Jump to text
On this page
Copy and source
Published source confirmed at last check

Source changed 2026-11-18 21:11:39 UTC · snapshot created 2026-11-18 21:12:35 UTC · last check 2026-11-18 22:00:10 UTC

C.35 - Assess Generated or Discovered Results for Architecture Use

Type: Architectural pattern Status: Stable Normativity: Normative unless marked informative

C.35:1 - Problem frame

Use this pattern when a generated, searched, clustered, queried, learned, transformed, simulated, or discovered result may seed or inform architecturing, and the practitioner must decide what that result is and whether it can enter architecture work before or around C.32 candidate admission.

Primary working reader: an architect, architecture researcher, AI-assisted architecture worker, model-based engineer, or reviewer receiving a result from DSM and MDM modularization, MBSE query and view generation, graph grammar, model transformation, NAS, DSE, QD, OEE, and NQD search, LLM-assisted architecture design, code-agent mapping, simulation, benchmark trace, or source discovery.

Typical entry phrases:

"The LLM generated an architecture diagram; can it seed synthesis?"
"The DSM clustering suggests modules; is this a candidate architecture yet?"
"The MBSE query produced a view; which organization is observed, proposed, or still unestablished?"
"NAS found a Pareto point; what architecture claim can use it?"
"A graph grammar transformed the model; what preservation and bearer boundary must be checked?"

Primary working object. The generated or discovered result on which the next architecture use would rely.

First useful move. Identify the result and the organization it concerns, then state its supported next use and any condition or limitation that affects reliance. Distinguish supported, refuted and unestablished claims about existing organization from proposals for a possible organization. Keep a proposal modal; add C.30 claim identity when the receiving use needs independent reidentification.

Keep a representation distinct from its represented object when correspondence or representation operations matter. Use C.30.AD or C.30.ASV for architecture descriptions or views, and C.29 for a remaining mathematical-representation choice, transfer, or reliance question. Publication detail enters only when availability or form changes the use.

Normal first result. A short sentence can suffice; reuse an adequate existing account. When separate lines help, use:

Result: <result relied on>
Organization: <organization concerned; whether the claim about it is established, refuted or unestablished, or presents a proposal>
Next use: <supported use and any unresolved condition>
Limit and return, if needed: <material limitation and how to resolve it>

Stop there when another practitioner can make the intended next move safely. Only when result identity, branch evidence, or a receiving claim must be reidentified independently, extend those four facts into the optional StructuralSynthesisDiscoveryAdequacyNote@Project C.2.1 episteme:

StructuralSynthesisDiscoveryAdequacyNote@Project:
  resultRef:
  organizationConcern:
  nextArchitectureUseAndCondition:
  forbiddenOverreadOrReturn?:
  resultKindAndIdentityRule?:
  admissibleUse?:
  unresolvedConditions?:
  representationRef?:
  representedObjectRef?:
  publicationFormRef?:
  publicationOccurrenceRef?:
  presentationCarrierRef?:
  projectWorkOccurrenceRef?: U.EntityRef constrained to U.Work
  structuralSynthesisAdequacyNoteProjectUseRelationRef?: U.RelationRef under a named architecture-use or work-use predicate when that relation identity is material
  groundedArchitectureQuestionRef?:
  resultBranch?: transformation | discovery | generative-proposal
  generationOrDiscoveryMethodRef?:
  generationOrDiscoveryWorkRef?: U.EntityRef constrained to U.Work
  generationOrDiscoveryWorkAttributionRefs?: refs to obtaining F.6 performedUnderAssignment relations only when the note or receiving use expressly represents attribution
  workToTransformationOrProductionClaimRefs?:
  transformationBranch?:
    exactSourceObjectRefs:
    exactResultObjectRefs:
    transformationTraceRef:
    preservedStructure:
    lostStructure:
    actualTransformationRefs?:
  discoveryBranch?:
    observationOrExtractionBasis:
    observedInferredUnknownStatus:
    coveredRegion:
    unexploredRegion:
    uncertainty:
    validation:
  generativeProposalBranch?:
    constraintRefs:
    proposedOrganizationContent:
    knownOmissions:
    validationNeeds:
    declaredBaselineComparison?:
      exactBaselineObjectRefs:
      preservedStructure:
      lostStructure:
  obtainingConstraintGovernedUnfoldingStructureRef?: A.22.CGUS reference only
  sourceLabelRecoveryRef?:
  obtainingStructureRefs?: A.22 U.Structure references only
  modalArchitectureClaimRef?: C.30 ArchitectureClaimRef or C.2.1 ClaimAddress to its claim
  candidateAdmissionCondition?:
  bearerOrRealizationBoundary?:
  obtainingRealizedHolonStructureRefs?: positive A.22 references only
  measurementOrEvalReturnRefs?:
  bearerFeasibilityQuestionRef?:
  nextClaimOrRuleRef?:
  receivingClaimKind?:

