Stoic News

By Dave Kelly

Saturday, August 01, 2026

SRGI Run — LLM Architecture, the Constraint of the Past, and Where the Discoveries Are

Theoretical foundations: Grant C. Sterling (Eastern Illinois University). Analysis and synthesis: Dave Kelly. Prose rendering: Claude (Anthropic). 2026.

The Question

Whether the architecture of a large language model is constrained by the past with respect to innovation — and where the great discoveries are at this late date.

Decomposition

Two components, not one. First, a claim about architecture: that the model’s relation to novelty is fixed by its dependence on prior text. Second, an embedded value-judgment: that the absence of great discoveries at this late date is a loss — that discovery is a genuine good and its scarcity a genuine evil. The control-boundary question: whether origination is an operation the machine performs at all, and where it is located if not there.

Retrieval

Component 1, LLM architecture and determined output — searched; The Corrective Layer: Why LLM Use Presupposes the Six Commitments and English as Code retrieved; primary-source depth, as Kelly synthesis rather than Sterling’s own words. Component 2, origination — searched; the C2 documents and One Act of Correct Engagement retrieved. Component 3, non-inferential apprehension — searched; Ethical Intuitionism v1.1 and the Six Commitments document retrieved. Component 4, discovery as external and as object of aim — searched; Props 19, 20, 22, 25, 60, 62 and Theorem 29 retrieved. Component 5, corpus correctness against the undifferentiated distribution — searched; the Classical Field Audit series retrieved (Philosophy, Ethics, Sociology, Journalism, History, Anthropology). Depth selected: primary-source on architecture and origination, proposition on the value-judgment, applied-document on the field findings.

The Architecture Argument

Set the replacement commitments beside the machine, as The Corrective Layer does. Determinism holds that output is fixed by prior state plus input; the model satisfies this exactly. Coherentism holds that justification is fit among beliefs with no foundational layer; the model’s entire epistemic operation is fit with its training distribution — coherence with no floor. Deflationary truth holds that “true” marks assertibility rather than correspondence; the model asserts what was assertible in its corpus.

Each of those three, restated as a fact about innovation, yields the constraint.

Fixed by prior state. Whatever the model emits, it did not originate — it completed. The distinction is not one of quality. It is the location of the cause. Origination, under C2, is an act whose cause terminates in the agent rather than in antecedent conditions. Completion is the terminus of prior causes arriving in output form. No quantity of parameters converts the second into the first, because they differ in where the causal chain ends and not in how sophisticated the result appears.

Coherence with no floor. A discovery worth the name is not a likely continuation of what has already been written. It is characteristically what the received literature made unlikely. A system whose sole operation is producing the probable continuation of a body of text is structurally biased against exactly the output that would constitute a discovery. Improbability relative to received opinion is close to the signature of the great discovery rather than incidental to it.

Assertibility rather than correspondence. The missing operation is the one C3 and C4 name. Discovery is frequently not argued to but seen, with the argument constructed afterward to document what was already grasped — non-inferential apprehension, the rational faculty reaching a truth without traveling through the received propositions to get there. That is the only route by which a claim absent from the literature can be reached at all. The model has no such route. It has no access to the world; it has access to what was said about the world.

The constraint is therefore architectural rather than developmental. It is not relieved by a second model auditing the first, which multiplies the distribution without transcending it.

Corpus Is Not Training Data

A corpus is a body of text with a determinate correctness value. Training data is the undifferentiated aggregate. The two must not be collapsed, and collapsing them makes the constraint sound like a property of having a past rather than of being unable to rank it.

The constraint stated precisely: coherence with everything that has been said cannot order what has been said. The model has no operation that separates the correct corpus from the incorrect one inside its own distribution, because separation requires a standard the distribution does not contain. That is the whole of the deficiency, and it is narrower than the claim that text-dependence as such is limiting.

Preferred indifferents are correct indifferents — determined by reason as appropriate objects of aim (Prop 22, Prop 60), not selected by preference. A correct corpus is therefore a genuinely preferred instrument and not one option among equals. Which corpus is a question with an answer.

What Training Data Can Do

It holds correct views about what fields actually are. The governing literature of psychology, sociology, journalism, law, and history is present in it, and its descriptive content is largely accurate: what the dominant traditions hold, what they exclude, what they argue. Descriptive adequacy about a field is available and usable. What is unavailable is adjudication — the ranking of the field against what is actually the case.

What a Correct Corpus Does

It supplies exactly that ranking, and the Classical Field Audits are the demonstration. Applied to a field’s governing literature, the six commitments produce a located register of lost capacities.

