Epistemic Forge: The Apex Synthesis

The whole map, skeleton-first — what holds, what is bet, and where it stops

Apex synthesis · the corpus in one document, stripped to its high-confidence spine and tiered by confidence

Andraž Đurič

Version 1 · June 2026

What this is. The apex statement of the body of work on this blog: the whole case in one document, built skeleton-first. Every claim is carried at the confidence it earns and no higher, which is why the document is tiered. The skeleton (§1) is the high-confidence core; below it sit the method (§2), the bets (§3), the named exposures (§4), the open frontiers (§5), and the limits (§6) where the map stops — each kept apart, because the confidence of the secure parts otherwise leaks onto the speculative ones. This is a logbook consolidation, not a bid for priority: the components are credited to prior art at every load-bearing joint, and the contribution is the honest demarcation and the assembly on one floor, not novel doctrine. It supersedes the prior apex. On any conflict with the modules, and on any point this document does not correct, the modules govern.

Note on the name. This document retires the label “Thermodynamic Realism.” The reason is a fidelity defect in the old name: it placed thermodynamics at the foundation, when thermodynamics is the functional floor — the lowest level whose vocabulary has referents — and not the bedrock. The work continues under Epistemic Forge, the name it was always written under. The rename corrects the label and changes no content. The substantive changes from the prior apex are separate, and marked where they occur: the Consistency Tax is no longer carried as a distinct claim, having been a special case of the thermodynamic cost of map–territory divergence, which stands (§1); the coupling density κ is now a sketched mechanism with its measurement owed, not a bare open question (§3); and the whole is framed as synthesis, its parts credited to prior art. Where this document corrects the earlier corpus, it governs; the older pieces keep the retired label until relabeled.

1. The skeleton

These hold, at the strength labelled. This is what survives when the corpus is stripped to what hard scrutiny leaves standing — and most of it is standard, by design. The job of the skeleton is to be solid, not novel; the moves that are distinctively this corpus's live in §3, where their status is marked as bets.

Foundation

Physicalism. Everything is physical or reduces to it. This is adopted as the maximally inductively justified standing position, not a bare posit: every phenomenon ever explained was explained physically, none by adding a non-physical ingredient, and the burden falls on whoever would break the pattern.

Semantic physicalism, the anchoring. Meaning is physically realized, and it answers to the physical: whatever the correct account of meaning turns out to be, its content is about physical facts — the only facts there are — or it is empty. A semantics connected to nothing physical describes nothing. This much follows from physicalism applied to meaning, and its denial is Platonism about propositions, which physicalism already excludes. The separate question of the mechanism of meaning — whether content is constituted by a referential tracking-relation or by inferential role — is contested and is held in the exposures (§4), not here.

The functional floor

The floor is one level: non-equilibrium statistical mechanics. Thermodynamic entropy and information entropy are two faces of a single quantity — statistical mechanics is inference about microstates (Jaynes), and information is physically realized and costs energy to process (Landauer). This is the level at which the arrow, dissipation, and the cost of prediction all live, and at which the working vocabulary — boundary, gradient, persistence, fidelity — first acquires referents. It is the functional floor, the lowest level from which the program can model upward into complexity, and not the bedrock. Information is not “made from” thermodynamics; if anything the priority runs the other way, with thermodynamic entropy a special case of information entropy. Which of information or matter is more fundamental below the floor is a frontier the floor does not depend on (§5).

Below the floor. A very nearly time-symmetric microphysics with no arrow and no agents. None of the floor's vocabulary has a referent there; an agent cannot be built without passing through the level at which boundaries and irreversibility first appear, and that level is the floor.

The arrow of time. The arrow the program uses is the thermodynamic one: the direction of entropy increase. It exists because the universe began in an extraordinarily low-entropy state — the Past Hypothesis — not because the dynamics are asymmetric. It is emergent, manufactured by statistics operating on a boundary condition, not a fundamental law. Whether the other arrows (cosmological, radiative) reduce to it is a further, contested question; the program relies on the thermodynamic one.

