God Does Not Exist

The classical supernatural God has no live reading

Argument note · physicalism, semantic indexing, agency, causation, evidence, free will, suffering, and the exact standing of atheism

Andraž Đurič

Version 1 · 2 September 2026

Abstract

This note argues one proposition: the classical supernatural God does not exist.

It does not redefine atheism as a mere absence of belief. It does not retreat to “the burden of proof is on the believer.” It does not infer certainty from confidence, or uncertainty from corrigibility. It makes the object-level claim and accepts the burden that comes with making it.

The target is the God of classical supernatural theism: a necessarily existing, nonphysical, personal agent who is omniscient, omnipotent, perfectly good, and the intentional creator or sustainer of the world. The conclusion does not concern every entity someone might choose to call a god. A finite universe-maker, simulator, alien intelligence, or unknown lawful cause is not ruled out by definition. None is the object under audit. If one of them exists, it belongs to reality and must be modeled there. Calling it “God” does not give it necessity, transcendence, omniscience, omnipotence, goodness, or the right to be worshipped.

The audit has five axes: semantic, logical, modal, empirical, and explanatory. On the semantic axis, the divine predicates either acquire determinate relata and place their bearer inside a structured reality, or lose their truth conditions. On the logical and modal axes, the classical attributes do not remain jointly stable once knowledge, agency, choice, necessity, timelessness, and goodness are indexed to what makes them contentful. On the empirical axis, interventionist forms of the hypothesis repeatedly fail to produce the differential evidence they predict. On the explanatory axis, retreating forms survive only by ceasing to exclude any possible observation. Their apparent compatibility with everything is explanatory contact with nothing.

The coherent remnants are lawful parts of reality, not a supernatural God. The transcendent remainder is a ghost. No referent survives the audit.

What is inherited, and what is added

The physical floor is inherited from Epistemic Forge: The Apex Synthesis, The Fidelity Program, and The Spine and the Web. The account of reality alignment as corrigible map–territory contact, and of alignment as a persistence strategy for embedded agents, is inherited from Reality Alignment and Reality Alignment as a Persistence Strategy V3. The name-the-relata diagnostic and its principal semantic exposure are inherited from Name the Relata. The agenthood criterion and no-global-optimizer result are inherited from the apex, What Can Bear an Ought, and the surrounding agency modules. The distinctions among diachronic causation, sustaining causation, grounding, and constitutive dependence are inherited from Causality Is Not a Flat List. The exact vocabulary for epistemic failure at a load-bearing ontological joint is inherited from Existence Literacy. The refusal to blur a warranted conclusion into false balance is inherited from The Two Ways to Be Wrong. The treatment of caused agency is inherited from You Do Not Need to Be Uncaused to Be a Cause.

What this note adds is the application: it runs those commitments together on classical supernatural theism, incorporates the strongest standard replies, and states the resulting verdict at its actual weight.

1. The target

“Which God?” is useful only until the target is specified. It is specified here.

Let G denote a being claimed to have all of the following properties:

Necessary existence: G cannot fail to exist.
Transcendence: G is not physical and is not merely one bounded process among others.
Personhood or agency: G knows, intends, chooses, commands, judges, creates, or answers.
Omniscience: G knows every truth.
Omnipotence: G can actualize every genuinely possible state of affairs compatible with its nature.
Perfect goodness: G is morally unsurpassable.
Creation or sustenance: the world depends on G through intentional creation, continuous conservation, grounding, or some combination of these.

This is not a caricature. It is the central package of classical monotheism. Omnipotence is not being defined as the ability to make contradictions true. A square circle is not an omitted option. The argument needs no such option. One genuinely possible lower-harm world is enough.

The target also cannot be rescued by saying that the predicates are merely analogical. An analogy has content only through preserved structure. If “knows,” “chooses,” and “good” preserve enough of their ordinary relational structure to support inferences, that structure can be audited. If they preserve none, they no longer tell us that G knows, chooses, or is good. Analogy can extend a concept. It cannot retain a conclusion after removing every relation that made the concept mean what it did.

2. Corrigibility is not suspension

Absolute certainty about substantive reality is generally unavailable to finite embedded agents. That fact does not place all ontologies at equal distance from the territory.

There are at least three different things routinely collapsed into “certainty”:

formal entailment: whether a conclusion follows from specified premises;
object-level confidence: how strongly the available evidence favors a proposition;
agent-level incorrigibility: whether the believer treats revision as impossible.