The note retains the result, organization, next use and any material limit or return; its fields are included when the selected branch or receiving use needs them. Its EntityOfConcern is the result designated by resultRef, not the reference value, representation, publication occurrence, form, or carrier used to reach it.

When publication detail is relied on, publicationFormRef, publicationOccurrenceRef, and presentationCarrierRef refer to the form, occurrence, and carrier, respectively; omit every one that the receiving use does not need.

The note’s ClaimGraph states the organization concerned, supported next use, and any material limitation or unresolved condition. A plausible mistaken inference is corrected when that changes reliance. Do not open or fill the dossier merely to prove completeness.

Here @Project is a compatibility and retrieval cue only. It establishes no project entity, composite-work identity, context, authority, viewpoint, or parthood.

When the note is used in one project, projectWorkOccurrenceRef identifies the composite U.Work and structuralSynthesisAdequacyNoteProjectUseRelationRef identifies the relation by which that project Work uses the note. The suffix or either reference alone establishes no project locality. The admission-note episteme, its result, and the composite project Work remain distinct.

When reliance depends on the result’s generation or discovery history, retain the operation and the material source, input, performer and condition facts needed to establish that history and its bearing on the result. For an express claim about a U.Work, generationOrDiscoveryWorkRef names that independently admitted occurrence: its performers have A.13 cores and A.15.1 supplies Work admission.

Include the Method reference when the receiving claim depends on that Method. generationOrDiscoveryWorkAttributionRefs appear only for precise assignment-bound F.6 attribution. Missing or failed F.6 attribution alone leaves independently admitted Work intact; recheck an admission premise questioned by the same finding and any conclusion depending on it. The operation, Work, attribution, admission-note episteme, result, representation and publication objects retain their distinct identities.

actualTransformationRefs cite independently identified A.3.4 bounded changes. A positive link from Work to an actual transformation or produced entity uses the declared predicate governing that link and its supporting facts, including A.15.PROD when its branch is selected. Return missing-governor under A.6.RCD when no predicate governs the link; leave missing case facts unresolved and reject a link contradicted by them.

An entry in obtainingStructureRefs resolves to an A.22 U.Structure with independently identified constituents, obtaining relation occurrences, applied constraints and a named use frame. An incomplete basis for a claim about existing organization leaves that claim unestablished.

A proposal remains modal until its proposed organization obtains; supply modalArchitectureClaimRef when that claim must be independently reidentified. A result, representation, publication item, graph, cluster or description supplies only the structural facts it establishes.

Adoption test. After using C.35, another practitioner can state the four-line minimum or an equivalent sentence: the result, the organization concerned and whether the claim about it is established, refuted or unestablished, or presents a proposal, the supported next use and any unresolved condition, material limitation or required return. That practitioner can also distinguish claim content, an obtaining A.22 structure, its representation, and a publication-side object when those distinctions affect use. Additional identity, branch, Work, bearer, publication, evaluation, or next-claim detail appears only when the receiving use relies on it.

What C.35 buys in practice. The practitioner can keep a useful generated or discovered result without treating it as authority for architecture claims. Architecture use attaches to the result; changing a rendering or file does not silently change the admitted claim, and admitting a carrier does not silently admit claim content.