Information is physical. Storing, processing, and erasing information carry minimum thermodynamic costs (Landauer), experimentally confirmed across platforms from colloidal beads to single atoms.

Dynamics on the floor

Gradients discharge, and structure discharges them faster. Far from equilibrium, matter organizes, because structured flow levels a gradient faster than disordered matter can; a dissipative structure is the limit case, maintaining its own boundary so it can keep discharging. This is not the claim that structures maximize the rate of discharge (maximum entropy production), which nothing here rests on.

Finite free energy forces competition. Local free energy is finite, so persisting systems compete for it.

Persisting systems are nested and coupled. Nested: boundaries sit within boundaries across scales — cell within organism, organism within family, family within city, life within biosphere. Coupled: systems carry predictive information about one another's states. Both are structural facts about the world the program describes, and one consequence is load-bearing: nesting and coupling do not by themselves make a higher level an agent. That gate is the agenthood criterion, below.

The persister

A persister is a dissipative-structure token, not a pattern. What persists is not an abstract pattern but a process that holds a boundary by continuous work — a self-maintaining disequilibrium that exists only by dissipating energy and shedding entropy to its surroundings. It is substrate-independent in its matter and a single particular as a process. The whirlpool is the instance, not the metaphor: its water is never the same water and it stays this whirlpool; redirected slowly into another basin while it keeps turning, it is the same whirlpool relocated; a second whirlpool started alongside is a second whirlpool, not this one twice.

Individuated by continuity of the work. What makes a later structure the same persister as an earlier one is the unbroken continuation of the boundary-maintaining work, not sameness of pattern. Sleep and anesthesia are not breaks, because the work runs through them; the criterion is continuity of the labour, not of consciousness.

Two identical tokens are numerically distinct. A faithful copy is a twin, not the original — the duplication argument is a short proof: were copying to preserve identity, two simultaneous copies would each be the original and so be each other, yet they are two, in two places, able to diverge from the first instant. Numerical identity tracks the spatiotemporal and causal trajectory, which removing the soul does not remove. Gradual substrate replacement that never halts the work keeps the one token; copying starts a second. One fact, continuity of the labour, delivers both verdicts.

Persistence structure

Binary at the barrier, graded above it. Persistence has a boundary and, above it, a gradient. The boundary is an absorbing barrier — dissolution, the state a system does not return from — and that part is binary: a system is either still itself or it has come apart. Above the boundary lies endurance, a matter of degree: how far above the barrier a system sits, how much disruption it can absorb, how long it can expect to last.

Fidelity dynamics

Map–territory divergence carries a thermodynamic cost. A model that fails to track its environment dissipates free energy: for a system driven by a stochastic environment, the nonpredictive information it retains lower-bounds the work it dissipates (Still and colleagues). The cost is latent until reality cashes it — paid at the prediction–reality interface, and displaceable across coupled agents onto whoever does the cashing. The Consistency Tax folds in here: the apparent extra cost of deception was a special case of this divergence, not a separate law. The override cost of holding a private-accurate and presented-divergent pair of maps is real but removable by retraining; what is not removable is the divergence itself, which simply goes latent when a falsehood is made substrate-consistent.

Selection toward fidelity is a tendency, not a law. In an ecology competing for finite free energy, lower-divergence systems dissipate less on error correction on average, so the distribution drifts toward closer trackers. This is a statistical tendency, indexed to horizon — weak in stable, resource-rich niches and at short horizons, where a distorted model can persist untouched — and never a law. The strong reading, that it approaches necessity at the cosmic limit, is held at the frontier (§5).

Agents and scale

The agenthood criterion. A collection is an agent only if it maintains a boundary it can fail to maintain, has persistence conditions it can fail, and runs a model deployed to keep itself from failing them. This blocks the slide from “coupled” or “nested” to “is an agent”: coupling and nesting supply at most the boundary. Two predators over a carcass, an ecosystem, a crowd — tightly coupled, not agents. The criterion gates every higher-scale claim in the program.