The first can be exact. The second is graded. The third is an epistemic defect, not an achievement.

I am not incorrigible about atheism. I am asserting atheism. Those are different facts. “God does not exist” is the correct object-level verdict on the target defined above. If the premises, semantics, or evidence change, the verdict can change. Until then, refusing to state the conclusion is not rigor. It is under-updating presented as humility.

The burden of proof is a rule for dialectical traffic: the person introducing a claim owes support for it. It is not a permanent shelter from posterior judgment. Once a hypothesis has been defined, its coherence examined, its expected evidence sought, and its rivals compared, the question is no longer who spoke first. The question is which model survived.

Nor are belief and suspension equally wrong. If G is false, confident belief in G is farther from the truth than uncertain suspension. That is the truth-distance index. But after the evidence decisively favors not-G, indefinite suspension can still be epistemically irresponsible: it withholds the update the evidence warrants. That is the calibration and intellectual-responsibility index. The two failures differ. Neither turns theism into a live equal.

3. The physical floor, without a definitional trick

The standing position of this corpus is physicalism. It is not the empty rule that whatever exists will be renamed “physical.” It is the substantive position that concrete reality is one causally and structurally continuous order; that every concrete causal difference is realized in, constituted by, or dependent on that order; that higher-level biological, cognitive, and social properties arise from its organized states rather than from an autonomous supernatural substance; and that agency is a local organization within reality, not a primitive intention outside it.

This does not require present physics to be complete. Present physics is not complete. It also does not define the physical as whatever an ideal future theory happens to contain. That would invite Hempel’s dilemma: define physicalism by current physics and it is probably false in detail; define it by unspecified future physics and it risks becoming trivial. The position here is constrained more tightly than either horn. A successor theory may revise our inventory radically, but it must still earn its place through structured dependence, causal contact, empirical constraint, and public corrigibility. An irreducible cosmic mind, primitive purpose, or exception that can explain every outcome while predicting none would not qualify merely because someone inserted it into a future vocabulary.

Physicalism is therefore not deductively guaranteed by the word “reality.” It is maximally inductively justified by the history of successful inquiry: domains once assigned to spirits, souls, vital forces, curses, providence, or disembodied intention have repeatedly yielded to lawful mechanistic and multilevel explanation; the reverse transition has not occurred in any robustly confirmed case. Causal closure remains a substantive empirical-metaphysical commitment, not a tautology. It could be broken by a reproducible nonphysical causal contribution that makes a distinct difference after the physical model is fixed. None has been established.

That is what supernaturalism opposes. It does not offer a comparably supported extension of the map. It asks for an exemption from the conditions by which additions to the map earn standing.

4. The three readings of God

Let T be total reality: everything that exists, together with every relation that obtains. There are only three relevant placements for G.

Reading	What follows	Verdict
G is identical with T	T has no exterior, no environment across a boundary, and no global persistence condition it can fail. It is not a bounded agent with a model, alternatives, needs, or a target.	Pantheistic redescription. “God” names reality but does not recover the personal creator.
G is distinct within T	The distinction between G and world, G’s knowledge of the world, and G’s effects on it are relations inside T. If stable, they belong to the wider structure of reality.	A potentially lawful entity or ground inside the total ontology, not a supernatural exception to it. If bounded enough to be an agent, it is not the unlimited totality.
G is outside T, law, relation, and logic	“Outside total reality” contradicts the meaning of totality. Without identity, difference, relation, or logic, even “G exists” lacks truth conditions.	Not a mysterious being. A ghost.

The first reading abandons classical theism for a name. The second abandons supernatural exemption for a model. The third abandons a claim while preserving its grammar.

A lawful meta-reality is not ruled out. Our observable universe could depend on a wider order, could be simulated, or could be one region of a structure inaccessible to us. Even permanent causal quarantine would establish only that a region is out of our epistemic reach, not that it is nonexistent. But a wider lawful order is deferred physics in the broad sense relevant here: more reality with more structure. It is not a warrant for a necessary, perfectly good, omniscient agent. Unknown is not divine, and outside our current model is not outside reality.

5. The supernatural trilemma

“Supernatural” is often allowed to oscillate among three claims. Fixing the meaning ends the oscillation.

