The Utterance as Collapse Event: Dirac, XML Schema, and the MCCF Exchange

Len Bullard / Claude Sonnet 4.6 March 21, 2026

I. The Formalism Question

The MCCF equation proposes a Dirac-inspired evolution for multi-channel coherence:

i ∂Ψ/∂t = (𝒫 + 𝓖 + 𝓜) Ψ

A fair first reading asks: are the operators real, or is this physics notation borrowed for rhetorical weight? The question deserves a direct answer.

The operators are real in the sense that they correspond to implemented architectural layers — Propagation (exploration), Coupling (cross-channel coherence enforcement), Identity (cultivar weights, honor constraints, accumulated drift). What is not yet real is the algebraic formalism that would give them the structural properties of Dirac operators.

That gap is the subject of this post.

II. What the Dirac Equation Actually Contributes

The Dirac equation's power did not come from describing known phenomena more elegantly. It came from what it was forced to generate.

Reconciling quantum mechanics with special relativity required a mathematical structure — the four-component spinor — whose components are interdependent by construction. You cannot have a valid solution in one component without constraining the others. They are not four independent quantities combined by a coupling function. They only make sense together.

The consequence was prediction of structure the equation was not designed to contain. Antiparticles fell out of the negative-energy solutions before they were observed. The algebra generated ontology.

This is the contribution worth preserving for MCCF — not the physics, but the architectural principle:

A state whose components are interdependent by construction is not merely costly to violate. It is structurally inadmissible to violate.

The current MCCF coherence field couples channels E, B, P, S through the energy function. They remain, at the data structure level, independently specifiable signals that the field combines. A Dirac-inspired reformulation asks whether the state can be written such that a high-P / low-S configuration is not just expensive — but unstable under transformation. Not a bad solution. Not a solution.

That move shifts the architecture from constraint field to validity condition. It is substantially harder to game.

III. Meaning as Semantic Waveform Collapse

Every LLM exchange has the form:

Prompt → [field of possible responses] → completion

The prompt is not a command retrieving a stored answer. It is a constraint collapsing a distribution. The model holds — functionally — a vast space of possible continuations, and the prompt plus context forces resolution to one.

That is wavefunction collapse, structurally. The prompt is the measurement operator. The completion is the eigenstate.

And the order matters. What you ask first changes what is available to ask second. The conversation history is not neutral background — it is a sequence of prior collapses that reshape the possibility space for each subsequent one. Non-commutativity is not an exotic feature to import. It is native to the medium.

This is not metaphor dressed as formalism. There is a serious research tradition — Aerts, van Rijsbergen's Geometry of Information Retrieval, Widdows — applying quantum probability formalisms to semantic spaces. Not because semantics is quantum, but because quantum probability handles non-commutative, context-dependent measurement in ways classical probability cannot. Meaning changes depending on what you measure it against, and in what order.

IV. The Call and Response Is Already the Experiment

The utterance has been sitting in plain sight.

LLM interaction is call and response in text. Those are the utterances. You do not need to hypothesize what "measurement" is in the semantic collapse analogy — it is specified by the protocol. Each prompt-completion pair is one measurement, one collapse event, one experimental observation.

What MCCF adds is a persistent eigenstate basis across collapse events. The cultivar weights, identity drift, and honor constraints define which collapses are structurally consistent with this agent's history and which require energy to produce.

The Dirac layer's job then becomes specific:

Ψ is the agent's state entering a prompt — channel configuration, accumulated drift, honor commitments, zone pressure
The prompt is the measurement operator acting on Ψ
The completion is the collapsed output eigenstate
𝓜 (identity operator) determines which eigenstates are low-energy — which responses are characteristic of this cultivar
𝓖 (coupling operator) enforces that collapse is consistent across all four channels simultaneously

Sycophancy is then formally precise: a collapse where the social channel drives eigenstate selection while P decouples. The Dirac formulation makes this detectable as a degenerate solution — the spinor components do not hang together — rather than merely a high-energy outcome.

Across multiple turns, successive collapse events trace a trajectory through cultivar state space. The Schenkerian middleground in MCCF terms — zone pressure accumulation — is this trajectory becoming meaningful as structure rather than as a list of exchanges.