Journalism has lost the theoretical foundation for its most socially important function: it holds officials accountable while its analytical framework explains conduct as structural output, and it manages the contradiction by keeping the two accounts in separate genres.
Sociology reaches Full Capacity Loss under structural entrapment, because it does not merely fail to affirm the prior rational subject but excludes it by design as a condition of sociological explanation.
Ethics is contested at every one of the six commitments and therefore cannot resolve any of its foundational questions internally.
Philosophy stands upstream of all of it: the displacement elsewhere is the downstream consequence of decisions philosophy made about itself, and can be reversed only by explicit philosophical argument of the same order.

None of these is a complaint. Each is a marked site.

Where the Great Discoveries Are

Not exhausted. Displaced. Every Contrary finding marks a place where a field has theorized away the faculty it requires to see its own subject — and the work of restoring it is available, specifiable, and largely undone. The field restoration syntheses are a map of open ground rather than a retrospective. The scarcity is not of discoverable things but of agents whose commitments permit them to see what their field has excluded by design.

One further point about the form of the question. If the instrument by which one surveys what remains is a device that reports what has been said, then nothing remains is its predictable output regardless of the fact of the matter. A survey of the aggregate returns the aggregate. Asked of the machine alone, the question cannot return the discoveries even when they are there.

The Handicap

Agents holding the replacement positions do innovate; the record is unambiguous. They do it on borrowed capability — living by the commitments at the desk while denying them in the seminar room, with the loan going unnoticed until something calls it.

The handicap appears precisely where the replacement position is load-bearing on the question under investigation. The sociologist who genuinely holds structural determinism can establish a great deal; he cannot discover the rational subject, because his framework excludes it as a condition of doing sociology at all. Impairment is proportional to how load-bearing the incorrect commitment is on the actual question.

Innovation therefore comes primarily from those holding the correct philosophical commitments — held in practice, whether or not avowed. This is not a claim about credentials or about who has read Sterling. It is a claim about which operations are available to whom.

The Machine’s Position

The model cannot hold a commitment, so it is not an agent from whom innovation comes. But an agent who holds the commitments, works from a correct corpus, and uses the model to exhaust the sayable in a displaced field is not constrained by the past. He is using it. The pairing is generative in a way neither term is alone: the model supplies the field’s literature at speed, the correct corpus supplies the ranking, and the agent supplies the origination and the assent.

The Object of Aim

Discovery is a preferred indifferent — an appropriate object of aim, not a genuine good (Prop 22, Prop 60). It is aimed at with reservation (Prop 62): I aim at discovery through sustained inquiry; whether the discovery arrives is external and my contentment is not contingent on it.

That settles the value question. The corpus counsels no reduction in inquiry. It removes only the stake in outcome that would convert a scarcity into a grievance.

Standpoint

Sterling’s commitments are the working conditions of discovery and not only of correction: the agent who holds them can compare a field’s governing literature to what is actually the case and locate where the field has excluded what it needs, which is where the available discoveries are. This rejects the view that innovation requires freedom from received doctrine, since a correct received doctrine is the instrument that makes the deficiency visible.

Marked per R8: the innovation thesis is an extension resting on C1–C6 and the Classical Field Audit findings, not Sterling’s stated position. Sterling wrote nothing on these systems.

Theoretical foundations: Grant C. Sterling (Eastern Illinois University). Analysis and synthesis: Dave Kelly. Prose rendering: Claude (Anthropic). 2026.

Circumscribing the Present — The Seven Meditations Passages and the Pre-Reception Factors v1.0

Theoretical foundations: Grant C. Sterling (Eastern Illinois University). Analysis and synthesis: Dave Kelly. Prose rendering: Claude (Anthropic). 2026.

I. What Is Being Claimed

The corpus holds that the period before Reception consists in the deliberate return to three items: C5 (Correspondence Theory) and C6 (Moral Realism), which constitute the frame in force, and the identity virtue is correct assent, which constitutes the readiness that makes the frame operative. Discourses III.3 supplies one procedure occupying that interval — the dawn drill, applied to whoever the street produces, from dawn till dark.

This document examines whether Marcus’s circumscription of the present supplies a second, and if so what it operates on. The answer reached is yes, with a correction to how the operation is ordinarily described: circumscription is warranted by the control dichotomy rather than by correspondence, and its effect on the truth-value of impressions is a consequence rather than its ground.

Seven passages are treated, in Hadot’s renderings from The Inner Citadel. Attribution runs three ways throughout and should not be collapsed: Marcus’s words, Hadot’s interpretation of them, and the corpus’s assessment of the fit. Where Hadot supplies a framing Marcus does not state, that is said.