Ordinary working move. Explain the organization concerned, its actual or proposed status, what remains unestablished, and the use the result supports. Stop when the next move and any material limit or return are clear. Claim an obtaining A.22 structure only when its four identity discriminators resolve. Preserve an unestablished claim about existing organization as unresolved and a proposal as modal. Add claim identity, representation, branch, Method, Work, bearer, publication or evaluation detail when the intended use depends on it; use C.29 for an unresolved mathematical-representation choice, transfer, or reliance question.

Not this pattern when. If the current question is how to search, choose, measure, decide, authorize, publish, govern a reusable generator, govern a cultural-evolution case, or run the work itself, use the pattern that defines or decides that question first, including C.36 for the cultural-evolution relation bundle. Use C.35 only when a generated or discovered result must be admitted or rejected before another architecture claim relies on it.

C.35:2 - Problem

Modern architecture work receives claim-bearing outputs, representations, publication items, and other results from DSM clusters, MDM slices, MBSE queries, generated views, graph grammars, model transformations, LLM architecture proposals, AI-assisted ADD, code-agent relation graphs, NAS graphs, DSE traces, Pareto fronts, QD archives, benchmark traces, simulations, and source-corpus mining.

These results can expose candidate decompositions, relation gaps, hidden invariants, feasible search regions, trade-off points, source labels, or overlooked organization. They are not automatically obtaining A.22 structures, realized holon structures, eval results, evidence sufficiency, or decision authority.

C.35 handles the gap between result identity and architecture use. It first asks what the result is; which claim, represented object, or proposed organization it carries; whether representation or publication distinctions matter; and which architecture use is proposed. It then asks only the questions belonging to the result’s branch:

  • A transformation names its source and result objects, trace, preservation, and loss.
  • A discovery names its observation or extraction basis, what is observed, inferred, or unknown, the covered and unexplored region, uncertainty, and validation.
  • A generative proposal names its constraints, proposed organization or claim content, known omissions, and validation needs; source and preservation enter only when a baseline is declared.

Method, performed Work, attribution, bearer feasibility, and publication detail enter only when the receiving use relies on them.

C.35:3 - Forces

ForceTension
Discovery value vs authority overreadGenerated and discovered outputs widen the candidate space; any selection, decision, proof, or realization claim still needs its own rule and support.
Result, proposal, obtaining structure, and representationClaim-bearing result, modal architecture proposal, obtaining A.22 structure, represented object, representation, publication form or occurrence, and presentation carrier have different identities; include only the distinctions on which the architecture use relies.
Search quality vs architecture adequacyA Pareto point, benchmark score, archive member, or cluster objective can guide synthesis only through the evidence required by its branch and the rule for the next synthesis claim.
Model transformation vs preservationGraph grammars and model transformations can return useful results, but the transformation claim needs its rules, source and result objects, trace and preservation/loss account. An express baseline comparison also needs the preservation relation it claims.
Bearer feasibilityA feasibility claim needs a viable bearer and applicable constraints. The bearer may be proposed; a claim that it actually exists requires its independent admission.
Reusable generator boundaryOne-case generated output stays with C.35 and its declared next use; reusable-generator or mechanism-suite claims require the patterns that define or constrain those claims.

C.35:4 - Solution

Ask what decision the result can settle. One candidate checked against every applicable constraint can establish feasibility within the model. It may complete a commission to find any suitable design. A claim that no design qualifies requires grounds that cover every admissible alternative, such as an exhaustive search or a proof of incompatible constraints. A search that stops without a candidate can still tell the next designer which alternatives were examined. When one design uses less energy and another takes less time, both can be nondominated. To call a design nondominated, exclude alternatives that do at least as well on every stated criterion and better on one. State the set over which this was established. A claim that a design is best needs a ranking or trade-off rule for the commission and grounds that the design ranks at least as high as every admissible alternative. A verified bound may settle that comparison. If the leading alternatives remain incomparable under that rule, report the choice that still calls for a preference. A unique optimum also excludes ties for the best value. A unique feasible design excludes every other feasible design regardless of preference.

