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 22:20:20 UTC

E.23 - Quality Improvement Loop Method

Type: Method-description pattern Status: Stable Normativity: Normative unless marked informative

E.23:1 - Problem frame

Use E.23 when an object version will be improved through repeated passes under the evaluation declared for that object. The object can be a pattern, DRR, FPF corpus object, engineering quality object, naming candidate, OEE and NQD candidate, archive or front member, selected set, parity report, refresh report, or declared transformation result, if the evaluation supplies values and stop meanings for that object kind.

First useful move: name the object version under improvement, the evaluation to use again on the changed version, the improvement aim, protected trade-offs, cost and risk account, and local stop condition.

For improvement of an operating work process whose useful next condition is clearer than the way to reach it, §4.4.1 develops the Improvement Kata branch: study current work, establish a near target condition, and use observations and bounded experiments to discover an effective change. Preparing that basis can finish with a useful next observation. An actual improvement pass needs the object, evaluation, protected qualities and continuation grounds above. Use the coached-practice application in HCD.9:4.10 when developing the learner’s or coach’s reasoning is also intended.

Not this pattern when one direct quality evaluation is enough. Run that evaluation; use E.22 first only when its question needs clarification. Use A.19.ECS first if the needed evaluation characteristic space does not exist.

When the intended improvement concerns how further improvements are obtained, name that way and the means needed to apply it as the object. Use C.40:4.7 and C.40:4.11 to construct alternatives and compare their subsequent development work; §5.1 shows the connection for people, organizations and AI. Changing an instruction, using the changed way in the next improvement, and obtaining better further development are different results. Keep an adequate improving method when changing it offers no worthwhile gain.

When the evaluation could report improvement without the claimed change, §4.1.1 constructs that absence through the same evaluation. An already adequate comparison needs no additional control merely because another improvement pass occurs.

Use E.23.CAE first when the apparent object to improve remains ambiguous because a holder’s capability may be outside its claimed envelope, unavailable through the current configuration, not selected as applicable, inaccessible or inactive, context-dependently unexpressed, unadapted, unenacted, or actually changed. That separate probe returns observations, a qualified disposition, surviving rivals, and candidate routes; it does not choose the object under improvement.

Use E.23.CDI instead when a separate applicable steering or choice result has made capability development current for one admitted holder System and named Work family, and the change must be checked in representative Work. That is a separate capability-development Method; E.23 points to its description without copying its actions. Use a population assessment when the question is a distribution across member capabilities, C.36 when generation, transmission, recognition, selection, retention, or loss across a cultural population is current, and C.32.MWA first only when the target-practice architecture itself must be recovered or compared.

What goes wrong if missed: teams close discharge rows instead of improving quality, retry blindly, optimize visible values while damaging protected qualities, stop merely because every value is 5, or treat a review recommendation as a decision, work, evidence, a selected-set result declaration, actual publication, parity, or refresh.

What this buys in practice: each pass has a declared object version, an intended evaluation-result change, a rerunnable evaluation, protected trade-offs, and a stop or switch condition. The team can then spend its effort on improving the object’s quality instead of accumulating review records, and stop when no non-dominated change is worth its cost.

Primary EntityOfConcern in plain terms: the repeated quality-improvement method for one object version under one declared evaluation.

E.23:2 - Problem

FPF often improves artifacts by repeated review, repair, and re-evaluation. The loop is useful only when the changed object is evaluated again under comparable conditions, or a change of evaluation is justified for its receiving use through §4.1a. Without that discipline, repeated passes become checklist closure, agentic retry, source citation, or process state.

The loop also avoids the maturity-ladder trap. Meeting the floor or receiving all 5s supplies quality evidence within the current scope. Stopping additionally requires the practical judgement in §4.3; neither result proves that the object cannot improve.

The loop also fails when an ordinal value becomes a work target. 5 is an assigned result after measurement, not an instruction to add apparatus until a 5 can be defended.

Below-floor values return a repair proposal or intended-work claim; they do not establish that Work occurred.

Above-floor improvement becomes a selected proposal when chosen under the applicable selection rule, but the target is a substantive content improvement: stronger positive action guidance, worked slice, case and countercase coverage, source-currentness carry-through, mature-content discharge, relation cleanup, deletion of displaced apparatus, split of overloaded content, or another named content gain. Stay at 4 or no proposal is admissible only after a by-value search finds no non-dominated content improvement worth its cost under protected qualities. Selection does not perform the Work or bring about the Transformation. Each occurrence needs its own grounds.