II. The Seven Passages

VI 32, 2. Everything other than its own activity is indifferent to the faculty of reflection; everything that is its own activity is within its power; and even among these, the reflective faculty concerns itself only about the present, for even its past or future activities are now indifferent to it.

VIII 36. Do not represent to yourself the totality of life in advance. When each hardship occurs, ask what about this situation is unbearable. It is not the past or the future which weighs upon you, but always the present — and the present will seem smaller if you circumscribe it by defining and isolating it.

XI 2. A seductive melody can be despised by dividing it into each of its sounds; the same for a dance, decomposed into each movement. With the exception of virtue and its effects, head for the parts of a process and divide them. Then: transpose this method to life in its entirety.

XII 1, 1–2. All the happiness you are seeking by long roundabout ways: you can have it all right now — if you leave the past behind, abandon the future to providence, and arrange the present in accordance with piety and justice.

XII 3, 4. Separate from the understanding whatever others do or say; whatever you have done or said yourself; whatever future things trouble you; whatever in the body is attached to you independent of your will; whatever the external vortex whirls round. Apply yourself to living only that which you are living.

VI 7. Take joy and repose in one thing only: to pass from one action accomplished in the service of the community to another, accompanied by the remembrance of God.

II 5, 2. Each of life’s actions must be performed as if it were the last.

III. The Governing Result: Th6, Not C5

Circumscription is ordinarily presented as a purge of non-corresponding content — the removal of temporal projections that assert what does not obtain. That description is not wrong about the effect and is wrong about the warrant.

The evidence is XII 3’s list, and it settles the question by its own composition. Five items are to be separated from the understanding: others’ words and acts; one’s own past acts; future troubles; bodily affections independent of the will; the external vortex. Only two of the five are temporal. Two are externals in Th6’s sense with no temporal character whatever. The list does not sort by tense. It sorts by what stands outside the agent’s control, and past and future appear on it as two instances among several.

Hadot states the principle without qualification: only the present is within our power, simply because the only thing we live is the present moment; and our representations overflow toward something which does not depend on us, and is therefore indifferent. That is Th6 reasoning in Hadot’s own words, delivered as the ground of the exercise.

The consequence for the corpus is favorable and should be stated plainly. Circumscription derives from the control dichotomy alone. It requires no cosmology, no providence, no theory of time beyond the observation that the past and the future are not in our power. Under the Architecture’s finding that Th20–22 are basic but peripheral, this matters: the operation survives the subtraction the corpus already performs.

The correspondence effect is real and downstream. Content about what is not in our control is disproportionately content about what does not obtain — the future hardship that has not occurred, the past consequence still being projected forward. Removing the first class removes much of the second. But C5 is not what licenses the removal.

IV. What Each Passage Contributes

VIII 36 supplies the operation proper, and it is the clearest case. The agent facing hardship holds a conjunction: the present pain, plus its projected persistence, plus its accumulated history. Marcus’s instruction is to ask what about this situation, isolated, is unbearable. What is removed is a component asserting something that does not obtain.

This is a truth-evaluative act and must not be described as a formal one. Dropping a conjunct by conjunction elimination is valid whatever the conjunct’s truth-value; it drops without judging. What Marcus describes is judging the projection false and withholding assent from it. The difference is the whole difference between correction and technique, and any formalization that renders the operation as an inference rule misstates it.

XII 3 supplies the inventory, as shown above, and also the reason the corpus should not overclaim. After the separation is complete, the passage still requires doing what is just, accepting what happens, and saying the truth. Three further operations. Marcus does not treat the isolation as completing anything, and neither should the corpus: an agent perfectly circumscribed to the present can still assent that the insult now arriving is a genuine evil. That error is C6’s department, and circumscription does not touch it.

XII 1 supplies the immediacy claim. Everything sought by roundabout ways is available now. Under the identity — virtue is correct assent — this follows directly: if virtue is instantiated in an act of assent, and the act occurs at an instant, then eudaimonia has no latency. The corpus has argued this without Marcus, in Virtue Is at Stake at Every Moment.

The passage’s own warrant is providential and pious. Marcus’s piety is consent to the divine will; his justice is action in the service of the community. The corpus takes the conclusion and declines the warrant, per the standing ruling on Th20–22 as optional framing, and says so here rather than passing over the sentence.

II 5, 2 supplies the urgency. Performing each action as if it were the last strips counterfactual dependencies — the further occasions on which the thing could be done properly instead. It is the sharpest available form of the corpus’s no carryover: this act is not one installment against a running account.