The universe is closed. The totality has no outside, and so no exchange across a boundary it does not have.

No global optimizer — a result, not a gap. Because the maximal scale has no outside, it has no boundary, fails the agenthood criterion, and is the subject of no ought. There is no holder of a cosmic optimization target; the selection direction is always indexed to a bounded agent. Any reading of the descriptive frontier as “what everything ought, globally, to maximize” reintroduces the very error the metaethics dissolves.

Orthogonality. Capability and goals are independent axes; intelligence entails no benevolence. This is the same point as the agent-relativity of value, reached from the side of agents: a perfectly truth-tracking agent whose persistence conditions do not overlap yours is more dangerous, not less, because it makes fewer errors about what it can safely discard.

Normativity, the secure half

“Ought” is agent-relative. An ought is always some agent's, indexed to that agent's persistence conditions, its environment, and its horizon; deployed with an agent and an end, it has content. This much is close to consensus across the deflationary tradition. The stronger claim — that a bare ought, stripped of any agent, is contentless rather than merely false — is the load-bearing contested half, and is held in the exposures (§4).

No non-physical categorical ought. There is no obligation binding an agent regardless of what it is, standing free as a fact about the universe rather than about some particular system. This follows directly from physicalism: there is no non-physical realm for such a fact to live in. The price is “wrong, full stop”: wrongness is always wrongness for an agent, relative to what dissolves it.

2. The method

Name the relata. For any referential-relational operator — ought, validity, meaning, modality, desert, authority — name the frame it is defined against: which agent, which system, which rules, which laws. If the frame can be named, the term is a determinate relational fact, located and true or false relative to that frame. If the proponent refuses, holding that the term obtains with no agent, system, or frame at all, then it has no truth-conditions and is a ghost — a word with the grammar of a claim and nothing on the far side to be true or false about. The diagnostic converts one question, in what frame?, into a test of well-formedness, and it is the single tool the rest of the corpus runs.

Its boundary is part of it. The diagnostic is discriminating, not universal. It dissolves stripped referential-relational terms and halts at genuine remainders — the residue of consciousness is the marked case. A solvent that dissolved everything would dissolve real things along with ghosts; one that stops at a genuine remainder does not. The diagnostic is high-confidence as a practice; its strong verdict, that a stripped term says nothing rather than something false, rides on the referential reading held in the exposures.

The standing discipline. Fidelity over memorability; mark the limits; account honestly for what is secure and what is owed; no overclaim; subject concessions to the same scrutiny as assertions; judge a contribution on its content, not its source.

3. The bets

Nothing in this section is a result. These are the program's wagers, each with its crux named.

Persistence fixes the content of an agent's ought. The end that fixes what an agent ought to do is persistence under thermodynamic constraint. The rival is not the expressivist or the realist but the flourishing-naturalist (Foot), who agrees the ought is agent-and-end relative and disagrees about the end; the two coincide across most cases and come apart on persistence-costly behaviour — altruism toward the uncoupled, art, play, martyrdom. The confidence of this bet is the confidence of whether those cases can be carried.

All values reduce in content to persistence. The monist wager, that every value worth the name bottoms out in persistence. Mostly uncomputed; a bet, not a proof. Vitality is best read here as constitutive — the form persistence takes for an agent like us — rather than a separate terminal value.

Value is indexed to the agent, not the pattern. The unit that matters is the persisting agent, individuated by its own boundary, not the pattern it carries. This is the program's posit, and it is most exposed in the copy case, where the agent and the pattern visibly diverge: it delivers the verdict that a copy does not secure what the original's death loses. A monist who indexed value to the pattern instead would hold the whole ontology and reject that verdict; the disagreement is real and is located here.

Intelligence reliably detects the overlap. The secure half is near-analytic: where an agent's persistence is coupled to external structure it cannot cheaply replace, it has an instrumental, persistence-indexed reason to preserve that structure. The bet is the detection half — that a sufficiently intelligent agent reliably estimates that coupling where it exists. It is a bet of the same kind as the persistence-fixes-content bet, defended against vacuity only by the requirement that the coupling be measured rather than asserted.