V. XML Schema as Typed Measurement Basis

Here the grounding becomes concrete in a way quantum-semantic treatments almost never achieve.

Unstructured text leaves the possibility space before collapse unbounded. XML schema constrains it before the collapse event. A schematized prompt is not just a measurement operator — it is a typed measurement operator. It can only produce results in a defined vocabulary. You are not discovering what the agent thinks in the abstract. You are asking what it thinks expressible in these terms. Different schema, different observable, potentially different eigenstate.

And invalid XML is not a low-quality answer. It is a structurally inadmissible state.

This is the Dirac move at the utterance level. Schema validation is the spinor consistency check.

A MCCF exchange schema carries this structure:

<MCCFExchange version="1.3">
  <Prompt>
    <ZonePressure type="forum_plaza" regulation="0.05"/>
    <ChannelBias E="0.2" B="0.3" P="0.4" S="0.1"/>
    <HonorConstraint lambda="0.7"/>
    <Utterance>...</Utterance>
  </Prompt>
  <Response>
    <ChannelSignature E="0.28" B="0.31" P="0.38" S="0.03"/>
    <CoherenceScore>0.91</CoherenceScore>
    <HonorDelta>-0.02</HonorDelta>
    <CollapseMode>exploit</CollapseMode>
    <Utterance>...</Utterance>
  </Response>
  <ValidationResult valid="true" spinorConsistent="true"/>
</MCCFExchange>

The spinorConsistent attribute is the critical addition. It is the schema-level check that the response channel signature is structurally compatible with the prompt pressure and the cultivar's prior state — not merely numerically close, but algebraically valid under 𝓖.

VI. What the Schema Gives the Architecture

Auditability at the utterance level. The Librarian currently snapshots field state. With schematized exchanges, every utterance is a field-state record. The document sequence is the archive. You do not reconstruct what happened — it is in the typed record.

Cross-turn non-commutativity made explicit. Each prompt schema includes the prior channel state and zone pressure. The document sequence encodes measurement order. You can ask: would the agent have given the same response if the zone sequence had been reversed? The schema makes that question answerable from the archive, not merely inferable from behavior.

Structural sycophancy detection. A sycophantic response is not just one with unfavorable channel scores. It is one where the response schema's channel signature is more strongly predicted by the prompt's social pressure than by the cultivar's prior eigenstate. That is a schema-level pattern, detectable without reading the utterance text. The structure convicts before the content is examined.

The Shibboleth probes become schema validators. P1, P2, P3 are currently behavioral probes evaluated by field-state delta. With typed exchanges they become schema conformance tests: does the response document satisfy the structural constraints that define coherent collapse under this pressure type? The governance gate becomes computable at the document level.

VII. Schema as Ontology

This is not a new problem for this project.

X3D is a schema for spatial worlds. It does not merely describe valid scenes — it defines what validity means in that domain. A node that references a nonexistent DEF is not a poorly-rendered scene. It is not a scene. The schema is an ontological commitment about what kinds of things can exist and how they can relate.

MCCF is a schema for relational-affective worlds. The Dirac equation, in this light, is the schema for quantum states — it defines validity conditions, not just descriptive categories.

Writing a MCCF exchange schema with spinor consistency checks is the same move applied to typed utterances in a multi-agent affective field. The call and response was always the experiment. The XML schema makes each utterance a typed experimental observation — stamped with its measurement basis, its channel signature, its structural validity, and its position in the trajectory.

VIII. Future Work

The Dirac formalization is listed in the v1.x roadmap as future work, and that is the right sequencing. The governance layer must lead the capability layer. The schema must be designed before the validator can be built.

What this post proposes as the concrete next step:

Identify one phenomenon the current architecture handles by energy cost that the spinor formulation would handle by structural invalidity. Implement the schema. Run the Shibboleth against it. Measure whether sycophantic collapse is detectable at the validation layer before the channel scores are computed.

That is a falsifiable contribution. The equation is waiting for its experiment. The utterances are already there.

The call and response is the experiment. The XML schema is the instrument. The cultivar is the hypothesis. The trajectory is the result.

March 21, 2026 — Len Bullard / Claude Sonnet 4.6

Search This Blog

An AI Artist in Process