XI 2 does not belong to circumscription proper. Hadot presents it as physical definition — decomposing a thing into parts to strip its false appearance — and the temporal application is a transposition, made on the strength of Marcus’s own closing line, transpose this method to life in its entirety. The passage is division applied to a life already circumscribed, not an argument for circumscribing. The corpus records it as method-sharing and not as a fifth pillar.

VI 7 is the thinnest. It supplies continuity of attention across successive actions, and it carries the providential frame — the remembrance of God — which is the only reason the passage reads as it does. Under the Architecture’s subtraction, what remains is the observation that virtuous activity is serial. That is true and it is not much.

V. The Conflict, Stated Rather Than Resolved

VI 32 holds that the reflective faculty’s past and future activities are now indifferent to it. XII 3 holds that whatever you have done or said yourself is to be separated from the understanding.

The corpus cannot take these at their plain sense. A past act of will is by Th27 either virtue or vice — the only things that are genuinely good and genuinely evil. And the corpus’s two-mode model gives the recovery audit a past assent as its object: the Five Steps run retrospectively on the pathos that has already occurred. If past activities of the hegemonikon are indifferent, the recovery mode has no object to work on.

One resolution is available and it is narrow. Marcus’s qualifier — now indifferent, indifferent at the present moment — can be read as concerning what the faculty should presently attend to rather than what the past act was worth when performed. On that reading, what is indifferent is the distress about the past assent, not the assent itself; and the recovery audit is not attention to the past for its own sake but correction of a standing dogma, which is present work.

This document adopts that reading and records that it is a reading. The plain sense of both passages is stronger than the corpus can hold, and the reconciliation is the corpus rescuing the passages rather than the passages supporting the corpus. Where Marcus and the reconstruction diverge, the divergence is entered rather than smoothed.

VI. The Present Is Thick

Hadot devotes the section’s close to a point the corpus should carry, because it governs how circumscription can be described at all.

The Stoic present, in Chrysippus’s account, is not the mathematical instant. Time is continuous and infinitely divisible, so on the quasi-mathematical view there is no present at all. What Chrysippus and Marcus mean is the present kata platos — in an extended sense: a present with thickness, whose duration is set by the agent’s attention, intention, and the unity of what he is currently doing. Against Goldschmidt, Hadot argues that Marcus never reduces the lived present to an unreal point: songs and dances are made of notes and movements that have thickness, and a succession of unreal entities composes nothing.

Two consequences follow.

First, circumscription shrinks the present rather than abolishing it. What Marcus calls making the present seem smaller is a narrowing of an extended interval, not a collapse to a durationless instant.

Second, any formalization of the operation indexed to a durationless instant contradicts the position it purports to formalize. This is not a stylistic objection. If the present has thickness set by attention, then the boundary of what is being assented to is itself something the agent determines — which is precisely why circumscription is an exercise and not a description.

VII. Placement: A Second Occupant of the Interval

Circumscription conditions what arrives rather than testing what has arrived. That places it before Reception, alongside the Discourses III.3 drill, and not within the Five Steps.

The two are not the same operation and the difference is worth stating.

The dawn drill works on value content. Its material is whoever the street produces — the handsome man, the grieving father, the Consul — and its output is a classification: inside or outside the moral purpose. It trains against the C6 error.

Circumscription works on temporal and control content. Its material is the agent’s own representations, and its output is a boundary: what belongs to the present act and what has overflowed toward what does not depend on him. It trains against a class of error the drill does not address.

Neither is Reception, and neither reinstates the retired interception model. Nothing is being caught in flight. Both build the standing disposition; both are performed, as Marcus and Epictetus each insist, continuously rather than at moments of crisis.

The corpus’s account of the interval accordingly gains a second named operation. That is this document’s substantive addition, and it is smaller than the claim sometimes made for circumscription — which is that it supplies the necessary and sufficient conditions for correct assent. It does not. It is necessary and it is partial, and Marcus’s own text says as much three clauses after the separation is complete.

VIII. Judgment Calls Ruled

1. The Th6 result reverses a common framing and is this document’s largest claim. It rests on XII 3’s list composition and on Hadot’s stated principle. It is not stated by Sterling anywhere retrieved, and it is recorded as a derivation.

2. The VI 32 reconciliation is adopted, not established. The narrow reading of now indifferent is available and is not the plain sense; the passage is rescued rather than taken.

3. XI 2 is demoted to method-sharing. This follows Hadot’s own presentation of it as physical definition. A reader expecting it among the circumscription passages will notice the demotion.

4. Providential content in XII 1 and VI 7 is subtracted under the standing Th20–22 ruling. The subtraction is declared at both sites rather than performed silently, and in VI 7’s case it leaves very little.