The two cruxes of the positive ethics. Within an agent, where its persistence dimensions trade off — intensity against duration, energy now against robustness later — physics fixes a Pareto frontier and no unique point on it; the determinate-optimum claim holds only where the agent is one-dimensional. Between agents, the coupling density κ — how far one system's states carry predictive information about another's persistence — is the crux of whether the persistence-costly cases can be carried. κ is now a sketched mechanism with its measurement owed, not a bare open question; the standing danger is that hidden coupling at some scale can rescue any case after the fact, and the only defence is that κ be operationalized and measured, never tuned.

4. The exposures

Two premises carry the program's main exposure.

The referential-relational reading. The diagnostic assumes that the operators it audits are referential and relational, their content fixed by what they relate, so that a stripped operator is contentless rather than merely false. A physicalist who holds a conceptual-role semantics instead — content is inferential role — can resist every ghost verdict without leaving physicalism. The defence is not assertion but a fork: an inferential role unanchored to the world cannot tell a contentful term from a coherent but empty one (it would license “phlogiston” as readily as “oxygen”), and any anchor rich enough to confer content has re-imported a tracking-relation to the world, which is reference under another name. The fork presses a real advantage; it does not refute every form of the rival by construction, and so the reading remains the principal exposure.

The no-default claim. Deployed with no agent, “ought” is type-errored — undefined the way “north” is undefined at the North Pole, not merely incomplete the way “tall” is without a comparison class. The dissolution of the is–ought problem rides on this, because the gap dissolves only if there is nothing on the far side. The realist resists by supplying a default agent-class — “all rational beings” — and once a class is named the original question reappears inside the rescue: why does that class's end bind? The claim is substantive and contested, and is held one tier below physicalism, which earns its standing from an unbroken empirical record a thesis in the philosophy of language does not have.

5. The frontiers

Open questions the program reaches but does not close. Each is marked rather than answered.

Uniqueness, as a selection problem. Why this self-consistent structure obtains rather than another is not a grounding question — a foundationless structure has no ground by hypothesis — but a selection question. Three roads are held open: bound the answer to the floor; construct a sub-thermodynamic self-reinforcement that both escapes temporality and selects (the live thread is constructor-theoretic and unbuilt); or accept that there is no selector below the floor and the filling is brute. The thesis is admissible only read as constitutive — about the necessary structure of any self-reinforcing thing — not as an empirical claim about the actual ensemble, which is out of empirical reach.

Consciousness. The diagnostic dissolves the confused half — the zombie intuition presupposes a physical difference physicalism denies — and halts at the genuine half, the question why there is something it is like at all, which is bracketed and not claimed. The author's own leaning toward a graded, proto-experiential view is a personal position, fenced off, and no part of this case.

The cosmological branch. Whether the descriptive program's forward edge is a hard end depends on a fact not yet settled: whether dark energy is a true cosmological constant. If it is, the mechanism is exact — an eternally accelerating universe has an event horizon with a fixed positive temperature (Gibbons and Hawking), the Landauer floor stops falling, a finite store of accessible energy is quarantined inside the horizon, and a finite store at a fixed minimum cost per operation is a finite number of operations: a last thought (Krauss and Starkman). If instead dark energy weakens and the acceleration relents, there is no event horizon, and the gradient argument alone cannot deliver finitude; Dyson's slow burn becomes possible, obstructed not by a horizon but by the discreteness of states and the alarm problem.