Judge the result and its production separately. A verified design can serve the commission whoever suggested it. To credit an agent with designing the organization, identify the decisions left open by the input, the method or generator, and the evidence that the agent made them. A fixed input that already specifies the organization can support an assessment of how the agent checked it. Claims about the installed organization depend on evidence about the installation. State any material limit or next check in the existing account when possible.

Use the smallest sufficient path:

  1. Give or reuse the sufficient account. If it lets the receiver make the intended next move, stop.

  2. When organization status affects use, distinguish an obtaining structure, a refuted or unestablished claim about existing organization, and a proposal. Apply A.22’s four discriminators to a positive structure claim. Add C.30 claim identity when the receiving use needs independent reidentification. Distinguish the representation when correspondence or representation operations matter; use C.29 for a remaining mathematical-representation choice, transfer, or reliance question.

  3. When the receiving use relies on how the result arose, select the branch for that claim. A complex result may need separate accounts for claims arising through different operations; do not infer one branch’s basis from another.

    • A transformation adds source and result objects, trace, preservation, and loss.
    • A discovery adds observation or extraction basis, what is observed, what is inferred, what remains unknown, the covered and unexplored region, uncertainty, and validation.
    • A generative proposal adds constraints, proposed organization content, known omissions, and validation needs; source and preservation enter only for a declared baseline.
  4. When reliance depends on how the result arose, establish the operation and the history facts material to that use. Add the Method, independently admitted dated U.Work, and production, discovery-use or Work-to-change relation only when those claims are being made; retain the premises and evidence each requires.

  5. Add bearer feasibility, realized structure, publication, archive or front policy, evaluation, measurement, decision, reusable-generator, or next-claim references only when the intended use relies on that claim. Use the pattern for that claim rather than copying its dossier into C.35.

  6. Stop when the receiver knows the supported next move and any unresolved condition, material limit or return. Materialize the optional full note only when those facts or their relied-on branch details need an independently reidentifiable result.

Use these patterns when their questions arise:

PatternQuestion or use
C.32Candidate-palette use.
C.32.ACEEval programs and results.
C.16Measurement.
C.29An unresolved mathematical-representation choice, transfer, or reliance question.
C.30.AD or C.30.ASVDescriptions and views.
G.5, C.18, and C.19Selected-set, archive, front, and pool claims.
E.24.PUBPublication occurrence, form, and carrier.
E.17A publication that is an episteme projection.
C.32.PAD and C.32.ADRProject architecture decisions and their ADR-like publication, respectively.
E.20, G.1, G.10, or G.11Only when the current question concerns a reusable generator or mechanism suite.

Use concerning CGUS. When a result is useful because it describes, compresses, or demonstrates a constraint-governed unfolding organization, C.35 admits only that result for the declared architecture use. Cite an A.22.CGUS structure only when its positive A.22 and A.22.CGUS basis already obtains; retain an unestablished claim about an existing organization as unresolved, and a proposal as modal claim content.

The structure itself, when it exists, remains with A.22.CGUS, E.18.3, C.32.P2S, E.23, or another pattern defining the structure. If the encountered item is only a route card, narrative sequence, demonstrative slice, generated publication form, or presentation carrier, recover its claim-bearing result and represented object before making any positive structure claim.

When the encountered item is a narrative sequence, A.6.3.NAR governs only the selected-source carry-through, ordering and connective account, loss, reader use, and return.

C.35:5 - Archetypal Grounding

C.35 is the pattern for admitting or rejecting a generated or discovered result before another architecture claim relies on it. The result may come from search, clustering, query, learning, transformation, simulation, or discovery. C.35 does not search, select, decide, or realize architecture. It asks what the result is, which organization it concerns, whether the claim about that organization is established, refuted or unestablished, or instead presents a proposal, what use it supports, and which conditions or limitations affect further reliance.

What a generated proposal establishes. An LLM returns proposal MedicalDeviceProposal-7 and diagram MedicalDeviceDiagram-7. An architect considering it for the candidate set records the following.

