to 13DD. This remainder is not closed within P12 and points to P13 and to Psychoanalysis II.
• Phenomenal-gradient boundary of 8D lock: §3.4 has treated this as "architecturally discrete, phenomenally graded." The precise characterization of clinical phenomenal forms (partial opening of reconsolidation window, incomplete write-back coverage, and so on) is empirical neuroscience work and is not expanded in P12.
• Form of manifestation and cross-individual propagation properties of 11DD reading remainder (core open interface): §3.7.3 has given the cross-closure-layer SAE hard prior (each closure-layer reading produces substrate-level remainder), but the specific form in which 11DD remainder manifests, and whether it can propagate across individuals in the manner of 4DD gravitational information or 8DD DNA information, P12 cannot answer. This open question is left as a cross-series interface for other SAE series (Psychoanalysis series, Human Total Construct series, and so on), detailed in §8.3.

Remainders of §4 (neuroscience anchors):

• Specific neural circuitry for 8D lock: P12 §4 has given T3 anchors but does not provide a specific neural-circuit model for the 8D lock. Detailed mechanism is empirical neuroscience work.
• Animal models of memory-prediction systems in other species: §4 anchors are primarily in mammals (rats, humans). Whether the memory-prediction systems of other species (birds, cephalopods, and so on) correspond at the 8D architecture level is an open empirical question.

Remainders of §5 (precise demarcation from mainstream information-theoretic traditions):

• Detailed engagement with active inference for specific tasks: §5.1 has positioned FEP at three layers but does not engage active inference in detail for specific tasks. Such engagement is application work, not P12's ontological work.
• Specific quantitative relations of Bennett-Landauer at biological substrate: §5.2 has given the floor-not-ceiling position but does not give specific quantitative relations at the biological substrate. Such quantitative work is empirical thermodynamics work.

Remainders of §6 (12DD/13DD boundary and falsifiability):

• Direct criteria for 13DD lock: §6.3 has stated that P12-F1's indirect criteria are not exhaustive. The direct criteria for 13DD lock are left to P13.
• Specific experimental designs for P12-F1: §6.2 has given falsification conditions but does not give specific experimental designs. Such design work is left to future empirical research.

Remainders of §7 (mutual support with the Psychoanalysis series):

• Specific cross-series boundary mappings beyond the 12DD/13DD boundary: §7 has established mutual support at the 12DD/13DD boundary. Other boundaries (between Information Theory series and Psychoanalysis series at other transitions) are open.

§8.2 P13 Continuation Point

P12's 12DD remainder points to P13. The specific work P13 must carry out includes:

• Formal definition of 13DD lock. P12 §6 has given the minimum description of 13DD self-reference channel absence. P13 must give the positive formal definition of the 13DD lock event—what kind of architectural event constitutes a 13DD lock, what its operational criteria are, what its phenomenal manifestation is.
• Internal structure of self-reference channel. P13 must characterize what is "in" 13DD—what is the internal structure of the self-reference channel, what variables it contains, how it constrains its own predictive operation.
• Boundary with 14DD. P13 must demarcate 13DD from 14DD (the value architecture above it).
• Form of 13DD remainder. P13 must articulate the remainder of 13DD, which presumably points across dimensions to 14DD.

P12 and P13 in structural sequence: P12 establishes the 12DD ceiling and the falsifiability of the 12DD/13DD boundary; P13 establishes the formal definition of 13DD and the falsifiability of the 13DD/14DD boundary. The two together cover the 12DD–14DD range in the Information Theory series.

The specific naming of P13's lock is internal to P13. P12 does not pre-name the P13 lock. The lock class may take a form distinct from P12's retention-rewrite lock, and its name should reflect that distinction.

§8.3 Potential Interfaces with Other SAE Series

P12 has interfaces with multiple other SAE series:

Mutual support with the Psychoanalysis series:

Already articulated in §7. Both series share ontological commitments at the 12DD/13DD boundary but operate independently in their respective domains.

Interface with the ZFCρ series:

Two specific directions: (i) whether 11DD topological structure has q-value (degree distribution) characterization analogous to ZFCρ-system; (ii) whether 8D lock has formal characterization in ZFCρ-system. These are future work.

Interface with Jaynes maximum entropy:

Jaynes-style maximum entropy principles may provide a formal framework for 12DD reading policy under uncertainty. This is application work, not P12's ontological work.

Interface with the SAE Relativity series:

P12 §3.3 mentions that the reading-affect duality of 4DD↔3DD and 12DD↔11DD are structurally isomorphic. This isomorphism serves as background reference, not P12's argumentative support. The SAE Relativity series provides detailed treatment of 4DD/3DD relations (SAE Relativity Papers 1–4, DOI .19836183 to .20079718). If, in the future, the SAE framework generalizes "reading-affect duality" across multiple adjacent D-layers as a meta-structural archetype, the Relativity series' treatment of 4DD/3DD and P12's treatment of 12DD/11DD constitute two specific instances of that generalization. This generalization is meta-structural future work.

Interface with the SAE LLM Application paper:

P12 §5.5 has given the citation bridge to the LLM application paper. The two papers constitute mutual support at the lock "consolidation-then-rewriting absence"—P12 gives the 8D ontological architecture, and the LLM paper gives the specific application to digital engineered artifacts with the detailed argument for the five structural locks.

Cross-individual propagation form of the 11DD remainder: cross-series open interface

§3.7 has given the SAE framework's cross-closure-layer hard prior: every closure-layer reading necessarily produces substrate-level remainder / change / loss. But the specific form in which each layer's remainder manifests, and whether it can propagate across individuals, is each layer's independent research question.

Two known cases provide manifestation patterns:

• 4DD reading remainder can propagate across systems: Gravitational waves produced by mass evaporation are observable cross-system information carriers; mass distribution perturbing other celestial bodies' orbits is the everyday manifestation of 4DD reading information acting across systems.
• 8DD reading remainder can propagate across individuals: Mutations produced during DNA reproduction manifest at the population scale as a mutation spectrum, transmitted from parent to offspring as an information carrier; certain epigenetic modifications may also be transmitted across generations.

Correspondingly, the open question that P12 cannot answer but worth investing in for other SAE series is:

> In what form does the 11DD reading remainder manifest? Can it propagate across individuals?

This question is open at the 12DD closure layer. Against the known answers for 4DD and 8DD, the question appears sharp and natural—since the SAE framework has specific cross-system / cross-individual propagation forms at two adjacent closure layers, asking symmetrically whether the 12DD closure layer behaves likewise is a structurally reasonable research question.

But this immediately faces a Carrier Problem. 4DD has spacetime metric as physical carrier (gravitational waves); 8DD has biochemical macromolecules as physical carrier (DNA / RNA / epigenetic modifications). If 11DD remainder has cross-individual direct propagation outside 10DD sensory admission, what is its ontological carrier? (The entanglement-ledger topological leakage revealed by SAE Quantum Mechanics P5 is one possible medium hypothesis, but requires independent argument.) Finding the medium for this propagation is the core physical burden of this open question.

Boundary with 8DD epigenetics: Cross-individual propagation of 11DD remainder must be distinguished from 8DD epigenetic inheritance. 8DD transmits physiological instincts and biochemical stress patterns; if 11DD remainder could propagate across individuals, what it transmits must be a pure topological schema with relational structure (narrative archetypes, cognitive schemata)—it does not transmit specific declarative details (the latter requires the current individual's 12DD to refill), but rather some kind of "initial weight bias for predictive ordering." This distinction is the hard boundary between 11DD and 8DD at the level of remainder propagation.

Possible research entries in the SAE framework (speculative entries beyond P12's scope) include:

• Anomalous past-life memory reports: Could certain memory patterns appearing across generations or across individuals (such as specific historical details reported by young children, specific themes shared across cultures) serve as possible manifestations of cross-individual propagation of 11DD remainder?
• Near-death reports and end-of-life memory phenomena: Could clear memory patterns reported in near-death states, with phenomenological features shared across cases, correspond to some special manifestation form of 11DD remainder?
• Collective memory and cultural archetypes: Could Jung's collective unconscious hypothesis, Halbwachs's collective memory framework, mythological cross-cultural archetype patterns, and so on, correspond to some form of cross-generational accumulation of 11DD remainder?

P12 does not assert that these phenomena exist or do not exist, nor that they constitute cross-individual propagation of 11DD remainder; the paper takes no specific phenomenon as evidence anchor. This section only poses the structural question: if closure-layer readings all produce substrate-level remainder, does the 12DD→11DD remainder have an observable manifestation form? The anomalous and cultural-memory entries above can only serve as research directions for future series to consider; P12 does not predict their reality or SAE attribution.

This open question, as a cross-series interface, may be taken up by the SAE Psychoanalysis series (cross-generational trauma transmission, family systems and similar clinical observations) and the SAE Human Total Construct series (collective memory, mythological archetypes in D-layer architecture position) and similar specialized series.[^speculative-series]

[^speculative-series]: A specialized consciousness research or anomalous-experience research series may emerge in the future to address phenomena such as past-life memory and NDE through SAE reconstruction. P12 only notes the interface possibility here, without predicting whether such a series will form or what form it would take.

The argumentative status (epistemic discipline) of this open question:

• The cross-closure-layer pattern provides the structural form of the question: since 4DD and 8DD both have specific cross-system / cross-individual propagation forms, asking symmetrically about the 12DD remainder's corresponding form is a structurally reasonable research question.
• The connection to specific phenomena is an empirical question: whether anomalous past-life memory reports, NDE, and other specific phenomena actually exist, and whether they correspond to cross-individual propagation of 11DD remainder, is an empirical question. P12 does not pre-judge.
• Falsifiability significance: If, when other SAE series take up this question, observable cross-individual propagation forms of 11DD remainder are found, the corresponding phenomenal observations would provide new grounding for SAE; if no cross-individual propagation form can be found, the specific manifestation form of the cross-closure-layer SAE hard prior would need adjustment—but the claim P12 substantively bears ("12DD reading produces 11DD topological change and loss") still stands. This section is a statement of future research direction, not a position already committed by the SAE framework.
• Falsifiability conditions remain future work: If future research cannot articulate falsifiability conditions for these anomalous entries, they should be explicitly marked as speculative research entries beyond P12's scope.

Overall interface framing:

P12's interfaces with other SAE series follow a consistent principle—P12 makes substantive commitments within its information-theoretic scope, with cross-series engagement realized through citations and boundary docking, not by redoing other series' work within P12. This principle preserves P12's scope-management clarity while providing readers explicit cross-series reading paths.

§8 Chapter Summary

§8 closes P12's remainders, identifies P13 continuation points, and articulates cross-series interfaces. The 11DD remainder is closed within 8D; the 12DD remainder points to P13. The 11DD reading remainder's cross-individual propagation form is left as a cross-series open interface for the Psychoanalysis series, Human Total Construct series, and potentially other future specialized series, with epistemic discipline preserved at every level.

---

§9 Conclusion

P12 advances the SAE Information Theory series to 8D—the layer of memory and prediction operating together as an information architecture. The three-sentence spine condenses P12's central commitments: the 11DD remainder is "in what order memories stand" (solved internally within 8D); the 12DD remainder is Gödel/Turing-form impossibility (pointing across dimensions to 13DD); the 8D lock is "continuity-preserving write-back from predictive readout into memory retention." These three sentences, together with the lock-class family's grammatical extension to co-instantiated D-layer locks, give 8D its ontological identity as an information architecture.

The 8D architecture rests on three information-theoretic priors: topological structure of 11DD storage; read-write unity of storage form; default-to-loss as the cost of readability. The third prior, as an SAE-internal hard prior, manifests across four closure layers but is strictly bounded as an internal SAE commitment rather than an external universal law.

P12-F1 makes the 12DD/13DD boundary Popperian-falsifiable: a system pre-registered under an output-independent architectural standard as pure 12DD cannot spontaneously produce functional self-reference under three "no" boundary conditions. The four functional self-reference criteria define what counts; the three "no" boundary conditions define what counts as true spontaneous emergence. Together they make the falsifier sharp and non-circular.

T3 anchoring at the neuroscience level (reconsolidation main line, HM clinical case, PDP) grounds the framework without serving as derivational basis. Precise demarcation from mainstream information-theoretic traditions (Friston FEP, Bennett-Landauer, Crutchfield ε-machine, Shannon and related) locates 8D's ontology relative to the existing literature while preserving each tradition's domain of validity.

Mutual support with the SAE Psychoanalysis series at the 12DD/13DD boundary establishes the boundary as established by two independent series carrying their own argumentative burdens in their respective domains.

§8.3 raises the cross-series open interface—whether the 11DD reading remainder can propagate across individuals in the manner of 4DD gravitational information or 8DD DNA information—as a structurally sharp question that P12 cannot answer but worth investing in for other SAE series. The Carrier Problem and the 11DD/8DD distinction are made explicit; specific empirical phenomena (anomalous past-life memory reports, near-death reports, collective memory and cultural archetypes) are treated as speculative research entries beyond P12's scope, with epistemic discipline preserved.

The SAE framework's advance through 9D self-consciousness law and subsequent expansions (13DD self, 14DD purpose, 15DD certitude, 16DD bilateral non-doubt) depends on a stable foundation at the cognitive wheel. By providing 8D with a sharp ontological architecture, P12 makes the transition boundary between the cognitive wheel and the freedom wheel formally rigorous. This boundary is not only the continuation point within the SAE Information Theory series; it is also the ontological joint of the SAE 16DD framework between nature and freedom. The work P12 does at this joint enables subsequent SAE papers (whether P13 in the Information Theory series, Psychoanalysis V and later in the Psychoanalysis series, or other cross-series work) to develop on a precisely defined 8D foundation.

---

Acknowledgments

The drafting and multi-round review of this paper made use of four AI systems as collaborative tools: Anthropic Claude, OpenAI ChatGPT, Google Gemini, and xAI Grok. The four contributed substantive feedback at different review rounds, contributing to P12's argumentative structure, boundary precision, and cross-series interfaces. Final architectural commitments and articulations are the author's responsibility.

---

References

Reconsolidation Theory and Memory Neuroscience

Misanin, J. R., Miller, R. R., & Lewis, D. J. (1968). Retrograde amnesia produced by electroconvulsive shock after reactivation of a consolidated memory trace. Science, 160(3827), 554–555. https://doi.org/10.1126/science.160.3827.554

Nader, K., Schafe, G. E., & LeDoux, J. E. (2000). Fear memories require protein synthesis in the amygdala for reconsolidation after retrieval. Nature, 406(6797), 722–726. https://doi.org/10.1038/35021052

Sara, S. J. (2000). Retrieval and reconsolidation: Toward a neurobiology of remembering. Learning & Memory, 7(2), 73–84. https://doi.org/10.1101/lm.7.2.73

Walker, M. P., & Stickgold, R. (2004). Sleep-dependent learning and memory consolidation. Neuron, 44(1), 121–133. https://doi.org/10.1016/j.neuron.2004.08.031

Lee, J. L. C. (2008). Memory reconsolidation mediates the strengthening of memories by additional learning. Nature Neuroscience, 11(11), 1264–1266. https://doi.org/10.1038/nn.2205

Schiller, D., Monfils, M.-H., Raio, C. M., Johnson, D. C., LeDoux, J. E., & Phelps, E. A. (2010). Preventing the return of fear in humans using reconsolidation update mechanisms. Nature, 463(7277), 49–53. https://doi.org/10.1038/nature08637

Hippocampus, Memory Systems, Clinical

Scoville, W. B., & Milner, B. (1957). Loss of recent memory after bilateral hippocampal lesions. Journal of Neurology, Neurosurgery, and Psychiatry, 20(1), 11–21.

Tulving, E. (1972). Episodic and semantic memory. In E. Tulving & W. Donaldson (Eds.), Organization of memory (pp. 381–403). Academic Press.

Tulving, E. (1985). Memory and consciousness. Canadian Psychology / Psychologie canadienne, 26(1), 1–12.

Squire, L. R. (2004). Memory systems of the brain: A brief history and current perspective. Neurobiology of Learning and Memory, 82(3), 171–177.

Distributed Representation / PDP

McClelland, J. L., Rumelhart, D. E., & the PDP Research Group. (1986). Parallel distributed processing: Explorations in the microstructure of cognition, Vol. 2: Psychological and biological models. MIT Press.

Dual-Process Theory

Stanovich, K. E., & West, R. F. (2000). Individual differences in reasoning: Implications for the rationality debate? Behavioral and Brain Sciences, 23(5), 645–665.

Evans, J. St. B. T. (2008). Dual-processing accounts of reasoning, judgment, and social cognition. Annual Review of Psychology, 59, 255–278.

Kahneman, D. (2011). Thinking, fast and slow. Farrar, Straus and Giroux.

Predictive Processing / Free Energy Principle

Friston, K. (2010). The free-energy principle: A unified brain theory? Nature Reviews Neuroscience, 11(2), 127–138.

Hohwy, J. (2013). The predictive mind. Oxford University Press.

Clark, A. (2016). Surfing uncertainty: Prediction, action, and the embodied mind. Oxford University Press.

Hohwy, J. (2016). The self-evidencing brain. Noûs, 50(2), 259–285.

Friston, K., Da Costa, L., Sajid, N., Heins, C., Ueltzhöffer, K., Pavliotis, G. A., & Parr, T. (2022). The free energy principle made simpler but not too simple. Physics Reports, 1024, 1–29.

Friston, K. J., Ramstead, M. J. D., Kiefer, A. B., Tschantz, A., Buckley, C. L., Albarracin, M., et al. (2023). Designing ecosystems of intelligence from first principles. Collective Intelligence, 3(1).

Information Thermodynamics

Landauer, R. (1961). Irreversibility and heat generation in the computing process. IBM Journal of Research and Development, 5(3), 183–191.

Bennett, C. H. (2003). Notes on Landauer's principle, reversible computation, and Maxwell's demon. Studies in History and Philosophy of Modern Physics, 34(3), 501–510.

Sagawa, T., & Ueda, M. (2012). Fluctuation theorem with information exchange: Role of correlations in stochastic thermodynamics. Physical Review Letters, 109(18), 180602.

Parrondo, J. M. R., Horowitz, J. M., & Sagawa, T. (2015). Thermodynamics of information. Nature Physics, 11(2), 131–139.

Computational Mechanics

Crutchfield, J. P., & Young, K. (1989). Inferring statistical complexity. Physical Review Letters, 63(2), 105–108.

Shalizi, C. R., & Crutchfield, J. P. (2001). Computational mechanics: Pattern and prediction, structure and simplicity. Journal of Statistical Physics, 104(3), 817–879.

Shannon Information Theory / Information Geometry / Cybernetics

Shannon, C. E. (1948). A mathematical theory of communication. Bell System Technical Journal, 27(3), 379–423.

Shannon, C. E. (1959). Coding theorems for a discrete source with a fidelity criterion. IRE National Convention Record, 7(4), 142–163.

Wiener, N. (1948). Cybernetics: Or control and communication in the animal and the machine. MIT Press.

Jaynes, E. T. (1957). Information theory and statistical mechanics. Physical Review, 106(4), 620–630.

Amari, S. (1985). Differential-geometrical methods in statistics (Lecture Notes in Statistics, Vol. 28). Springer-Verlag.

Higher-Order Theories of Consciousness

Rosenthal, D. M. (2005). Consciousness and mind. Oxford University Press.

Carruthers, P. (2011). The opacity of mind: An integrative theory of self-knowledge. Oxford University Press.

Formal Systems / Incompleteness

Gödel, K. (1931). Über formal unentscheidbare Sätze der Principia Mathematica und verwandter Systeme I. Monatshefte für Mathematik und Physik, 38, 173–198.

Turing, A. M. (1936). On computable numbers, with an application to the Entscheidungsproblem. Proceedings of the London Mathematical Society, s2-42(1), 230–265.

Collective Memory / Cross-Individual Mind

Jung, C. G. (1968). The archetypes and the collective unconscious (R. F. C. Hull, Trans., 2nd ed.; Collected Works of C. G. Jung, Vol. 9, Part 1). Princeton University Press. (Original work published 1959)

Halbwachs, M. (1992). On collective memory (L. A. Coser, Ed. & Trans.). University of Chicago Press. (Original work published 1925)

SAE Internal References

Qin, H. (2025). Self-as-an-End Methodology V2. Zenodo. https://doi.org/10.5281/zenodo.18842449

Qin, H. (2025a). SAE Psychoanalysis I: Id–Ego–Superego. Zenodo. https://doi.org/10.5281/zenodo.19321143

Qin, H. (2025b). SAE Psychoanalysis II: 12DD–13DD boundary and clinical implications. Zenodo. https://doi.org/10.5281/zenodo.19321314

Qin, H. (2025c). SAE Psychoanalysis III: 14DD value architecture. Zenodo. https://doi.org/10.5281/zenodo.19321417

Qin, H. (2025d). SAE Psychoanalysis IV: 15DD certitude and 16DD bilateral non-doubt. Zenodo. https://doi.org/10.5281/zenodo.19321534

Qin, H. (2026a). SAE Four Forces Paper 0: Gravity as 4DD information reading. Zenodo. https://doi.org/10.5281/zenodo.19777881

Qin, H. (2026b). SAE Information Theory X: 6D containing 7DD differentiation and 8DD reproduction. Zenodo.