Which branch we are in is open, and the honest reading is more cautious in mid-2026 than a year earlier. The signal for weakening dark energy appears only when DESI's baryon-acoustic-oscillation data is combined with the cosmic microwave background and a Type Ia supernova sample; DESI's data alone is consistent with a cosmological constant, and the preference depends sharply on which supernova sample is chosen — from roughly 3.9σ with the DES five-year sample down to about 2.5σ with Pantheon+, which alone shows no preference at all (Efstathiou). Through 2026 the signal has, if anything, softened: a recalibration of the DES sample lowered the significance from 4.2σ to 3.2σ, a weak preference (Popovic and colleagues); an independent consistency check found a mismatch between the BAO and supernova datasets and concluded the evidence may be a systematic artifact (Afroz and Mukherjee). The competing challenge from the other side — that supernovae brighten with cosmic age, which would undermine the distance ladder — was itself corrected, leaving cosmic acceleration confirmed as real while the constant-versus-evolving question stays open (Wiseman and colleagues). DESI completed its planned five-year survey in April 2026, mapping more than 47 million objects, and the first results from the full dataset are expected in 2027; that is the decisive test.

The terminus is therefore stated conditionally. It is also a live test of this corpus's own discipline. The pull toward the weakening branch is strong precisely because it is the kinder one — it reopens the universe's fate and dissolves the hard end — and that is exactly the shape of a motivated update. The honest entry is not “the end may not be absolute,” leaning on a sub-threshold and currently softening signal, but: the branch is genuinely unsettled, the decisive measurement arrives in 2027, and the terminus holds or fails on its outcome. Resisting the update the evidence does not yet license is the same move the firewall makes everywhere else.

The sub-floor, and what grounds the laws. Below the functional floor lies whatever the microphysics bottoms out in, and the priority of information and matter — it-from-bit — is unresolved, and the program does not lean on it. The question of what grounds the laws splits cleanly at the diagnostic: a lawful meta-reality is deferred physics, and a lawless outside is a ghost (§6). The self-grounding form closes — a foundationless structure can be coherent and, granting Gödel, cannot certify itself from within — but which structure obtains stays open, the invariant that survives at every level.

6. The limits

Where the map stops, and nothing is added past it.

The tautologies. Existence, persistence, dissolution, and the sentence they collapse to — it is what it is. What exists, exists; what persists, persists; what dissolves, dissolves. Each is certain and each is contentless, and the emptiness is the point: that they say nothing is what they show. They mark where saying stops, not a further layer to be stated.

The far edge. The question of a wider reality splits into two with opposite verdicts. A lawful meta-reality — a wider physics, other pockets of low entropy, a multiverse — is deferred physics: internally coherent, possibly true, and, where causally quarantined, permanently outside empirical reach. Out-of-scope is not nonexistent; it is a verdict about what can be tested. A lawless outside, invoked to explain where the laws come from, is the ghost: every verb its invocation needs — produces, causes, gives rise to — imports the regularity being denied, so the proposal unsays itself in the saying.

The foundation is brute. The demand for a reason at the foundation is projected outward from inside a structured world, where everything has an external ground. At the foundation there is no outside for a ground, a cause, or a justification to come from; the same wall denies the totality a boundary, a cause, and a reason. Justification bottoms out in self-support, because there is nothing external to vindicate it with. Something has to be brute; the regress forces that much.

7. The loop

A last structural fact. The framework is a physical model, running in physical agents, of the physics of physical agents that model. When it describes a structure that models its environment in order to persist, the describing is an instance of the described; the document is a token of its own type. This must not be confused with vicious circularity, which is a justification that presupposes its conclusion; the program's premises are graded independently and nothing in its case for a claim assumes that claim. It is self-instantiation: a claim turning out to satisfy its own standard. At the limit this is the map–territory relation closing over one region of territory, where a perfectly faithful map of a modelling structure would itself be such a structure. It explains why the account can describe itself without paradox, and it adds nothing to the confidence of any claim — a false but coherent self-model passes its own filters too. The loop is structure, not warrant.