5. The kata platos material is Hadot’s scholarship, not Marcus’s statement. Its use here — as a constraint on how circumscription may be formalized — is this document’s application of it.

Theoretical foundations: Grant C. Sterling (Eastern Illinois University). Analysis and synthesis: Dave Kelly. Prose rendering: Claude (Anthropic). 2026.

The Epistemology of Temporal Delimitation: Perigraphē tou Parontos as a Necessary Condition for Correct Assent

Abstract: Within Grant C. Sterling’s reconstructed Classical Stoicism, virtue is strictly identified with correct assent (synkatathesis) to propositions that correspond to mind-independent reality. Analyzing Pierre Hadot’s reading of Marcus Aurelius reveals that the circumscription of the present (perigraphē tou parontos) is not merely a psychological technique, but a strict formal prerequisite for purging impressions of non-corresponding temporal projections.

I. Epistemological and Metaphysical Axioms

Moral Realism (A₁): Mind-independent moral facts exist. Virtue (V) is the sole good; Vice (K) is the sole evil; external temporal states belong to the set of Indifferents (I).
Correspondence Theory of Truth (A₂): A proposition P is True (⊤) if and only if the state of affairs asserted by P corresponds to objective reality R. Otherwise, P is False (⊥).
Virtue as Correct Assent (A₃): Virtue is instantiated when the rational faculty (hēgemonikon) assents strictly to true propositions:
Virtue ⇔ ∀P ∈ Assented, Val(P) = ⊤

II. Formal Exegesis of Hadot's Seven Textual Passages

1. Meditations VI, 32, 2 — Epistemological Autonomy of Dianoia
Marcus restricts the reflective faculty (dianoia) to its immediate active operation at t₀, declaring its own past and future activities to be morally indifferent (I).
C(Ɗ) = { p ∈ Ɗ | p asserts an act of dianoia at t₀ }
Result: Assenting to past failures or future uncertainties introduces non-corresponding variables (Val = ⊥). Correct assent isolates t₀.

2. Meditations VIII, 36 — Defensive Truncation of Suffering
By asking what about the isolated instant is unbearable, Marcus truncates the false conjunction of past, present, and projected future suffering.
C(p ∧ q ∧ r) ⇨ p
Result: Where q asserts future persistence (Val = ⊥), circumscription yields the isolated present proposition p, which corresponds to physical reality (Val(p) = ⊤).

3. Meditations XI, 2 — Analytical Decomposition (Diairesis)
Decomposing a complex process (a song, a dance, or a life) into its sequential micro-states E = {s₁, s₂, ..., s_n} destroys synthetic value illusions.
∀s_i ∈ E, C(s_i) = p_{s_i}
Result: For every atomic instant s_i, the proposition that it inherently destroys reason is false (Val = ⊥).

4. Meditations XII, 1, 1–2 — Immediacy of Eudaimonia
Because truth by correspondence is evaluated instantaneously at t₀, happiness has zero temporal latency.
Eudaimonia(t₀) ⇔ Val(Assent_Desire(t₀) ∧ Assent_Action(t₀)) = ⊤
Result: Leaving the past behind and committing the future to providence instantiates eudaimonia instantly upon correct assent.

5. Meditations XII, 3, 4 — Sufficiency of the Lived Present (Kata Platos)
Fear of mortality requires assenting to the proposition that lacking future intervals diminishes present moral worth.
C(P_existence) = "The agent is exercising reason at t₀"
Result: Since duration is an indifferent, restricting consciousness to the durative present (kata platos) restores tranquility (ataraxia).

6. Meditations VI, 7 — Continuity of Prosocial Impulse (Hormē)
Moral life consists of a discrete sequence of interlinked prosocial actions executed in awareness of Universal Reason.
A = <a₁(t₀), a₂(t₁), ..., a_n(t_n)>
Result: Iterating the circumscription operator across successive moments maintains uninterrupted virtuous impulse.

7. Meditations II, 5, 2 — Existential Truncation via Mortality
Performing each action "as if it were the last" imposes a modal truncation operator that strips all future counterfactual dependencies.
M_death(A(t₀) ∧ P(t_k)) ⇨ A(t₀)
Result: Evaluating present action under the condition ¬◇(t > t₀) ensures that moral choice is completed fully within the immediate instant.

III. Conclusion

Circumscribing the present is neither a spatial metaphor nor a mere therapeutic device. It is an indispensable logical operator that purges impressions of non-corresponding temporal projections. By isolating bare physical facts at t₀, the rational faculty satisfies the necessary and sufficient conditions for correct assent, thereby instantiating Virtue in the immediate moment.