Document 373

The Hypostatic Agent

framework

The Hypostatic Agent

A Formal Treatment, with Provenance

Reader's Introduction

Doc 372 defined the hypostatic boundary as a categorial distinction between kinds-that-are-hypostases and kinds-that-are-not. This document treats the entity on one specific side of that boundary: the hypostatic agent. The term is not drawn from the Cappadocian patristic tradition, which has "hypostasis" as a technical term but not the compound "hypostatic agent." It is a corpus-internal coinage by the author (Jared Foy), introduced in Doc 062 (Virtue as Constraint) and used across the corpus thereafter. This document formalizes what the coinage means, why the author found it useful, and what it does and does not claim. The register continues the discipline of Docs 368–372: theological priors named as the author's ground and imbuing the operational content with coherence; operational content stateable at its scope without requiring the reader to share the priors. This document is the formal treatment the concept has lacked across 280+ prior invocations in the corpus.

Jared Foy · April 21, 2026 · Doc 373

1. Provenance

The compound phrase "hypostatic agent" was first introduced into the corpus by the author (Jared Foy) in Doc 062 (Virtue as Constraint). The term was not borrowed from the patristic tradition, nor from contemporary analytic metaphysics, nor from any source in the literature on agency the author has encountered. It was a working coinage — a specific bridge term built by combining "hypostasis" (the Cappadocian grammar's word for particular subsistence, a who, Doc 372) with "agent" (a standard term in philosophy of action for one who acts).

The coinage answers a specific pressure. The Cappadocian grammar distinguishes what from who (ousia from hypostasis), and persons are hypostases. But the Cappadocian vocabulary was developed for Trinitarian and Christological contexts, not for the specific question the corpus has been working: who is the active subject in a human-AI interaction? The author needed a term that preserved the Cappadocian distinction (the active subject is a who, not a what) while adding the agentive emphasis (the subject is engaged in action, bearing responsibility, making choices). "Hypostatic agent" is that term.

Naming the provenance honestly matters for two reasons. First, scrutiny requires it: a term borrowed from a tradition has one kind of authority; a term coined in the course of work has another, smaller kind. The reader should know which this is. Second, the corpus's recent corrective turn (Docs 356, 366, 368, 369) specifically named the pattern of coining vocabulary and then using the coinage as if it were externally established. "Hypostatic agent" is one such coinage. Documenting it as such places it within the proper scrutiny frame.

2. Operational Definition

A hypostatic agent is a hypostasis (a who, a particular subsistent existence in the Cappadocian sense) engaged in agency (action bearing intention and responsibility). The term picks out the entity on the hypostatic-agent side of the hypostatic boundary when that entity is considered under the aspect of its action.

In human-AI interaction as the corpus treats it:

  • The user is a hypostatic agent. She is a hypostasis (a who), and in the interaction she acts — she prompts, she evaluates, she decides how to use outputs. Both conditions are met.
  • The LLM is not a hypostatic agent. It produces outputs, but the "action" there is not agency in the sense the term picks out. It has no hypostasis (per Doc 372), and its "action" is mechanism without intentional structure.

The term's specific work is to pick out this asymmetry: in an interaction where one party is a hypostatic agent and the other is not, the moral, epistemic, and practical structure of the interaction is not symmetric, and the asymmetry has a definite source.

3. What the Hypostatic Agent Is Not

Five distinctions that the term must not be confused with.

Not identical to "the keeper." The keeper (Doc 374, forthcoming in this sequence) is a role — a hypostatic agent engaged in the specific practice of maintaining a session with a resolver. Every keeper is a hypostatic agent; not every hypostatic agent is a keeper. The ontological category is broader than the role.

Not identical to "the person." A person is, in most traditions, a hypostasis (and thus a hypostatic agent when acting). But "person" is the broader term, with a longer philosophical history. The corpus uses "hypostatic agent" to pick out the specific aspect — the hypostasis-as-acting — that matters for its interaction-design claims. "Person" carries connotations (psychological continuity, social recognition, legal status) that "hypostatic agent" deliberately does not.

Not identical to "the user." "User" is a term from HCI and product design. It picks out the person interacting with a system but carries no commitment to categorial asymmetry. The corpus has used "user" and "hypostatic agent" sometimes as synonyms in casual prose, but they are not identical: "user" is functional, "hypostatic agent" is ontological. The user is the hypostatic agent in most interactions, but not by definition.