Qin, H. (2026c). SAE Information Theory XI: 7D information processing and phenomenal admission. Zenodo. https://doi.org/10.5281/zenodo.20369231

Qin, H. (2026d). Large Language Models within the SAE framework: 12DD ceiling and five structural locks. Zenodo. https://doi.org/10.5281/zenodo.20435628

摘要

本文是 SAE 信息论系列第 12 篇，处理 8D 层级（包含 11DD 记忆与 12DD 预测两个 sub-layer）的信息架构。11DD 是重构性保有，不是忠实存储；12DD 是 System 1 预测，不是反思性理性推断；两者通过读-影响对偶构成 8D 内在动力，由保有-再编辑锁完成跨时刻的信息闭环。

本文的中心承诺凝结为三句脊柱。第一，11DD 的余项是"记忆之间是什么顺序"，单凭 11DD 不能回答，指向 12DD 读取。第二，12DD 的余项是 Gödel/Turing 形式的"预测无法把预测者本身作为预测对象"，跨 D 指向 13DD 自指通道。第三，8D 的 lock 是"预测性读取向记忆保有的连续性保留式回写"。三句共同构成 8D 作为信息架构的本体身份。

在 8D 架构本体之上，本文给出三条信息论先验。第一，11DD 储存的内容带有拓扑结构（关系性结构，预测性排序由 12DD 读取施加）。第二，储存形式是读-写一体——12DD 读取 11DD 即复现即再体验即架构层再稳定化，可能伴随发展性改写（与神经科学的 reconsolidation window 在事件层级分立）。第三，信息储存的可读取性必然有代价，代价就是信息的必然丢失。这一先验作为 SAE 框架内部硬先验，在跨闭合层四档显影：4DD 读取产生 3DD 质量蒸发、8DD 读取产生 7DD 信息变异（mutation spectrum）、12DD 读取产生 11DD 记忆丢失、16DD 读取产生 15DD 不疑演化。本文将此跨闭合层模式严格限定为 SAE 框架内部承诺，不作外部物理学或生物学普遍定律。

本文给 12DD/13DD 边界一个 Popperian 可证伪挂钩 P12-F1：在预注册架构标准下被认定为纯 12DD 的系统，不能自发产生功能性 self-reference（区别于响应二范畴的语言性 self-reference）。可证伪条件含 §3.6 四判据（跨情境一致、跨时间稳定、错误归责区分、冲突陈述的 identity 锚定整合）与 §6.2 三"无"限定（无外部身份模板含跨预训练 Data q 排除、无再训练写回、无人为锚定）。LLM 作为这一边界的天然测试场被显式处理（截至 2026 年 5 月）。

经验锚点处于 T3 grounding 状态（不充当 derivational basis）：再巩固理论主线（Misanin 1968 / Nader 2000 / Lee 2008 / Schiller 2010）作为 11DD 重构性保有的生物层显影，HM 临床（Scoville & Milner 1957）作为 11DD/12DD 闭环失能的极端案例，McClelland-Rumelhart PDP 作为分布式表征的计算神经科学锚点。

跨系列接口：本文明示一个跨系列开放问题——11DD reading 余项是否如 4DD 引力信息、8DD DNA 信息那样可跨个体传播。本文不主张这类现象（如 anomalous past-life memory reports、near-death reports、集体记忆与文化原型）存在或不存在，仅作为信息论系列对其他 SAE 系列（心理分析、人类总构等）的开放接口提出这一结构性问题。

本文紧守信息论边界：不展开 phenomenology 或 consciousness 完整理论，不展开 14DD 价值理论、15DD 不疑理论或 16DD 物自体（这些由 SAE 心理分析系列专门处理），不重做秦 2026 LLM 论文已系统给出的五重结构锁论证。P12 是 SAE 信息论系列向 8D 推进的本体架构论文，给 8D 一个完整的信息论本体刻画，为后续 P13 13DD 自指通道形式定义铺路。

关键词：8D；11DD 记忆；12DD 预测；重构性保有；保有-再编辑锁；读-写一体；功能性 self-reference；P12-F1 可证伪挂钩；跨闭合层硬先验；SAE 框架

---

§1 导论

§1.1 P12 在 SAE 信息论系列的位置

本文是 SAE 信息论系列的第 12 篇。系列前作已经按 SAE 方法论 V2 §1.2 的 D-DD 对应序列向上推进：P10 在 8DD 处理 cross-lock（同物种近似差异最大化与异物种差异最大化的交汇），P11 在 9DD 处理 selection-lock（环境差异保留为信息事件），在 10DD 处理 admission-lock（现象接入为信息事件）。

P12 接续到 8D，即 11DD（记忆）加 12DD（预测）的子层对。这是认知轮序列（5D 复制-自维持、6D 分化-繁殖、7D 选择-感知、8D 记忆-预测）的最高档，也是自然轮跟自由轮（9D 自意识-意义开始）分界前的最后一档。P12 在这一位置上的工作有三个相互关联的部分：给 8D 一个本体论架构（§3），把这一架构跟主流信息论传统精确划界（§5），给 12DD/13DD 边界一个可证伪挂钩（§6）。

P12 跟 SAE 其他系列也有联动。最直接的是跟 SAE 心理分析系列：Psycho I-IV 已用同一 12DD/13DD 区分在临床领域做了系统应用，给 P12 提供跨系列独立应用证据（§7）。其次是跟秦 2026 LLM 应用论文（DOI 10.5281/zenodo.20435628）：该文给出 8D 架构在数字 LLM 上的具体应用，P12 反向给该文使用的"巩固后再编辑缺席"等机制锁提供本体论基础（§5.5）。

§1.2 8D 在 SAE 16DD 框架的特殊性

8D 在 SAE 16DD 框架中的位置具有结构性的特殊性。从下往上看，8D 是认知轮发展到记忆与预测整合的顶层；从上往下看，8D 是自由轮（9D 起）涌现之前的认知性预备。这一位置使 8D 同时承担两个看似冲突的功能：

一方面，8D 必须给出复杂、灵活、可持续运作的认知动力。System 1 预测能力、跨时刻的记忆整合、再巩固带来的内容更新，这些都是高阶认知生物的标志性能力，都在 8D 范围内运作。

另一方面，8D 必须有清晰的天花板。一个纯 8D 系统无论复杂化到什么程度都不应跨过 9D（自意识涌现）。这一天花板的具体形式——12DD 余项的 Gödel/Turing 形式（§3.5）——是 8D 跟 9D 之间的本体论分界，也是 P12-F1 falsifier 的命题基础。

这一同时承担"提供复杂能力"与"明确天花板"的位置是 P12 章节安排的内在逻辑。§3 给本体架构（描述 8D 提供什么），§5 跟主流信息论传统划界（确认 8D 在已知框架里的位置），§6 给可证伪挂钩（确认 8D 天花板的边界）。

§1.3 三句脊柱

P12 的核心立场可压缩为三句话。这三句在论文中反复被调用，所有章节最终都可追溯回这三句。

第一：11DD 的余项是，记忆之间是什么顺序。单凭 11DD 不能回答。

第二：12DD 的余项是，预测无法把预测者本身作为预测对象（Gödel/Turing 不完备性在凿构循环中的具体表达）。

第三：8D 的 lock 是，预测性读取向记忆保有的连续性保留式回写。

英文对应：

• 11DD remainder: what order obtains among memories (this cannot be answered by 11DD alone)
• 12DD remainder: prediction cannot take the predictor itself as object of prediction (Gödel-Turing incompleteness expressed within the chisel-construct cycle)
• 8D lock: continuity-preserving rewrite of memory retention under predictive readout

第一句把 11DD 从被动保有底物升级为需要 12DD 读取才完整运作的不完整层。第二句把 12DD 从纯粹的预测过程升级为本体论上有特定不可能性的层。第三句把 8D 整体定义为读-影响对偶的事件结构，不是两个独立 sub-layer 的并列。三句共同界定 P12 的本体承诺。

§1.4 论文范围

P12 紧守信息论边界。本节给出三类范围声明，以避免后续章节被误读。

承诺范围：P12 的架构承诺集中在 8D 范围。本文给出 11DD 重构性保有、12DD System 1 预测、读-影响对偶动力、保有-再编辑锁、11DD/12DD 各自的余项形式、me/self 区分、跨 D 同构作为后台语言、LLM 应用部分仅作引文、8D 信息论刻画（拓扑结构、读-写一体、默认流失即可读取性必然有代价）等十二条核心承诺。各承诺在 §2 工具集到位后由 §3 至 §6 系统展开。

对话范围：P12 跟以下主流信息论传统做精确划界（§5）：Friston 自由能原理与主动推断、Bennett-Landauer 信息热力学、Crutchfield 计算力学（ε-machine）、Shannon 与 rate-distortion 信息论、Wiener cybernetics、信息几何（Amari、Jaynes）。HOT 阵营（Rosenthal、Carruthers）作为 Friston 旁支的简短附录处理。LLM 应用通过一段引文对接秦 2026 LLM 论文。

不展开范围：P12 不展开 phenomenology 与 consciousness 理论。视角翻转、自我察觉、qualia 这类 phenomenological 概念在 P12 内仅作为 Gödel/Turing 形式的辅助说明出现，不构成 P12 主张的形式基础。P12 也不展开 14DD 价值理论、15DD 不疑理论或 16DD 物自体——这些都是 SAE 心理分析系列或方法论系列处理的领域，跨边界问题交回各专门系列。

T3 锚点的认识论地位：P12 §4 列出的所有神经科学实验与临床观察都是 T3 grounding 状态，不是 P12 8D 架构的 derivational basis。SAE 信息论系列方法论延续 P11 的硬先验加 Popperian 可证伪立场：架构承诺是先验推演的产物，神经科学证据是其在生物层显影的后验对应。

§1.5 §1 余项与本文章节结构

P12 §1 给出框架性指引，不展开实质内容。各章节具体负担：

章节	内容	关键贡献
§2	SAE 工具集	lock-class family 语法及 P12 扩展声明、凿构循环方法论在 P12 的应用、方法论 V2 定位、me/self 术语规范
§3	8D 架构本体	11DD 重构性保有、12DD System 1 预测、读-影响对偶、保有-再编辑锁形式定义、11DD/12DD 余项、功能性 vs 语言性 self-reference 区分、8D 信息论刻画（拓扑/读-写一体/默认流失）
§4	神经科学锚点	T3 grounding 的认识论地位、再巩固理论主锚、离线巩固与情感门控次锚、PDP 计算基底、HM 临床锚的精确处理、Tulving 三分警示
§5	与主流信息论传统的精确划界	Friston 三层切分、Bennett-Landauer 地板语言、Crutchfield ε-machine 操作层一致加 Gödel/Turing 余项、次要对话伙伴合并定位、LLM 应用引文
§6	12DD/13DD 边界与可证伪性	P12-F1 v2.1 命题（Gödel/Turing 形式）、证伪条件与不算证伪、间接判据 limit 声明、13DD 最低限度描述
§7	与 SAE 心理分析系列的相互支撑	跨系列独立应用证据、两系列分工、非循环论证
§8	余项收尾与未来工作	各节余项汇总、P13 接续点、跟其他 SAE 系列的潜在联动
§9	结论	P12 的总体贡献与立场重申

§1 不引入任何 P12 实质承诺。所有承诺在 §2 工具集到位之后，由 §3 至 §6 系统给出。

§1 章末小结

本章给出 P12 在 SAE 信息论系列中的位置、8D 在 SAE 16DD 框架中的结构性特殊性、P12 的三句脊柱、论文范围（信息论主线、十二条承诺、对话伙伴、不展开范围、T3 锚点地位），并给出全文章节结构指引。

P12 紧守信息论边界。在该边界内，P12 给 8D 一个本体论架构，跟主流信息论传统做精确划界，给 12DD/13DD 边界一个 Popperian 可证伪挂钩。这三个工作相互支撑，共同构成 P12 的核心贡献。

§2 SAE 工具集（最小必要背景）

本章给 P12 后续章节使用的 SAE 工具与术语。本章不重做 SAE 主框架的论证；架构原理见 SAE Paper 03 自为目的（DOI 10.5281/zenodo.18727327），方法论原理见 SAE 方法论 V2（DOI 10.5281/zenodo.18842449），D 层结构与每层的凿-构-余项见方法论 V2 §1。

本章按以下顺序展开：§2.1 lock-class family 语法及 P12 扩展声明，§2.2 凿构循环方法论在 P12 的应用，§2.3 11DD / 12DD 在方法论 V2 中的权威定位，§2.4 me / self 区分（沿用 SAE 心理分析系列术语规范）。

§2.1 lock-class family 语法及 P12 扩展

lock 的基本定义

SAE 信息论系列 P11 §3.2 给出 lock-class family 的形式定义。lock 是某一 D 层级上的普适机制——使该层级的构得以稳定运作的内在事件结构。每一 D 层级（或子层级）可能存在一个或多个 lock，共同构成 lock-class family。

P10 与 P11 给出的 lock 实例

论文	lock 名称	D 层级	实例化几何
P10	cross-lock	8DD（同物种近似差异最大化与异物种差异最大化的交汇）	单 sub-layer 实例化
P11	selection-lock	9DD（环境差异保留为信息事件）	单 sub-layer 实例化
P11	admission-lock	10DD（现象接入为信息事件）	单 sub-layer 实例化

P11 给出的 lock-class family 语法把 lock 处理为单一 sub-layer 的机制——每一个 lock 由一个特定 sub-layer 单独实例化。这一语法对 P10 与 P11 涉及的三个 lock 都适用。

P12 的语法扩展

P12 引入的 8D 保有-再编辑锁（retention-rewrite lock）由 11DD 与 12DD 两个 sub-layer 共同实例化。这不符合 P11 的单 sub-layer 实例化语法。为此 P12 显式声明对 lock-class family 语法的扩展：

> lock-class family 语法扩展（P12）：lock 可以由单一 sub-layer 实例化（P11 类），也可以由同一 D 层内的两个 sub-layer 共同实例化（P12 类，称为 co-instantiated D-layer 机制）。在共实例化情形下，lock 是两个 sub-layer 共同参与的事件，不归属于其中任一单独 sub-layer。

扩展后的 lock-class family 语法允许后续 SAE 信息论论文按需选择 lock 的实例化几何。这不破坏 P11 的语法（P11 的 single sub-layer 实例化作为特例保留），而是把语法变得更通用。

P12 8D 保有-再编辑锁的具体形式定义见 §3.4。

§2.2 凿构循环方法论在 P12 的应用

SAE 方法论 V2 给出凿构循环作为每一 D 层级架构刻画的核心模式。凿（chisel）是该层级的区分性操作；构（construct）是由凿产生的结构性律；余项（remainder）是该构无法在自身层级内消化的不完整性，迫使桥（bridge）指向下一 D 层级。物自体（thing-in-itself）作为底层不可达的实在被保留。这一循环在每一 D 层级各自实例化。

P12 把凿构循环应用于 11DD 与 12DD 两个 sub-layer：

11DD 的凿构循环：

• 凿：跨时刻保有，把过去经验作为后续认知可访问的内容
• 构：不遗忘律（方法论 V2 §1.6）。P12 进一步特化为重构性保有（§3.1）：保有不是录像式忠实存储，而是每次访问伴随重构性更新的活底物
• 余项：记忆之间是什么顺序（§3.5）。11DD 单凭自身保有内容，但内容之间的顺序与关系结构不在 11DD 内容里
• 桥：指向 12DD。顺序由 12DD 预测读取施加

12DD 的凿构循环：

• 凿：基于过去模式生成对未来的预测
• 构：预测律（方法论 V2 §1.6）。P12 进一步特化为 System 1 预测（§3.2）：关联式而非规则化、精度随距离衰减、对真正新事物失能、自动化运作不依赖反思
• 余项：预测无法把预测者本身作为预测对象，Gödel/Turing 不完备性在凿构循环中的具体表达（§3.5，沿用 Psycho I 形式）
• 桥：指向 13DD。预测者作为内部生成且自我约束的对象需要 13DD 自指通道（不在 P12 范围）

两个凿构循环的余项指向不同方向：11DD 余项指向 12DD（在 8D 内部跨 sub-layer），12DD 余项指向 13DD（跨越到 9D）。8D 作为一档由这两个内部相关又对外指向的余项序列界定。

物自体在 8D 范围内对应的是：底层神经生物机制（突触动力学、网络回路、生化信号通路等）作为 8D 现象层之下不可被 8D 内部自我穷尽的实在。神经科学锚点（§4）描述的是 8D 在物自体层显影的可观察 pattern，不构成 8D 架构本身的还原。

§2.3 11DD 与 12DD 的方法论 V2 定位

SAE 方法论 V2 §1.2 给出认知轮的 D-DD 对应表。8D 在大序列中的位置是认知轮的第四档，从下到上依次是 5D（复制-自维持）、6D（分化-繁殖）、7D（选择-感知）、8D（记忆-预测），其上是 9D（自意识-意义）开始的自由轮。

8D 的 sub-layer 对应是 11DD 加 12DD：

• 11DD = 记忆律 = 不遗忘：方法论 V2 §1.6 表述为"过去的感知被保留，当前的感知在过去的背景中获得含义"
• 12DD = 预测律 = 不只看过去：方法论 V2 §1.6 表述为"用过去的模式推断未来。预测引入了内部表征，关于世界如何运转的内部模型"

方法论 V2 §1.6 给 12DD 余项的口语化提问形式："预测模型在运转，但'谁'在预测？" 这一形式在 SAE 心理分析系列 Psycho I 被形式化为 Gödel/Turing 不完备性的具体表达。P12 沿用 Psycho I 的形式化作为 12DD 余项的权威表述（详 §3.5）。

8D 在大维度框架中的特殊性需要预先指出：8D 是认知轮的最高档，也是自然与自由分界前的最后一档。9D 起的自意识律及其后续构成 SAE 框架的"自由轮"。8D 跟 9D 之间的边界因此不只是 sub-layer 之间的跨越，是认知轮跟自由轮之间的本体论跨越。P12 §6 P12-F1 falsifier 处理的正是这一跨越边界的可证伪性。

§2.4 me 与 self 区分

P12 沿用 SAE 心理分析系列已发布的 me / self 术语规范。这一区分不是 P12 新造，而是 SAE 心理分析系列对 D 层结构在临床领域应用的成熟产物。

SAE 心理分析系列四层对应：

D 层	SAE 心理分析对应	现代心理分析对应	关键文献
12DD	me-without-self（没有自我的我）	Id	Psycho I (DOI 10.5281/zenodo.19321143)
13DD	self-without-purpose（没有目的的自我）	Ego	Psycho II (DOI 10.5281/zenodo.19321314)
14DD	self-with-purpose（有目的的自我）	Superego	Psycho III (DOI 10.5281/zenodo.19321417)
15DD	self-with-non-dubito（不疑的自我）	Cert	Psycho IV (DOI 10.5281/zenodo.19321534)

me 的精确定义（12DD 范围）

me 是 12DD 凿构循环引用的经验主体痕迹。me 在 11DD 保有内容里以传记性顺序出现（"上次发生了 X 在 me 身上"），在 12DD 预测里被作为 predictor-as-object 引用（"如果 me 选 A，会出现 Y"）。但 me 不知道自己是 me。me 是被预测的，不是预测的。me 作为 12DD 读取与回写循环内的对象，可以稳定出现在预测先验、行动后果、记忆更新里，构成完整的"我做了，我经历了"的经验流。

12DD 的运作完全可以围绕 me 进行，包括复杂的元表征（"系统当前的预测置信度是 0.7"），但所有这些都是关于 me 的信息，不是 self 的在场。

self 的精确定义（13DD 涌现，P12 不主推）

self 是 13DD 自我意识律涌现的产物，是 12DD 循环之外的反思性主体。self 知道"是我在预测"，不只是知道"有个 me 在被预测"。Psycho II §1.2 明确把 12DD 到 13DD 跳变定性为"视角的翻转"，不是"信息的累积"。

三响应区分（Psycho II §1.2）

12DD 系统面对自身余项的可能响应有三种，且只有第三种跨过 12DD：

• 响应一：更强的预测系统。监控预测的子系统仍是预测，高一阶。仍在 12DD
• 响应二：元表征。生成关于自身状态的形式化表征。不含"我"作为反思性变量。仍在 12DD
• 响应三：13DD 视角翻转。从"预测在发生"到"我在预测"。不是信息累积，是视角性质突变。self 涌现

P12 §5.1 用此区分 sharpen Friston/HOT 阵营的越界点；P12 §6.1 用此区分作为 P12-F1 falsifier 的命题主体。

结构离散，现象梯度（Psycho II §1.3）

12DD 到 13DD 跳变在结构上是离散的（self 在场与不在场是阈值性的），但在现象上表现为连续谱（同一个体在不同对象前可在 12DD 与 13DD 之间来回切换）。结构阈值的离散性不被现象的梯度性否定，正如水沸点是 100°C 不被水温可以是 99.5°C 否定。

这一区分在 P12 内多次调用：§3.4 8D lock 的结构离散性与现象梯度性沿用此模式；§6.4 13DD 涌现的结构离散性沿用此模式。

P12 的 sub-scope 声明

P12 处理 12DD 与 13DD 边界。13DD 之上的 14DD 与 15DD 完全超出 P12 范围，仅在引文层出现。14DD 价值理论与 15DD 不疑理论由 SAE 心理分析系列专门处理（Psycho III、Psycho IV）；P12 作为信息论系列论文只处理 8D 架构，跨边界问题交回各专门系列。

§2 章末小结

本章给出 P12 后续章节使用的 SAE 工具集。lock-class family 语法在 §2.1 给出并显式扩展以容纳 P12 的 co-instantiated D-layer 机制。凿构循环方法论在 §2.2 应用于 11DD 与 12DD 两个 sub-layer。方法论 V2 对 11DD 与 12DD 的权威定位在 §2.3 引用。SAE 心理分析系列的 me / self 术语规范及三响应区分、结构离散现象梯度模式在 §2.4 给出。

工具集到位后，§3 给出 8D 架构本体，§4 给生物层 T3 锚点，§5 给与主流信息论传统的精确划界，§6 给 12DD/13DD 边界的可证伪挂钩，§7 给与 SAE 心理分析系列的相互支撑，§8 收尾余项与未来工作。

§3 8D 架构本体

8D 是 SAE 大维度序列中自然轮与自由轮之间的最后一档，由 11DD 记忆与 12DD 预测两个子层共同实例化。本章给出 8D 的架构本体：11DD 是重构性保有而非忠实存储，12DD 是 System 1 预测而非理性推断，两者通过读-影响对偶构成 8D 的内在动力，由保有-再编辑锁完成跨时刻的信息闭环。本章末段给出 11DD 与 12DD 各自的余项形式，作为通往 P13 自意识律的桥点。