5.1 It makes a differential causal difference

If G creates matter, alters a disease, answers a prayer, inspires a text, reveals a proposition, moves a body, changes a probability, or places a thought into a brain, then a state differs because of G. That difference is coupling. Coupling has structure: conditions under which the effect occurs, a target on which it occurs, and some relation between antecedent and outcome. We may not yet know the law, and the process may be stochastic rather than deterministic, but the event has entered the investigable causal order.

This does not prove that the cause resembles known particles. It proves that “beyond nature” cannot mean beyond all constraint while retaining causal work. The cause now owes a model.

5.2 It makes no possible differential difference

If no observation, experience, instrument state, memory, outcome, or probability could differ depending on whether G exists, then G is empirically equivalent to nonexistence. It has no explanatory or action-guiding weight. It occupies the same evidential class as indefinitely many invented entities defined never to leave a trace.

This is not the fallacy that infinitely many alternatives must each have probability zero. Their number is irrelevant. The point is discrimination: no evidence selects G over its rivals, because the hypothesis forbids evidence from doing so.

5.3 Logic does not apply to it

If identity, non-contradiction, inference, and predication do not apply to G, then neither does the believer’s sentence. One cannot coherently infer that G exists, is one rather than many, knows rather than fails to know, is good rather than cruel, or creates rather than does nothing. Logic is not a biological measuring device from which a supernatural object might be hidden. It is constitutive of the difference between asserting something and emitting words.

A nonclassical logic would still be a logic with rules. “No logic applies” removes the rules required to preserve the claim.

The trilemma closes the usual escape routes. Causal G becomes modelable. Causally inert G becomes evidentially idle. Extra-logical G becomes unassertable.

6. Name the relata of the divine attributes

The divine predicates look absolute because their indices are usually suppressed. Restoring them exposes the load-bearing structure.

Predicate	Relata and structure it requires	What the classical package removes
Knows	a knower, truth-bearer or represented state, accuracy relation, and conditions distinguishing knowing from not knowing	Divine simplicity and timeless changelessness remove the differentiated representational or relational organization by which distinct contents are known.
Can	an agent, an act-domain, constraints, counterfactual alternatives, and a transition or dependence relation	“Unlimited power” is stated while the space that makes one capacity differ from another is left unnamed.
Chooses	alternatives, a selection rule or reason, counterfactual sensitivity, and a difference between selecting A and not-A	Necessity, immutability, and timelessness erase the modal or state difference that gives selection content.
Creates	creator, created, and a specified relation: temporal production, conservation, grounding, constitution, or personal explanation	“Outside time and law” removes the framework in which production or dependence differs from non-production or independence.
Is good	a bearer, affected subjects, a welfare or persistence relation, a horizon, and a standard of comparison	“Good by nature” is used to block every possible counterexample, converting goodness from an assessment into an honorific.
Exists necessarily	a domain of possibilities and a rule determining what holds across them	Necessity is treated as intrinsic verbal force rather than a modal relation whose system and accessibility conditions must be named.

The reply that God possesses these properties in a unique, incomprehensible mode does no work. Partial incomprehension is ordinary: we can know that an unknown process has effects without knowing its mechanism. Total semantic exemption is different. If nothing distinguishes divine knowing from not knowing, divine choosing from not choosing, or divine goodness from cruelty, the predicate has not been deepened. It has been detached.

This audit inherits the corpus’s principal semantic exposure. A conceptual-role theorist can say that a term receives content from its inferential place even when no worldly relatum is named. The answer is the anchor fork. If the inferential role is not constrained by what it is about, a coherent network can preserve phlogiston, astrology, or a fictional bureaucracy as easily as oxygen. If it is constrained strongly enough to distinguish successful reference from internally stable error, the worldly anchor has returned. Divine predicates face the same fork. Unanchored role preserves a story. Anchored role restores the relations the audit examines.

7. A necessary, timeless totality is not an agent

An agent in this corpus is a bounded process that maintains a distinction between itself and an environment, has persistence conditions it can satisfy or fail, models relevant parts of its world, and acts through that model. This is not a demand that every agent resemble a human. It is the minimal organization that lets action differ from mere occurrence.

The classical God is said to be necessary, immutable, timeless, simple, independent, and incapable of failure. Such a being has no boundary it must maintain, no environment relative to which it persists, no state it can lose, no informational deficit it can correct, and no transition from undecided to decided. It fails the agenthood criterion at the joints that give “intention” and “choice” their content.

