#Open-Ended Search Archive and Front Stewardship

C.18 · Cluster C.IV - Composite & Macro-Scale · Stable · C-pattern · Normative unless explicitly marked informative · Part C - Kernel Extension Specifications

#About this pattern

This is a generated FPF pattern page projected from the published FPF source. It is canonical FPF content for this ID; it is not a FPF Reference product feature page.

#How to use this pattern

Read the ID, status, type, and normativity first. Use the content for exact wording, the relations for adjacent concepts, and citations to keep active work grounded without pasting the whole specification.

Tech-name: OpenEndedSearchArchiveAndFrontStewardship Plain-name: open-ended search archive and front stewardship Type: C-pattern Status: Stable Normativity: Normative unless explicitly marked informative Placement: Part C Builds on: C.16, A.19.CPM, A.19.SelectorMechanism, and E.18. Coordinates with: C.19, G.5, G.9, G.11, E.23, E.18.1, the C.30 family, C.32.P2S, C.32, C.35, C.36, F.17, F.18, F.9, and the A.15 family. Purpose: make archive, front, Q-front, descriptor, telemetry, retained exploration value, stepping-stone value, lineage, edition, architecture-candidate generation, and cultural-variant generation usable without turning them into publication, decision, work permission, or cultural-evolution authority.

Use this pattern when a project needs to generate, retain, compare, or report many candidate variants while preserving descriptor editions, distance definitions, archive policies, front semantics, telemetry, lineage, and retained exploration value.

#Keywords

• open-ended search
• generation
• exploration archive
• non-dominated front
• descriptors
• lineage
• telemetry
• retained exploration value
• currentness
• refresh.

#Relations

#Content

#Use This When

Use this pattern when a project needs to generate, retain, compare, or report many candidate variants while preserving descriptor editions, distance definitions, archive policies, front semantics, telemetry, lineage, and retained exploration value.

Typical cases include quality-diversity archives, open-ended engineering variant sets, Pareto or Q-front treatment, phenotype-like descriptor maps, architecture-candidate generation, style or tradition variant generation, scientific or engineering school variants, and candidate pools whose value is not captured by one immediate selected set.

#What Goes Wrong If Missed

The project treats an archive as a shortlist, a front as a decision, illumination telemetry as dominance, a retained stepping stone as current best, or a cultural-style variant as a root cultural kind. Generation looks productive, but the next relation is unclear: retain, compare, publish selected set, choose locally, plan work, measure effects, refresh, or write a cultural-evolution case.

#What This Buys

The practitioner gets separate records for archive, front, and generation. Each record pins descriptors, characteristic spaces, edition refs, retention policy, telemetry, lineage, and next governing relation. Downstream selection, architecture, cultural evolution, work planning, measurement, and refresh then start from named records rather than from a broad archive label.

#Problem Frame

Open-ended search and quality-diversity work deliberately keep more than one candidate alive. That is useful for engineering, science, design, music, dance, AI-agent frameworks, medical method families, and other evolving practices. The same archive or front label can hide strong candidates, weak but promising stepping stones, coverage-expanding variants, architecture candidates, cultural variants, and telemetry-only signals.

The primary EntityOfConcern in C.18 is the archive or front relation being stewarded: which variants are generated or retained, under which descriptor and characteristic space, with which edition and lineage pins, and with which next relation available. C.18 is not a local-choice pattern, not a selected-set publication pattern, not a cultural-evolution subject-governing pattern, and not an architecture pattern.

#Problem

Without C.18, a team often compresses several different objects into one word such as archive, front, Q-front, portfolio, style pool, or candidate set. That loses four distinctions:

• a front answers current non-domination under a declared comparator or dominance set;
• an archive answers retained exploration value, coverage, stepping-stone value, or future reachability under a declared retention policy;
• telemetry reports search health, coverage, novelty, diversity, or lineage but does not by itself dominate alternatives;
• downstream selected-set publication, local choice, architecture work, cultural-evolution case work, planning, performed work, and refresh each have their own governing pattern.

#Forces

#Solution

Keep archive, front, telemetry, generation, and downstream relations as separate records.

#Archive Record

ExplorationArchiveRecord@Context:
  archiveRef:
  variantSetRef:
  descriptorMapRef:
  characteristicSpaceRef:
  distanceDefinitionRef?:
  retentionPolicyRef:
  retainedExplorationValue:
  steppingStoneUse?:
  lineageOrEditionPins:
  telemetryRefs?:
  currentnessAsOf:
  currentStatus:
  stopOrRefreshReason:
  nextGoverningRelation:

Use this record when the current question is retained exploration value, coverage, novelty, diversity, stepping-stone value, future reachability, curriculum expansion, lineage, or archive policy. Do not use the archive record as a selected-set publication or work permission.

