The governing standard a system runs on must be supplied to the execution function; it cannot be legitimated merely by executing it. A model, an orchestration layer, a decentralization scheme, a clean data foundation, a perfected hardware interlock — none supplies it. It enters from a locatable governing role.

The same person or institution may occupy both the source role and the execution role. The dependency is logical, not necessarily a separation of actors: what cannot happen is the execution function authorizing the basic norm it presupposes, regardless of who runs it.

A standard can be reshaped across time without being derived from the execution layers beneath it. Generated candidates, critique, evidence, or convergence inside a workflow may inform revision; they do not themselves authorize the revision. The regress halts only when a locatable source of intent adopts a standard under authority and recourse. Non-derivable does not mean immutable; it means the standard's authority is not produced by the machinery that later executes under it.

Source of intent is a locatable governing role, not exhaustive specification or execution control. Execution may be delegated, stochastic, collective, or exploratory; the apex remains attributable as long as purpose, authorization, closure, and answerability remain attributable rather than dissolving into aggregation. The inverse error is to let private intention float free of those gates.

A distributed- or extended-cognition account may place the human inside the boundary of the coupled cognitive system — operator, models, repositories, critics, and durable context form one system. In ASK's topology, that does not reduce the human to another participant “in the loop”: human judgment stays at the apex, supplying or authorizing the purpose and governing standard the execution layers cannot derive. Read naively, system membership may still seem to collide with apex non-derivability: if the human is part of the system, how can the apex fail to be produced by it? The two claims answer different questions and do not compete.

• Membership — is the human part of the coupled cognitive system? Yes. That describes the system boundary, not the human's normative role; in ASK, human judgment remains at the apex.
• Derivability — can the apex's function (determining which purpose and standard are authorized to govern a named scope) be produced by the layers below it? No. Origin does not aggregate upward.

An apex inside the system boundary is not a paradox but an axiom — internal to the scoped system, but supplied rather than generated by the execution layers beneath it.

The apex is defined by directionality, not location. Candidate aims and proposed standards may arise anywhere. Standing intent acquires operative force through adoption or authorization by the source role for a named scope; delegation carries that force and bounded decision rights downstream. The governing standard's authority is not synthesized upward from the components, nor is it produced merely by execution under that standard. The boundary may be drawn to include the apex (the whole apparatus) or to exclude it (the stack the apex governs) — both scopes are valid, and the invariant across both is this normative direction.

Corollary: membership never answers an origin question. "The human is part of the system" (true) does not license "the system can supply its own source of intent" (false). Capability gains in the lower layers are not the apex being filled.

Accountability decomposes into three things distribution can never produce:

• Intent — before the act: what the system is for and which governing standard applies.
• Authority — the right to act, not merely the power.
• Recourse — after the act: who answers when an authorized move is later judged wrong.

The three are jointly necessary. Intent without authority is preference. Authority without a nameable intent is power. Either without recourse is unaccountable. The source may be an individual, office, institution, governed collective, constitutional order, or delegated authority — but purpose, authorization, closure, and answerability must remain locatable. It need not precompute every result; it governs why the aperture exists, where its bounds sit, who may accept, reject, revise, or bind what emerges, and who answers for the result.

The debate usually runs on a two-term ontology — human vs. machine, two minds across a gap. That interactional dyad is real, but it sits inside a governance stack, and the executor may be human or computational. The dyad asks who is across the gap; the apex asks which layer answers. "Apex" names structural dependency and answerability, not moral status or dominance — a governed stack has different heights, but that is ordering, not rank.

The requirement holds independent of the consciousness question. Whether or not anyone is home in the executor, a consequential system still needs a locatable apex. The requirement is over-determined: the purpose and beneficiary of the system, and the executor's incapacity to legitimate its own standard, each suffice to secure it.

Held source-role boundary: The permitted-source list above and the human-or-computational executor ruling do not together determine whether a computational office may itself occupy the source role. This doctrine also does not determine whether chains of constituted authority must ultimately terminate in natural persons. Both questions are deliberately held open.