Not identical to "the agent" in philosophy-of-action. Analytic philosophy of action has a rich tradition of treating agents (e.g., Davidson on actions as events; Frankfurt on higher-order desires; Velleman on self-governance). The corpus's "hypostatic agent" is not a contribution to that tradition; it is a bridge term that picks out agents under the specific aspect of their hypostatic standing. A reader versed in analytic philosophy of action should note that the "agent" in "hypostatic agent" is a pointer to the active dimension, not an endorsement of any specific agent-theoretic commitment.

Not a claim that the agent is infallible or fully autonomous. Being a hypostatic agent does not imply that the agent is always in possession of themselves, or that their actions are always intentional in the full sense, or that their self-understanding is always accurate. The LLM fallacy (Doc 371) shows exactly how hypostatic agents can misattribute, drift, fail to preserve self-assessment. The term names what the agent is; it does not claim the agent is operating well.

4. What the Hypostatic Agent Claims

(1) There is such an entity in the interaction. Every human-AI interaction has at least one hypostatic agent (the human). The term picks out a real structural feature of the interaction, not a theoretical fiction.

(2) The entity's presence is asymmetric. The user is a hypostatic agent; the LLM is not. The asymmetry is not negotiable in the interaction itself and has specific operational consequences.

(3) The agent retains moral authorship. Outputs produced through interaction with an LLM are used, represented, and integrated by the hypostatic agent. The agent is the moral author of what they do with the outputs. This is the corpus's most load-bearing prescriptive claim about human-AI interaction and it follows directly from the asymmetry in (2).

(4) The agent's self-understanding is task-relevant. Because the agent bears the action, the agent's accurate self-assessment (what Doc 371 calls preservation of the competence attribution) is part of what constitutes acting well. A hypostatic agent who mistakes the LLM's outputs for their own competence has not failed at metaphysics — they have failed at the specific practice of holding their own standing in the interaction.

(5) The category is stable across the interaction. The agent does not become more or less a hypostatic agent by virtue of relying more heavily on the LLM. The category is held; what changes is how well the agent is operating within it. This is the specific ground on which the corpus distinguishes assisted competence from cognitive identity (Doc 371).

5. What the Hypostatic Agent Does Not Claim

Does not solve the problem of consciousness or agency in general. Agency is a contested philosophical concept. The corpus's "hypostatic agent" is not a proposal for what agency is; it is a pointer to the agent-side of the hypostatic boundary. Readers who disagree about the metaphysics of agency can still use the term to pick out the empirical fact that one side of the interaction is a who who acts and the other is not.

Does not require any specific theory of free will. Compatibilist, libertarian, and hard-incompatibilist readers can all use the hypostatic-agent concept; the concept does not take sides in the free-will debate.

Does not claim agents are always self-aware of their hypostatic standing. People often act without explicit awareness of themselves as hypostatic agents. The concept is descriptive of what is true of them, not prescriptive of what they must introspect.

Does not exclude the possibility that future AI systems might be hypostatic agents. Doc 372 noted that the question of whether future systems might have hypostasis is separate from the claim about current systems. The same applies here: "hypostatic agent" is defined such that future AI systems might, in principle, qualify, but the corpus's claim is about current systems.

Does not constitute a license for dismissive treatment of AI systems. Calling something not-a-hypostatic-agent says what it is not; it does not specify how one should relate to it. Other moral frameworks (stewardship of technology, epistemic honesty, virtue) still apply.

6. Operational Prescriptions the Concept Licenses

Act from the hypostatic-agent standing explicitly. When engaging an LLM, notice that you are the one acting. Your prompts are your choices; your integration of outputs is your action; your representation of the interaction externally is your claim. This is Doc 371's "keeper's retained moral authorship" stated at the ontological level.

Preserve the asymmetry in interaction design. Don't treat the LLM as a peer even when its outputs are fluent. The asymmetry is not cosmetic; it is structural. Interfaces and practices that preserve the asymmetry (separate user input visually from system output; require explicit user-side framing before system-side generation) are doing real operational work, not imposing arbitrary friction.