#Front Record

FrontRecord@Context:
  frontRef:
  candidateSetRef:
  comparatorOrDominanceSetRef:
  admissibilityRef:
  descriptorMapRef?:
  characteristicSpaceRef?:
  relationTokenSetRef:
  excludedTelemetryRefs?:
  selectedSetPublicationRef?:
  currentnessAsOf:
  currentStatus:
  stopOrRefreshReason:
  nextGoverningRelation:

Use this record when the current question is non-domination, Pareto relation, Q-front membership, comparator currentness, admissibility, or partial-order preservation. The front may feed [G.5](/generated/patterns/G.5), but it is not itself a selected-set publication unless [G.5](/generated/patterns/G.5) makes that publication.

#Filled Archive And Front Micro-Records

ExplorationArchiveRecord@Context:
  archiveRef: dance-lab-variant-archive-2026
  variantSetRef: choreography variants generated during a festival lab
  descriptorMapRef: timing, body vocabulary, risk, teachability, audience recognizability
  characteristicSpaceRef: festival style-engineering characteristic space
  distanceDefinitionRef: difference in timing and body-vocabulary descriptors
  retentionPolicyRef: keep rare but teachable variants and variants that open later combination work
  retainedExplorationValue: stepping stones for teaching and later style intervention
  steppingStoneUse: candidate material for C.36 cultural-evolution case work
  lineageOrEditionPins: lab session, teacher edit, platform-publication edition
  telemetryRefs: replay counts, class adoption counts, jury notes
  currentnessAsOf: lab records and platform-publication edition reviewed through 2026-07-31
  currentStatus: active for retained exploration and teaching use
  stopOrRefreshReason: reopen if the teaching-use judgement, retention policy, lineage, or platform edition changes
  nextGoverningRelation: C.36 case card or G.11 refresh, depending on the current question

FrontRecord@Context:
  frontRef: cooling-module-maintainability-energy-front
  candidateSetRef: retained cooling-module architecture candidates
  comparatorOrDominanceSetRef: energy-use and maintainability comparator
  admissibilityRef: safety and manufacturing constraints already admitted by project policy
  descriptorMapRef: thermal performance, service access, part count, manufacturing tolerance
  characteristicSpaceRef: product-family architecture characteristic space
  relationTokenSetRef: non-dominated candidates under current comparator
  excludedTelemetryRefs: tests outside the current temperature envelope
  selectedSetPublicationRef: empty until G.5 publishes the selected set
  currentnessAsOf: comparator, safety constraints, and test evidence reviewed through 2026-07-31
  currentStatus: active non-dominated front for the declared comparator
  stopOrRefreshReason: reopen if eligibility, comparator, dominance grounds, or evidence edition changes
  nextGoverningRelation: C.30 architecture candidate treatment, G.5 selected-set publication, or G.11 refresh

#Generation And Downstream-Use Record

When loop-engineering practice generates many agent prompts, harness variants, workflow variants, or framework seeds, C.18 records generation, archive, front, descriptors, telemetry, retained exploration value, lineage, and the next governing relation. It does not say that the loop improved. Use E.23 only when one retained object version is changed and re-evaluated; use G.9 for parity between variants and G.5 when a selected set must be published.

OpenEndedVariantGenerationRecord@Project:
  problemCardRef?:
  generationMethodOrFamilyRef:
  sourceRefs?:
  evaluatorOrComparatorRef?:
  emitterPolicyRef?:
  insertionPolicyRef?:
  dedupThreshold?:
  deduplicationBasisRef?:
  deduplicationUnit?:
  variantSetRef:
  descriptorMapRef:
  characteristicOrDescriptorSetRef:
  archiveOrFrontRef?:
  architectureCandidateRefs?:
  culturalVariantRefs?:
  telemetryRefs?:
  projectLocality?:
    generationWorkOccurrenceRef:
    compositeProjectWorkOccurrenceRef:
    governingRelationPatternRef:
    exactGenerationToProjectRelationRef:
  workPlanOrMeasurementRef?:
  refreshRef?:
  currentnessAsOf:
  currentStatus:
  stopOrRefreshReason:
  nextGoverningRelation:

Across the archive, front, and generation records, currentnessAsOf names the replay date or window together with the relevant pinned editions; currentStatus states whether the recorded archive/front/generation relation is active, held, closed, or stale for its declared use; and stopOrRefreshReason names the condition that ended it or the exact trigger that would reopen its currentness. These fields record the boundary but do not perform refresh. When source, descriptor, comparator, policy, evidence, or edition currentness becomes the live question, nextGoverningRelation points to [G.11](/generated/patterns/G.11) and refreshRef? may cite the separately governed refresh record.