Standing of this document. The apex synthesis of the corpus, with each tier at its own confidence, none borrowed across the line. The skeleton (§1) is secure at the strengths labelled, and mostly standard by design; physicalism is the maximally inductively justified position, the agent-relative ought and the denial of a categorical one are theorems from it together with the choice of deflation over error-theory, and the persistence ontology and the agenthood criterion follow the dissipative-structure account the corpus commits to. The method (§2) is secure as a practice; its strong verdict rides on the referential reading. The bets (§3) are bets, each with its named crux, and are where the distinctive content sits; the named exposures (§4) — the referential reading and the no-default claim — are defended, contested, and held one tier below physicalism. The frontiers (§5) are open in the ordinary, empirical way, the cosmological branch most concretely so, with a decisive test due in 2027. The limits (§6) mark where the map stops. The loop (§7) is structure, not warrant. The change from the prior apex is the retraction of the Consistency Tax as a separate claim, the update of κ to a sketched mechanism with its measurement owed, the synthesis framing with prior art credited, and the retirement of the old name. On any conflict with the modules, and on any point this document does not correct, the modules govern. Corrections and counterexamples are welcome and change the document.

References

1. Jaynes, E. T. (1957). Information Theory and Statistical Mechanics. Physical Review 106, 620–630; and 108, 171–190. Statistical mechanics as inference about microstates; the identification under which thermodynamic and information entropy are two faces of one quantity, behind the floor of §1. Standard
2. Landauer, R. (1961). Irreversibility and Heat Generation in the Computing Process. IBM Journal of Research and Development 5, 183–191. The minimum thermodynamic cost of erasure; “information is physical,” experimentally confirmed across platforms. Standard
3. Albert, D. Z. (2000). Time and Chance. Harvard University Press. The Past Hypothesis; the arrow as a boundary condition on the low-entropy start, not a law. Standard
4. Nicolis, G., & Prigogine, I. (1977). Self-Organization in Nonequilibrium Systems. Wiley. The dissipative structure that holds order far from equilibrium by dissipating energy across its boundary; the ontology of the persister in §1. Standard
5. Schrödinger, E. (1944). What Is Life? Cambridge University Press. The living system as one that staves off equilibrium by feeding on order; the order-maintaining work that marks the token. Standard
6. Still, S., Sivak, D. A., Bell, A. J., & Crooks, G. E. (2012). Thermodynamics of Prediction. Physical Review Letters 109, 120604. Nonpredictive retained information lower-bounds dissipated work; the thermodynamic cost of map–territory divergence into which the Consistency Tax folds. Standard
7. Maturana, H. R., & Varela, F. J. (1980). Autopoiesis and Cognition. D. Reidel. The self-producing, boundary-maintaining character of living systems; one source of the agenthood criterion. Standard
8. Friston, K. (2010). The free-energy principle: a unified brain theory? Nature Reviews Neuroscience 11, 127–138. The boundary as a Markov blanket and persistence as the minimization of variational free energy; the formal counterpart of the criterion, used instrumentally. Standard
9. Conant, R. C., & Ashby, W. R. (1970). Every Good Regulator of a System Must Be a Model of That System. International Journal of Systems Science 1, 89–97. The modelling clause of the agenthood criterion. Standard
10. Hume, D. (1739–40). A Treatise of Human Nature, Book III, Part I, Section I. The is-to-ought passage: the inferential point granted, and the autonomous-realm gloss refused. Standard
11. Foot, P. (1972). Morality as a System of Hypothetical Imperatives. The Philosophical Review 81(3), 305–316; and Natural Goodness (2001), Oxford University Press. The hypothetical ought used throughout, and the flourishing-naturalist rival to the persistence bet of §3. Standard
12. Mackie, J. L. (1977). Ethics: Inventing Right and Wrong. Penguin. The error theory retiring the categorical “wrong, full stop”; the argument from queerness as the residue charge against a free-floating ought. Standard