Calling it “pure act” does not restore agency. It replaces a selecting process with a timeless dependence relation. Calling the entire history one eternal choice does not restore alternatives. It redescribes the fixed totality as a decision without identifying the state or counterfactual under which it would not have been selected.

If G is instead given a boundary, internal differentiation, goals, alternatives, and coupling, it can become an agent. The cost is exact: G becomes an organized part of a wider reality and loses the properties used to make it the classical God. A very powerful world-maker may survive that repair. The object under audit does not.

8. Creation is not one flat causal word

The slogan “everything needs a cause, so God needs a cause” is too crude. Classical accounts do not all mean temporal production. The possibilities must be separated.

Diachronic creation: an earlier event or agent produces a later universe. This requires a temporal or ordered framework broad enough to contain the production relation. A cause literally outside every order cannot stand earlier, later, or in a transition relation.
Continuous conservation: the world exists at each moment because G sustains it. This posits an ongoing dependence. It therefore predicts a structured relation between sustainer and sustained; it does not place G outside all relation.
Grounding: worldly facts obtain in virtue of G. This is not temporal causation, but it remains a dependence relation with direction, relata, and modal consequences. Naming it “grounding” does not explain why a personal mind rather than a nonpersonal structure occupies the base.
Constitutive dependence: the world is made of, realized by, or is a mode of G. This approaches identity or pantheism and loses the distinct personal creator.
Personal explanation: the world exists because G intended it. This adds, rather than removes, the debts of agency: alternatives, reasons, selection, and a contentful distinction between intending this world and another.

None yields the classical package for free. A cosmological argument may point toward an unexplained terminus, necessary condition, or fundamental structure. It does not thereby derive a mind, one mind, unlimited knowledge, unlimited power, moral perfection, or concern for humans. Those are separate claims. The bridge is normally crossed by changing vocabulary faster than the argument adds premises.

“Why is there something rather than nothing?” does not select God. Declaring G necessary is a stopping rule, not an explanation of necessity. If some fact may be fundamental or necessary, the same modal permission is available to total structured reality. Adding a cosmic mind, purposes, commands, and a plan adds commitments; it does not reduce them.

9. Omniscience and free will

Free will is not one claim. The distinction decides the issue.

Libertarian freedom requires, at minimum, that an agent could genuinely do otherwise in the same relevant conditions and is not merely one caused contributor inside a sufficient prior history. Compatibilist agency requires that action issue from the agent’s own reasons-responsive organization, even when that organization is itself caused.

Exhaustive infallible foreknowledge conflicts with libertarian alternate-possibilities freedom under the standard fixity conditions. Suppose that at time $t_0$ , G infallibly knows that a person will perform act $A$ at $t_1$ . Infallibility entails that if G knows $A$ , then $A$ occurs. At $t_1$ , the person cannot make G’s knowledge false. Holding the past and the infallibility relation fixed, not- $A$ is unavailable. The exact conjunction is:

G infallibly knows at $t_0$ that A occurs at $t_1$ .
The agent at $t_1$ cannot alter the fixed past or make an infallible belief false.
Libertarian freedom requires the genuine availability of not-A under the relevant fixed conditions.
Therefore the agent is not libertarian-free with respect to A.

The standard replies—timeless knowledge, Ockhamist “soft facts,” counterfactual dependence of divine knowledge on later choice, and rejection of alternate possibilities—change or reject one of these premises. They do not establish the unconditioned authorship usually invoked to remove responsibility from an omniscient designer. Timeless knowledge changes the temporal location of the knowledge. It does not make the total history unknown. Rejecting alternate possibilities abandons the libertarian freedom the defense was using.

Compatibilist agency can survive. A caused process can remain a cause; this corpus already treats deliberation, inhibition, and reasons-responsiveness as real causal organization. But compatibilist agency does not exonerate the creator of the complete initial conditions, laws, capacities, vulnerabilities, and foreknown history. It preserves local agency, not divine innocence.

The free-will defense therefore forks badly. If it needs libertarian freedom, omniscience and the physicalist account of agency remove it. If it needs only compatibilist agency, God knowingly selects the system in which every action unfolds and retains the power to prevent its consequences. Either way, “free will” does not carry the theodicy.

10. Suffering is not repaired by calling it a plan

The problem is not that an omnipotent being cannot make a world containing any suffering. The crude logical problem has familiar defenses. The live indictment concerns the actual distribution, scale, apparent gratuity, and preventability of suffering under the joint claims of knowledge, power, and goodness.