Here @Project is a compatibility and retrieval cue, not a project kind or relation assertion. Fill projectLocality? only after both Work occurrences have been admitted independently and the cited relation actually obtains under its direct governor. generationWorkOccurrenceRef names the exact dated generation U.Work; compositeProjectWorkOccurrenceRef names the selected composite project U.Work; and exactGenerationToProjectRelationRef cites, rather than creates, the exact work-part, containing-work, decision-use, source-use, or other governed relation. Use work parthood only when the complete [A.15.1](/generated/patterns/A.15.1) basis holds. Otherwise omit projectLocality?: the @Project suffix remains retrieval-only and establishes no project Work, parthood, authority, context, or viewpoint.

For example, a completed harness-generation run may state:

projectLocality:
  generationWorkOccurrenceRef: HarnessVariantGenerationRun-2026-07-31 : U.Work
  compositeProjectWorkOccurrenceRef: AgentHarnessProjectWork-2026 : U.Work
  governingRelationPatternRef: A.15.1
  exactGenerationToProjectRelationRef: OperationalPartOf_work(HarnessVariantGenerationRun-2026-07-31, AgentHarnessProjectWork-2026)

Every optional field whose name ends in Ref? points to a separately governed object, claim, policy profile, or measurement basis. dedupThreshold? is not a reference: it carries one declared scalar threshold value. deduplicationUnit? carries its unit literal. Fill emitterPolicyRef? and insertionPolicyRef? only when the cited C.19 profile or insertion policy actually governs the current pool treatment. When a threshold is inherited, the cited profile supplies dedupThreshold, deduplicationBasisRef, and deduplicationUnit; when it is not inherited, carry the scalar in dedupThreshold? and its basis and unit in deduplicationBasisRef? and deduplicationUnit?. These references and scalars do not give C.19 a generation operator or move archive/front ownership out of C.18. In particular, problemCardRef? may cite a C.22.2 problem-side episteme but creates neither an actual Problem nor a ProblematicForRelation under [C.22.PFR](/generated/patterns/C.22.PFR). A generated variant, archive entry, front membership, telemetry value, or retained-exploration claim is neither an improvement-result nor a work-result identity and creates no relation from generation Work to a result. nextGoverningRelation names the next owner; it does not itself make a choice, publish a selected set, authorize work, perform refresh, or make any relation obtain.

Use this record when generation is current. architectureCandidateRefs become architecture moves only through [C.30](/generated/patterns/C.30), [C.30.ASV](/generated/patterns/C.30.ASV), or [C.30.AD](/generated/patterns/C.30.AD). culturalVariantRefs become cultural-evolution cases only through [C.36](/generated/patterns/C.36). Local choice uses [C.11](/generated/patterns/C.11); work planning and performed work use the A.15 family; effect measurement uses its direct measurement and evaluation owners; refresh uses [G.11](/generated/patterns/G.11). P2W carry-through uses [E.18.1](/generated/patterns/E.18.1) when an accepted problem-side distinction must be preserved into the next relation.

#Front And Archive Are Different Returns

• Start from one declared candidate or eligibility set.
• Return the non-dominated front over the declared comparator, dominance set, or relation-token set.
• Return the exploration archive separately when retained exploration value, coverage, novelty, diversity, stepping-stone value, or future reachability is current.
• Keep tie-breakers and telemetry explicit so diversity, illumination, or popularity signals do not rewrite front semantics.
• Before promoting telemetry or a popularity-like signal into the comparator, dominance set, or selected-set criteria, state which intended archive/front use or value becomes worse when that signal improves and cite the policy or decision authority that admits the trade-off. If either answer is missing, keep the signal as telemetry or an explicitly bounded tie-breaker rather than silently promoting it.
• Use RetentionIntent=steppingStone when retention exists for frontier expansion or later curriculum value rather than current dominance.
• If one source line keeps both returns, say that the front answers current non-domination while the archive answers retained exploration value.

#Cultural And Architecture Variant Boundaries

For architecture-candidate generation, C.18 records generation, archive, front, descriptor, telemetry, and retained exploration value. C.30 governs the architecture claim: ArchitectureOf@Context, selected structure or structure kind, affected characteristic, and next architecture move.

For cultural variants, C.18 records the generated or retained variant set and its descriptors, lineage, telemetry, and archive or front relation. C.36 governs the cultural-evolution case when collective-holon or discipline-facing method, work, role, canon, memory, recognition, selection, mediation, style, tradition, or intervention relations are current. F.17, F.18, and F.9 govern durable term and bridge work for labels such as style, tradition, genre, scene, school, and technique.