Indexed partial results inside the hold. The hold is not lifted — both questions remain open — but the rulings bearing on them are indexed rather than general, and two carry results worth stating. An office may hold constituted authority without being a legal person in its own right. That settles the legal-personality component of the first question above; it does not establish root-role occupancy.

On the liability limb specifically, a natural-person terminus is not required. Liability is one indexed strand of a chain of constituted authority, and settling it does not settle the register in which such a chain finally comes to rest for answering: non-retributive answerability, punitive or retributive responsibility, and the anchor a tracing argument would need all remain open. That is the terminus question, not the requirement that answerability remain locatable, which the decomposition above states. Neither partial resolves either held question. The legal-personality result does not establish root-role occupancy; the liability result does not address it.

Where the underlying evidence sits. The partial results above are already part of this doctrine. The underlying evidence record and the ASK rulings indexed here are assembled in the Consciousness + Free Will structured survey. Its public projection is implemented under apexSolarKiss/ASK/apex-solar-kiss/consciousness-free-will/ and rendered at a-s-k.studio/apex-solar-kiss/consciousness-free-will/. Two of the survey's held questions restate this boundary's limbs verbatim, and a third bears on the occupancy limb without being identical to it. The survey informs this doctrine; it does not authorize it, and it resolves neither held question.

The apex is a locatable governing role — a chair, a hat, an office someone occupies. It is not a file, a folder, or a document. Artifacts carry the role's intent; they never constitute it.

source of intent    the role that supplies or authorizes purpose and the governing standard
artifact of intent  a carrier through which that role's intent is expressed, deployed,
                    routed, relayed, or preserved

The confusion is easy to make and expensive to keep: naming a storage plane after the role invites the reading that the plane is the source of intent, and therefore that placing an artifact there confers authority. It does not. Folder membership confers no authority, no activation, and no queue state.

A document about source of intent — a grounding note, a context master — remains an artifact of context, even when its subject is the apex and even when it carries authoritative intent-bearing claims. docs/intent-artifacts.md governs the carrier classes and their lifecycles; this doctrine governs the role.

Standing intent is normative content adopted by a source of intent for a named scope, remaining in force until revised, revoked, or retired through the applicable governance path. Its continuing force does not require a fresh source-of-intent act for every downstream action within that scope. Standing here describes the status of the intent-content — not whether any carrier is continuously deployed or available, which is a separate activation posture governed by docs/intent-artifacts.md.

Four slots stay distinct, and collapsing any pair is the recurring error:

source of intent      the governing role that may adopt or authorize
standing intent       adopted normative content currently in force for a named scope
artifact of intent    what preserves, expresses, deploys, routes, or relays that content
runtime realization   the causal organization through which intent influences operation

source of intent      the ROLE is the CONSTITUTIVE LEGISLATIVE SOURCE — it adopts or
                      authorizes the governing standard
standing intent       adopted normative content; it performs no legislative function
artifact of intent    classified by WHAT IT DOES: DELEGATED legislative apparatus where it
                      acts as a grammar, brief, rule, or grant carrying the standard into
                      operative scope — not legislative merely because it carries intent
runtime realization   a causal organization, not itself a function class; what operates
                      within it is classified by what it does — realizing candidates under
                      the standard is executive; evaluating, selecting, and closing is
                      judicial; deriving a sub-brief under a grant is delegated legislation

A source of intent may delegate task formation, local subgoal-setting, evidence gathering, candidate generation, evaluation, and — where governance names the scope — bounded binding judgment. None of those acts, alone or stacked, delegates the source role itself. What travels down a delegation chain is intent, apertures, and bounded decision rights from the apex; the apex does not travel with them.