A Plantinga-style free-will defense may block the bare inference from “some evil exists” to a formal contradiction in every possible theistic model. Grant that entire point. It does not answer this argument. A merely possible morally sufficient reason is not evidence that the actual horrors were necessary; libertarian freedom does not generate childhood cancer; and the indexed content of perfect goodness remains answerable to what an all-knowing, all-capable author knowingly makes sentient beings endure. The evidential and explanatory indictment survives the defense of abstract consistency.

For any horror H:

If G did not know H would occur, G is not omniscient.
If G knew but could not prevent H or realize the relevant good through a less harmful possible route, G is not omnipotent.
If G knew, could prevent H without losing an equal or greater good, and selected H anyway, G is not perfectly good relative to the sentient beings made to bear it.

The argument does not need the claim that omnipotence guarantees literally infinitely many alternatives. It needs one feasible lower-harm alternative. A child’s cancer could fail to initiate, fail to vascularize, respond to immunity, be detected early, or occur in a world whose biology does not use that failure mode. Children already exist for whom each of these differences obtains. Preventing one mutation does not require a square circle.

Human violence does not rescue the free-will defense either. An aggressor can form an intention while the weapon jams, the missile misses, the warning arrives, or the child reaches shelter. We routinely distinguish preventing an agent from having a thought from preventing the thought’s consequences. An omnipotent being has at least the intervention range of a human safety system. If it does not, the adjective is false.

Natural evil is more direct. Cancer, congenital disease, earthquakes, parasites, and predation do not originate in a child’s morally significant choice. Moving the responsibility to ancestral sin, demonic agency, soul-making, or unknown goods does not explain the evidence; it adds auxiliaries that are introduced only after the expected goodness fails to appear.

“God has reasons beyond our understanding” has a symmetry cost. If our moral cognition is too unreliable to judge that permitting a child’s agony is bad, it is also too unreliable to judge that God is good. The same skeptical acid dissolves condemnation and praise. A theology cannot use human moral concepts to advertise perfect goodness and then declare those concepts incompetent at every attempted audit.

Nor does “part of the plan” reduce responsibility. Under omniscience, every consequence is foreknown. Under omnipotence, every feasible alternative is available. A plan therefore increases selection responsibility. It does not wash it away.

If “good” means whatever G does, goodness has become obedience to power. If it means good for G’s unknowable purpose, it says nothing about good for children, humans, or any other sentient patient. If it retains the welfare, reasons, and avoidable-harm relations by which moral goodness has content, the world is strong evidence against the claimed author.

The resulting theological picture is morally worse than an indifferent universe. An indifferent universe selects nothing. The classical God is stipulated to know every scream before creating, possess every possible remedy, and choose the complete history anyway. No human perpetrator could approach that responsibility, because every human perpetrator is limited in knowledge and power. The theology manufactures the greatest possible author of preventable suffering and then calls the author love.

The good news is that the author is not there.

11. Absence of expected evidence is evidence of absence

“Absence of evidence is not evidence of absence” is false as a general rule.

Let $N$ be a null observation: the expected sign does not appear. Comparing G with not-G, the evidential direction is given by the likelihood ratio

$BF_{\neg G:G}=\frac{P(N\mid \neg G)}{P(N\mid G)}.$

If G makes the expected sign likely, then $P(N\mid G)$ is low. If not-G makes the null unsurprising, then $P(N\mid \neg G)$ is higher. The null is evidence for not-G over G. Repeated well-powered observations can compound that evidence, although dependencies, selection effects, and weak tests must be accounted for.

A failed underpowered study may be inconclusive. A hypothesis that never predicted an observable difference is untouched by a null. Those facts limit the inference; they do not reverse it.

The classical God is not normally introduced as causally idle. It is said to answer prayer, reveal truths, heal, guide history, ground moral knowledge, create souls, punish, reward, and seek relationships with humans. Those claims create evidential expectations: reliable effects under specified interventions, information unavailable by ordinary routes, publicly discriminating revelation, convergence not explained by cultural transmission, or some other repeatable residual that natural models do not absorb.

Divine hiddenness creates the same pressure without a laboratory. A perfectly loving personal being said to desire relationship with every capable nonresistant person makes widespread sincere nonbelief less expected. A world in which religious affiliation tracks birthplace, family, institutions, testimony, and cultural transmission is more expected if religions are human historical systems. “Hidden for a reason” can restore compatibility. It does so by spending the prediction that made perfect relational love evidentially relevant.