Result: MedicalDeviceProposal-7.
Organization: proposed arrangement of device components. The proposal leaves open which parts of that arrangement are already installed.
Next use: C.32 may examine the proposed arrangement as an alternative now. Its feasibility remains open.
Limit and return: before relying on feasibility, establish the applicable constraints and that components capable of meeting them can be provided under the intended conditions. To establish what is installed, use the device's applicable as-built evidence.

This is enough for the immediate candidate-use boundary. Add the generative-proposal branch only if the receiver relies on its constraints, omissions, or validation needs.

Treat MedicalDeviceDiagram-7 as a separate representation when graphical correspondence matters, using C.30.AD or C.30.ASV for its architecture description or view and C.29 when a mathematical-representation choice, transfer, or reliance question remains. A source or preservation account requires a declared baseline.

DSM and MDM clustering. A team groups the rows of a dependency matrix using co-change observations and interface data. Check whether the reported groups satisfy the grouping rule on that matrix. If the rule allows several groupings, the saved output identifies the one returned by this run. Reproduction also suffices when the procedure and its inputs determine the same choice, including any tie-breaking or random choices that mattered. To infer that the corresponding components perform one function, the team must relate the observed dependencies to that function and check what the matrix omits. A different decision is to reorganize the components into those groups. That proposal needs a reason the new allocation would serve the design. Use the discovery or proposal branch for the claim actually made.

Retain the matrix’s mapping to the represented dependencies and the validity conditions of its operations; use C.29 when choosing, transferring, or relying on that representation remains unresolved. A comparison with an earlier modularization may add a declared baseline and preservation account, but clustering alone does not require one.

NAS result. A multi-objective neural architecture search returns a proposed neural organization and its graph. No evaluated alternative matched or improved on the proposal on every criterion and improved on it on at least one. Report that comparison over the evaluated set. To report the same conclusion over the full search space, establish coverage or a bound that excludes a dominating alternative. Keep the search constraints, hardware conditions and validation needed for deployment with the proposal.

The graph represents the proposal; C.29 applies when a separate mathematical-representation choice, transfer, or reliance question remains. The graph does not prove that the proposed relation occurrences obtain. C.35 does not invent a preserved-dataflow claim unless the run explicitly transforms or compares against a declared baseline.

C.32 may consume the proposal as candidate input; C.32.ACE handles evaluation results.

Graph grammar or model transformation. A graph-grammar Method is applied in dated generation Work and returns a claim-bearing result plus a graph representation for a product-line model. This is the transformation branch: C.35 names the Method, Work when performed-work reliance is current, source and result objects, rules, preserved interfaces, lost manufacturing constraints, and transformation trace.

If the resulting organization is only proposed, it remains modal content in an architecture claim and the graph remains a representation of that proposal, with C.29 used for an unresolved mathematical-representation choice, transfer, or reliance question. Source or result A.22 structures are cited only when their four discriminators independently resolve.

If the use additionally asserts an actual formal or world-side change, it cites the A.3.4 occurrence and the separately governed Work-to-change or A.15.PROD claim; otherwise model transformation remains the Method or operation-family label and no U.Transformation is inferred. C.34 may check preservation; C.32 may admit the proposal without actualizing it.

C.35:6 - Bias-Annotation

BiasHow C.35 counters it
Output authority biasIdentify the result, organization, claim status (established, refuted, unestablished, or proposal), and supported next use, with any material condition, limitation or return. Add representation, publication, branch, or other detail only when the use depends on it.
Pareto-point admission biasTreat a Pareto point, benchmark score, archive member, or search trace as a candidate input cue until its branch-specific basis and the candidate-use rule are named.
Reusable-generator collapseKeep one-case output admission in C.35; handle reusable-generator, mechanism-suite, model-family, or production-pipeline claims with E.20, G.1, G.10, G.11, or another pattern that defines or constrains those claims.
Bearer-free synthesis biasRequire bearer or realization boundary before treating a discovered function, relation, or candidate form as architecturally feasible.
Eval substitution biasHandle eval programs and eval results under C.32.ACE; handle measurement under C.16; do not let good eval numbers act as candidate admission or decision authority.
Currentness freezeReopen when result identity or claim content, represented object or correspondence, source publication edition or source-use record, search space, query rule, validation trace, bearer constraints, realized structure, or eval return changes. A carrier-only change reopens C.35 only when availability or form changes the intended use.