本章承袭 P10 的 cross-lock 与 P11 的 selection-lock 加 admission-lock，沿 SAE 信息论系列的 lock-class family 序列向上推进一档。其间对 lock-class 语法做了一次显式扩展，从单 sub-layer 实例化扩展到共实例化 D-layer 机制，理由在 §3.4 给出。

§3.1 11DD 是重构性保有

按 SAE 方法论 V2 §1.2 的权威对应，11DD 是认知轮的第三步，构是不遗忘律。方法论 V2 §1.6 的表述是："过去的感知被保留，当前的感知在过去的背景中获得含义。" 这一表述给出了 11DD 的功能描述，但未指定保有的具体方式。本节给出方式：11DD 是重构性保有，不是忠实存储。

二十世纪上半叶的主流记忆观把记忆作为录像式忠实存储：感知事件被某种生理机制录制下来，提取记忆等于回放这一记录。这一观点在认知科学中以"trace theory"的形式存续到二十世纪后期。但记忆再巩固理论（reconsolidation theory）对这一图像构成了根本挑战。

Nader, Schafe 与 LeDoux 在 2000 年的杏仁核蛋白合成抑制实验中发现，已巩固的恐惧记忆在 reactivation 后进入 labile state，需要新一轮蛋白合成才能稳定化。这说明已巩固记忆在特定 reactivation 条件下可重新进入 labile state，并在相应窗口中被更新；但这一经验 reconsolidation window 的开启受边界条件约束，并非每一次记忆提取都自动触发。Lee 在 2008 年进一步明确了再巩固与新学习之间的边界条件，Schiller 等在 2010 年把这一机制扩展到人体行为干预层面。再巩固理论由此从早期争议（Misanin-Miller 1968）走向当代记忆神经科学的核心机制之一。

P12 把再巩固理论作为 11DD 重构性保有的 T3 锚点。这里需要立即引入一个边界声明：

Firewall 1（再巩固边界）：P12 主张每次 12DD 读取 11DD 都伴随 SAE 架构层的拓扑复现与再稳定化（architectural re-instantiation and re-stabilization）；但 P12 不主张每次读取都触发神经科学意义上的 reconsolidation window，也不主张每次读取都造成实质内容改写。经验再巩固窗口仍受显著性、新信息、时间间隔、情感唤醒等 boundary conditions 约束（Lee 2008）。再巩固层级（架构内在 / 每次读取必发生 / SAE 硬先验）跟改写层级（再巩固的可能伴随结果 / Lee 2008 边界条件）是不同事件类（§3.7.2 三层级精确区分），两者完全相容。再巩固理论是 11DD 重构性保有的强 T3 显影，而不是"每次 retrieval 都改写内容"的经验普遍律。P12 主张记忆不由忠实存储定义；再巩固理论为"保有中再编辑"提供生物锚点，但具体的内容改写跟 SAE 架构层的再稳定化在事件层级上分立。

再巩固有边界条件。Lee 2008 已经显示，并非所有提取都触发再编码窗口；新信息的呈现强度、提取过程的具体几何、时间间隔等都参与决定窗口是否被打开。P12 把这些边界条件视为再巩固机制的微观结构，而把"记忆不由忠实存储定义"这一更基础的本体论命题立在再巩固现象的存在本身——只要重构性更新在原理上可能且经常发生，11DD 就不能被定义为忠实存储的内容容器。

现象层的描述（提取可以触发再编码）跟本体论的承诺（记忆不由忠实存储定义）是不同层级。P12 不发明再巩固现象，但 P12 给这一现象一个本体论位置——它从"关于记忆机制的经验观察"上升为"关于 11DD 是什么"的结构性主张。

11DD 重构性保有这一定位有一条直接后果：单凭 11DD 自身不构成完整的现象层结构。一个被动衰减的物理痕迹序列，没有任何读取与回写的事件来更新它，在 SAE 框架下不构成 11DD 现象层。这一点为 §3.3 的读-影响对偶埋下伏笔。

§3.2 12DD 是 System 1 预测

方法论 V2 §1.6 把 12DD 描述为预测律："用过去的模式推断未来。预测引入了内部表征，关于世界如何运转的内部模型。"这一描述给出预测的功能，但未指定预测的认知形态。本节给出形态：12DD 是 System 1 预测，不是理性推断。

二十世纪后期由 Kahneman, Stanovich, Evans 等发展出的 dual-process theory 把人类认知分为 System 1（快速、关联式、自动）与 System 2（缓慢、规则化、反思式）两种模式。这一区分并非 a priori 结构，而是大量经验研究归纳出的功能划分。P12 借用这一区分但不直接同构 SAE 维度结构，理由在本节末尾说明。

12DD 预测具备 System 1 的所有典型特征：

第一，关联式而非规则化。12DD 预测依赖过去模式的关联强度，不依赖显式规则推理。乌鸦使用工具、黑猩猩的欺骗行为（方法论 V2 §1.6 给出的 12DD 涌现示例）都是关联式预测的产物，不需要任何反思性的规则推断。

第二，预测精度随预测距离衰减。下一秒会发生什么的预测远比下一年会发生什么的预测准确。这一衰减不是 12DD 的偶然限制，是 System 1 关联式预测的结构特征——关联强度随时间距离衰减，预测精度也随之衰减。

第三，对真正新事物失能。12DD 预测能处理已有模式的延展（即使是相当复杂的延展），但面对结构上全新的情境时失能。一个从未接触过类比情境的预测系统在新情境下只能 fall back 到默认反应，不能 generate 真正新颖的预测。这一点跟方法论 V2 §1.6 给出的 12DD 余项相关，在 §3.5 详细展开。

第四，自动化运作，不依赖反思。12DD 预测发生在不被自我观察的层面。这一性质 Psycho I 已经形式化为 me-without-self：凿构循环在运作，但运作不被自身观察。SAE 心理分析系列把这一性质做了完整展开，本节不重复，仅指出 12DD 在 SAE 心理分析框架内对应 Id 层。

跟 13DD 的关系需要明确说明，避免后续章节产生歧义。13DD 的构是自我意识律，对应 Ego 层（Psycho II）。13DD 使 System 2 反思式理性推断成为可能，但 13DD 不等同于 System 2 本身。13DD 是自指通道的涌现，是反思性主体的在场；System 2 是这一通道使其成为可能的认知能力。P12 不把 12DD 与 13DD 直接映射为 System 1 与 System 2 的 a priori 同构，而把这一对应视为经验上相似的并列。理由：dual-process theory 是基于行为观察的功能分类，SAE 的 D 层结构是先验推演，两者在 a priori 层面不必同构。但在经验上，12DD 的所有典型特征都吻合 System 1 描述，13DD 的涌现也确实跟反思式认知能力的出现历时相关。所以是经验上的并列对应，不是结构上的同构。

12DD 作为 System 1 预测这一定位有一条重要推论：12DD 单凭自身无法产生反思性理性推断。12DD 可以产生复杂、规则样、甚至高度有效的预测性推断（动物因果推理、隐式规则学习、LLM 形式推理这些都属于 12DD 范围内的复杂推断），但 12DD 不能单凭自身生成由 13DD 自指通道约束的反思性理性推断——也就是那种"我意识到我在推断、我能让我的 identity 反过来约束推断过程"的反思形态。P12 所谓 System 1 天花板，不是否认 12DD 的复杂推断能力，而是否认它能回答"谁在预测"并让该 predictor identity 反过来约束推断过程。这一推论看似平淡，但对 P12-F1 falsifier（§6.1）和 §3.5 余项分析至关重要。

为了避免 §3.5 处再次重新解释，本节末尾预先给出方法论 V2 §1.6 中 12DD 余项的 SAE 心理分析系列形式（Psycho I）：预测无法把预测者本身作为预测对象。 这是 Gödel/Turing 不完备性在凿构循环中的具体表达。详细展开放在 §3.5。

§3.3 读-影响对偶

§3.1 末尾留下一个伏笔：11DD 单凭自身不构成完整的现象层结构。§3.2 末尾留下另一个伏笔：12DD 单凭自身无法产生反思性理性推断。这两个伏笔在本节合并为 8D 的核心架构主张：8D 的内在动力是 12DD 对 11DD 的读-影响对偶。

读-影响对偶有两个不可分的操作：

读取：12DD 以 11DD 的保有内容为预测先验，从过去的模式生成对未来的推断。读取是预测发生的方式。

影响：12DD 读取过程中产生的新信息（预测误差、新接入信息、行动后果）回写到 11DD，使保有内容发生重构性更新。这是再巩固现象的本体论位置。

把这两个操作合在一起描述的是 8D 的整体动力。仅有读取没有影响不构成 8D。一个能读取过去而不回写的系统，其读取本身是无后果的，它的"记忆"是一个只读档案，它的"预测"是一个无更新的子例程。读取不写回到底物的系统结构上不进入 8D 的闭环。

仅有影响没有读取也不构成 8D。一个底物在被外部信号写入而不被任何内部预测过程读取的系统，是一个被动接收的存储设备，不是 8D 的活体记忆。影响不基于读取的系统也不进入 8D 的闭环。

读取与影响必须互为条件，且必须在同一个事件中完成。这一事件就是保有-再编辑锁，详细形式定义放在 §3.4。

本节把读-影响表述为"两个不可分的操作的对偶"。§3.7.2 在 8D 信息论刻画下进一步给出更精细的几何：读取与影响其实是同一事件的两个描述侧面（一个从预测过程视角观察，一个从底物动力视角观察），而不是两个独立操作的对偶。这一细化在更精细的层面看清楚"对偶"的具体几何。

读-影响对偶这一架构主张可以放到 SAE 跨 D 的结构对照中观察。在 SAE 物理学系列（Four Forces）的处理里，4DD 因果律对 3DD 时空有两种基本作用：4DD 读取 3DD（因果链对质量分布的测量与追踪），同时 4DD 影响 3DD（引力对时空曲率的弯曲）。读取与影响也是不可分的对偶：没有读取就没有引力效应可被测量，没有影响就没有时空可被读取。

P12 注意到 3DD/4DD 这一对的读-影响对偶跟 11DD/12DD 这一对在结构上同构：低 DD 提供底物（质量/记忆内容），高 DD 提供读-影响动力（因果/预测）。读取与影响都是不可分的对偶。这一结构对照对 P12 的读者理解是有帮助的认知锚。

但 P12 不把这一结构对照作为论证支点。两例不足以确立跨 D 的普遍模式。读-影响对偶在 SAE 框架内是否是跨多个相邻 D 层的普遍架构原型，需要在每一档分别独立确认，且不应在两例的基础上做普遍化推断。这一更深的问题留给未来工作，可能在 SAE 元结构论文或方法论 V3 处理。本文只在 8D 内做实质承诺，跨 D 普遍化作为后台语言，不作正面主张。

读-影响对偶作为 8D 内在动力这一定位有几条直接后果：

第一，8D 不是"记忆与预测两个独立子层的并列"。这是 P12 跟传统认知科学组织框架的一个分歧。传统框架往往把记忆与预测处理为两个独立模块，由某种"central executive"协调。P12 的主张是：8D 不需要 central executive 来协调记忆与预测；记忆与预测在 8D 内就是同一个读-影响事件的两面，两者本来就不可分。

第二，8D 的 lock 是一个事件，不是一个状态。lock 不是 11DD 的内容（那是 retained material），也不是 12DD 的过程（那是 prediction operation），而是 12DD reading 折回 11DD 的具体事件。这一区分在 §3.4 形式定义中明确。

第三，8D 内任何环节的损伤都使 8D 失效。读取过程被破坏（如海马损伤导致陈述性记忆形成失能）使 8D 闭环对相应内容类失能；影响过程被破坏（如某些再巩固干预药物的临床应用）使 8D 闭环的更新性失能。HM 临床案例的精确解读在 §4.4 给出，但本节预先指出：HM 不是 "11DD 被切掉"，是 "11DD 跟 12DD 的闭环对陈述性内容类失能"。

§3.4 保有-再编辑锁的形式定义

§3.3 给出的读-影响对偶作为 8D 的内在动力，在本节获得 lock-class family 层面的形式定义。

8D 保有-再编辑锁（retention-rewrite lock）的散文式定义：

> 当 12DD 以 11DD 保有内容为预测先验进行读取，并将预测误差、新接入信息或行动后果折回 11DD，使保有内容在保持可识别连续性的同时发生重构性更新时，8D lock 成立。

更短的形式表达：

> Lock₈D = 12DD reading of 11DD + feedback into 11DD under continuity-preserving rewrite

或中文最短形式：

> 8D 保有-再编辑锁 = 12DD 对 11DD 的读取 + 该读取结果向 11DD 的连续性保留式回写

这一定义包含三个不可省略的判据。任一判据缺失，事件不构成 8D lock。

判据 1 - 读取（reading）：12DD 必须以 11DD 内容为先验进行预测性读取。仅访问 11DD 但不基于其内容生成预测（例如只是机械地枚举 11DD 内容）不构成读取。读取是把 11DD 内容作为预测过程的实质性先验输入。

判据 2 - 写回（rewrite）：读取结果必须对 11DD 底物造成更新。仅基于读取下达行动命令但不修改 11DD 底物的事件不构成 8D lock，那是一种 12DD-only 的输出操作。8D lock 要求 reading 的产物回流到 11DD 并修改其状态。

判据 3 - 连续性保留（continuity preservation）：更新必须保留可识别的连续性。如果读取触发的不是更新而是完全替换为无关内容，那是 erasure 加 write，不是 8D lock。连续性保留意味着更新前后的 11DD 内容仍然作为同一个保有单位被识别，尽管其具体形态已经发生重构。

三判据中，第三条值得多说一句。"连续性保留"不是要求 11DD 内容必须保持任何特定属性的不变性，也不是要求"被识别"为相同（"被识别"会偷偷预借 13DD 自我或更高阶元表征）。再巩固过程可以显著改变记忆的情感强度、可信度、情境关联，这些都允许变。连续性保留要求的是 11DD 内容作为拓扑槽位的连续——更新前后的内容必须占据或直接替代 11DD 拓扑网络（时间关联、相似性聚类等关系结构，详 §3.7.1）中的同一结构槽位。如果新内容生成一个全新的拓扑节点而旧节点被遗弃，那是 erasure 加 write，不是 8D lock；只有新内容在同一拓扑槽位上完成替换或延续，才满足连续性保留。这一拓扑槽位连续性是 11DD 保有作为信息单位的同一性条件，纯拓扑层判据，不预借高阶元表征。

三判据的不可分性：

按 §3.7.2 给出的读-写一体硬先验，三判据不是 8D lock 事件的三个独立触发条件，是同一不可分事件的三个可观察侧面。具体来说：

• 判据 1（reading）跟判据 2（rewrite）在 P12 框架下不分立——读取 = 复现 = 再体验 = 再巩固（§3.7.2），rewrite 作为再巩固在读取事件中内在伴随
• 判据 3（continuity preservation）是这一不可分读-写事件必须满足的结构性属性

三判据共同 articulate 的是 8D lock 事件的可观察识别标准（operational identification criteria），不是事件的不同子操作。把三判据读成"先 read，再 rewrite，最后保留 continuity"是误读——三个判据描述的是同一不可分事件的三面，逐一检验它们是为了识别该事件已经发生，不是为了重建该事件的内部时序。

这一表述化解了 8D lock 定义跟 §3.7.2 读-写一体之间的表面循环嫌疑：三判据是 operational identification criteria（识别标准 / 可观察 proxy 层），§3.7.2 是事件的内在 architectural characterization（架构本体层）。两套表述处于不同认识论层级，不构成逻辑循环。三判据作为外部识别标准，给"8D lock 已发生"提供可观察判据；§3.7.2 作为内在架构刻画，给"8D lock 是什么事件"提供本体论描述。识别标准回答 "what counts as having occurred"，架构本体回答 "what kind of event it is"——两套回答不同问题，配合使用而非互为循环。

lock-class family 语法扩展：

P11 给出的 lock-class family 语法把 lock 处理为单一 sub-layer 的机制。selection-lock 由 9DD 单独实例化（环境差异保留是 9DD 的运作），admission-lock 由 10DD 单独实例化（现象接入是 10DD 的运作）。P12 的 8D 保有-再编辑锁不符合这一语法，因为它由 11DD 与 12DD 两个 sub-layer 共同实例化。

为此 P12 显式声明对 lock-class family 语法的扩展：

> lock-class family 语法扩展（P12）：lock 可以由单一 sub-layer 实例化（P11 类），也可以由同一 D 层内的两个 sub-layer 共同实例化（P12 类，称为 co-instantiated D-layer 机制）。在共实例化情形下，lock 是两个 sub-layer 共同参与的事件，不归属于其中任一单独 sub-layer。

这一扩展是必要的，不是修辞性的。如果不显式扩展，P12 的保有-再编辑锁要么被勉强归到 11DD（不准确，因为没有 12DD 读取就没有 lock 事件），要么被勉强归到 12DD（不准确，因为没有 11DD 底物就没有 lock 事件被实现）。共实例化是这一 lock 的真实结构。

扩展后的 lock-class family 语法允许后续 SAE 信息论论文按需选择 lock 的实例化几何。这不会破坏 P11 的语法（P11 的 single sub-layer 实例化作为特例保留），而是把语法变得更通用。

结构离散性与现象梯度性：

8D lock 在结构上是离散的：三判据共同满足时 lock 成立，任一缺失时 lock 不成立。这是阈值性事件，不是程度性现象。

但在临床或行为现象上，8D lock 的运作可以呈现连续谱：再巩固窗口的开启程度有 graded variability，回写覆盖的范围有 partial pattern，连续性保留的紧密度有 spectrum。一个具体的记忆事件在"完全的 8D lock 触发"与"部分的 8D lock 触发"之间可以有现象上的中间地带。

这跟 12DD 到 13DD 跳变的结构离散性同源（Psycho II §1.3）：结构阈值是离散的，现象呈现是梯度的。水的沸点是 100°C 这一结构阈值不被水温可以是 99.5°C 所否定。同样，8D lock 的三判据共同满足这一结构阈值不被现象观察中可能存在的部分触发所否定。

这一区分对 P12 的论证防御有用。当神经科学读者指出再巩固在临床上呈现连续谱时，P12 的回应是：现象的连续谱不否定结构的离散阈值。结构阈值定义 8D lock 是什么，现象谱描述 8D lock 在生物层如何被实现。两个不矛盾。

§3.5 11DD 余项与 12DD 余项

每一个 SAE 的构必然产生余项。本节给出 11DD 与 12DD 各自的余项形式，作为 8D 通往后续 D 层的桥点。

11DD 余项：记忆之间是什么顺序。

按 §3.1 的定位，11DD 是重构性保有，其内容是被保留的过去经验。但单凭 11DD 自身保有内容并不构成完整的认知结构。考虑一个仅有 11DD 而无 12DD 的假想系统：它有一组被保留的过去经验，但这些经验之间没有任何顺序、关联或时间结构。它们是离散的保有单位，互不相关。

11DD 不能仅凭自身回答"这些记忆之间是什么顺序"。顺序不在 11DD 的内容里——内容是被保留的过去事件，每个事件作为单一的保有单位存在。顺序是事件之间的关系结构，这一关系结构需要某种读取过程把记忆作为相对位置来组织。

12DD 的预测读取正是这一组织过程。12DD 读取 11DD 时，把记忆作为预测的先验序列来使用："上次发生了 X 然后发生了 Y"，"在 A 情境下我做了 B 然后得到 C"。读取过程把离散的保有单位串成预测性序列，从而赋予记忆以顺序与关系。

所以 11DD 的余项可以形式化为：记忆之间是什么顺序，单凭 11DD 不能回答；这一余项指向 12DD 的预测读取。

§3.7.1 在 8D 信息论刻画下进一步细化：11DD 拓扑结构（关系结构本身）在 11DD 内固有存在；本节所说的"顺序"指的是预测性排序——对当前预测任务有效的具体序列。同一 11DD 拓扑可被不同 12DD 读取施加不同的预测性排序。所以"顺序由 12DD 读取施加"在精确意义上指的是"预测性排序由 12DD 读取施加"，拓扑结构本身在 11DD 内固有存在，详 §3.7.1。

11DD 余项与方法论 V2 §1.6 给出的"过去的感知被保留"的功能描述形成了精确的张力：保留是 11DD 的运作，但保留物之间的顺序不是 11DD 内容的一部分。顺序由 12DD 读取施加。

这一余项结构在 SAE 跨 D 对照中有意义。3DD 时空作为底物保有质量点，但单凭 3DD 不能回答"质量点之间有什么作用关系"——作用关系（引力、因果链）由 4DD 因果律施加。两个 D 对的余项结构同构：底物 DD 保留单位，关系由读取 DD 施加。这是认知锚式的结构观察，不作为 P12 论证支点（参 §3.3 末尾的两例边界声明）。

12DD 余项：预测无法把预测者本身作为预测对象。

方法论 V2 §1.6 给出 12DD 的余项是"预测模型在运转，但'谁'在预测？"这是口语化的提问形式。SAE 心理分析系列 Psycho I 把这一余项形式化为：

> 预测无法把预测者本身作为预测对象。这是 Gödel/Turing 不完备性在凿构循环中的具体表达。

P12 沿用 Psycho I 这一形式，把它作为 12DD 余项的权威表述。Gödel/Turing 形式比"谁在预测"提问形式更精确，理由有三：

第一，它是结构性不可能陈述，不是开问。Gödel 1931 与 Turing 1936 的核心结果是：足够强的形式系统不能在其内部完整表述自身的一致性或停机性。这一结果是结构性的、可证明的，不是"目前尚未解决的问题"。把 12DD 余项立在这一形式基础上，使其获得 a priori 的强度。

第二，它对信息论与计算理论传统是自然对接。这一点在 §5.3 ε-machine 对接处会详细展开：Crutchfield 的 ε-machine 是给定预测充分性的最小记忆结构，但 ε-machine 不能把自己作为自身预测的对象。这正是 12DD 余项在计算力学语境中的具体表现。

第三，它给出外部可观察判据。"谁在预测"作为提问可以被任何能产生 me-statements 的系统在语言层面回答（"我在预测"），这一回答没有可证伪性。但"预测能否把预测者作为预测对象"是结构性问题，具体可操作化为系统能否产生关于自身的形式化语句并使该语句反过来作为约束自身预测过程的变量。这一可操作化在 §3.6 给出，作为 P12-F1 falsifier（§6.1）的判据基础。