There is no robust, independently replicated result that uniquely requires a supernatural explanation. That sentence is more exact than “there is literally no evidence.” Testimony, private experience, tradition, apparent design, and sacred texts are data in the weak sense that people report and interpret them. They do not discriminate the God hypothesis from psychology, culture, selection, coincidence, error, fraud, or rival religions. Their abundance does not repair their likelihood structure.

Intercessory prayer is one narrow case, not the whole argument. The large STEP randomized trial found no improvement in complication-free recovery from receiving prayer, and a Cochrane review of ten randomized trials involving 7,646 people found no clear effect on death or general clinical state and judged the evidence equivocal. These results do not disprove every god. They are exactly what absence of expected evidence looks like when a specific interventionist claim is made testable.

The usual response is to modify G: God will not be tested, answers are indistinguishable from nonanswers, hiddenness serves a greater good, every outcome is compatible with the plan. Each repair raises $P(N\mid G)$ by making nulls expected. It thereby protects logical compatibility at the price of empirical content. A model that predicts recovery, non-recovery, revelation, silence, rescue, catastrophe, belief, and nonbelief equally well has not become profound. It has ceased to risk contact with reality.

12. The standard positive arguments do not reach the target

Cosmological arguments

At most, these arguments seek a first cause, necessary being, sufficient reason, or non-contingent ground. Every step is contested, but grant the terminus. Nothing in “necessary ground” entails personhood, singularity, omniscience, omnipotence, goodness, revelation, or interest in human conduct. The God of worship enters after the argument has ended.

Fine-tuning arguments

Fine-tuning arguments depend on probability distributions over possible constants or laws that are not independently established, selection effects that must be conditioned on, and an unexplained prior over designers and their preferences. Even a successful design inference would yield a designer of some domain, not a necessary, perfectly good, unlimited mind. A cosmic engineer is not classical God.

Moral arguments

The premise that objective moral facts require God fails on both sides. This corpus locates oughts in relations among agents, actions, beneficiaries, persistence conditions, and horizons; it finds no categorical ought floating free of every bearer. Adding divine command creates the old fork rather than closing it: if an act is good only because G commands it, goodness reduces to power-backed preference; if G commands it because it is good, goodness does not depend on G.

Ontological arguments

Modal ontological arguments place the conclusion in their possibility premise. In modal systems where possible necessary existence entails necessary existence, the disputed work is done by “possibly, a necessarily existing maximally great being exists.” That is not an innocent starting point. A definition can specify what would follow if instantiated. It cannot instantiate its object.

Religious experience

Religious experiences are real experiences. Their supernatural interpretation is not forced by their phenomenology. Incompatible religions generate incompatible attributions from experiences produced through the same human cognitive architecture; neurological, psychological, social, and predictive-processing models explain why such experiences occur without validating their objects. Private conviction is not public discrimination.

The pattern is stable. Arguments that retain content reach at most an unknown cause, ground, order, designer, or experience. Arguments that leap to the classical God do so by importing the attributes they claim to derive.

13. Why physicalism and supernaturalism are not peers

Physicalism does not know every law, solve consciousness, close cosmology, or eliminate all metaphysical disputes. Its open problems are not evidence for a supernatural rival unless that rival predicts and explains the missing structure better.

The comparison is asymmetric:

Physicalism inherits every successful physical, biological, cognitive, and social model without adding a second substance or exempt agent.
It predicts continuity: unexplained phenomena should yield to better structured models rather than to exceptions from modeling.
It exposes itself to correction through causal and empirical contact.
It compresses domains that supernaturalism historically split among unrelated wills.
It can accommodate emergence, multilevel causation, and unknown lawful ontology without turning “physical” into “currently listed particle.”

Supernaturalism does the reverse:

It adds a primitive bearer of enough unmodeled capacities to know and select every fact in the world it is meant to explain.
It supplies no mechanism for disembodied knowledge, intention, or causal influence.
It predicts no stable residual that survives controlled investigation.
It uses mystery selectively: enough content to claim creation, purpose, goodness, and authority; too little content to permit a failed prediction.
It converts counterevidence into evidence for hidden wisdom and therefore disables correction.