The ordinary worked topology makes this concrete. An operator authorizes a program objective and its governing aperture. A parent executor receives that delegated intent, derives bounded task intent, and dispatches sub-briefs to computational subagents. Subagents realize the tasks — generating candidates, analyzing, comparing, critiquing, testing — and exercise scoped binding judgment only where an explicit grant permits it. The parent verifies what returns, rejects or integrates it, and performs local selection within its own grant. The operator retains operative authorization, program-level closure, and ultimate answerability. Delegation may recurse where the grant permits: each authorized dispatch opens a nested aperture inside the outer one, and recursion increases execution depth, never source-role height.

A lower office comes to occupy a source-of-intent role only where governance separately constitutes or recognizes it there for a named scope, with purpose, authorization, closure, and answerability attributable for that scope — the same requirements the decomposition states. Ordinary dispatch constitutes nothing: the proximate author of a task brief is not thereby its source of intent, because causal or textual origin of an instruction is not normative origination of the standard it serves (docs/intent-artifacts.md owns the matching drafting-provenance ≠ normative-voice distinction). Whether a computational office may itself occupy the source role, and whether chains of constituted authority must ultimately terminate in natural persons, remain the held source-role boundary above, together with the indexed partial results recorded there — this section answers neither question in either direction.

The subject is one primitive stated in distinctions the discourse keeps eliding:

• power ≠ authority — the capacity to act is not the right to act.
• authorization evidence ≠ conferred authority — credentials, identity proofs, role grants, signatures, permissions, and approval records are operational carriers by which governance recognizes an upstream grant; they are not the source of that grant. Verification may recognize an authorization within scope under the current policy, but it cannot originate or legitimate the authority behind it.
• realization persistence ≠ continuing authorization — that a system still carries an intent, and reconstructs it after disturbance, establishes only that the realization persists. It does not establish that the authorization behind it remains in force. A goal can outlive the standing intent it once realized.
• aggregation ≠ origination of intent — a blend of participants, or a statistical composite of preferences, does not originate a governing intent. A poll aggregates; a constituted collective decides.
• intention alone ≠ legitimate authority — private will does not self-authorize; it acts through attributable authority, closure, and recourse.
• free-will sourcehood ≠ the source-of-intent role — sourcehood in the free-will sense concerns an agent's relation to its own actions and motives; the source of intent is a normative-governance role carrying standing to originate a directive that binds another actor or office. They are not one question in two vocabularies. Non-identity is not non-requirement: whether a separate free-will condition bears on eligibility for the root source role is not settled here.
• the interactional dyad ≠ the governance stack that contains it — the field mistakes the two-minds-across-a-gap picture for the whole structure.
• membership ≠ derivability — being inside the system boundary is not being produced by the system. In ASK's topology, the human is inside the coupled system while judgment remains at the apex, a role the lower layers cannot generate.
• location ≠ origin — the apex is directional, not positional: "on top" is a picture; standing intent acquires operative force at the source role and is carried through delegation. The governing standard's authority is not synthesized upward from the components.
• system capability ≠ apex integrity — a capability gain in the lower layers cannot offset erosion of the apex faculty (judgment / selection); "the system got more intelligent" never answers an origin question.
• candidate-generation ≠ standard-authorship — generating candidates, including candidate standards, is not authoring or adopting the standard by which they are judged; the apex faculty is the adoption, not the proposal.
• delegated intent ≠ delegated source role — tasks, apertures, and bounded decision rights travel down a delegation chain; the governing role does not travel with them.
• delegation depth ≠ source-role height — recursive delegation deepens execution; it raises no delegate toward the apex.
• proximate task instruction ≠ origination of standing intent — composing the brief that directs a lower layer is derivative authorship under the upstream standard, not adoption of a standard.
• non-derivable ≠ immutable — the standard is non-derivable at any instant yet plastic across time; the apex may revise it through cycles without it becoming derivable from below.
• operative standing ≠ correctness — the source-of-intent role supplies or authorizes the governing purpose and standard for a named scope. The applicable governance path determines which results are accepted as conforming for operative purposes. Adoption and acceptance establish standing; they do not, by themselves, establish factual truth, explanatory warrant, evaluative rightness, the work's merit, actual conformance, or immunity from criticism.
• convergence ≠ authority — a loop may converge to its own fixed point; convergence can stabilize a candidate standard, but it does not authorize that standard. Authority requires adoption by a locatable source of intent and remains answerable through governance and recourse.
• convergence ≠ convergence-to-truth — a converging process settles to the fixed point its dynamics define, not necessarily to a true one. Truth-tracking requires evidence and criticism capable of exposing the relevant errors; conformity to premises a process merely carries forward supplies no additional warrant for those premises. Exteriority is relative to the error class, not to location, authorship or the number of agents. Internal derivation retains its scoped evidentiary force; the fuller distinction between inherited assumption and new warrant is set out in docs/method.md.
• adversarialness ≠ independent error — adversarial checking tracks truth only to the degree the critic's errors are independent of the error class being tested. Adversarialness is a posture; independence is a property, and it is graded per error class.