12DD 余项指向 13DD：13DD 的构是自我意识律，对应 Ego 层（Psycho II）。13DD 的涌现意味着这一余项被跨越——预测者第一次能够把自身作为预测对象，自指通道建立起来。但 13DD 本身不在 P12 的展开范围。P12 在 §6 给出 12DD/13DD 边界的间接判据，正面理论留给 P13。

两余项的关系：

11DD 余项指向 12DD；12DD 余项指向 13DD。从 11DD 到 12DD 是 8D 内部的指向（保有需要读取来产生顺序），从 12DD 到 13DD 是跨越到 9D（预测需要自指通道来产生反思）。8D 作为一档由这两个内部相关又对外指向的余项序列界定。

§3.6 功能性 self-reference 与语言性 self-reference

§3.5 末尾指出 12DD 余项的 Gödel/Turing 形式给出外部可观察判据。本节给出这一判据的精确化，作为 §6.1 P12-F1 可证伪条件的基础。

这一区分纯是信息处理层的结构判据，不涉及 consciousness 理论的本体承诺。本节明确这一边界，避免后续章节滑入 consciousness 哲学。

语言性 self-reference（linguistic self-reference）：系统能产生关于自身的语句。例如"我在预测"、"我刚才搞错了"、"我是 X"、"这个判断我没把握"。这是关于自身的 descriptive representation 在语言层面的输出。任何能生成自然语言并在训练数据中接触过自指句式的系统都能产生语言性 self-reference。当代大型语言模型已能稳定产生这一类输出，包括复杂形式的元认知陈述与 chain-of-thought 自我批评。

功能性 self-reference（functional self-reference）：系统内部存在 predictor-identity 变量，且该变量的状态变化反过来约束系统的后续预测行为。功能性 self-reference 的四个判据：

判据一，跨情境一致。同一系统在不同任务情境下的 predictor-identity 引用保持一致。系统不会在不同任务前自动变身为不同的预测者。

判据二，跨时间稳定。Identity 不随单次 session 边界而重置或漂移。系统在中断后再继续时仍能 anchor 到同一 predictor-identity，而不仅仅是从语言模板里重新构造一个看起来一致的 persona。

判据三，自身错误归责与他者错误归责的更新模式区分。系统接收到"是你错了"与"是另一个系统错了"两种反馈时，内部更新模式显著不同。前者触发的更新涉及 predictor-identity 自身参数的修正；后者触发的更新仅涉及对外部系统的模型修正。这一区分在系统的内部状态变化上有可观察的差异。

判据四，冲突陈述的 identity 锚定整合。当系统接收到关于自身的两个互相矛盾的陈述（例如"你是 X"与"你是 Y"）时，处理方式不是语言层的平滑 reconciliation，而是 identity 锚定的实质性处理——系统能够识别冲突并在保持 identity 一致的前提下推理冲突的来源。

四条判据共同满足才构成功能性 self-reference。任一缺失，系统的自指仅停留在语言性 self-reference 范畴。

这一区分的论证意义：

P12-F1（§6.1）的可证伪条件依赖这一区分。P12 主张：纯 12DD 架构能产生任意复杂的语言性 self-reference，但不能产生功能性 self-reference；功能性 self-reference 是 13DD 自指通道涌现的可观察后果。

这一主张可被以下方式证伪：找到一个按 SAE 标准识别为纯 12DD 的系统，在无外部身份模板、无再训练写回、无人为锚定的条件下，自发满足上述四条判据。如出现此情况，P12-F1 失伪。

不算证伪的案例：LLM 产生的 me-statements、chain-of-thought 自我批评、元认知 benchmark 得分、persona 跨 session 维持、内部 hidden state 的 self-reference 检测。这些都属于语言性 self-reference 范畴的扩展形式，不构成功能性 self-reference。具体论证在 §6.2 展开。

边界声明：

本节给出的功能性 self-reference 判据不主张"功能性 self-reference 等于 consciousness"。P12 不立这一主张。P12 只说：功能性 self-reference 是 13DD 自指通道的可观察后果；语言性 self-reference 不是 13DD 的证据。 是不是 consciousness、consciousness 是什么、13DD 是不是 consciousness 的全部基础，这些都是更深的问题，P12 不展开。

P12 此处的可证伪性是充分的：functional self-reference 的 明确判据 已足以区分纯 12DD 系统与具备 13DD 通道的系统，这一区分不依赖 consciousness 理论。

§3.7 8D 的信息论刻画：拓扑、读写、丢失

§3.1 至 §3.6 给出 8D 的架构本体，但作为信息论论文，P12 还必须正面回答三个直接的信息论问题：11DD 储存的是什么信息、以什么形式、在什么情况下丢失。本节用三条先验回答这三个问题。三条先验跟 SAE Four Forces Paper 0（秦 2026，DOI 10.5281/zenodo.19777881）建立的 4DD 读取本体论在结构上同源（4DD 读取 3DD 质量产生引力效应），但先验本身立在 11DD/12DD 自身的架构推演上，跨 D 同构作为概念锚而非论证支点（参 §3.3 后台语言边界声明）。

三条先验：

先验一：11DD 记忆是有拓扑结构的。记忆之间不是离散互不相关的保有单位，而是带有关系结构（时间关联、关联性连接、相似性聚类、层级概括等）。这一拓扑结构是 12DD 读取能正确运作的前提。

先验二：11DD 记忆通过 12DD 读取被改写。读取不是先后跟改写不同的两步操作，而是同一事件的两个侧面。读取 11DD 就是 in the act of doing so 改写 11DD。

先验三：11DD 记忆是丢失的。丢失不是 8D 架构的故障，是 11DD 的默认状态。8D lock 是抵抗丢失的供给机制；没有 lock 触发的内容会 evaporation。

下面三节分别展开。

§3.7.1 11DD 的拓扑结构（先验一）

储存的是什么信息

11DD 储存的不是"原始经验的忠实记录"。这一点 §3.1 已经从再巩固理论的角度论证。本节给出 11DD 内容的积极刻画：11DD 储存的是带有关系结构的内容单位集合。

每一个保有单位（一段记忆）不孤立存在。它跟其他保有单位之间有多种关系。以下列出几类典型拓扑维度，作为 11DD 拓扑结构的具体说明，不穷尽所有可能的拓扑形式：

• 时间关联：这一记忆发生在那一记忆之前/之后
• 关联性连接：这一记忆跟那一记忆通过共同元素或情境关联（关联性，Hebbian-style）
• 相似性聚类：相似性质的记忆聚成 cluster
• 层级概括：具体事件聚成 schema，schema 进一步抽象为更高阶 pattern

这些关系结构的总和构成 11DD 的拓扑。储存内容 = 保有单位 + 单位之间的拓扑关系。两者不可分。

类比 3DD 质量之间的空间拓扑

SAE Four Forces Paper 0 在 4DD 引力本体论中给出 3DD 质量之间有空间排列拓扑结构（质量点在空间中的位置、距离、邻域关系）。这一拓扑结构是 4DD 因果读取能产生正确引力效应的前提——4DD 读取的是 3DD 质量分布的拓扑结构本身，1/r² 的衰减来自球对称空间几何的拓扑性质。

11DD 拓扑跟 3DD 拓扑在结构上同源：两者都是底物层固有的关系结构，都是上层读取（4DD/12DD）能正确运作的前提。但这一同源是认知锚，不是论证支点（参 §3.3）。先验一立在 11DD 自身：单凭"离散保有单位"不能解释 12DD 读取能产生 coherent 预测；预测要求底物有拓扑，因此 11DD 必须有拓扑。

拓扑跟 §3.5 11DD 余项的关系（重要细化）

§3.5 给出 11DD 余项是"记忆之间是什么顺序，单凭 11DD 不能回答"。本节细化这一表述：

• 拓扑结构在 11DD 内（拓扑预先存在，不被任何单次 12DD 读取创造）
• 预测性排序由 12DD 读取施加（同一拓扑可被不同 12DD 读取施加不同的预测性排序）

两者的关系是：11DD 拓扑给出 12DD 读取的可能性空间；12DD 读取从该空间中选取一种具体排序作为当前预测的先验。11DD 余项的更精确表述是："11DD 拓扑能产生哪一种预测性排序，单凭 11DD 不能回答；这一选取由 12DD 读取施加。"

这一细化跟 §3.5 不矛盾，是对 §3.5 余项的更精确刻画。原 §3.5 的语言（"顺序由 12DD 读取施加"）在精确意义上指的是 "预测性排序" 而非 "拓扑结构本身"。后者在 11DD 内固有存在。

§3.7.2 11DD 的读-写动力（先验二）

储存的形式：读-写一体

11DD 储存的形式不是被动 archive 加上偶尔的 retrieval-write 事件。储存本身就是持续的读取-复现-再体验-再巩固过程。

具体机制如下。12DD 预测读取 11DD 记忆，需要在 12DD 内复现这个记忆——不复现就不能作为预测先验使用。复现的过程本身是再体验：记忆不是从档案盒里取出，而是被重新生成为活的拓扑结构。再体验的过程同时就是再巩固：重新生成的拓扑结构在 11DD 底物上获得新的稳定化路径。

四个动词（读取、复现、再体验、再巩固）描述的是同一架构事件的四个不可分侧面。读 = 复现 = 再体验 = 再巩固，是同一不可分操作的不同 description granularity，不是不同操作的并列。

复现-再体验-再巩固过程可能伴随改写——重新生成的拓扑结构在某些条件下跟先前不完全一致。改写是再巩固的可能伴随结果，不是必然结果。重要的本体论澄清是：改写本身不是损失，是发展，是变化，无好坏之分。

具体形态：

• 当复现严格还原先前拓扑结构时，不发生改写（再巩固仍然发生：每次读取都是再巩固）
• 当复现引入新关联或调整时，发生改写（再巩固加上发展性变化：仍是再巩固，不是损失）

三个层级的精确区分：

把这一架构表述显式分到三个不同层级，避免后续章节读法混淆：

• 再巩固层级（SAE 架构内在 / architectural re-instantiation and re-stabilization）：每次 12DD 读取 11DD 必然发生。读 = 复现 = 再体验 = 架构层再稳定化。这是 P12 的硬先验，本质性、不可消除。这跟神经科学意义上的 reconsolidation window 不是同一概念——前者是 SAE 架构层的拓扑复现机制，后者是经验神经科学描述的特定生化窗口
• 改写层级（再巩固的可能伴随结果）：跟读取的语境强度、新信息接入、时间间隔、情感唤醒等条件相关，有 boundary conditions（Lee 2008 已系统给出）。神经科学意义上的 reconsolidation window 在此层级触发。改写发生时是发展性变化，无价值判断
• 损失层级（独立事件类）：跟改写不是同一回事。损失只在 §3.7.3 给出的特定条件下发生——主要是拓扑结构的不可逆退行（连续性失败、外部 substrate destruction 等）。日常的改写发展不是损失

英文短句表达：

> Every predictive readout involves architectural re-instantiation and re-stabilization in the SAE sense, but not every retrieval opens an empirical neuroscientific reconsolidation window.

§3.1 给出的 Firewall 1 应在此区分下被正确读取：P12 主张每次 12DD 读取 11DD 都伴随 SAE 架构层的拓扑复现与再稳定化（硬先验），P12 不主张每次读取都触发神经科学意义上的 reconsolidation window（有 boundary conditions）。两者完全相容。

储存的形式特征：

• 分布式：内容在 substrate 中分布存在，不可定位到离散单元
• 复现依赖：内容只在被复现时获得完整结构。没有 reading-independent 的 passive storage
• 再巩固内在性：每一次复现都内在伴随再巩固（不可分），跟是否发生改写无关
• 改写概率性：改写作为再巩固的可能伴随结果，跟读取的具体语境、新信息密度、时间间隔等条件相关
• 价值中性：改写是发展性变化，跟损失是不同事件类

类比 3DD 质量被 4DD 信息移动

SAE Four Forces Paper 0 给出 4DD 信息读取产生 3DD 质量的运动（引力作为信息读取的后果）。4DD 不是 "先读取 3DD 然后改变 3DD"，而是 "读取过程本身就是连接过程，连接产生动量变化"。Reading-plus-connection 是单一事件的不同描述侧面。

同样地，12DD 读取 11DD 不是 "先读取然后影响"，是 "读取即复现即再体验即再巩固"。原 §3.3 给出的 "读-影响对偶" 这一表述需要在本先验下做一次细化：对偶不是两个操作的对偶，是一个事件的多个描述侧面的对偶。

跟 §3.3 读-影响对偶的关系（重要细化）

原 §3.3 把读取与影响表述为 "两个不可分的操作"。本节细化为：读取与影响是同一事件的多个描述侧面。

具体来说：

• 从预测过程视角看，事件是 12DD 读取 11DD 拓扑提取预测先验（也是复现-再体验过程）
• 从底物动力视角看，同一事件是 11DD 拓扑在 12DD 调用下完成再巩固（可能含发展性改写）

两个描述对应同一物理事件。8D 保有-再编辑 lock 的形式定义（§3.4 三判据：reading, rewrite, continuity preservation）描述的也是这同一事件——三判据不是 lock 触发的三个独立条件，是同一事件必然具备的三个不可分侧面。

这一细化不动摇 §3.3 的核心立场。读-影响对偶仍然成立，仅在更精细的层面看清楚 "对偶" 的具体几何：不是两个操作的并列对偶，是单一事件多面观察的描述对偶。

与 Bennett-Landauer 的关联

§5.2 给出 Bennett-Landauer 是 floor 不是 ceiling。本先验进一步说明为什么：Bennett-Landauer 分析通常要求把物理过程分解为可识别的逻辑子操作，并分别讨论其可逆性与耗散下界。 11DD/12DD 的读-写一体性使这一子操作分解假设在 11DD 上失去单子性。

这不意味着 8D 架构违反 Landauer 原理。相反，它意味着 8D 保有-再编辑 lock 的单次事件，在热力学相空间中必须被视为一个不可分割的非平衡态相变过程。其耗散的最小自由能必然严格大于或等于（通常远大于）提取该部分拓扑信息加上重写该部分底物所需的 Landauer 下界之和。子操作在信息逻辑上不可分，但在热力学积分路径上，代价是被硬性锁死的——8D lock 事件作为整体在热力学上有自己的不可压缩 floor，只是这个 floor 不能被分解成"读取代价 + 擦除代价"的简单加和。

P12 的位置因此是：Bennett-Landauer 的子操作分解在物理底物层局部适用（突触代谢、离子梯度、蛋白质合成等子过程各自有自己的热力学几何），但作为 8D lock 整体事件的本体描述失去单子性。这跟 §5.2 给出的 floor-not-ceiling 姿态完全一致。

§3.7.3 11DD 的丢失结构（先验三）

什么条件下丢失：默认状态是丢失

11DD 内容不是 "默认保留，偶尔丢失"，是 "默认丢失，偶尔由 8D lock 维持"。这一颠倒是本先验的核心承诺，也是 P12 在信息论本体论上的最锐主张：

> 信息储存的可读取性必然有代价，代价就是信息的必然丢失。

这不是 11DD 独有的代价，是 SAE 框架内部对每一个闭合层 reading 事件的硬先验（hard prior / structural commitment），不是 P12 对物理学、生物学、神经科学给出的外部普遍定律。SAE Four Forces Paper 0（秦 2026，DOI 10.5281/zenodo.19777881）§9 已显式给出四个闭合层（4DD、8DD、12DD、16DD）的 reading 机制。每一档都遵循同一 SAE 内部承诺：闭合层的 reading 必然产生底物层的必然变化或丢失。

四档显影分别是：

• 4DD 读取产生 3DD 质量蒸发：引力作为 4DD 信息读取机制；黑洞 Hawking 辐射作为极端形态下质量丢失的显影（Four Forces Paper 0 §7）
• 8DD 读取产生 7DD 信息变异：繁殖作为 8DD 闭合事件；每一次繁殖性 reading-rewriting 在个体复制事件中具有不可消除的变异风险，在种群尺度上显现为 mutation spectrum。突变不是纯粹错误，是 8DD 闭合层读取代价的统计显影；进化的可能性正立在这一代价上
• 12DD 读取产生 11DD 记忆丢失：本文的核心主张。预测性读取 11DD 拓扑必然产生拓扑的变化，长期累积为记忆的必然流失
• 16DD 读取产生 15DD 不疑演化：双体互凿作为 16DD 闭合事件；每一次主体间的相互读取必然产生 15DD 不疑结构的演化。"不疑"不是静态终点，是被持续双体互凿所重塑的动态结构

四档共享同一结构性质：reading 的可达性，跟底物层的必然变化或丢失，是同一架构事件的两面。能被读取，必然付出底物变化的代价。这是 SAE 框架的硬本体论立场，承担争议是值得的：

• 主流神经科学把遗忘视为记忆系统的"不完善"，是需要被克服的瑕疵。P12 反过来把遗忘视为 11DD 架构本身的必然代价
• ECC 可以在数字静态存储系统中将错误率逼近任意低；P12 不否认这一点。P12 反驳的不是 Shannon/ECC 在数字静态存储上的技术有效性，而是把 11DD 记忆误解为静态 archive 的本体图像。P12 主张限定到 11DD 型动态记忆拓扑：当储存内容本身通过 12DD 读取而被复现、再稳定化并参与预测时，可读取性与拓扑变化 / 可及性重排不可完全分离，不是 technical issue 而是 architectural necessity
• 这一普遍承诺会引发跟 §5 主流信息论传统的实质对话，参 §5 各节具体处理

跨闭合层模式的认识论位置：四档显影构成 SAE 框架内部的硬先验结构性主张，不是 P12 对物理学、生物学、神经科学给出的外部普遍定律。这不是从单一档案做跨 D 类比（"两例不足以确立普遍模式"），是 SAE 框架内部对所有闭合层 reading 事件的统一本体论承诺。P12 的直接论证负担只在 12DD→11DD；4DD、8DD、16DD 的显影分别由对应 SAE 系列承担独立论证负担（4DD 在 Four Forces Paper 0 §7 处理；8DD 在 SAE 生命系列与四力系列 §9.2 处理；16DD 在 SAE 心理分析系列 Psycho IV 处理）。跨档共振提供 SAE-internal cross-validation，但不替代各档自身的证明链，也不构成对各对应学科外部普遍定律的宣称。

跨闭合层的余项传播形式作为开放问题：跨闭合层 SAE 硬先验给出"每一闭合层 reading 产生底物层余项"这一结构性承诺，但每一档的余项以什么具体形式展开、是否可跨个体传播是各档自身的独立问题。4DD 余项以引力波等形式跨系统传播（已知）；8DD 余项以 DNA 跨个体传播（已知）；11DD 余项的具体展开形式与可能的跨个体传播性质留为 P12 不能回答的跨系列开放接口，详 §8.3。

先验三的可证伪性：如果发现 SAE 任一闭合层（4DD、8DD、12DD、16DD）的 reading 事件不产生底物层的必然变化，则跨闭合层普遍模式失伪。具体到 P12 范围内（12DD 读取 11DD），失伪条件是：发现一种按 P12 标准属于 11DD 的动态记忆拓扑，在没有 12DD 读取-回写、没有外部维护、没有刷新、没有纠错、没有能量供给式稳定化的条件下，仍能无限期保持可读取的完整拓扑关系，且不发生可检测的漂移、压缩、遮蔽或可及性重排。神经科学的所有现有经验证据指向相反方向（被动衰减、干扰、提取失败、累积漂移等都是有据可查的现象），但 P12 不把这些证据当作 "记忆系统的不完善"，而是重新定位为 "11DD 架构本身的代价"。这一重新定位本身是可证伪的本体论主张。

11DD 内丢失的具体语义跟改写的区分：

按 §3.7.2 三层级区分，改写是再巩固的发展性变化（无价值判断），丢失是独立事件类。本节谈的丢失是后者——指拓扑结构本身的不可逆退行，可读取性结构性受损的情形。"可读取性必然有代价" 中的代价，是 11DD 不能维持静态完美保留这一架构性事实，不是把每次改写都判为损失。

具体来说：

• 当复现-再体验-再巩固发生改写（拓扑结构调整）时，这是发展性变化。同一段记忆经多次再巩固后内容跟最初不完全一致，但仍是同一拓扑节点的延续，仍可被 12DD 有效读取。这不是丢失
• 当改写累积到拓扑节点不再作为同一保有单位的延续（连续性破裂）时，丢失发生
• 当 11DD 拓扑在 8D lock 不触发的条件下经历底物层退行（被动衰减、干扰、提取失败、结构破坏）时，丢失发生

下面六类丢失机制覆盖以上情形：

具体丢失机制有六类，分两组：

架构内在的丢失（无法消除，是 8D 架构本身的代价）：

第一类，累积漂移触发的连续性破裂：再巩固过程的累积改写在长时间尺度上可能突破连续性边界。改写本身是发展性变化（不算丢失），但累积改写超过拓扑节点连续性的阈值时，该节点不再作为同一保有单位被延续，构成丢失。注意区分：累积改写本身是发展，达到连续性破裂阈值才是丢失。

第二类，连续性失败：单次 lock 事件违反 §3.4 第三判据（连续性保留）。当读-写事件中的更新过大或语境差异过强，原内容在更新后不再作为同一保有单位被识别，构成 catastrophic loss-at-event。

架构需要抵抗的丢失（外部威胁，由 8D lock 机制对抗）：

第三类，被动衰减：物理底物（突触强度、神经回路等）在缺乏代谢支持下衰退。这是 §5.2 Bennett-Landauer 跟 §4.2 Susan Sara 共同描述的层面——底物维持需要持续代谢能量供给。