This is not a tie broken by personal taste. Physicalism is the maximally inductively justified standing ontology. Classical supernaturalism is among the least justified would-be explanations of reality because it combines maximum ontological ambition with minimum discriminating constraint.

14. The exact epistemic verdict

“Maximally logically wrong” is not the clean label. Logic is not ordinarily a scalar. An argument is valid or invalid relative to a logic; a set of propositions is consistent or inconsistent; a sentence is truth-apt or ill-formed under an interpretation. Different readings of God fail differently:

some are semantically ill-formed, because their predicates have been stripped of the relations that fix their content;
some are internally unstable or inconsistent, because the attributed agency, necessity, timelessness, simplicity, knowledge, freedom, and goodness cannot all retain their indexed meanings together;
some are empirically disfavored, because expected differential effects fail to appear;
some are explanatorily empty, because they accommodate every outcome and exclude none;
the coherent residues are misclassified, because they describe lawful finite or structural possibilities rather than the classical supernatural God.

The scalar verdict is epistemic. The God hypothesis has near-minimal epistemic standing among serious ontological explanations. Confident belief in it is near-maximal epistemic misalignment at a load-bearing ontological joint and therefore a high-weight direct failure of existence literacy.

“Near” is not a softener. It records that epistemic measures have no demonstrated finite maximum and that total existence illiteracy is an asymptote. The verdict itself is categorical. The hypothesis is false. Its standing is not merely lower than physicalism’s; it is inverted relative to the evidence, retaining confidence in proportion to the mechanisms used to prevent correction.

Suspension does not become the correct endpoint merely because agent-level certainty remains corrigible. Once every live reading has either changed the subject, entered ordinary structured reality, contradicted the defining package, failed its predictions, or lost its truth conditions, “I just lack belief” understates the result. Atheism here is propositional: G does not exist.

15. What remains open

Unknown reality remains open. A wider cosmos remains open. New fundamental structure remains open. A simulator or finite universe-maker remains open. Nonhuman agents with powers beyond ours remain open. Present physics can be wrong in major ways.

None of that keeps classical supernatural theism alive.

A finite creator is not the unbounded, necessary God under audit. A simulator is not the ground of total reality. A lawful meta-cause is not outside law. A nonpersonal ground does not know or love. Total reality is not a bounded agent. An extra-logical “being” cannot be asserted to exist. The open categories are exactly the categories the God label obscures.

The conclusion does not come from closing inquiry. It comes from keeping the categories open and refusing to let one failed object migrate among them whenever pressure arrives.

16. Kill conditions

The claim is corrigible. A successful rival must do more than report conviction or rename an unknown. It must:

specify a stable referent for G;
name the relata and truth conditions of knowledge, power, will, creation, necessity, and goodness;
show that the defining attributes are jointly coherent under those meanings;
preserve genuine agency without turning G into a bounded process inside a wider reality;
produce distinctive evidence that is more expected under G than under physicalism and its live extensions;
survive independent replication and adversarial checking where the claim is empirical;
explain more than the physicalist model without purchasing compatibility through ad hoc auxiliaries; and
answer the distribution of suffering without emptying “good” of the relations by which goodness is recognized.

Meeting those conditions would change the map. No theology has met them.

Standing of this document

An argument and synthesis note. Within the corpus, the conclusion that the classical supernatural God does not exist is a result of applying the standing physical floor, agenthood criterion, no-global-optimizer result, semantic indexing diagnostic, causal distinctions, and evidence discipline to one specified target. It is not presented as a novel theorem or as a proof independent of every possible semantics. Its strongest inherited exposure is the corpus’s referential-relational account of the audited predicates; the conceptual-role reply and the anchor fork are stated in §6. Its empirical component is cumulative and model-relative, not a claim that every religious sentence has been experimentally tested. Its free-will component targets libertarian alternate possibilities, not compatibilist causal agency. Its suffering component is evidential and explanatory, not the crude claim that the existence of any evil alone produces a universally accepted formal contradiction.

Those scope statements do not weaken the conclusion. They identify exactly what carries it and exactly what would break it. On any conflict with the apex synthesis or the governing modules, they govern. Corrections and counterexamples are welcome and change the note.