C.35:7 - Conformance checklist

CheckPass condition
CC-C35-1A sentence or four-line statement names the result, the organization concerned and whether the claim about it is established, refuted or unestablished, or presents a proposal, the supported next use and any unresolved condition, material limitation or required return. This is a conforming first result.
CC-C35-2An optional materialized note has the result as its EntityOfConcern and repeats the readable minimum. It adds identity, representation, publication, project-use, branch, Method, Work, attribution, bearer, evaluation, or next-claim fields only when the receiving use relies on them. Any present publication-form, publication-occurrence, or presentation-carrier reference identifies the corresponding form, occurrence or carrier; absent publication apparatus creates no field-filling duty.
CC-C35-3When branch evidence is relied on, the branch is selected for each claim; distinct operations in a complex result retain their separate bases. Transformation supplies source and result objects, trace, preservation, and loss; discovery supplies observation or extraction basis, what is observed, what is inferred, what remains unknown, coverage, uncertainty, and validation; generative proposal supplies constraints, proposed organization content, omissions, and validation needs, with source and preservation only for a declared baseline.
CC-C35-4A positive structure reference passes all four A.22 discriminators. An unestablished claim about existing organization remains unresolved; a refuted claim stays refuted; a proposal stays modal. C.32 may use a qualified provisional result without treating an unestablished or proposed organization as obtaining architecture.
CC-C35-5Bearer or realization detail appears only when the intended use relies on feasibility or realization; any such question uses the rule that defines or tests it.
CC-C35-6Use the pattern for each archive, front, pool, publication, eval, measurement, mathematical-lens, decision, evidence, assurance, gate, release, Method, or Work claim; an absent claim creates no C.35 field-filling duty.
CC-C35-7The first result states what it warrants for the receiving decision. Feasibility has a qualified witness or other sufficient grounds. Claims of impossibility, optimality or uniqueness identify the alternatives covered or excluded and the grounds for that conclusion.

C.35:8 - Common Anti-Patterns and How to Avoid Them

Anti-patternWhy it failsRepair move
LLM output as architecturePlausible prose and a diagram may denote a modal architecture claim and its representation; neither supplies obtaining relation occurrences, an A.22 structure, bearer feasibility, decision, or realization.Recover the proposal, adding claim identity when its use needs independent reidentification; distinguish the diagram and its correspondence when used. Use C.29 for an unresolved mathematical-representation choice, transfer, or reliance question. State the admission condition and return; C.32 can consume the modal proposal without actualizing it. Use C.32.PAD and C.32.ADR for decision and ADR claims.
Pareto point as admissionA Pareto result records trade-off position under chosen criteria; its graph, table, or file is a neighboring representation or publication item, not architecture adequacy.Name the result and the current next-use condition. Add search space, criteria, constraints, bearer boundary, and eval return only when the candidate use relies on them; then handle that use under C.32.
One output as reusable-generator governanceA single generated output does not by itself establish claims about the reusable generator’s method, mechanism suite, dataset, prompt policy, or refresh process.Keep the one-case output in C.35 and open E.20, G.1, G.10, G.11, or another pattern that defines or constrains the reusable-generator claim.
Cluster as module architectureA cluster claim can expose co-change or dependency pressure while leaving functional-bearer semantics, interface substitutability, and obtaining relation occurrences unknown; its matrix or file does not settle that gap.Recover the cluster result, extraction basis, observed and inferred content, unknowns, coverage, uncertainty, validation, and the representation correspondence required by its use. Keep unestablished claims about the existing organization unresolved and any proposed reorganization modal; handle modularity and reuse under C.31 and candidate use under C.32.
Transformation output as feasibility proofA graph grammar or model-transformation Method can return a useful claim and representation while proving neither an actual U.Transformation nor an obtaining A.22 result structure.Recover the result, source and result objects, transformation trace, preservation and loss, and the bearer boundary needed by the use. Include the Method, Work and attribution when those claims are current. Retain a relied-on representation correspondence and use C.29 for a remaining choice, transfer, or reliance question. Keep a proposed result organization in its architecture claim; cite A.22 only after its four discriminators resolve, and cite A.3.4 plus the Work-to-change or A.15.PROD claim for any actual change.
Bypassing eval and measurement governanceA search score, benchmark, ablation, or validation trace can look like proof of architecture quality.Use C.16 for measurement readings, C.25 for Q-bundle use, C.32.ACE for eval programs and eval results, and C.32.PAD for project architecture decisions.