A below-floor value, finding, improvement aim, or repeated-evaluation need is not by itself an actual Problem. If one improvement use relies on an actual Problem, cite one current C.22.PFR ProblematicForRelation occurrence with its actual-condition and criterion-applicability participants and its maximal continuous adverse-episode identity. Evaluation Work, result epistemes, evidence, and loop records may support a claim about that occurrence; they neither create nor split it.

E.23:3 - Forces

ForceTension
Improvement ambition vs costExceptional improvement can be valuable while ordinary floor work stays affordable.
General adaptive methods vs specialized cyclesBroad loops scale, while specialized cycles can be cheaper when the characteristic space fits.
Feedback vs self-confirming retryFeedback helps only when re-evaluation checks changed quality.
Operation hardening vs bureaucracyVerification, memory, decomposition, and supervision are admitted only when their expected improvement effect justifies cost.
Visible improvement vs protected trade-offsOne coordinate can rise while use, source preservation, locality, or ecology worsens.
Proposal portfolio vs selector overreadProposals can guide improvement without becoming selected results or work plans.

E.23:4 - Solution

E.23 guides repeated improvement of one object version under a settled evaluation question. Reuse an adequate ordinary request; use the E.22 QualityEvaluationQuestionFrame and QualityEvaluationUseDeclaration when a receiving use needs those declarations. The evaluation pattern defines the evaluation; restoration and subject patterns supply the relevant rules or guidance. A pattern body or locator does not by itself establish a semantic U.Method or qualify as its description. Claim Method or MethodDescription identity only after A.3.1 and A.3.2 admit it.

When an evaluation or improvement pass is claimed as actual A.15.1 U.Work, identify every actual performer through A.13 and independently identify the occurrence, time, Method, and containing System through A.15.1. Add A.2.1 assignment-occurrence identity and F.6 only when the loop record or receiving use expressly represents precise assignment-bound attribution through the same obtaining A.13 assignment; missing or failed attribution alone leaves independent Work admission intact.

If the same finding invalidates a premise consumed by A.13 or A.15.1, apply A.15.1’s premise-recheck rule to that premise and its dependent conclusions, preserving independently grounded history and relations. A proposed pass, intended performer, planned condition, or A.15.2 WorkPlan remains modal until its predicates obtain.

Keep a returned value, durable result episteme, changed object and actual Transformation separate. Connect a returned value through its A.6.1 binding or the evaluation’s direct result relation. Connect Work to a result or change only through a declared direct relation or supported local claim that obtains.

Return missing-governor when the required predicate or governing condition is unavailable. If the governor exists, apply its criterion to the case: retain a supported positive, negative or inapplicable result, or name the missing facts and how they can be obtained.

The repeated organization changes the object, re-evaluates the changed version through the same declared method and quality model, checks trade-offs and cost, and identifies the admissible continuations: stop, continue, switch methods, open a new frame, wait for sufficient information, branch, or return to a subject pattern.

When the receiving use needs its structure, admit that organization under A.22.CGUS through §4.2a; use E.18 only when an independently selected transformation-flow structure is the EntityOfConcern. Neither the method, record, visible cycle, nor selected continuation is an enduring Work occurrence or context container.

E.23:4.1 - Ordinary loop method

For one quality-improvement loop:

  1. Name the object version and the evaluation that will judge it. Reuse the current QualityEvaluationQuestionFrame and QualityEvaluationUseDeclaration when they still fit the purpose and receiving use. Keep the evaluator, evaluation pattern or Method, characteristic space, evidence basis, result form, ClaimScope, and qualification window separate.

  2. State the content change sought, protected trade-offs, cost and risk account, and local stop condition. Do not use 5, all-5, or 5-defensible as the target; say what should become better in practice.

  3. Resolve any remaining ambiguity in purpose, scope or result-consuming work or decision through E.22. Keep an adequate question; constitute a separate E.22 frame only when its declaration is needed.

  4. Run the declared evaluation. When E.21 is the selected evaluation, retain its complete result: every coordinate, ShortRationale, PrecisionRestorationProfile, evidence basis, coordinate payload, and status. A loop note, blocker summary, or “no blockers” statement is not a substitute. If dated evaluation Work is asserted, identify it and its result binding or direct result relation; keep any durable result episteme separate.

  5. Record each returned finding or proposal separately. A grouped memory summary does not close skipped items, and a proposal remains a proposed next action rather than performed Work.

  6. Select the next change. Selection does not perform it. When the account asserts dated improvement Work, identify that Work and connect it to a returned value or changed object only through an obtaining A.6.1 binding or declared Work-to-result or Work-to-change relation. If that relation is unavailable, keep proposal, Work, changed object, and Transformation separate and return the missing relation.