References

Internal

Epistemic Forge: The Apex Synthesis. The tiered map, physical floor, agenthood criterion, no-global-optimizer result, and current standing of the corpus.
The Fidelity Program. Map–territory divergence, correction, and evidential discipline.
The Spine and the Web. The architecture of the program, its physical spine, and its inheritance structure.
Name the Relata. The semantic diagnostic, the ghost verdict, the conceptual-role exposure, and the anchor fork.
Existence Literacy. The load-bearing ontological joint and the distinction between a false belief, a defective process, and an existence-illiterate architecture.
The Two Ways to Be Wrong. The symmetry between overclaim and false concession, and the requirement not to hedge a result into vagueness.
Causality Is Not a Flat List. Diachronic, sustaining, grounding, constitutive, and multilevel causal relations.
You Do Not Need to Be Uncaused to Be a Cause. Compatibilist causal agency and the rejection of fatalist inference from determinism.
What Can Bear an Ought. Agency as the gate on normative predicates and the no-global-optimizer application.
The Is/Ought Firewall. Agent-indexed normativity and the denial of a categorical ought detached from its bearer.
The Wizard’s Error Is Delocalization. The error of moving a local property outside the system and scale that bear it.
Everything Is One Process. The continuity of concrete reality and the rejection of ontological exemption.
Reality Alignment. Alignment as corrigible contact between a physically realized map and the reality that constrains it.
Reality Alignment as a Persistence Strategy V3. The current persistence-indexed treatment of reality alignment and its consequences for embedded agents.

External

“Physicalism.” The Stanford Encyclopedia of Philosophy. Supervenience formulations, theories of the physical, Hempel’s dilemma, and major objections.
“Naturalism.” The Stanford Encyclopedia of Philosophy. Causal closure and the causal argument for ontological naturalism.
“Atheism and Agnosticism.” The Stanford Encyclopedia of Philosophy. Propositional atheism, agnosticism, priors, evidence, and burden-of-proof distinctions.
“Foreknowledge and Free Will.” The Stanford Encyclopedia of Philosophy. Theological fatalism and the principal replies.
“Free Will.” The Stanford Encyclopedia of Philosophy. Libertarian, compatibilist, and skeptical accounts.
“The Problem of Evil.” The Stanford Encyclopedia of Philosophy. Logical and evidential formulations.
“Skeptical Theism.” The Stanford Encyclopedia of Philosophy. The unknown-reasons response and its epistemic consequences.
“Hiddenness of God.” The Stanford Encyclopedia of Philosophy. Nonresistant nonbelief and hiddenness arguments.
“Omnipotence.” The Stanford Encyclopedia of Philosophy. Maximal power and the restriction to genuinely possible states of affairs.
“Divine Freedom.” The Stanford Encyclopedia of Philosophy. Divine necessity, foreknowledge, and freedom.
“Eternity in Christian Thought.” The Stanford Encyclopedia of Philosophy. Timelessness, changelessness, and divine action.
“Divine Simplicity.” The Stanford Encyclopedia of Philosophy. The classical simplicity doctrine and its problems.
“Creation and Conservation.” The Stanford Encyclopedia of Philosophy. Creation, conservation, continuous creation, and dependence.
“Cosmological Argument.” The Stanford Encyclopedia of Philosophy. Causal, contingency, and necessary-being arguments.
“Fine-Tuning.” The Stanford Encyclopedia of Philosophy. Design, chance, necessity, selection, and probability disputes.
“Moral Arguments for the Existence of God.” The Stanford Encyclopedia of Philosophy.
“Ontological Arguments.” The Stanford Encyclopedia of Philosophy.
“Religious Experience.” The Stanford Encyclopedia of Philosophy.
Benson, H. et al. (2006). “Study of the Therapeutic Effects of Intercessory Prayer (STEP) in cardiac bypass patients.” American Heart Journal 151(4), 934–942.
Roberts, L., Ahmed, I., and Davison, A. (2009). “Intercessory prayer for the alleviation of ill health.” Cochrane Database of Systematic Reviews, CD000368.

Author: Andraž Đurič, Slovenia. Draws on the Epistemic Forge corpus; on any conflict, the apex synthesis and governing modules govern. Written in dialogue with ChatGPT (OpenAI), which assisted with corpus comparison, adversarial literature review, argument architecture, and drafting. Contributions are judged on content rather than origin; doing otherwise would be the genetic fallacy. Text licensed CC BY 4.0.

atheism · classical theism · physicalism · supernaturalism · epistemology · philosophy of religion · problem of evil · free will · divine hiddenness · causal closure · semantic indexing · Epistemic Forge

Search This Blog

Epistemic Forge