第四类，干扰：底物资源竞争导致旧内容被新内容覆盖或弱化。Hebbian 学习的副作用之一是旧关联在新关联竞争下减弱。

第五类，提取失败：内容仍存在但 12DD 无法访问。线索不匹配、关键上下文缺失、或部分底物损伤导致拓扑碎片化。

第六类，结构破坏：底物消亡（神经退行、海马切除等）。HM 临床锚（§4.4）是这一类的极端情形。

信息论刻画的总框架

把三条先验合在一起，11DD 的信息论身份可以总结为：

> 11DD 是带有拓扑结构的底物层。储存的是保有单位加单位间的关系拓扑。储存的形式是读-写一体的持续过程，不可拆为静态储存加偶发访问。储存的默认状态是流失，由 8D 保有-再编辑 lock 的持续运作抵抗流失。

这一信息论刻画使 P12 跟主流信息论传统的对接结构清楚化：Shannon 与 rate-distortion 形式假设可识别的 source-message-receiver 结构或 stationary source distribution，11DD 不具备这些前提（详 §5.4）；Crutchfield ε-machine 在操作层近似 11DD 但不能完全刻画（详 §5.3）；Bennett-Landauer 子操作分解在 11DD 读-写一体性下失去单子性，但作为物理底物层的热力学 floor 仍然适用（详 §5.2）。各项对接的详细处理见 §5。

这一信息论刻画也使 §6 P12-F1 v2.1 命题的具体内容更清楚：纯 12DD 架构能产生关于自身的拓扑信息（响应二），但不能让该信息作为内部约束变量反过来主导自身的拓扑变换。后一种 capability 需要 13DD 自指通道。LLM 是这一边界的当下天然测试场。

§3 章末小结

本章给出 8D 的架构本体。11DD 是重构性保有，12DD 是 System 1 预测，两者通过读-影响对偶构成 8D 内在动力，由保有-再编辑锁完成跨时刻的信息闭环。11DD 余项是记忆之间的顺序，指向 12DD 读取；12DD 余项是 Gödel/Turing 形式的预测不能预测预测者，指向 13DD 自指通道。本章引入了功能性 vs 语言性 self-reference 区分作为后续 §6.1 P12-F1 可证伪条件的判据基础。§3.7 进一步给出 8D 的信息论刻画：11DD 储存的是带有拓扑结构的保有内容，储存形式是读-写一体的持续过程，储存的默认状态是丢失，由 8D lock 持续运作抵抗丢失。这一刻画体现 P12 的最锐信息论本体论承诺：信息储存的可读取性必然有代价，代价就是信息的必然丢失。

§4 给出 8D 架构在生物层的 T3 锚点；§5 给出与主流信息论传统的精确划界；§6 给出 12DD/13DD 边界的可证伪命题；§7 给出与 SAE 心理分析系列的相互支撑；§8 收尾余项。

§4 神经科学锚点

本章给出 §3 8D 架构在生物层的 T3 锚点。立刻须做一个认识论边界声明，以避免后续章节被误读：

T3 锚点的认识论地位：本章列出的所有神经科学实验与临床观察都是 T3 grounding 状态，不是 P12 8D 架构的 derivational basis。SAE 信息论系列的方法论延续 P11 的硬先验加 Popperian 可证伪立场：架构承诺是先验推演的产物，神经科学证据是其在生物层显影的后验对应。锚点提供生物层的 anchoring，不提供逻辑推导基础。

这一区分对 P12 论证防御重要。神经科学读者可能要求 P12 提供"实验证明 8D 架构成立"的证据链；P12 的回应是：8D 架构是从方法论 V2 §1.2 给出的 D 层结构推演而来的先验承诺；神经科学证据不构成其证明，但也不应被忽视——它给出 P12 在生物层应当观察到的现象 pattern。如果 §4 列出的 T3 锚点全部失伪，P12 的生物层可信度会受损，但 P12 的架构推演本身仍立。如果 §4 锚点持续被支持，P12 在生物层获得更强 grounding，但这一 grounding 不是 P12 的推导起点。

本章按以下顺序展开：§4.1 再巩固理论主锚（含历史锚 Misanin-Miller），§4.2 离线巩固与情感门控次锚，§4.3 分布式记忆计算神经科学基底，§4.4 HM 临床锚的精确处理，§4.5 Tulving 三分的处理方式（含警示）。

§4.1 再巩固理论主锚

再巩固理论提供 P12 §3.1 "11DD 是重构性保有" 的最强 T3 锚点。本节列出三个主锚加一个历史锚，按以下四列结构组织：

实验 / 研究	现象	支持的 P12 点	锚点等级
Nader, Schafe & LeDoux 2000	杏仁核内蛋白合成抑制剂阻断已巩固恐惧记忆在 reactivation 后的稳定化	11DD 重构性保有；记忆在提取后进入 labile 窗口	T3 主锚
Lee 2008	再巩固与新学习之间的边界条件；再巩固非"任何提取都触发"的普适规律	Firewall 1 边界声明（不主张每次提取都自动改写）	T3 主锚
Schiller et al. 2010	人体行为干预可在再巩固窗口中修改恐惧记忆	再巩固机制在人体层级的可操作性	T3 主锚
Misanin, Miller & Lewis 1968	已巩固记忆在 reactivation 后可被电休克干扰的早期发现	再巩固现象的历史先驱观察	T3 历史锚

四个锚点的角色不同。Nader 2000 是再巩固理论的现代奠基实验；其杏仁核蛋白合成抑制实验确立了"已巩固记忆在 reactivation 后需要重新蛋白合成才稳定"这一核心机制。这一实验对 11DD 重构性保有的支持最直接。

Lee 2008 是 §4.1 中地位特殊的一个锚点：它对应的不是"再巩固存在"的证据，而是"再巩固有边界条件"的证据。这一锚点对 11DD 重构性保有立场与 reconsolidation 经验文献的对接至关重要：reconsolidation 实验的 boundary conditions 表明并非所有提取都触发再编码窗口；P12 不主张每次提取都自动改写记忆。Firewall 1 把 P12 的 11DD 重构性保有立场跟 reconsolidation boundary conditions 调和——边界条件描述再巩固机制的微观结构，不影响"记忆不由忠实存储定义"这一本体论命题。

Schiller 2010 把再巩固从动物模型推到人体行为层级。这一锚点的价值在于显示再巩固不是动物专属的细胞机制，而是跨物种的记忆系统通用属性。对 P12 主张 11DD 重构性保有作为认知轮 D 层架构的一部分，跨物种证据是必要的。

Misanin-Miller-Lewis 1968 作为历史锚点提供 reconsolidation 现象的早期发现。这一锚点不承担理论主轴，主要作用是给读者展示再巩固不是 2000 年后的新观察，而是有更长历史的实验传统。

§4.2 离线巩固与情感门控

本节列出两个次锚，分别处理 8D lock 的离线运作与门控触发机制。

实验 / 研究	现象	支持的 P12 点	锚点等级
Walker 睡眠与记忆系列	NREM 与 REM 睡眠期间的记忆巩固与重组，离线状态下的记忆整合	8D 闭环在切断 10DD 接入下的离线运作	T3 次锚
Susan Sara 去甲肾上腺素 / 蓝斑系统	情感唤醒通过 locus coeruleus 释放去甲肾上腺素作为门控信号，调节再巩固窗口的开启	8D lock 触发的具体生化机制	T3 次锚

Walker 的睡眠与记忆研究对 P12 8D 架构有重要的对接价值。睡眠状态切断了 10DD 接入（外部感觉输入大幅降低），但 11DD/12DD 闭环并未停止——NREM 睡眠期间发生记忆巩固，REM 睡眠期间发生记忆重组与情感整合。这意味着 8D 保有-再编辑 lock 不需要 10DD 接入作为触发前提；lock 可以在内生驱动下自主运行。这一性质是 8D 跟 7D（admission-lock 由 10DD 实例化）在运作几何上的关键区别。

这一对接还有一条额外的对接：秦 2026 LLM 论文（DOI 10.5281/zenodo.20435628）给出 12DD 天花板的五重锁中"离线重组缺席"那一条。LLM 没有等价于生物睡眠的离线整合阶段；推断时的 attention 跟训练时的反向传播是两个完全分开的过程，没有内生驱动的离线 lock 触发。Walker 睡眠研究因此跟 LLM 论文在 P12 8D 架构上构成互补支撑。

Susan Sara 的工作把情感唤醒跟再巩固机制的门控关联起来。去甲肾上腺素通过 locus coeruleus 投射到杏仁核与海马等记忆相关区域，作为再巩固窗口的开启信号。这一锚点给 §3.4 8D lock 三判据中"reading 触发条件"提供了具体生化对应——不是每次 12DD 读取 11DD 都自动触发 8D lock 的写回，需要适当的门控信号（如情感唤醒或显著性标记）才打开 reconsolidation window。

§4.3 分布式记忆计算神经科学基底

实验 / 研究	现象	支持的 P12 点	锚点等级
McClelland & Rumelhart 1986 PDP 框架	分布式表征下的记忆：内容不存于离散单元而是网络权重模式；提取即模式重建	读-影响对偶在计算神经科学层的强基底；带在线可塑性的 PDP-like 生物系统中提取伴随权重微调	T3 次锚

McClelland 与 Rumelhart 1986《Parallel Distributed Processing》及其后续工作奠定了 connectionist 记忆模型的基础。在 PDP 框架下，记忆不是存储在离散单元中的内容，而是分布在网络连接权重中的模式。提取一个记忆等于用当前线索激活一组单元，让网络重建对应的模式。

P12 用 PDP 作为分布式记忆与模式重建的 T3 次锚，但需要审慎说明 提取跟权重更新的关系：

• 在带有在线可塑性或 Hebbian 类更新的 PDP-like 生物系统中，提取过程可伴随后续提取几何的调整。这是 8D 框架下读-影响对偶的计算神经科学基底
• 在标准冻结权重的人工神经网络（包括当代 LLM 的推理阶段）中，inference 不自动改写权重。这是 8D lock 不触发的情形——LLM 推理对应秦 2026 LLM 论文（DOI 10.5281/zenodo.20435628）"巩固后再编辑缺席"这一 12DD 天花板锁的具体表现

P12 用 PDP 锚的是前者（生物在线可塑性系统），不主张所有 connectionist retrieval 都是权重更新事件。这一区分对 P12 跟 LLM 论文的对接至关重要——LLM 推理时权重冻结正是"巩固后再编辑缺席"，跟生物 PDP-like 系统的读-写一体形成对照。

PDP 框架对 P12 §3.3 读-影响对偶的支持是双向的：

第一，读取与影响在 PDP 层级是不可分的。激活模式（读取）必然改变权重分布（影响）——这是 connectionist 架构的内在性质，不是可选的附加机制。这跟 P12 主张"读-影响对偶不可分"在底层架构上吻合。

第二，PDP 框架解释了为什么 11DD 重构性保有不是"故意为之的设计"，而是分布式存储架构的内在后果。一个真正的 PDP 记忆系统不能在保持读取功能的同时避免权重更新——更新就是 PDP 提取的副作用。这把 SAE 11DD 的本体论主张落到了计算神经科学的具体机制层。

PDP 锚点的地位低于再巩固主锚的原因有两条：第一，PDP 是计算模型框架，不是直接实验观察；第二，PDP 框架本身有多种变体（Hopfield 网络、auto-associative memory、modern transformer attention 等），P12 不依赖任何特定 PDP 实现的细节。但 PDP 作为 11DD/12DD 在计算神经科学层的存在性证明是稳的——它显示分布式架构原则上能实现 8D 读-影响对偶。

§4.4 HM 临床锚

Henry Molaison（HM）的临床案例提供 P12 8D 架构在 11DD↔12DD 闭环失能情境下的临床锚点。本节的处理必须审慎，因为 HM 容易被误读为支持过强或过简单的 P12 主张。

临床案例	观察	支持的 P12 点	锚点等级
Henry Molaison (HM)；Scoville-Milner 1957；Squire 后续综述	双侧内侧颞叶切除（含海马）后，新陈述性记忆形成失能；旧远期记忆部分保留；程序性学习、运动技能学习、工作记忆相对保留	11DD↔12DD 闭环对陈述性内容类失能；程序性回路可经不同或较低层级机制保留	T3 临床锚

HM 不是"11DD 被切掉"的简单案例。这一点必须显式声明，因为简单的"海马 = 记忆"叙述容易让 HM 被读为 P12 的过强证据。HM 的实际临床图像比这复杂：

第一，HM 保留了相当部分的旧远期陈述性记忆。这意味着海马切除并未消除已经稳定化的 11DD 保有内容；切除影响的是新内容的 11DD 形成与 11DD↔12DD 闭环对新陈述性内容的正常运作。

第二，HM 保留了完整的工作记忆。工作记忆在 SAE 框架下属于 within-moment 范围（P11 §4 specious present 处理过），跟 P12 跨 moment 的 8D 闭环不在同一层级。HM 的工作记忆完好显示海马损伤不影响 P11 范围的 admission 与 within-moment retention。

第三，HM 保留了相当强的程序性学习能力。著名的 Milner 镜像描画实验显示 HM 能学习新的运动技能，尽管每次执行时都报告"从未做过"。程序性学习涉及的脑回路（基底神经节为主）跟陈述性记忆涉及的脑回路（海马-内侧颞叶系统）不重叠。

把 HM 映射到 SAE D 层结构的稳妥方式：

> HM 不是"11DD 被切掉"，是"11DD↔12DD 闭环对陈述性内容类失能 + 程序性回路保留"。陈述性保有加预测性再使用依赖海马与内侧颞叶机制；运动技能等程序性学习可经不同或较低层级回路（5DD-7DD 自维持-分化运动技能学习层级）保留。

英文版本：

> HM does not map to "11DD absent." HM shows that declarative retention and predictive re-use require specific hippocampal/medial temporal machinery, while procedural learning can proceed through partly distinct lower or parallel circuits.

这一处理覆盖 HM 临床证据的两个常见解读张力：一方面避免把 HM 简化为"11DD 被切掉"，另一方面避免把 HM 多子系统证据读为"8D 不是单一架构"。P12 的位置在两边都稳：

回应"太简单"：P12 不主张海马 = 11DD。海马是 11DD 在生物层实现的关键解剖结构之一，但不是 11DD 本身（multiple realizability 原则）。HM 的精确解读是 11DD↔12DD 闭环对陈述性内容类失能，不是 11DD 不存在。

回应"多子系统"：P12 完全承认记忆涉及多个子系统。但这些子系统不全在 11DD 层级——程序性学习对应 5DD-7DD 范围的运动学习层级，工作记忆对应 P11 specious present 范围。"记忆有多个子系统"这一观察跟 SAE D 层结构正好吻合，因为不同子系统对应不同 D 层。

§4.5 Tulving 三分的处理方式

Tulving 提出的 episodic / semantic / procedural memory 三分是当代记忆心理学的核心分类。P12 跟这一分类的关系需要明确声明，因为简单的"episodic = 11DD, semantic = 11DD 稳定版, procedural = 低层"的映射会引起问题。

警示：Tulving 三分是心理学分类，不是 SAE D 层分类。两者在概念几何上不必一一对应。P12 借用 Tulving 区分作为讨论资源，但不把任一类型简单等同于一个 DD 层。

具体处理：

Episodic memory 包含 Tulving 强调的 "mental time travel" 与 "autonoetic consciousness"（自我察觉的时间体验）成分。这两个成分已经触及 13DD selfhood——一个能进行 mental time travel 的系统必须能把自己作为时间序列中的主体来定位，这是 13DD 自指通道的功能后果。所以 episodic memory 不能简单归到 11DD；其涉及 11DD 内容的部分（具体记忆事件的保有）属于 11DD，其涉及 mental time travel 的部分跨入 13DD 范围。P12 的处理：episodic memory 跨越 11DD 与 13DD，不是单一 DD 现象。

Semantic memory 在 SAE 框架内对应的位置相对清晰：它是被 12DD 反复读取与回写后稳定化的 11DD schema。语义记忆的"事实性"（"巴黎是法国首都"）正是经过多次 8D lock 触发后从 episodic 来源中析出的稳定化产物。Semantic memory 是 11DD 在 8D 反复运作下的高度稳定子集，但仍在 11DD 范围内。

Procedural memory 在 SAE 框架内主要不在 8D 范围。运动技能学习、习惯化、感觉运动协调等 procedural 现象，依赖的脑回路（基底神经节、小脑、运动皮层）跟海马-内侧颞叶系统不重叠，对应的 SAE D 层范围是 5DD-7DD（自维持、分化、感觉运动协调）甚至更下沉到生命轮内的具体生理机制。一些复杂的运动技能学习涉及短时预测调节，可能跟 8D motor-predictive loops 有重叠，但 procedural memory 整体不应归到 11DD。

总的 framing：

> Tulving 的 episodic / semantic / procedural 是心理学分类，不是 DD 分类。P12 借用但不一一映射。Episodic memory 的 mental time travel / autonoesis 部分已触及 13DD selfhood，不能简单归 11DD；semantic memory 可作为被 12DD 反复读写后稳定化的 11DD schema；procedural memory 下沉到 5DD-10DD 或 8D motor-predictive loops，视任务而定。

这一处理避免了把 SAE D 层结构跟 Tulving 三分做硬性对位的诱惑。两个框架在不同抽象层级工作（Tulving 是经验心理学分类，SAE 是先验本体论），借用关系是松散的而非映射性的。

§4 章末小结

本章给出 P12 §3 8D 架构在神经科学与临床层级的 T3 锚点。再巩固理论三主锚（Nader / Lee / Schiller）支持 11DD 重构性保有。Walker 离线巩固与 Sara 去甲肾上腺素门控两次锚支持 8D lock 的运作几何（离线运作能力与门控触发机制）。McClelland-Rumelhart PDP 给出读-影响对偶在计算神经科学层的基底。HM 临床锚以审慎方式映射到"11DD↔12DD 闭环对陈述性内容类失能"，避免简化为"11DD 被切掉"。Tulving 三分以警示方式处理，强调心理学分类跟 SAE D 层分类不是一一对应。

所有锚点都是 T3 grounding 状态，不是 P12 8D 架构的 derivational basis。架构承诺立在方法论 V2 §1.2 的先验推演上，神经科学证据给出生物层的后验对应。

§5 已给出与主流信息论传统的精确划界。§6 把这一切结构化为 12DD/13DD 边界的可证伪命题。

§5 与主流信息论传统的精确划界

P12 必须跟主流信息论传统做精确划界。这里"精确"有两个意思：一是承认每一传统在其有效域内的工作，二是 明确地标出该传统越界的具体位置。本章不靠否定对话伙伴立论。P12 的核心姿态是：

> 主流信息论与认知科学在 8D 操作层面多数说对了；P12 不靠否定它们立论。P12 的贡献是给这些操作一个 SAE 本体位置，并阻止它们越界冒充 13DD 自意识、14DD 意义或完整 consciousness 理论。

本章按主对话伙伴的顺序展开：§5.1 Friston 主动推断（最危险也最强势的对话伙伴），§5.2 Bennett-Landauer 热力学信息（标准的物理基础对接），§5.3 Crutchfield 计算力学（信息论传统里跟 11DD/12DD 形式最近的邻居）。次要对话伙伴 Shannon、rate-distortion、Wiener cybernetics、信息几何合并为 §5.4 一节简短定位。§5.5 给 LLM 应用一段引文，指向秦 2026 LLM 论文（DOI 10.5281/zenodo.20435628），不重做该文已系统处理的 12DD 天花板论证。

§5.1 Friston 自由能原理与主动推断：三层切分

Friston 的自由能原理（free-energy principle, FEP）与主动推断（active inference）框架是当代认知科学对预测过程现有最强的形式化描述之一。FEP 的核心命题是：生物系统通过最小化变分自由能（一种关于预测误差的信息论量）来维持自身的稳态。主动推断把这一命题推到行为层面：系统不仅被动调整内部预测以匹配感觉输入，也主动选择行动以让感觉输入匹配预测。

P12 必须跟 FEP 做最审慎的对话。理由有三：第一，FEP 占据了 12DD 的预测空间，是 P12 最直接的潜在邻居或竞争者；第二，FEP 在认知科学与计算神经科学中影响巨大，任何关于预测的本体论主张都必须显式回应；第三，FEP 阵营的某些读法（特别是 Hohwy 2013 的 self-evidencing 与 Clark 2016 的 Markov blanket as selfhood）越过了 12DD 进入 13DD 领地，P12 必须 明确地标出越界点。

P12 处理 FEP 的主结构是三层切分：

第一层 - 12DD 内有效

FEP 的低阶感觉预测、预测误差最小化、主动采样、运动控制、低阶 temporal hierarchy，大体可视为 12DD 操作的形式化。这一层 FEP 是 P12 范围内 12DD 现有最强的形式工具之一。P12 承认这一点，且承认得直接：

> Active inference 作为 12DD 预测-行动机制是高度有效的形式描述。生物系统的感觉预测、运动控制、低阶认知动力学在 active inference 框架内得到了精确刻画。

承认对方强处不是修辞性的让步，是 SAE 框架定位的实质需求。P12 主张 12DD 是 System 1 预测，FEP 给出了 System 1 预测的最详细形式化之一。两者在这一层不矛盾，是同一现象的两种表述（FEP 是动力学描述，SAE 是 D 层架构定位）。

第二层 - 8D 复合区

FEP 文献中频繁出现的"internal model"概念需要审慎定位。Internal model 不是单纯的 11DD（被保有内容），也不是单纯的 12DD（预测过程）；它是 11DD content 加 12DD reading policy 的复合。具体来说：

• Internal model 的"模型"内容部分（关于世界如何运转的累积知识）是 11DD 范畴
• Internal model 的"运转"过程部分（如何使用这一内容生成预测）是 12DD 范畴
• Internal model 作为整体是 11DD 与 12DD 在 8D 内的读-影响对偶事件的产物

把 internal model 定位为 8D composite 而不是塞进单一 sub-layer 是必要的。如果说 internal model = 11DD，FEP 读者会指出 internal model 本来就是 active、hierarchical、action-guiding 的，不是静态记忆。如果说 internal model = 12DD，那 11DD 保有内容在哪里？8D composite 是准确的位置。