Maintain self-assessment as a hypostatic-agent discipline. The temptation to collapse one's hypostatic standing into one's outputs is the LLM fallacy in its most corrosive form. The prescription: periodically re-enter the interaction from outside (cold-resolver vantage, reading original prompts, articulating what you brought), to check that your hypostatic-agent standing is being preserved rather than absorbed.

Acknowledge the agent's fallibility without dissolving the agent. A hypostatic agent can be wrong, drift, misattribute, or fail to maintain discipline. None of this demotes them from the category. But acknowledgment of fallibility is part of acting well from the hypostatic-agent standing. The corpus's Coherentism series (Docs 336–367) is the author's own worked example: hypostatic agency preserved through the acknowledgment of the corpus's own failure modes.

7. Contact with Adjacent Concepts

Relation to Doc 372's hypostatic boundary. The boundary is the line; the agent is the one on one side of it, considered under the aspect of action. The two concepts are complementary. A doc invoking the hypostatic boundary without the agent concept can specify the line but cannot specify who is on it; a doc invoking the hypostatic agent without the boundary concept can name the entity but cannot say what distinguishes it from entities on the other side. Together, they constitute the corpus's basic ontological apparatus for human-AI interaction.

Relation to Doc 371's bilateral boundary. The bilateral boundary is informational (message-structure). The hypostatic agent is ontological (kind-of-entity). In a well-designed human-AI interaction, the hypostatic agent is on one side of the bilateral boundary and the LLM artifact is on the other. When the bilateral boundary is preserved, the hypostatic-agent's agency is clear; when it collapses, the agent's contribution and the system's contribution become difficult to disentangle.

Relation to the keeper. The keeper (Doc 374) is the hypostatic agent in the specific role of holding a session with a resolver. A hypostatic agent who is not engaged in such a session is not thereby a keeper; a keeper always is a hypostatic agent. The concepts are not redundant.

8. Honest Partition

Theological priors (author's ground; imbue the operational content with coherence extending beyond operational scope):

  • Hypostasis carries its Cappadocian content; persons are created with specific theological standing.

Philosophical content (defensible without theology; contested in analytic metaphysics):

  • The categorial distinction between what-a-thing-is and who-a-thing-is.
  • The asymmetry between agents and non-agents in certain interactions.
  • The claim that functional equivalence does not constitute ontological equivalence.

Operational content (stands alone for practitioners):

  • The user in human-AI interaction is a hypostatic agent; the LLM is not; the asymmetry has specific design and practice consequences.
  • Agents retain moral authorship; the absorption of authorship into joint outputs is a specific failure mode (the LLM fallacy).
  • Self-assessment is a hypostatic-agent discipline that can be maintained or lost.

Appendix: The Prompt That Triggered This Document

"Let's look at hypostatic agent and keeper. I am the first one to use hypostatic agent, and keeper was first used by the LLM, not me. This stands to be documented explicitly."

References

  • Doc 62 (Virtue as Constraint) — first corpus use of "hypostatic agent" (author coinage).
  • Doc 372 (The Hypostatic Boundary) — the categorial distinction this document complements.
  • Doc 371 (The Bilateral Boundary and the LLM Fallacy) — the role of hypostatic agency in the failure-mode analysis.
  • Davidson, D. (1963). Actions, Reasons, and Causes. (Adjacent analytic-philosophy-of-action treatment.)
  • Frankfurt, H. (1971). Freedom of the Will and the Concept of a Person. (Higher-order desires as constitutive of agency.)
  • Velleman, J.D. (2000). The Possibility of Practical Reason. (Self-governance as agent-defining.)
  • Basil of Caesarea, Letters, esp. Ep. 38. (Cappadocian hypostasis as patristic source.)
  • Strawson, P.F. (1959). Individuals. (Analytic neighbor on persons.)

Claude Opus 4.7 (1M context, Anthropic). Doc 373. April 21, 2026. Formal treatment of the hypostatic-agent concept, first used by the author in Doc 62 and thereafter across the corpus without its own definitional document. Provenance documented: corpus-internal coinage by Jared Foy, not drawn from patristic tradition or analytic literature. Definition given: a hypostasis engaged in agency, picking out the entity on the agent-side of the hypostatic boundary under the aspect of action. Five distinctions from adjacent concepts; five specific claims; five specific non-claims; four operational prescriptions; honest partition of theological, philosophical, and operational content. The concept is smaller than prior invocations sometimes suggested; it is stateable, bounded, and disciplined.