C.35:9 - Consequences

Positive consequences:

  • Generated or discovered results can enter architecture work without becoming authority. Ordinary use can stop with a sufficient account; additional detail appears only for a receiving dependency.
  • C.35 keeps the result-use return visible. If the result cannot support the next architecture claim, repair returns to that result, the organization it concerns and any unresolved or refuted claim about it, the missing condition, and the required rule or test; the return never turns the proposal into an A.22 member.
  • C.32 continues to define candidate-palette admission. C.35 supplies the result-use boundary that C.32 may need.
  • Search, query, transformation, and AI-assisted results become auditable without conflating claim-bearing content, represented objects, representations, publication items, and carriers.
  • Reusable generator governance stays outside C.35 until opened, which prevents one-case result review from becoming a hidden method or mechanism-suite pattern.

Costs and trade-offs:

  • Assessing a result takes attention. Reuse a sufficient account; obtain further detail when it can change the warranted next use.
  • Some outputs will be useful but not yet admissible. Supply the missing branch-specific basis, bearer boundary, validation, or next claim plus its required rule; keep useful outputs and leave irrelevant preservation rows empty.
  • The pattern is narrow. It does not choose among alternatives, manage archives, define eval programs, or authorize work.

C.35:10 - Rationale

Architecture synthesis receives results from search, model transformation, LLM proposal, code-agent mapping, DSM modularization, NAS, simulation, benchmark and source discovery. A short account can identify the result, the organization it concerns and the supported next use, with any condition or limitation that affects reliance. The larger record is justified by a receiving use that depends on its additional distinctions.

The result can carry a proposal, describe an obtaining structure or serve as a representation. Its selected use determines which distinctions and supporting results matter: representation correspondence, feasibility, evaluation, publication or decision authority, for example. Obtain each required result under its own subject rule; retain a mathematical representation’s mapping and validity conditions, using C.29 when a separate choice, transfer, or reliance question remains.

The source families explain why the branches differ. MBSE query practice, DSM and MDM work, and code-agent mapping expose discovery questions about extraction basis, observation, inference, unexplored regions, uncertainty, and validation. Graph grammars and model transformations expose the distinct need for source and result objects, trace, preservation, and loss. Multi-objective NAS and LLM-assisted design expose proposal questions about constraints, proposed organization, omissions, and validation without inventing a source structure. The GonzoML intake supplies neural operation vocabulary; resolve a source label when its meaning is unclear and changes the branch-specific claim or use.