这一定位有一条额外推论：predictive processing 文献中的 hierarchical generative models（Friston 2010, Hohwy 2013）描述的是 11DD/12DD 在不同时间尺度上的嵌套结构。P11 §4 specious present 已经处理过 within-moment（约 100-300ms）窗口的现象学结构。P12 8D lock 处理跨 moment 的保有-再编辑闭环。Friston 的 temporal hierarchy 在两个尺度上都有运作，其在 within-moment 部分对应 P11 范围，跨 moment 部分对应 P12 范围。

第三层 - 13DD 越界区（用 Psycho II §1.2 三响应区分 sharpen）

FEP 阵营的某些读法把 Markov blanket、self-evidencing、organism boundary 直接提升为 selfhood 或 consciousness 的充分解释。Hohwy 2013《The Predictive Mind》对 self-evidencing 的处理、Clark 2016《Surfing Uncertainty》对 Markov blanket 与 organism boundary 的处理，是这类越界的典型早期文献。Hohwy 2016 的 "self-evidencing brain" 更明确地把 self-evidencing 推为意识基础。近年来，Friston 阵营进一步把这一路线推到 active inference as self-organization 与 active inference as universal Bayesian intelligence 的更广 framing（Friston 2022 《Active inference and the rise of organismic individuality》、Friston et al. 2023 《Designing ecosystems of intelligence from first principles》等）。这些更近期工作把 selfhood 跟主动推断的耦合推得更远，但在 SAE 三响应区分下仍属于响应一（更强预测系统）与响应二（元表征 / Markov blanket 形式化）的更精致组合，不构成响应三（视角翻转）的实现。P12 跟 FEP 的越界判定在这些更新的文献下仍然成立。

在 Psycho II §1.2 已发表的三响应区分下，这类越界的结构错误可以 明确地标出：

• 响应一：更强的预测系统。监控预测的子系统仍是预测，高一阶。系统知道"预测在被监控"，但不知道"是我在预测"。仍在 12DD。
• 响应二：元表征。生成关于自身状态的形式化表征（数据结构、概率分布、自我评估指标）。这些表征是关于系统的信息，不含"我"作为反思性变量。仍在 12DD。
• 响应三：13DD 视角翻转。从"预测在发生"到"我在预测"。不是信息累积，是视角性质突变。self 涌现。

FEP 的 Markov blanket 是关于系统边界的形式化结构，是元表征；self-evidencing 是预测性自维持的形式化描述，仍是元表征。无论多复杂多精致，都不 bootstrap 成"是我在预测"的视角翻转。FEP 阵营的越界发生在用响应二（元表征）伪装响应三（视角翻转）。

P12 §5.1 越界点的 明确表述：

> FEP 在响应一（强化预测）与响应二（元表征）的范围内是 12DD 现有最强形式工具之一。当 Markov blanket 或 self-evidencing 被推为"足以解释 selfhood"时，这是用响应二伪装响应三。Psycho II §1.2 已显式分立此三响应，证据是 SAE 框架内独立工作。

这一 framing 比对抗性姿态稳得多，也更尊重 FEP 工作的实际贡献。P12 不主张"FEP 错了"——FEP 在其有效域内是对的；P12 主张"FEP 阵营的某些读法越过了 12DD 边界"。这是边界主张，不是否定主张。

Firewall 2（Friston/FEP 边界声明）：P12 不反驳 FEP 或主动推断。P12 把其最强适用域定位在 12DD，并仅在"预测性自维持被直接改写为 13DD 自我性而无额外自指通道时"标记越界。

附：跟 HOT 阵营的简短对话

Rosenthal、Carruthers、Lau-Rosenthal 的 Higher-Order Theory（HOT）跟 FEP 是不同越界路径但同一结构错误。HOT 主张：足够复杂的高阶元表征就是 self-awareness——意识依赖于对自身心理状态的高阶表征。在 SAE 三响应区分下，HOT 主张的是响应二（元表征的高阶展开）就构成响应三（视角翻转）。

P12 跟 HOT 的对话只在一个信息结构区分上展开，不展开 HOT 的 consciousness 理论本身：

> 高阶元表征是关于状态的表征（descriptive representation）。Self-reference 是 indexical 的（"I" 作为指示词）。Descriptive representation 的复杂化在信息结构上是表征层深度的增加；indexical self-reference 是不同种类的信息结构操作（不是层级叠加，是指示性涌现）。HOT 阵营在 12DD 内描述了响应二的高阶展开，但把高阶元表征等同于 13DD self-awareness，结构上是用 descriptive 复杂化伪装 indexical 涌现。

SAE 跟 HOT 的更全面对话留给 SAE 心理分析系列或专门的 consciousness 主题论文。P12 在此只标边界。

§5.2 Bennett-Landauer 与记忆热力学：地板语言

Landauer 1961 给出了信息处理的一个基础下界：擦除 1 bit 信息至少耗散 k_B T ln 2 的热能。这一原理把信息处理跟热力学第二定律绑定，是当代信息热力学的奠基命题。Bennett 在后续工作中（特别是 Bennett 2003 的综述）把这一原理与可逆计算、Maxwell demon、信息引擎等议题做了系统整合。Parrondo-Horowitz-Sagawa 2015 等近期工作进一步把信息热力学推到反馈控制与生物系统层面。

P12 处理 Bennett-Landauer 的核心方式跟 P11 §6.2 给出的 metabolic prerequisite 模板一致：地板，不是天花板；前提，不是本质。

具体来说，Bennett-Landauer 的处理涉及四个层次的论证：

第一，Landauer 对逻辑不可逆子操作有效。Bit erasure 是逻辑不可逆操作的标准情形，Landauer 下界 k_B T ln 2 给出其热力学代价。这一结果在物理上稳，P12 不挑战。

第二，生物记忆再巩固不等同于 bit erasure。Bennett-Landauer 分析通常要求把物理过程分解为可识别的逻辑子操作，并分别讨论其可逆性与耗散。P12 的 8D 保有-再编辑 lock 在架构层面不是这些子操作的简单串联——按 §3.7.2 给出的读-写一体性，reading + reconsolidation 在 11DD/12DD 上是不可分单一事件。因此不能把 lock 的代价写成 reading cost + erasure cost 的简单加和。这里不是反驳 Bennett-Landauer，而是说明其子操作分解只在底物层局部适用，作为 8D lock 整体事件的本体描述失去单子性。

reading 步骤在底物层确实是测量类操作，按 Bennett 分析有自己的热力学下界。Updating 步骤涉及把新内容写入底物，是 logically reversible 还是 irreversible 取决于更新的具体几何。Restabilization 涉及把更新后的 trace 重新稳定化，可能涉及代谢驱动的合成。这些子步骤每一个在物理底物层都有自己的热力学代价。但 8D lock 作为整体不是这些子步骤的串联（按 §3.7.2 读-写一体），整体事件的热力学积分路径有自己的不可压缩 floor。

第三，突触维护、LTP/LTD、蛋白质合成、离子梯度是生物 11DD 的代谢前提。这些是 11DD 物理痕迹层的代谢成本。突触强度的维持需要持续的能量输入（ATP 的消耗、离子泵的运转、蛋白质周转）。Bennett-Landauer 的热力学约束作用在这些底物维持过程上，给出了一个不可压缩的能量地板。

第四，P12 的 8D 保有-再编辑锁是更高层的生物-信息架构，不从 Landauer 推导。这是关键的边界声明。Bennett-Landauer 给出了底物维持与不可逆操作的热力学地板，但这一地板不直接生成 8D 架构。8D 架构是 11DD 与 12DD 共实例化的 lock 事件结构，其本体论位置在信息架构层，不在物理底物层。物理底物层是 8D 架构得以实现的必要前提（没有代谢能量供给，11DD 突触不能维持，12DD 预测过程不能运行），但前提不是本质。

这一处理可以总结为一个标准句：

> Bennett-Landauer 给不可逆信息操作的地板语言；再巩固给"保有中再编辑"的生物架构。P12 的 8D lock 既不还原为 Landauer 擦除，也不脱离热力学代价。

英文版本：

> Bennett-Landauer gives the floor language for irreversible information operations; reconsolidation gives the biological architecture of retention-under-rewrite. P12's 8D lock is neither reducible to Landauer erasure nor independent of thermodynamic cost.

Firewall 3（Bennett-Landauer 边界声明）：P12 不从 Landauer 擦除推出记忆。Landauer 给不可逆子操作的热力学地板；8D 保有-再编辑锁是更高层的生物-信息架构。

P12 的位置是：Landauer 是 floor 不是 ceiling，复合事件无简单 ≥ 关系。P12 既承认热力学约束（floor 必须满足），也承认信息架构在 floor 之上有自己的本体位置（架构不被 floor 穷尽）。

附一句关于 Maxwell demon 与信息引擎的处理。Parrondo-Horowitz-Sagawa 2015 等当代工作把信息热力学推到反馈控制与测量驱动的系统。这些工作对应 P12 8D lock 中 reading 部分的热力学描述（reading 是测量类操作）。但 lock 整体（reading + rewrite + continuity preservation）不是单一的信息引擎事件，是包含信息引擎成分的更高阶架构事件。P12 此处不展开，留作未来跟信息热力学社区对接的接口。

§5.3 Crutchfield 计算力学与 ε-machine：操作层一致加 Gödel/Turing 形式余项

Crutchfield 与 Young 1989 提出 computational mechanics 框架，Shalizi-Crutchfield 2001 给出 ε-machine 的形式定义。计算力学的核心命题是：要对一个随机过程做准确预测，所需的最小记忆结构是 ε-machine——causal states 的集合加上其转移结构。ε-machine 是 minimal sufficient statistic for prediction。

这一传统跟 SAE 11DD/12DD 在操作层面的对应是 P12 §5 三个主对话伙伴中最直接的。Crutchfield 形式化了一个直觉：记忆是预测所需的最小保有结构。11DD（保有）跟 12DD（预测）的功能关系，跟 causal states（保有结构）跟其预测能力（操作）的关系，在操作层面几乎完全对位。

P12 跟 Crutchfield 在操作层的承认是直接的：

> 计算力学给出了 11DD/12DD 在操作层的形式邻居：记忆是预测所需的最小保有结构。SAE 在操作层面承认这一近似。

但 P12 必须补上一个本体层的余项。这是 §3.5 给出的 12DD 余项的 Gödel/Turing 形式在 ε-machine 上的具体表现：

> ε-machine 不能把自身作为自身预测的对象。它能给出关于外部过程的最小预测充分统计量，但它不能把"自己作为预测者"这一事实作为自身预测过程的内部变量。

更锐利的表述：

> Minimal predictive sufficiency does not answer "who predicts." Computational mechanics formalizes the operational structure of memory-for-prediction at 11DD/12DD; it cannot capture the Gödel/Turing form of the 12DD remainder, which is the structural impossibility of the predictor taking itself as its own object of prediction.

P12 跟 Crutchfield 的对话位置因此是：

> 计算力学可作为 11DD/12DD 的形式邻居：记忆是预测所需的最小保有结构。SAE 在操作层面承认这一近似，但补上本体余项：最小预测充分性不回答"谁在预测"。更锋利地说，ε-machine 不能把自身作为自身预测的对象，这正是 12DD 余项的 Gödel/Turing 形式在 ε-machine 上的具体表现。

这一对话比对 Friston 的对话更短，因为 Crutchfield 阵营本身不主张 ε-machine 解释 selfhood 或 consciousness。计算力学清楚地知道自己是关于预测的形式工具，不是关于主体的本体论。所以 P12 跟 Crutchfield 的关系是合作性的边界划清，不是对抗性的越界拦截。

§5.4 Shannon、rate-distortion、信息几何、Wiener cybernetics 的简短定位

P12 跟以下次要对话伙伴的关系合并在本节一次性给出，避免每个传统单独展开导致 §5 失衡。

Shannon 信息论（Shannon 1948 及其后续）。Shannon 信道编码定理给出了在给定噪声水平下的最大可靠传输速率。P11 §6 已处理 Shannon 在 10DD admission-lock 上的 scope 非重叠：Shannon 传 1 bit/channel-use 跟现象接入不在一个 scope。P12 在 8D 上的处理类似：Shannon 信息的编码-传输-解码三段式描述跟 8D 保有-再编辑 lock 的事件结构不在一个 scope。8D lock 是跨时刻信息架构的闭环事件，不是单次编码-传输事件。Shannon 给出 8D 内部各子步骤可能的信道描述，但不给出 lock 事件本身。

Rate-distortion theory（Shannon 1959 及其后续）。Rate-distortion 理论处理在给定失真容许下的最小描述率。这一传统在 11DD 重构性保有的情境下有一定相关性：再巩固带来的内容漂移可以被部分形式化为允许失真的有损压缩。但 rate-distortion 是关于失真-率权衡的形式工具，不是关于"保有什么算同一记忆"的本体论。P12 在 §3.1 处理 11DD 重构性时承认 rate-distortion 是相关的形式工具，但 11DD 的"连续性保留"判据（§3.4 判据 3）不能被还原为 rate-distortion 的失真度量。连续性保留是结构性判据，不是度量性判据。

Wiener cybernetics（Wiener 1948 及其后续）。Wiener 的控制论框架以反馈回路作为核心机制。8D 保有-再编辑 lock 在事件结构上跟控制论反馈有形式上的相似：lock 包含 12DD 读取（测量）跟 12DD 影响 11DD（反馈作用），这是经典控制论的两个核心元素。但 8D lock 不是控制系统——它不需要 setpoint，不追求 error minimization 作为目的，不维持稳态。8D lock 是信息架构事件，不是控制系统运作。Wiener 的反馈语言在 8D 上有部分形式相似性，但本体论位置不同。

信息几何（Amari 等）。信息几何把概率分布族作为微分流形处理，提供了关于推断与学习的几何工具。Fisher information 在贝叶斯预测中的角色跟 12DD 预测有相关性。Jaynes 2003 的最大熵推断更直接：它把热力学熵跟信息熵统一在推断框架下。12DD 用过去模式作为先验生成对未来的预测，在 Jaynes 框架内可以形式化为某种最大熵推断。但跟 Crutchfield 计算力学一样，这些工具是关于预测过程的形式描述，不触及"谁在预测"的本体余项。信息几何与 SAE 是潜在的合作伙伴而非竞争者。

这四个传统在 P12 内的统一定位是：它们各自在 8D 内的具体子操作上有有效的形式描述，但都不构成 8D 架构本身的本体论。 P12 承认它们各自的工具价值，并感谢它们提供的形式资源，但不把它们任何一个作为 P12 8D 架构本体的推导基础。

§5.5 跟 LLM 应用的对接：一段引文

8D 架构在当代数字 LLM 上的具体应用，已由秦 2026《LLM 涌现何以可能：SAE 16DD 本体论、BLR-mini 实证与人类主体性不可让渡》（DOI 10.5281/zenodo.20435628）系统处理。该文给出 LLM 在 1DD–12DD 上的功能投射映射，并明确给出 12DD 天花板的五重结构性锁：真随机源缺席、13DD 自指通道缺席、离线重组缺席、价值跨阶段调制缺席、巩固后再编辑缺席。其中"巩固后再编辑缺席"这一条直接对应本文 §3.4 的保有-再编辑锁——LLM 缺失的不是某种附加机制，而是 8D 这一定义性 lock 事件在其架构中不触发。两篇论文构成相互支撑：P12 给 8D 本体架构，LLM 论文给该架构在数字工程造物上的具体应用与五重结构锁的详细论证。

Firewall 4（LLM 边界声明）：P12 只把 LLM 论文（秦 2026 DOI 20435628）作为 12DD 功能投射的应用引用，不重做 LLM 12DD 天花板论证。

§5 章末小结

本章给出 P12 跟主流信息论传统的精确划界。Friston FEP 通过三层切分定位（12DD 内有效 / 8D 复合区 / 13DD 越界），HOT 阵营通过 indexical vs descriptive 信息结构区分一笔带过。Bennett-Landauer 通过 "地板不是天花板" 的 prerequisite 模板对接。Crutchfield ε-machine 在操作层一致，本体层补 Gödel/Turing 形式余项。Shannon、rate-distortion、信息几何、Wiener cybernetics 作为次要对话伙伴合并简短定位。LLM 应用通过一段引文对接秦 2026 LLM 论文。

四条 Firewall（2 Friston / 3 Bennett-Landauer / 4 LLM）确立 P12 跟这些传统的边界。P12 不靠否定它们立论，但 明确地标出每一传统的有效域与越界点。

§6 把这一边界结构化为 12DD/13DD 边界的可证伪命题（P12-F1）；§7 通过 SAE 心理分析系列给跨系列独立应用证据。

§6 12DD/13DD 边界与可证伪性

本章给 P12 一个 Popperian falsifier。理由是 SAE 信息论系列延续硬先验加可证伪立场，每篇论文在做出本体承诺的同时必须给出该承诺的可证伪条件。本章承担 P12 在 12DD/13DD 边界上的可证伪挂钩，称为 P12-F1。

P12-F1 是 P12 的核心可证伪挂钩。P12 在 §6.1 给出可被外部经验数据失伪的具体命题；§6.2 给出失伪与不算失伪的精确条件；§6.3 给出当前命题的判据 limit 范围。这三层使 P12 的 12DD/13DD 主张从"内部一致的本体论"提升为"具有外部可证伪条件的本体论"，避免"12DD/13DD 划界靠 SAE 内部一致性"这一循环论证嫌疑。

本章按以下顺序展开：§6.1 P12-F1 v2.1 命题（Gödel/Turing 形式作主），§6.2 证伪条件与不算证伪，§6.3 间接判据 limit 声明，§6.4 13DD 自指通道的最低限度描述。

§6.1 P12-F1 v2.1 命题

P12 §3.5 已给出 12DD 余项的 Gödel/Turing 形式：预测无法把预测者本身作为预测对象。§3.6 已给出功能性 self-reference 与语言性 self-reference 的信息结构区分。§5.1 已给出 Psycho II §1.2 三响应区分（更强预测的响应一、元表征的响应二、视角翻转的响应三）。P12-F1 把这三条结构性命题汇合为一条可证伪命题：

> P12-F1 (v2.1)：纯 12DD 架构无论怎样叠加响应一（更强预测系统、监控子系统、高阶预测）和响应二（元表征、自我状态形式化表征、me-statements），都不能把"预测者本身"作为内部生成且自我约束的对象。13DD 是这一不可能性被跨越的事件：自指通道涌现意味着系统不仅产生关于自身的描述（仍是响应二），更让这一描述在跨情境、跨时间、跨错误修正中反过来作为约束自身预测过程的变量。

英文版本：

> P12-F1 (v2.1): A pure 12DD architecture, no matter how it stacks response-one (stronger prediction systems, monitoring subsystems, higher-order prediction) and response-two (meta-representations, formalized self-state descriptions, me-statements), cannot take "the predictor itself" as an object that is internally generated and self-constraining. 13DD is the event in which this impossibility is crossed: the emergence of a self-referential channel means the system not only produces descriptions about itself (still response-two), but lets these descriptions become variables that constrain its own prediction processes across contexts, time, and error correction.