    Repair below-floor findings first. Above the floor, prefer a substantive gain—such as clearer action, a missing case or countercase, current source support, restored predecessor content, cleaner relations, or a split of overloaded material. Do not add guards, catalogues, or quality proof merely to defend a higher score. Close with no change only after the evidence shows that no feasible non-dominated improvement remains worth its cost under the protected trade-offs.

    For a precision-restoration defect, apply F.19 and open a restoration or subject pattern for an unresolved FPF-specific meaning. Claim a Method or MethodDescription only when A.3.1 and A.3.2 admit it. Keep locator, Method, description, performer, assignment, Work, result, and responsibility separate. Run one bounded KindRestorationCheck when the changed expression can alter FPF-governed meaning; otherwise F.19’s local revalidation completes the ordinary repair.

  7. Re-evaluate the changed object as a separate pass through the same declared evaluation with comparable evidence and conditions. If the evaluation itself needs to change, use §4.1a before comparing values across that change. Keep the later Work, application or direct result relation, evidence, returned value, and result episteme distinct from the first pass.

  8. Record what improved, what stayed at the floor, what was unchanged by value, what became worse, and which findings moved outside this evaluation. Compare the earlier and later evaluation results. When result epistemes or dated Work are asserted, preserve their separately governed identities.

  9. Decide stop, continue, switchMethodFamily, openNewFrame, or holdUntilInformationBasisSufficient. When later replay depends on alternatives and guards, use the conditional structure block below. A decision or selected continuation neither authorizes nor performs the next Work.

  10. Leave an account that lets the next reader recover the object versions, evaluation, proposals, performed passes, result or change bases, evidence, trade-offs, cost and risk, continuation, stop and return boundaries, and the reason for the decision. Use the structured QualityImprovementLoopRecord only when a named replay, handoff, audit, or machine-facing use needs that form; otherwise a short result with the same recoverable facts is enough.

Stop here when this route answers the current use. Open the names, record schemas, and unfolding-structure block below only when a named receiving use depends on that added assurance detail.

E.23:4.1.1 - When evaluation can imitate improvement

Use this branch when sampling, repeated assessment, selection or the calculation of a reported result could produce the apparent gain while the claimed change is absent. The practical question is not just how variable the score is, but what this procedure would report without that change. Keep an existing comparison when it already answers that question adequately.

Name the claim and the reported quantity. Better average performance, newly successful cases, fewer losses and faster subsequent learning are different quantities. State the object, cases, conditions and change that the comparison concerns. A control for one quantity needs a separate basis to support a claim about another: a stable average does not prevent cases from moving in both directions between success and failure.

Construct the absence of the claimed change. Use an applicable model, a matched control, or both. For an unchanged stochastic solver, the model can describe repeated attempts under unchanged conditions. For a grading procedure, reassess the same archived answers without revealing their earlier grades. Say what remains fixed and what can vary. If a changed task population, assistance or operating condition can explain the difference, include it in the comparison or narrow the claim. Use C.16.MR to construct the measurement relation, C.16.IR to determine what it can resolve, and C.16.RM when the arrangement needs repair; the relevant mathematical or empirical method supplies the particular calculation or study.

Carry that absence through the whole procedure. Apply the same inclusion rule, thresholds, selection of a best attempt or version, recording, calculation and stopping rule used for the reported gain. If only initially unsuccessful cases are counted, reproduce that conditioning too. Preserve shared cases and other dependence: all ordered pairs of a few repeated runs are not that many independent experiments. An idealized model is sufficient only where its assumptions support the intended use; an empirical control can expose effects omitted from it, such as altered batching or assessor behavior.

Use the resulting behavior to decide the stated question. Obtain the relevant distribution or control estimate and its uncertainty, then choose a comparison rule for the intended claim and tolerated error. A mean control value is not itself a threshold, and subtracting it is not a universal correction. Check a rule tuned on control results on material not used for that tuning. Carry the uncertainty of the control into the final comparison. If the arrangement cannot distinguish a worthwhile change from these effects, obtain more informative evidence, improve the arrangement or leave the claim unresolved. Failure to detect a difference does not establish equivalence or absence of improvement. A difference beyond the control still needs any additional evidence required by a causal or wider capability claim.

Worked calculation. Suppose an unchanged solver has probability 1/2 of success on each of 20 cases, with independent attempts and independent cases. In two evaluations, failure followed by success has probability (1/2) × (1/2) = 1/4; success followed by failure has the same probability. The expected counts are five new successes and five losses, while expected net change is zero.