C.35:11 - SoTA-Echoing

Source or practice lineUse of the contributionContribution to the assessmentBoundary of use
MBSE query and view generationIdentify the source, selection rule and coverage of the query.A complete, validated query can settle a claim about the model.Claims about the installed architecture depend on evidence connecting that model to the installation. Reopen when the model edition, represented object, query or coverage changes.
Google OR-Tools, CP-SAT solver statusesRead a feasible witness, a proof of infeasibility, an optimum or an inconclusive stop according to the stated result.Helps decide whether a search has already fulfilled the commission or whether coverage or a bound is still needed.Solver conclusions apply to the encoded constraints, objective and stopping conditions.
Graph grammars and model transformationsAdapt rule-governed transformation as the transformation branch.Identifies source and result objects, transformation trace, preserved structure, lost structure, and C.34 preservation exit.Grammar or transformation output does not prove adequacy, feasibility, or realization. Reopen when transformation rules, source object, result object, trace, or constraints change.
DSM, MDM, and modularization practice including Jiang and Luo, arXiv:2604.28018Adapt modularization and LLM-assisted DSM work as discovery.Identifies extraction basis, observed dependencies, inferred clusters, unknown functional-bearer semantics, coverage, uncertainty, validation, and C.31 plus C.32 exits.Cluster, partition, or MDM slice is not candidate architecture adequacy. Reopen when relation matrix, covered region, modularity objective, functional prior, validation, or solution pool changes.
Multi-objective NAS and Sukthanker et al., arXiv:2402.18213Adapt multi-objective search as a generative-proposal source.Identifies constraints, proposed neural organization, known omissions, validation needs, search criteria, bearer boundary, eval return, and C.32 admission condition.A Pareto point or neural graph is not holonic architecture adequacy. Preservation is claimed only against a declared baseline. Reopen when search space, constraints, criteria, hardware target, proposal content, or eval trace changes.
DSE, QD, OEE, NQD, and evolutionary architecture practice inherited through C.32Adapt retained alternatives and stepping-stone pressure as candidate-input practice.Connects candidate input and its use conditions to the archive, front and pool policy for that use and C.32 candidate-palette rules.These practices do not make C.35 a second candidate-set admission rule. C.18, C.19, and G.5 define archive, front, and pool policy; C.32 defines candidate-palette admission.
AI-assisted architecture design and AI-assisted ADDAdapt generated decompositions, relation graphs, and decision proposals through the generative-proposal branch.Identifies constraints, proposed organization or claim content, known omissions, source-label recovery, validation needs, and candidate-admission boundary.An LLM proposal, ADD suggestion, benchmark trace, or agent consensus is not decision authority, evidence sufficiency, realization, or architecture adequacy by itself.
Sapunov, Theory of Code Space, and code-agent architecture-map practiceAdapt partial-observability mapping through the discovery branch.Identifies observation or extraction basis, separate observed, inferred, and unknown content, confidence, covered and unexplored regions, active-passive comparison, and validation.A code-agent map, JSON probe, benchmark score, dependency F1, invariant F1, or active-passive gap is not architecture adequacy, internal-state proof, safe-change authority, evidence sufficiency, gate passage, or release authority.
GonzoML neural-network architecture intakeAdapt neural architecture operation language as source-label recovery for discovery or generative proposals.Identifies recovery for dataflow change, routing, gating, memory placement, cache placement, block substitution, pruning, distillation, NAS, ablation, and compute, memory, and latency trade-offs without choosing a branch by label.Use clear source-local meanings directly. Obtain the branch-specific basis, bearer or characteristic needed by the architecture claim; ablation or benchmark gains alone do not establish it.

C.35:12 - Relations

  • Builds on: C.30, C.30.AD, C.30.ASV, A.22, C.32.P2S, and C.32.
  • Uses: C.34 when the transformation branch or an explicit baseline comparison must preserve selected source structure; C.33 when capture and loss in the output are the current issue; C.29 when choosing, transferring, or relying on a mathematical representation remains unresolved.

Coordinates with:

PatternCoordination
A.3.4Each actual bounded change.
A.13, A.15.1, A.2.1, and F.6A.13 supplies actual performer assignment bases, then independent A.15.1 governs performed generation or discovery Work; A.2.1 concerns the assignments, and F.6 applies only when precise assignment-bound attribution is current.
A.15.PROD and A.6.RCDProduction or Work-to-change claims.
C.36A cultural-evolution case supplies the generated or discovered result; C.36 retains governance of that case.

C.35 also coordinates with C.30.STRAT, C.30.TFS-REL, A.6.M, C.31, C.31.ASAP, C.32.ACS, C.32.ACE, C.16, C.25, G.5, C.18, C.19, E.18, C.32.PAD, and C.32.ADR.

  • Boundary: C.35 governs admission of a generated or discovered result for architecture use before or around C.32 candidate admission. It does not build the candidate palette, select from alternatives, govern reusable generators, define eval programs, measure values, decide projects, supply evidence or assurance, authorize work, publish the result, or prove realization.

C.35:End

Referenced in the corpus

72 literal mentions in other sections. Read their context to establish the relation.