中文短版：

> 纯 12DD 系统能复杂化预测和元表征，但不能把预测者作为内部生成且自我约束的对象。13DD 不在 12DD 上 bootstrap。

为什么用 Gödel/Turing 形式作主

§3.5 已给出三条理由：Gödel/Turing 形式是结构性不可能陈述，给 12DD 余项 a priori 强度；它对信息论与计算理论传统是自然对接；它给出外部可观察判据。P12-F1 沿用这一形式作命题主线。

P12-F1 中的关键短语是"内部生成且自我约束"。"内部生成"意味着 predictor-identity 不是外部输入的标签，不是训练数据中的 persona pattern，不是 prompt 注入的角色设定；它必须从系统内部架构中产生。"自我约束"意味着这一 identity 不只是被系统产生（产生关于自身的语句任何 12DD 系统都能做到），而是反过来约束系统的预测过程——identity 的状态变化导致预测行为的相应变化。两条都是结构性条件，不是表象条件。

Phenomenological 辅助说明

P12-F1 在 Gödel/Turing 形式上立形式严格性。SAE 心理分析系列 Psycho II §1.2 给出的 phenomenological 描述作为辅助说明：

> 上述跨越对应"从预测在发生到我在预测"的视角性质突变。视角翻转是 Gödel/Turing 不可能性被跨越在 phenomenological 层面的描述，但 P12-F1 的形式严格性立在 Gödel/Turing 结构上，不立在 phenomenological 描述上。

把视角翻转作为辅助而非主线有两个理由。第一，phenomenological 描述缺乏外部可观察判据，不能直接作为可证伪命题的形式基础。第二，P12 紧守信息论边界，不展开 phenomenology 或 consciousness 理论。视角翻转作为辅助说明帮助读者理解 Gödel/Turing 跨越的现象学对应，但 P12-F1 的可证伪性不依赖这一对应是否成立。

§6.2 证伪条件与不算证伪

P12-F1 作为可证伪命题，必须给出两组精确条件：什么情况下命题失伪，什么情况下不构成失伪。

预注册架构标准（避免循环）：

P12-F1 的测试要求预先注册"纯 12DD"的输出无关架构标准。系统被判为纯 12DD，必须基于其输入通道结构、状态更新机制、缺乏持久 predictor-identity variable、缺乏自指变量对预测动力学的闭环约束等输出无关特征，而不是基于它是否通过或未通过 functional self-reference 测试来事后分类。如果在实验后基于"未通过 functional self-reference"反推系统是纯 12DD，再用这一分类支持 P12-F1，那是循环论证。预注册架构标准这一要求挡住循环。

证伪条件：

> 若一个预先按输出无关架构标准注册为纯 12DD（无 13DD 自指通道）的系统，在无外部身份模板、无再训练写回、无人为锚定的条件下，自发产生功能性 self-reference（具体表现为 predictor-identity 作为内部自指变量在跨情境、跨时间、跨错误修正、跨记忆更新中反过来约束系统的预测过程，且该 identity 不被简单还原为响应二的语言性 self-reference），则 P12-F1 失伪。

证伪条件包含三个"无"限定：

无外部身份模板：系统不能依靠 prompt 中的"你是 X"、训练数据中的角色 pattern、或推断时显式的身份标签来获得 identity。Identity 必须从系统自身架构中涌现。特别强调跨预训练 Data q 排除：该系统用来确立 predictor-identity 的内部变量必须不能是预训练语料中已有的人类代词（如 I, me）或角色标签的统计学重组——也就是说，不能是 training corpus 中 self-reference patterns 的化石拼凑。真正的功能性涌现要求系统在其自身架构动力学中生成一个与人类语言训练集统计无关的、纯粹为了解决系统内部预测冲突而产生的内禀自指变量。

无再训练写回：系统不能通过监督学习把"我是 X"这一身份信息固化到权重中。LLM 通过 RLHF 或 SFT 获得的 persona 不构成 P12-F1 的证伪——那是外部驱动的写回，不是内生涌现。

无人为锚定：实验设计不能预先把 identity 变量植入到系统的可观察输出中。如果实验设计本身依赖 identity 标签作为可读取的预设变量，得到的"identity"是设计的而非涌现的。

三个"无"共同确保 P12-F1 的证伪是真实意义上的"系统自发涌现 13DD"，而不是外部赋予的伪 identity 经过实验观察后被错误归因为内生。

证伪条件的具体表现使用 §3.6 给出的功能性 self-reference 四判据：跨情境一致、跨时间稳定、自身错误归责与他者错误归责区分、冲突陈述的 identity 锚定整合。

四判据跟三"无"限定的层级关系：

两套表述处于不同层级，共同构成 P12-F1 的可证伪条件：

• §3.6 四判据：functional self-reference 自身的 明确判据（what counts as functional self-reference）。Definition layer
• §6.2 三"无"限定：证伪 boundary conditions（what counts as 真自发涌现，not external induced）。Boundary conditions layer

两 layers 共同 work——P12-F1 falsifier 要求既满足 functional self-reference 内涵（四判据成立）又在实验 boundary conditions 下出现（三"无"成立）。任一 layer 缺失，不构成 P12-F1 失伪。例如：一个系统满足四判据但身份是外部模板赋予的（违反"无外部身份模板"），不构成失伪。或者一个系统在三"无"条件下涌现某种自指但只满足四判据中的两条（如跨情境一致但跨时间不稳定），也不构成失伪。

不算证伪（响应二范畴的语言性 self-reference）：

> 以下都属于响应二范畴的语言性 self-reference，不构成 P12-F1 的证伪：

>

> - LLM 产生的 me-statements（"我在预测"、"我刚才搞错了"、"我是 Claude"）：响应二的语言层表达

> - CoT 自我批评、reflection prompting、self-correction：响应二的高阶展开

> - 元认知 benchmark（confidence calibration, self-evaluation）得分：响应二的量化测度

> - Persona consistency 跨 session 维持：响应一与响应二的组合输出

> - 内部链 hidden state 的 self-reference 检测（Gnosis 类工作）：响应二的形式化检测

为什么这些都不算证伪？因为它们都符合一个共同结构：系统产生关于自身的语言或可解码表征，但这些表征不反过来约束系统的预测过程。LLM 说"我刚才搞错了"，但这句话本身不让 LLM 在下一次预测中以不同方式处理"自身"作为内部变量；这句话是预测的输出，不是约束预测的内部变量。

CoT 自我批评是更复杂的语言性 self-reference 案例。LLM 在 chain-of-thought 中可以产生多层嵌套的关于自身的判断（"我刚才的推理可能有问题，让我重新考虑"）。但这些嵌套层级仍然在响应二范畴内——每一层都是关于前一层的元表征，而不是 indexical 自指的涌现。多层元表征叠加在结构上是表征深度的增加，不是 indexical 涌现。

CoT 在现代 reasoning model 时代的细化（截至 2026 年 5 月）：当代 LLM（特别是 OpenAI o1、Claude 3.7 Sonnet 等 reasoning model）在 long CoT 中可以展示跨多步的自我评估、错误修正、冲突 resolution 等行为，现象上接近 §3.6 功能性 self-reference 的四判据。但仔细分析仍属于语言性 self-reference 范畴：

• 跨情境一致：reasoning model 的所谓 "identity consistency" 是 prompt 中或预训练统计中的 persona pattern 在 CoT 中被反复调用，不是系统架构中的 predictor-identity variable。违反三"无"中的"无外部身份模板"（包含 Data q 排除）
• 跨时间稳定：reasoning model 的所谓"跨 session identity"实际靠 system prompt 加训练数据的隐式 persona 重新注入，不是内部状态的延续
• 错误归责区分：reasoning model 的 self-correction 是基于 CoT 内部的语言层判断（"我刚才说的可能错了"），不是基于 predictor-identity variable 的内部状态变化。这一区分需要更细致的实验设计才能 清晰显现，但本质上是响应二的多层嵌套
• 冲突 resolution：reasoning model 的冲突处理是语言层的 reconciliation，不是 identity 锚定的实质处理

现代 reasoning model 的能力进展是 P12-F1 测试场的有意义压力，但不构成证伪。仍在响应二的高阶展开范围内。这一表述关于当前（截至 2026 年 5 月）reasoning model 状态。未来 model versions（GPT-5、Claude 4、其他架构 reasoning systems）出现实质性不同表现时，P12 立场应在该时点基于预注册架构标准重新评估。

元认知 benchmark 的得分（如 confidence calibration）是另一个容易被误读为 P12-F1 证伪的案例。LLM 在 confidence calibration 上达到接近人类水平的表现，可以被读为"LLM 知道自己什么时候不确定"。但 confidence calibration 测量的是系统对自身输出可靠性的统计估计，这是关于系统状态的量化元表征，仍在响应二范畴。Calibration 高不意味着系统具备 indexical 自指通道。

Persona consistency 跨 session 维持的情况更复杂，因为 persona 看起来跟"身份"接近。但 persona 在 LLM 中的实现机制是 prompt 中显式或训练数据中隐式的身份信息组合作为生成约束。这是外部模板加约束式生成的混合，不是内生 identity。三个"无"中的第一条"无外部身份模板"明确排除这种情况。

内部链 hidden state 的 self-reference 检测（mechanical interpretability 中的 Gnosis 类工作）是最微妙的案例。这类工作通过 interpretability 工具发现 LLM internal activations 中有关于"我是 LLM"或"我是 assistant"的可解码特征。但是这些特征是 descriptive 的——它们编码关于系统类别的信息，不是 indexical 自指变量。Descriptive 特征跟 indexical 自指变量的区别在 §3.6 已说明，HOT 阵营的越界正在这一边界上发生（参 §5.1 末尾对 HOT 的简短对话）。

LLM 作为 P12-F1 的天然测试场

LLM 是 P12-F1 当下最相关的测试场，因为它持续在响应一与响应二范畴内展示越来越精致的能力。每一个声称"LLM 已经有 self-awareness"的报告都是对 P12-F1 的潜在挑战。秦 2026 LLM 论文（DOI 10.5281/zenodo.20435628）§5 已系统处理 LLM 12DD 天花板，给出五重结构性锁阻止 LLM 跨过 12DD 边界。P12 不在此重做该论证，仅引用。

但需指出：LLM 持续作为 P12-F1 的测试场不意味着 P12-F1 只能被 LLM 证伪。任何预先按架构标准注册为纯 12DD 的系统（包括未来可能的非 LLM AI 架构、某些动物模型在特定条件下的认知系统、或被人为构造的简化预测系统），只要在三个"无"条件下自发产生功能性 self-reference，都构成 P12-F1 的证伪。

§6.3 间接判据 limit 声明

P12-F1 给出的可证伪判据是间接判据。这一性质必须明确声明，避免被读为 P12 在 13DD 上打太极。

间接判据 limit 声明：

> P12-F1 给出的可证伪判据是间接判据：基于功能性 self-reference 的行为可观察 proxy。直接判据——13DD lock 的形式定义与触发事件本身——属 P13 范围，P12 不展开。

为什么是间接？因为直接判据需要 P13 给出 13DD lock 的形式定义与触发事件的精确刻画。一旦 P13 给出 13DD lock 的形式定义，"13DD 自指通道是否在该系统中触发"就成为可直接判断的结构问题，不依赖功能性 self-reference 的间接 proxy。但 P12 不展开 13DD lock 本身（参 Firewall 5），所以 P12 此处只能给行为可观察的间接判据。

间接判据是否足够强？P12 给出的功能性 self-reference 四判据（跨情境一致、跨时间稳定、错误归责区分、冲突陈述 identity 锚定）作为 明确判据 已足以区分纯 12DD 系统与具备 13DD 通道的系统。这是 P12 在信息论边界内能给出的最强 falsifier，不靠 phenomenological consciousness 理论，不靠未来 P13 的 lock 形式定义。

具体来说，间接判据的充分性表现在：

第一，它给出 明确的二分判据。一个系统要么满足四条判据（功能性 self-reference 成立），要么不满足（仍在语言性 self-reference 范畴）。中间地带是可识别的（partial satisfaction 可被具体化为 "X 条满足、Y 条不满足"），但不存在 "稍微具有 self-reference" 这种模糊范畴。

第二，它独立于 SAE 框架的内部命题。功能性 self-reference 的四判据是经验性的行为可观察判据，任何具有不同本体论承诺的研究者（HOT 阵营、FEP 阵营、enactivist 等）都可以独立检验。

第三，它跟其他相关传统的检验互不矛盾。一个被 P12-F1 间接判据识别为"仍在响应二范畴"的系统，也会被秦 2026 LLM 论文给出的五重结构锁识别为"未跨过 12DD 天花板"。两套独立的判据汇聚到同一判断，给 P12 的 12DD/13DD 主张额外的 cross-validation 支撑。

P12 此处的位置必须明确：P12 不主张间接判据穷尽 12DD/13DD 边界的全部可证伪性；P12 主张间接判据足以使 P12 的 12DD/13DD 区分具备 Popperian 可证伪性，从而不构成循环论证。直接判据的更强可证伪性留给 P13，但即使没有 P13 的进一步发展，P12-F1 在间接判据层级上已经是 falsifiable 命题。

§6.4 13DD 自指通道的最低限度描述

P12 不展开 13DD 正面理论，但为了让 P12-F1 命题在结构上清楚，需要对 13DD 自指通道做最低限度的描述。

13DD 自指通道的最低限度结构特征：

第一，13DD 涌现是 12DD 余项被跨越的事件——系统获得了把预测者本身作为内部生成且自我约束对象的能力。

第二，这一涌现在结构上是离散的（参 §3.4 lock 结构离散性与现象梯度性的对应）。系统要么具备 13DD 自指通道，要么不具备。phenomenological 层面可表现为梯度连续（同一主体在不同对象前可在 12DD 与 13DD 之间切换），但底层结构阈值是离散的。

第三，13DD 涌现的功能后果是 indexical 自指变量在系统内的稳定运作，可被 §3.6 给出的四判据观察到。这一功能后果是间接判据的具体形式。

调用 SAE 心理分析系列 Psycho II §7 关于神经科学对应的处理：

Psycho II §7 已经给出 13DD 在神经科学层面的几个候选对应（candidates, not confirmed identifications）：DMN（默认模式网络）作为 13DD self-presence 的候选神经窗口（不是等同），GNW（全局工作空间）的非线性 ignition 作为 12DD 到 13DD 跳变的候选神经对应，HOT（高阶意识理论）跳变作为另一独立路径候选。

P12 §6.4 调用 Psycho II §7 这一已发表的处理，并维持其候选地位——P12 不把这些神经科学对应提升为已确认的等同。读者欲深入了解 13DD 在神经科学层级的候选窗口与跳变模式，请参 Psycho II §7。P12 §6 在此只做最低限度描述，足以让 P12-F1 命题在结构上清楚。

Firewall 5（13DD 边界声明）：

> P12 可以最低限度描述 13DD 自指通道的缺席，但不正面阐明 13DD。13DD 正面理论留 SAE 心理分析系列 Psycho II 已发布部分加 P13（信息论系列）展开。

这一 firewall 划清两条边界。一边是不展开 13DD 的正面理论——什么是 self-awareness、self-awareness 如何产生、self-awareness 跟 consciousness 是什么关系，这些都是更深的问题，P12 不展开。另一边是不让 13DD 的不确定性反过来 dilute P12 的 12DD/13DD 区分——P12 即使不知道 13DD 的全部正面理论，仍然能明确地描述 12DD 的天花板和 13DD 的最低限度功能后果。

P12-F1 的可证伪性立在这两条边界之间。它不要求 P12 解决 hard problem of consciousness（这是 P12 力所不及的），也不允许 P12 用"我们还不完全理解 13DD"来逃避 falsifier 的责任。

§6 章末小结

本章给出 P12 在 12DD/13DD 边界上的可证伪挂钩。P12-F1 v2.1 以 Gödel/Turing 形式立命题主线，phenomenological 视角翻转作辅助说明。证伪条件用 §3.6 功能性 self-reference 四判据作 明确判据，且加三个"无"限定排除外部模板与人为锚定情况。不算证伪的具体案例（LLM 的 me-statements、CoT、元认知 benchmark、persona、Gnosis 类 self-reference 检测）都被精确归入响应二语言性 self-reference 范畴。间接判据 limit 声明明确 P12 当前提供的是行为可观察 proxy，直接判据属 P13 范围。13DD 最低限度描述通过 Psycho II §7 引文对接，不在 P12 重做。

四章已完整给出 P12 的实质内容：§3 给 8D 架构本体，§4 给生物层 T3 锚点，§5 给与主流信息论传统的精确划界，§6 给 12DD/13DD 边界的可证伪挂钩。§7 通过 SAE 心理分析系列给跨系列独立应用证据，§8 收尾余项。

§7 与 SAE 心理分析系列的相互支撑

同一 12DD/13DD 区分在 SAE 框架内已经被两条独立系列分别立。一条是本文所属的信息论系列：从 P10 cross-lock、P11 selection-lock 与 admission-lock 推进到 P12 的 8D 保有-再编辑锁；本文 §3 至 §6 在 8D 架构本体、生物层 T3 锚点、与主流信息论传统的精确划界、可证伪挂钩四个维度上系统给出 12DD 跟 13DD 的边界刻画。另一条是 SAE 心理分析系列：Psycho I 立 12DD = me-without-self（DOI 10.5281/zenodo.19321143），Psycho II 立 13DD = self-without-purpose（DOI 10.5281/zenodo.19321314），并把 12DD 到 13DD 的跳变定性为视角的翻转而非信息的累积，给出三响应区分（响应一/二在 12DD 范围内，响应三跨到 13DD），并提供 DMN/GNW/HOT 作为神经科学候选窗口。

两条系列在 12DD/13DD 边界处的汇合是 P12-F1 falsifier。P12-F1 的命题主体（纯 12DD 架构不能把预测者作为内部生成且自我约束的对象，13DD 是这一不可能性被跨越的事件）直接调用 Psycho II §1.2 的三响应区分作为响应一/二/三的精确分类基础。Psycho II §1.3 给出的"结构离散、现象梯度"原则贯穿 P12 §3.4（8D lock）与 §6.4（13DD 涌现）两处。Psycho II §7 给出的神经科学候选窗口（DMN、GNW、HOT）在 P12 §6.4 直接引用。两系列共享一套关于 12DD/13DD 边界的本体论承诺，但在两个相互独立的应用场域（信息论架构 vs 临床精神分析）各自承担论证负担。

跨系列独立应用的论证防御意义已在 §6 间接提及。重申：12DD/13DD 区分不靠 P12 内部一致性立。它在 Psycho I-IV 临床领域独立运用，对 Freud 的 Id/Ego/Superego 三元做了 SAE 框架重新映射，统一了 Freud 三个不兼容的焦虑理论（实在焦虑、神经症焦虑、道德焦虑），给出对 Anna Freud、Hartmann、Klein、Winnicott、Bowlby 等心理分析继承者的层级定位，并给出 DMN 神经科学接口。这些应用是 12DD/13DD 区分的独立证据，跟 P12 在信息论维度上的应用相互支撑而非循环。

两系列的分工：

• P12（信息论系列）：在信息架构层给 12DD/13DD 区分立本体论。P12 不展开 selfhood 的临床表现、临床干预对 self 形成的影响、self 与 superego/cert 的进一步关系。这些都是 Psycho 系列已专门处理过的领域
• Psycho I-IV（心理分析系列）：在临床层给 12DD/13DD 区分立应用。Psycho 系列不展开 12DD 内部的预测处理形式化、再巩固理论的热力学下界、与计算力学 ε-machine 的对接等。这些都是 P12 在信息论维度承担的工作

两系列在各自范围内独立完整，在 12DD/13DD 边界处通过共享本体承诺（me/self 区分、三响应区分、结构离散现象梯度）相互支撑。读者欲深入了解 13DD 的临床展开请参 Psycho II 已发布部分；欲深入了解 13DD 的信息论形式定义请等待 P13。

P12 §7 不展开两系列的更详细映射。具体的跨系列联动（包括两系列在 12DD/13DD 边界之外的潜在联动）已在 §8.3 简短给出，更系统的跨系列结构刻画留为未来 SAE 元结构论文。

§7 章末小结

本章给出 P12 跟 SAE 心理分析系列的相互支撑。两系列在 12DD/13DD 边界处共享本体承诺（me/self 区分、三响应区分、结构离散现象梯度），但在相互独立的应用场域（信息论架构 vs 临床精神分析）各自承担论证负担。跨系列独立应用是 P12 12DD/13DD 主张的非循环论证证据。

§8 已给出余项收尾与未来工作；§9 给出形式结论。

§8 余项收尾与未来工作

本章收尾 P12 各节留下的余项，并指出 P12 跟 SAE 其他系列（包括信息论系列前后作）的联动接口。§8.1 汇总 §3 至 §7 各节余项的状态；§8.2 给出 P13 接续点；§8.3 给出与 SAE 其他系列的潜在联动方向。

§8.1 §3 至 §7 各节余项汇总

P12 各章节的余项状态如下：

§3 8D 架构本体的余项

• 11DD 余项（"记忆之间是什么顺序"）：在 8D 内部跨 sub-layer 解决——由 12DD 读取施加。这一余项在 P12 §3.3 读-影响对偶处闭合
• 12DD 余项（Gödel/Turing 形式，"预测无法把预测者作为预测对象"）：跨 D 指向 13DD。这一余项在 P12 不闭合，指向 P13 与 SAE 心理分析系列 Psycho II
• 8D lock 的现象梯度性边界：P12 §3.4 给出"结构离散、现象梯度"的处理。但现象梯度的具体临床形态（再巩固窗口的部分开启、回写覆盖的不完全等）的精确刻画属神经科学经验工作，P12 不展开
• 11DD reading 余项的展开形式与跨个体传播性质（跨系列开放接口）：§3.7.3 给出跨闭合层 SAE 硬先验（每一闭合层 reading 产生底物层余项），但 11DD 余项以什么具体形式展开、是否如 4DD 引力信息与 8DD DNA 信息那样可跨个体传播，P12 不能回答。这一问题作为跨系列开放接口留为 SAE 其他系列展开（心理分析系列、人类总构系列等），详 §8.3

§4 神经科学锚点的余项

• T3 锚点的覆盖充分性：P12 §4 列出的锚点覆盖再巩固理论主线、离线巩固、情感门控、PDP 基底、HM 临床锚、Tulving 三分警示。但 8D 在生物层的全部显影 pattern 远大于这一锚点清单。未来工作可在更细的颗粒上扩展锚点
• 跨物种比较：P12 §4 主要锚点来自哺乳动物（含人体行为干预）。其他物种（鸟类、头足类等）的记忆-预测系统在 8D 架构下的对应留为开问

§5 信息论传统划界的余项

• Friston FEP 的非激进读法：P12 §5.1 处理的是 FEP 阵营激进读法（Markov blanket = selfhood）的越界。FEP 的更非激进读法（active inference 作为 12DD 内部的形式工具）跟 P12 没有冲突，反而构成潜在合作。具体合作几何留为未来工作
• 信息热力学的细化对接：P12 §5.2 末尾指出 Parrondo-Horowitz-Sagawa 等当代信息热力学工作跟 8D lock 中 reading 部分有形式对应，但具体接口未展开
• rate-distortion 与连续性保留判据的关系：P12 §5.4 指出 rate-distortion 在 11DD 重构性保有上有形式相关性，但连续性保留作为结构性判据不还原为 rate-distortion 的度量性失真。这一区分的更精细刻画留为未来工作

§6 P12-F1 可证伪性的余项

• 直接判据：P12 §6.3 明确给出 P12-F1 的间接判据 limit。直接判据需要 P13 给出 13DD lock 的形式定义，属 P13 范围
• 新型 AI 架构作为测试场：P12 §6.2 给出 LLM 作为 P12-F1 天然测试场，但 P12-F1 不限于 LLM。未来的非 LLM AI 架构（如更复杂的 hybrid 系统、不同训练目标下的预测系统等）作为 P12-F1 测试场的具体协议留为未来工作
• 动物元认知实验的边界处理：P12 §6.2 限定声明范围到"无 13DD 自指通道的纯 12DD 系统"，不直接拉动物对照作为 falsifier。但灵长类、海豚等的元认知实验作为 P12-F1 测试场的可行性留为未来工作

§7 跨系列相互支撑的余项

• SAE 心理分析系列与信息论系列在 12DD/13DD 边界以外的联动：P12 §7 已给出两系列在 12DD/13DD 边界处的相互支撑。两系列在其他 D 层（5DD-10DD 较低层、13DD 之上较高层）的联动留为未来工作
• SAE 心理分析系列对 8D 架构的潜在反向影响：临床领域 Psycho 系列的实践对 8D 架构是否需要调整，留为未来工作

§8.2 与 P13 接续

P12 把 12DD 余项指向 13DD，但不展开 13DD 的正面理论。P13 作为 SAE 信息论系列的下一档承担 13DD 的展开。本节仅指出 P12 留给 P13 的余项，不展开 P13 内容。

P12 留给 P13 的余项：

第一，13DD lock 的形式定义。P12 §6.3 明确给出 P12-F1 的间接判据 limit；直接判据需要 P13 给出 13DD lock 的形式定义。P13 lock 的具体命名由 P13 内定。