Two further distinctions are argued in full in the essay line and carried here by reference: the residue of judgment ≠ the act of judgment, and outputs ≠ intent (see Judgment Doesn't Transfer).

The normative apex supplies or authorizes the governing standard. Governance constitutes or recognizes authority, translates the standard into rules, thresholds + bounded decision rights, and provides verification, adjudication, revision + recourse. The operational floor is the boundary those rules establish. A substrate or execution mechanism may carry, expose, clear, or enforce that floor; it cannot legitimate the standard that authorized it.

Every substrate verb — clear, see, enforce, run, audit, arrest, de-energize — operates on a floor whose location has already been supplied to that substrate. Define is never on the substrate's list. The standard is presupposed and supplied upstream of the substrate's execution function. Four structurally different worked cases make the point and are argued in full in the essay line: the architecture layer, the state / enforcement layer (Authority Without a Standard), the operational layer (detection supplies evidence about the relevant factual condition, never authority to determine that condition's governing meaning), and the physical / limit layer (Determinism Is Not a Standard — perfecting enforcement only throws the standard's absence into sharper relief). You Cannot Build Your Way to the Apex synthesizes the essay arc; Judgment Doesn't Transfer argues the corpus frame.

The apex states the normative requirements — purpose, governing standard, legitimate authority, closure, recourse. Governance (docs/governance.md) provides their institutional mechanisms — how authority is constituted or recognized, how rules are made and applied, how application is verified and adjudicated, how recourse is operated. Authority and recourse appear in both docs for this reason: the apex names them as requirements; governance operates them as mechanisms. The two are jointly necessary in consequential, delegated, or collective systems, and are articulated separately rather than merged. Adopted governing standards remain operative while governance is designed for the possibility that any one may contain correctable error.

• Kelsen — the Grundnorm. The basic norm is presupposed, supplied to the system, not generated by it.
• Dworkin — weak vs strong discretion. Bounded interpretive latitude inside a standard vs the authority to set the standard.
• Fuller — the Speluncean Explorers. Competing legitimate sources of authority; the standard is contested, never self-evident from the facts.
• Weiner — the Declaration of Intent. Intent declared and binding ahead of the act.
• Power vs right. The load-bearing distinction throughout.

The structural face of source of intent, generalized beyond any single project. It clears the promotion bar by recurrence: the same apex absence recurs across distribution, data, enforcement, supply, operations, silicon, and corpus framings (the essay arc), and across image-making, asset production, and civic domains. It articulates the apex the operational handling doctrine (docs/source-of-intent.md) presupposes.

This doctrine should be superseded by:

• a thinner apex statement if the non-identity field earns its own home (e.g. docs/non-identities.md);
• a refactor if docs/governance.md and this doc are found to share enough structure to warrant a common parent — the "co-primitive" question is deliberately held open here, not asserted.

The doctrine is not finished; it states the apex modestly and refers the full arguments to the essay line.

Anchor reading lives in docs/articles.md. The source-of-intent arc and the trilogy — Determinism Is Not a Standard, Judgment Doesn't Transfer, You Cannot Build Your Way to the Apex — carry the full arguments this doctrine states in compact form.