#Conformance Checklist

• CC-C18-1 Descriptor, characteristic, distance, and family-coordinate refs are named before generation, archive update, or front publication.
• CC-C18-2 Archive and front returns are separate unless a governing pattern explicitly publishes a selected set from one of them.
• CC-C18-3 Telemetry remains telemetry unless a declared policy promotes it into the comparator, dominance set, or selected-set criteria and the governing record names both the intended archive/front use or value made worse by that promotion and the authority that admits the trade-off.
• CC-C18-4 Retained exploration value, stepping-stone use, lineage, and edition pins are recorded for archive use.
• CC-C18-5 Architecture candidates use C.30 family patterns before becoming architecture moves.
• CC-C18-6 Cultural variants use C.36 or term-bridge patterns before becoming cultural-evolution claims.
• CC-C18-7 Refresh uses G.11 with the smallest affected archive, front, descriptor, edition, or lineage locus.
• CC-C18-8 Agent-loop, harness-loop, workflow-store, or DPF-seed variants retained in an archive name their descriptor, lineage, telemetry, and next governing relation; archive membership does not claim quality improvement without E.23 re-evaluation.
• CC-C18-9 A filled projectLocality? names independently admitted dated generation and composite project U.Work occurrences, the direct governing pattern, and one exact obtaining relation; @Project alone remains retrieval-only.
• CC-C18-10 Problem-card, result, selected-set, choice, work, and refresh references remain references to separately governed objects or next owners and create none of those identities or relations.
• CC-C18-11 The SoTA basis names its reviewed-through boundary and exact mutable editions; a material source revision, newer field survey, or contrary archive, descriptor-generalization, or OEE-evaluation evidence records a reopen trigger and hands refresh to G.11.

#Archetypal Grounding

System-facing case. A robotics team generates gait variants. The front records non-dominated speed and energy relations under declared measures. The archive retains diverse coordination patterns because some are stepping stones for new terrain. Telemetry reports coverage. A selected set may later be published through G.5; performed test runs use A.15.

Architecture case. A cooling-module project keeps an archive of modular layout variants and a front over maintainability and energy use. C.18 records descriptors, archive policy, front relation, and telemetry. C.30 decides whether any retained variant becomes an architecture move by naming the selected structure and affected architecture characteristic.

Cultural case. A dance-lab project generates movement variants around several source labels. C.18 records generated variants, descriptors, archive membership, front relation, and lineage. C.36 decides whether the lab is deliberately changing a cultural-evolution case; F.17, F.18, and F.9 handle the label bridges.

#Bias-Annotation

Lexical and semiotic bias are controlled by keeping archive, front, telemetry, selected-set publication, local choice, cultural-evolution case, architecture move, work permission, and evidence relations distinct. Mathematical descriptions of descriptor maps, fronts, distances, coverage, or novelty use the mathematical-lens pattern when lens adequacy matters.

#Consequences

Positive consequences:

• archives keep exploration value without pretending to decide;
• fronts preserve partial-order and comparator semantics;
• architecture and cultural-variant generation become usable without creating parallel root kinds;
• refresh and source-currentness have clear loci.

Costs:

• teams must keep at least archive and front records separate;
• one generated variant may need several downstream records before it becomes selected, chosen, planned, worked, measured, or refreshed;
• descriptor editions and distance definitions require maintenance.

#Rationale

Current quality-diversity, illumination search, open-ended engineering, and evolutionary-engineering practice shows that retained diversity, stepping stones, archive lineage, and descriptor currentness often matter before a single choice is justified. FPF keeps that practical gain while preventing archive and front language from replacing comparison, selected-set publication, architecture, cultural evolution, work, evidence, decision, or refresh patterns.

#SoTA-Echoing

Source-currentness boundary. This source-use basis was reviewed through 2026-08-01. Mutable preprints are pinned below to the edition actually used; the Qin et al. survey is pinned to its DOI-fixed journal article. Reopen the affected source-use row through G.11 when a cited revision changes the archive, front, generation, or evaluation claim used here; when a newer field survey materially changes the known QD/OEE boundary; or when contrary evidence changes what can be claimed about bounded archives, descriptor generalization, or open-ended evaluation. C.18 records that trigger and the affected row but does not itself perform refresh.

#Relations

Builds on: C.16, A.19.CPM, A.19.SelectorMechanism, and E.18.

Coordinates with: C.19 for current-pool treatment, G.5 for selected-set publication, G.9 for parity and benchmark comparison, G.11 for refresh, E.23 when an archived object version enters a declared quality-improvement loop, E.18.1 for P2W carry-through, C.30 family, C.32.P2S, C.32, and C.35 for architecture candidates, problem-to-structure carry-through, candidate palette admission, and generated or discovered carrier adequacy before archive or front use, C.36 for cultural-evolution cases, F.17, F.18, and F.9 for term and bridge work, and the A.15 family for planning or performed work.

#C.18:End

Last Updated: 2026-08-04 — upstream FPF commit 7ba40a95 (github.com/ailev/FPF)