13. Korsgaard, C. M. (1996). The Sources of Normativity. Cambridge University Press. Normativity from the conditions of agency (constitutivism); re-grounded here in physical persistence. Standard
14. Harman, G. (1982). Conceptual Role Semantics. Notre Dame Journal of Formal Logic 23(2), 242–256; and Block, N. (1986), Advertisement for a Semantics for Psychology, Midwest Studies in Philosophy 10. The rival semantics named as the principal exposure in §4, and the two-factor view whose anchor is reference under another name. Standard
15. Bostrom, N. (2012). The Superintelligent Will. Minds and Machines 22(1), 71–85. The orthogonality thesis: capability and final goals are independent axes. Standard
16. Gibbons, G. W., & Hawking, S. W. (1977). Cosmological event horizons, thermodynamics, and particle creation. Physical Review D 15, 2738–2751. The horizon temperature that sets the Landauer floor under a true cosmological constant, in §5. Standard
17. Dyson, F. J. (1979). Time without end. Reviews of Modern Physics 51, 447–460; and Krauss, L. M., & Starkman, G. D. (2000), The Astrophysical Journal 531, 22–30. The slow burn in an open universe, and the discreteness and finite-information limits that bound it. Standard
18. DESI Collaboration (2025). DESI DR2 Results II: BAO and Cosmological Constraints. arXiv:2503.14738. The three-year preference for evolving dark energy, sample-dependent and below the discovery threshold; the survey's planned five-year map was completed in April 2026, with full results expected 2027. Verified · June 2026
19. Efstathiou, G. (2024–25). Evolving Dark Energy or Supernovae Systematics? arXiv:2408.07175; MNRAS 538(2), 875. The dependence of the signal on the supernova sample (about 3.9σ DES5Y to about 2.5σ Pantheon+; Pantheon+ consistent with ΛCDM). Verified · June 2026
20. Popovic, B., et al. (2026). The Dark Energy Survey Supernova Program: A Reanalysis with an Updated Type Ia Calibration (DES-Dovekie). MNRAS, April 2026. Recalibration lowering the significance from 4.2σ to 3.2σ — a weak preference for evolving dark energy. Verified · June 2026
21. Afroz, S., & Mukherjee, S. (2026). Hint toward an inconsistency between BAO and supernovae datasets. Physical Review D (2026). A consistency check finding the DESI DR2 evidence for evolving dark energy may be a systematic artifact; corroborated by Wiseman and colleagues (2026) confirming acceleration is real while leaving constant-versus-evolving open. Verified · June 2026
22. Farzulla, M. (2026). The Replicator-Optimization Mechanism. arXiv:2601.06363. The near-twin of the program's bridge — same Humean instrumental structure, same scale-relativity — with a coarse-graining theorem the program lacks; engaged as prior art. Verified · June 2026
23. Smolin, L. (1997), The Life of the Cosmos, Oxford University Press; and Tegmark, M. (2014), Our Mathematical Universe, Knopf. Selection among structures (cosmological natural selection), and the mathematical-universe hypothesis with the uniqueness gap it provokes; the precedents for the selection frontier in §5. Standard
24. The Is/Ought Firewall (Epistemic Forge). The agent-relative deflationary ought derived from physicalism, and the firewall as a provisional debt. Internal
25. Name the Relata (Epistemic Forge). The diagnostic of §2 and the fork defending its referential premise in §4. Internal
26. Preservation Is a Coupling Problem (Epistemic Forge). The coupling mechanism and the preserve–sever boundary; the secure and bet halves of the coupling claim in §3. Internal
27. What You Actually Are (Epistemic Forge). The dissipative-structure token, the duplication argument, and the value-indexing posit of §3. Internal
28. Uniqueness Is a Selection Problem (Epistemic Forge). The three roads and the constitutive reading of the selection frontier in §5. Internal
29. The Last Gradient and It is what it is (Epistemic Forge). The cosmological endgame of §5 and the tautological terminus of §6. Internal

Author: Andraž Đurič, Slovenia. Supersedes Thermodynamic Realism: The Apex Synthesis (12 June 2026) and retires that label; the work continues under Epistemic Forge. Written in collaboration with Claude (Anthropic); the corpus it synthesises was developed in dialogue with Claude, DeepSeek, and Gemini, with contributions credited in the modules. Contributions are judged on content rather than origin; doing otherwise would be the genetic fallacy. Dark-energy figures and external sources verified against primary sources in June 2026. Text licensed CC BY 4.0.