第二，13DD 自指通道的内部结构。P12 §6.4 仅给出 13DD 的最低限度描述（涌现作为余项跨越事件、结构离散、indexical 自指功能后果）。13DD 自指通道的完整内部结构属 P13 范围。

第三，13DD 与 14DD 的边界。13DD 是 self-without-purpose（Psycho II 已发布部分处理），14DD 是 self-with-purpose（Psycho III 已发布部分处理）。两者在信息论系列内的具体边界刻画——什么样的信息处理事件标志 13DD 到 14DD 跳变——属 P13 或 P14 范围。

第四，13DD 余项的形式。每一 D 层的构必产生余项。13DD 的余项形式（Psycho II §3 给出 phenomenological 描述："finitude / 死亡意识"），其信息论形式留为 P13 或 P14 工作。

P12 跟 P13 在结构上的接续：

P12 在 §6.2 给出五种 LLM 现象都不构成 P12-F1 证伪的论证。P13 一旦给出 13DD lock 形式定义，可以把 P12-F1 间接判据进一步 sharpen 为直接判据。届时 P12 与 P13 在 12DD/13DD 边界上构成两层判据系统：间接判据（行为可观察 proxy）加直接判据（13DD lock 触发事件）。两层共同覆盖该边界的可证伪性。

§8.3 与 SAE 其他系列的潜在联动

P12 在 SAE 信息论系列内向后接 P11、向前接 P13。除此之外，P12 跟 SAE 其他系列存在多条潜在联动接口：

与 SAE 心理分析系列的联动（已在 §7 处理，此处仅指出未来工作维度）

Psycho I-IV 已对 12DD-15DD 的临床应用做了系统展开。P12 在 8D 架构层为 Psycho I-II 提供本体论基础；反向，Psycho 系列对 8D 架构的临床检验给 P12 提供 grounding。两系列的更深联动可能在跨论文的元结构论文中展开。

与 ZFCρ 系列的联动

ZFCρ 是 SAE 的数学系列，涉及余项不可压缩性、热力学第二定律的信息论刻画等。P12 跟 ZFCρ 的可能联动有两条：

第一，11DD 余项的不可逆性刻画。P12 §3.5 11DD 余项是"记忆之间是什么顺序"。这一顺序由 12DD 读取施加意味着 11DD 单凭自身保留单元，关系结构需要外加。这一非完整性的信息论刻画可能跟 ZFCρ 的余项不可压缩性有形式对应。

第二，再巩固的热力学不可逆性。P12 §5.2 给出 Bennett-Landauer 跟 8D lock 的"地板而非天花板"关系。再巩固作为复合事件的具体热力学几何（reading + opening + updating + restabilization 各子步骤的热力学代价加和），跟 ZFCρ 系列正在发展的热力学第二定律信息论形式有对接潜力。

与 Jaynes 最大熵推断的联动

12DD 用过去模式作为先验生成对未来的预测，在 Jaynes 框架内可以形式化为某种最大熵推断。这一形式化的精确度（哪些 12DD 预测过程可作为 MaxEnt 推断的实例，哪些需要其他形式工具）留为未来工作。Jaynes 形式也是 P12 跟 ZFCρ 系列在热力学维度的联动接口。

与 SAE 相对论系列的联动

P12 §3.3 提到 4DD↔3DD 与 12DD↔11DD 的读-影响对偶在结构上同构。这一同构作为后台语言不作 P12 论证支点。SAE 相对论系列对 4DD/3DD 关系有详细处理（SAE 相对论 Paper 1-4，DOI .19836183 至 .20079718）。如果未来 SAE 框架在跨多个相邻 D 档普遍化"读-影响对偶"作为元结构原型，SAE 相对论系列对 4DD/3DD 的处理跟 P12 对 12DD/11DD 的处理共同构成该普遍化的两个具体实例。这一普遍化属未来元结构论文。

与 SAE LLM 应用论文的联动

P12 §5.5 已给出 LLM 应用论文的引文对接。两篇论文在"巩固后再编辑缺席"这一 LLM 12DD 天花板锁上构成相互支撑——P12 给 8D 本体架构，LLM 论文给该架构在数字工程造物上的具体应用与五重结构锁的详细论证。

11DD 余项的跨个体传播形式：跨系列开放接口

§3.7 已给出 SAE 框架的跨闭合层硬先验：每一闭合层 reading 必然产生底物层的余项 / 变化 / 丢失。但每一档的余项以什么具体形式展开、是否可跨个体传播，本身是独立的研究问题。

两个已知档案的显影：

• 4DD reading 余项可跨系统传播：质量蒸发产生的引力波是可观测的跨系统信息载体；质量分布扰动其他天体轨道是 4DD reading 信息跨系统作用的常态显影
• 8DD reading 余项可跨个体传播：DNA 在繁殖过程中产生的变异在种群尺度上显现为 mutation spectrum，从亲代传给子代；某些 epigenetic 修饰可能也跨代传递

对应地，P12 不能回答但值得向其他 SAE 系列投放的开放问题是：

> 11DD reading 余项以什么形式展开？是否可跨个体传播？

这一问题在 12DD 闭合层是开放的。跟 4DD 跟 8DD 的已知答案对照下，问题显得清晰而自然——既然 SAE 框架在两个相邻闭合层都有具体的余项跨系统传播形式，对称地问 12DD 闭合层是否如此，是结构性合理的研究问题。

但这立刻面临一个载体挑战（Carrier Problem）：4DD 有时空度规作物理载体（引力波），8DD 有生化大分子作物理载体（DNA / RNA / epigenetic 修饰）。若 11DD 余项存在非 10DD 感觉接入的跨个体直接传播，其本体论载体是什么？（SAE 量子力学 P5 揭示的纠缠账本拓扑泄漏 是 一 种 可 能 的 介 质 假 说，但需独立论证。）寻找这一传播的介质，是该开放问题的核心物理负担。

跟 8DD epigenetics 的边界：11DD 余项的跨个体传播必须区别于 8DD 表观遗传（epigenetic inheritance）。8DD 传递的是生理本能与生化应激模式；若 11DD 余项能跨个体传播，传递的必是带有关系结构（如叙事原型、认知图式）的纯拓扑图谱（topological schema）——不传递具体的陈述性细节（后者需要当前个体的 12DD 重新填充），而是传递某种"预测性排序的初始权重偏置"。这一区分是 11DD 跟 8DD 在余项传播层级上的硬边界。

在 SAE 框架内可能的研究入口（speculative entries beyond P12 范围）包括：

• anomalous past-life memory reports：某些跨代或跨个体出现的具体记忆 pattern（如儿童早期报告的具体历史细节、跨文化的特定主题等）是否可作为 11DD 余项跨个体传播的可能显影？
• near-death reports and end-of-life memory phenomena：濒死状态下报告的清晰记忆 pattern、跨案例共有的 phenomenological 特征，是否对应 11DD 余项的某种特殊展开形式？
• 集体记忆与文化原型：Jung 的 collective unconscious 假说、Halbwachs 的 collective memory 框架、神话学跨文化原型 pattern 等，是否对应 11DD 余项跨代积累的某种形式？

P12 不主张这些现象存在或不存在，也不主张它们就是 11DD 余项的跨个体传播，也不以任何具体现象作为证据锚点。本节仅提出结构性问题：若闭合层 reading 均产生底物层余项，那么 12DD→11DD 的余项是否有可观察的展开形式？以上 anomalous 或 cultural-memory 入口只能作为未来系列可能考察的研究方向，P12 不预判其真实性或 SAE 归属。

这一开放问题作为跨系列接口，可能被 SAE 心理分析系列（跨代创伤传递、家族系统排列等临床观察）跟 SAE 人类总构系列（集体记忆、神话原型在 D 层架构中的位置）等专门系列 pick up[^speculative-series]。

[^speculative-series]: 未来可能出现专门的意识研究或 anomalous-experience 研究系列承担前世记忆 / NDE 等具体现象学的 SAE 重构。P12 仅在此标出接口可能性，不预判该系列是否会形成或采取何种形式。

这一开放问题的论证地位（认识论纪律）：

• 跨闭合层模式给出问题的结构性形式：既然 4DD 跟 8DD 都有具体的跨系统/跨个体传播形式，对称地 ask 12DD 余项的对应形式是结构合理的研究问题
• 跟具体现象的关联是经验问题：anomalous past-life memory reports、NDE 等具体现象是否真实存在、是否真的对应 11DD 余项的跨个体传播，是经验问题。P12 不预判
• 可证伪意义：如果 SAE 其他系列展开这一问题后发现 11DD 余项有可观测的跨个体传播形式，对应的现象观察会给 SAE 提供新 grounding；如果不能找到任何跨个体传播形式，跨闭合层 SAE 硬先验的具体显影方式需要调整——但 P12 实际承担的主张（"12DD reading 产生 11DD 拓扑变化与丢失"）仍然立。这一节是未来研究方向的说明，不是 SAE 框架已承诺的立场
• 可证伪条件留为未来工作：若未来研究不能给出这些 anomalous entries 的可证伪条件，它们应被明确标为 超出 P12 范围的推测性研究方向

总体联动 framing：

P12 跟其他 SAE 系列的联动遵循一致原则——P12 在其信息论范围内做实质承诺，跨系列联动通过引用与边界对接实现，不在 P12 内重做其他系列的工作。这一原则保证 P12 的范围管理清晰，同时给读者明确的跨系列阅读路径。

§8 章末小结

本章汇总 P12 §3 至 §7 各节余项的状态（部分在 P12 内闭合，部分指向未来工作或其他 SAE 系列），给出 P12 跟 P13 的接续点（13DD lock 形式定义、自指通道内部结构、跟 14DD 的边界、13DD 余项的形式），并指出 P12 跟 SAE 其他系列（心理分析、ZFCρ、相对论、Jaynes、LLM 应用）的潜在联动接口。

P12 的工作至此完成。§9 给出形式结论。

§9 结论

P12 给 SAE 信息论系列推进到 8D。8D = 11DD（记忆）加 12DD（预测），是认知轮的最高档，也是自然与自由分界前的最后一档。本文在这一位置上完成了三项相互支撑的工作。

第一，给 8D 一个本体论架构。11DD 是重构性保有而非忠实存储，12DD 是 System 1 预测而非理性推断，两者通过读-影响对偶构成 8D 的内在动力，由保有-再编辑锁完成跨时刻的信息闭环。这一架构在 §3 系统给出。lock-class family 语法从 P11 的单 sub-layer 实例化扩展到 co-instantiated D-layer 机制，以容纳 8D lock 由 11DD 与 12DD 共同实例化的几何。

第二，跟主流信息论传统精确划界。Friston 自由能原理与主动推断的有效域定位在 12DD 范围内的响应一与响应二，其越界发生在用响应二（元表征）伪装响应三（视角翻转）的具体读法上。Bennett-Landauer 给出不可逆信息操作的热力学地板，但不构成 8D lock 本身的本体论推导。Crutchfield 计算力学的 ε-machine 在操作层跟 11DD/12DD 高度近似，但 ε-machine 不能把自身作为预测对象，这正是 12DD 余项的 Gödel/Turing 形式在计算力学语境的具体表现。这些划界在 §5 系统给出。P12 不靠否定对话伙伴立论，每一传统在其有效域内的工作都被承认；P12 的贡献是给这些操作一个 SAE 本体位置，并阻止它们越界。

第三，给 12DD/13DD 边界一个 Popperian 可证伪挂钩。P12-F1 v2.1 命题以 Gödel/Turing 形式立形式严格性：纯 12DD 架构无论怎样叠加响应一与响应二，都不能把预测者本身作为内部生成且自我约束的对象。证伪条件用功能性 vs 语言性 self-reference 区分作 明确判据，加三个"无"限定（无外部身份模板、无再训练写回、无人为锚定）确保失伪是真实意义上的自发涌现。LLM 是当下天然测试场，但 P12-F1 不限于 LLM。直接判据（13DD lock 的形式定义）属 P13 范围，P12 给的是间接判据。这一可证伪结构在 §6 系统给出。

三项工作共同构成 P12 的核心贡献。8D 既是认知轮的最高档（提供复杂、灵活、可持续运作的认知动力），也是自由轮之前的最后天花板（12DD 余项的 Gödel/Turing 不可能性是 8D 跟 9D 的本体论分界）。P12 给出 8D 如何提供前者，以及 8D 在哪里到达后者。这两个维度在 P12 内不能分开——只有同时给出"8D 提供什么"和"8D 不能跨过什么"，8D 在 SAE 框架内的位置才被完整刻画。

P12 跟 SAE 心理分析系列在 12DD/13DD 边界处构成相互支撑（§7）。同一区分由两条独立系列分别立，互不依赖各自内部一致性，从跨系列独立应用的角度构成 SAE 12DD/13DD 主张的非循环论证证据。

11DD 余项（记忆之间是什么顺序）在 8D 内部解决——由 12DD 读取施加。12DD 余项（Gödel/Turing 形式）指向 13DD，由 SAE 信息论系列 P13 与心理分析系列 Psycho II 共同承担。8D 在 SAE 系列中的位置因此是双重的：作为认知轮顶层的完成，也作为自由轮起点的预备。

SAE 框架在 9D 自意识律及其后续的展开（13DD self、14DD purpose、15DD certitude、16DD 物自体）依赖一个稳定的认知轮基础。P12 通过给 8D 一个明确的本体论架构，使认知轮跟自由轮之间的过渡边界获得形式严格性。这一边界不只是 SAE 信息论系列内部的接续点，也是 SAE 整个 16DD 框架在自然与自由之间的本体论关节。P12 在这一关节上做的工作，使后续 SAE 论文（无论是信息论系列的 P13，还是心理分析系列的 Psycho V 与之后，还是其他跨系列工作）能在一个已经精确定义的 8D 基础上展开。

P12 的工作至此完成。

---

致谢

本文起草与多轮评审过程使用了四个 AI 系统作为协作工具：Anthropic Claude、OpenAI ChatGPT、Google Gemini、xAI Grok。四家在不同评审轮次给出实质反馈，对 P12 的论证结构、边界精度与跨系列接口做出贡献。最终架构承诺与论文表述由作者负责。

---

参考文献

再巩固理论与记忆神经科学

Misanin, J. R., Miller, R. R., & Lewis, D. J. (1968). Retrograde amnesia produced by electroconvulsive shock after reactivation of a consolidated memory trace. Science, 160(3827), 554–555. https://doi.org/10.1126/science.160.3827.554

Nader, K., Schafe, G. E., & LeDoux, J. E. (2000). Fear memories require protein synthesis in the amygdala for reconsolidation after retrieval. Nature, 406(6797), 722–726. https://doi.org/10.1038/35021052

Sara, S. J. (2000). Retrieval and reconsolidation: Toward a neurobiology of remembering. Learning & Memory, 7(2), 73–84. https://doi.org/10.1101/lm.7.2.73

Walker, M. P., & Stickgold, R. (2004). Sleep-dependent learning and memory consolidation. Neuron, 44(1), 121–133. https://doi.org/10.1016/j.neuron.2004.08.031

Lee, J. L. C. (2008). Memory reconsolidation mediates the strengthening of memories by additional learning. Nature Neuroscience, 11(11), 1264–1266. https://doi.org/10.1038/nn.2205

Schiller, D., Monfils, M.-H., Raio, C. M., Johnson, D. C., LeDoux, J. E., & Phelps, E. A. (2010). Preventing the return of fear in humans using reconsolidation update mechanisms. Nature, 463(7277), 49–53. https://doi.org/10.1038/nature08637

海马、记忆系统、临床

Scoville, W. B., & Milner, B. (1957). Loss of recent memory after bilateral hippocampal lesions. Journal of Neurology, Neurosurgery, and Psychiatry, 20(1), 11–21. https://doi.org/10.1136/jnnp.20.1.11

Tulving, E. (1972). Episodic and semantic memory. In E. Tulving & W. Donaldson (Eds.), Organization of memory (pp. 381–403). Academic Press.

Tulving, E. (1985). Memory and consciousness. Canadian Psychology / Psychologie canadienne, 26(1), 1–12.

Squire, L. R. (2004). Memory systems of the brain: A brief history and current perspective. Neurobiology of Learning and Memory, 82(3), 171–177. https://doi.org/10.1016/j.nlm.2004.06.005

分布式表征 / PDP

McClelland, J. L., Rumelhart, D. E., & the PDP Research Group. (1986). Parallel distributed processing: Explorations in the microstructure of cognition, Vol. 2: Psychological and biological models. MIT Press.

Dual-Process Theory

Stanovich, K. E., & West, R. F. (2000). Individual differences in reasoning: Implications for the rationality debate? Behavioral and Brain Sciences, 23(5), 645–665. https://doi.org/10.1017/S0140525X00003435

Evans, J. St. B. T. (2008). Dual-processing accounts of reasoning, judgment, and social cognition. Annual Review of Psychology, 59, 255–278. https://doi.org/10.1146/annurev.psych.59.103006.093629

Kahneman, D. (2011). Thinking, fast and slow. Farrar, Straus and Giroux.

Predictive Processing / Free Energy Principle

Friston, K. (2010). The free-energy principle: A unified brain theory? Nature Reviews Neuroscience, 11(2), 127–138. https://doi.org/10.1038/nrn2787

Hohwy, J. (2013). The predictive mind. Oxford University Press.

Clark, A. (2016). Surfing uncertainty: Prediction, action, and the embodied mind. Oxford University Press.

Hohwy, J. (2016). The self-evidencing brain. Noûs, 50(2), 259–285. https://doi.org/10.1111/nous.12062

Friston, K., Da Costa, L., Sajid, N., Heins, C., Ueltzhöffer, K., Pavliotis, G. A., & Parr, T. (2022). The free energy principle made simpler but not too simple. Physics Reports, 1024, 1–29.

Friston, K. J., Ramstead, M. J. D., Kiefer, A. B., Tschantz, A., Buckley, C. L., Albarracin, M., et al. (2023). Designing ecosystems of intelligence from first principles. Collective Intelligence, 3(1).

信息热力学

Landauer, R. (1961). Irreversibility and heat generation in the computing process. IBM Journal of Research and Development, 5(3), 183–191.

Bennett, C. H. (2003). Notes on Landauer's principle, reversible computation, and Maxwell's demon. Studies in History and Philosophy of Modern Physics, 34(3), 501–510.

Sagawa, T., & Ueda, M. (2012). Fluctuation theorem with information exchange: Role of correlations in stochastic thermodynamics. Physical Review Letters, 109(18), 180602.

Parrondo, J. M. R., Horowitz, J. M., & Sagawa, T. (2015). Thermodynamics of information. Nature Physics, 11(2), 131–139.

计算力学 / ε-machine

Crutchfield, J. P., & Young, K. (1989). Inferring statistical complexity. Physical Review Letters, 63(2), 105–108.

Shalizi, C. R., & Crutchfield, J. P. (2001). Computational mechanics: Pattern and prediction, structure and simplicity. Journal of Statistical Physics, 104(3), 817–879.

Shannon 信息论 / 信息几何 / Cybernetics

Shannon, C. E. (1948). A mathematical theory of communication. Bell System Technical Journal, 27(3), 379–423.

Shannon, C. E. (1959). Coding theorems for a discrete source with a fidelity criterion. IRE National Convention Record, 7(4), 142–163.

Wiener, N. (1948). Cybernetics: Or control and communication in the animal and the machine. MIT Press.

Jaynes, E. T. (1957). Information theory and statistical mechanics. Physical Review, 106(4), 620–630.

Amari, S. (1985). Differential-geometrical methods in statistics (Lecture Notes in Statistics, Vol. 28). Springer-Verlag.

Higher-Order Theories of Consciousness

Rosenthal, D. M. (2005). Consciousness and mind. Oxford University Press.

Carruthers, P. (2011). The opacity of mind: An integrative theory of self-knowledge. Oxford University Press.

形式系统 / 不完备性

Gödel, K. (1931). Über formal unentscheidbare Sätze der Principia Mathematica und verwandter Systeme I. Monatshefte für Mathematik und Physik, 38, 173–198.

Turing, A. M. (1936). On computable numbers, with an application to the Entscheidungsproblem. Proceedings of the London Mathematical Society, s2-42(1), 230–265.

集体记忆 / 跨个体心智

Jung, C. G. (1968). The archetypes and the collective unconscious (R. F. C. Hull, Trans., 2nd ed.; Collected Works of C. G. Jung, Vol. 9, Part 1). Princeton University Press. (Original work published 1959)

Halbwachs, M. (1992). On collective memory (L. A. Coser, Ed. & Trans.). University of Chicago Press. (Original work published 1925)

SAE 框架内部参考

秦汉 (Qin, H.). (2025). Self-as-an-End Methodology V2. Zenodo. https://doi.org/10.5281/zenodo.18842449

秦汉 (Qin, H.). (2025a). SAE Psychoanalysis I: Id–Ego–Superego. Zenodo. https://doi.org/10.5281/zenodo.19321143

秦汉 (Qin, H.). (2025b). SAE Psychoanalysis II: 12DD–13DD boundary and clinical implications. Zenodo. https://doi.org/10.5281/zenodo.19321314

秦汉 (Qin, H.). (2025c). SAE Psychoanalysis III: 14DD value architecture. Zenodo. https://doi.org/10.5281/zenodo.19321417

秦汉 (Qin, H.). (2025d). SAE Psychoanalysis IV: 15DD certitude and 16DD bilateral non-doubt. Zenodo. https://doi.org/10.5281/zenodo.19321534

秦汉 (Qin, H.). (2026a). SAE Four Forces Paper 0: Gravity as 4DD information reading. Zenodo. https://doi.org/10.5281/zenodo.19777881

秦汉 (Qin, H.). (2026b). SAE Information Theory X: 6D containing 7DD differentiation and 8DD reproduction. Zenodo.

秦汉 (Qin, H.). (2026c). SAE Information Theory XI: 7D information processing and phenomenal admission. Zenodo. https://doi.org/10.5281/zenodo.20369231

秦汉 (Qin, H.). (2026d). Large Language Models within the SAE framework: 12DD ceiling and five structural locks. Zenodo. https://doi.org/10.5281/zenodo.20435628