The new-success count has distribution Binomial(20, 1/4): the probability of eight or more is about 0.1018, and of nine or more about 0.04093. For a single comparison whose rule was fixed in advance, nine would therefore meet a chosen 5% upper-tail rule under this model; eight would not.

Choosing the best of several comparisons changes that calculation. Neither threshold establishes a cause, a general capability or improvement on a different task population. These are calculated properties of the constructed example, not intervention data.

Human and organizational limits. Regrading unchanged answers tests variation in assessment, not whether their author learned. Reprocessing unchanged service records tests the reporting procedure, not whether the service improved. A person can retain practice, fatigue or changed expectations, and an organization retains its history even if a document is restored. Construct a comparison appropriate to the claimed change in their work and support; do not manufacture an unchanged person or organization by copying an instruction. If only the reporting artifact can be held fixed, preserve that useful bounded result and leave the stronger claim open.

E.23:4.1a - When the evaluation needs to change

Use this branch when a missed defect, false objection, harmful proposed repair, changed use or excessive evaluation cost gives reason to reconsider the evaluation. An ordinary object repair can keep using the existing adequate evaluation.

First identify what changed. A new object version, a new requirement, a different task population, a different measuring procedure and a different evaluator implementation can require different responses. Fixing an implementation error can preserve the intended evaluation while invalidating the results affected by that error. A threshold that varies with load under a fixed rule still uses that rule; it does not by itself change the evaluation.

Keep the evaluation meanings and comparison conditions stable within one comparison. To change them between cycles, use A.19.ECS:4.5: compare the old and proposed evaluations on independently grounded defect and admissible cases, inspect harmful repairs and full cost, and use a new independent case when known cases were used for tuning. Decide whether to adopt the change for a stated use, retain it on trial, revise it, reject it or keep the current evaluation. A stricter result or higher evaluator agreement is not the adoption criterion.

After adoption, name which earlier conclusions depend on the changed rule. Re-evaluate affected preserved cases where cross-evaluation comparison is needed; otherwise retain earlier results with their original scope. Declare a bridge when it is justified, and leave unlike values incomparable when it is not. New task populations and requirements create new questions without retroactively falsifying supported old answers.

Recover how the evidence was obtained. The old criterion may have decided which variants to generate, which cases to examine, which branches to continue and which observations to retain, as well as how to score an answer. Changing a score does not undo these earlier choices. Start from the receiving decision under the adopted evaluation and trace the evidence it uses.

For each affected result, recover the object or response, case and conditions, observation or raw material, rule used for its assessment, and any old-criterion dependence in its generation, selection or retention. Use existing provenance when sufficient; add only the distinctions needed to settle the present use.

Distinguish three kinds of repair. When the same intended evaluation had an implementation error and the saved observations contain the required information, correct the affected calculation and its dependents.

When a changed evaluation can be applied to a preserved object, obtain the new judgement and retain the old one under its original question; evaluating an old artifact now is not rerunning its former performance.

When the new question needs outcomes never observed, a different population or consequential conditions absent from the record, rescoring cannot provide them. Obtain suitable new evidence, use a justified inference under the relevant subject Method, or restrict the conclusion. A large archive can still lack precisely the needed observation.

Rebuild the basis that directs the next choice. Follow the affected results into the summaries and estimates used to choose further work: success proportions, branch or method rankings, accumulated utility, fitted predictors, uncertainty estimates and resource priorities. Suspend their affected use until their basis has been repaired. Do not mix old- and new-criterion values in one current estimate just because their field names agree. Recompute the needed dependents from still-applicable evidence and newly obtained judgements, keeping population, denominator, conditions and uncertainty meaningful. A narrow correction requires only the dependent part; independent observations and unrelated conclusions remain usable.

Keep raw artifacts, observations and earlier qualified results recoverable. Withdrawing an old contribution from a current utility estimate is not deleting the historical event or declaring it false. Where a record contains both an unaffected observation and a criterion-dependent judgement, separate their uses instead of discarding both. Conversely, retaining an artifact does not qualify an old ranking built from it. If dependency cannot be recovered, mark that basis unresolved for the affected decision rather than assuming independence.

The repaired result should let the practitioner make the next warranted choice: continue with a valid comparison, test a promising alternative under the new conditions, retain a limited old answer, or return an explicit unresolved question. Include the work of obtaining new evidence and rebuilding estimates in the decision. No automatic instruction to reevaluate the whole archive follows. C.40:4.8.1 supplies a bounded history replay when its recorded responses and conditions suffice; changing the evaluator may require repairing both those responses and the selection mechanism that chose them.

Worked selection change. An organization received ten project proposals. A low-upfront-cost criterion admitted six to pilots; four were never piloted. Complete pilot records permit a later, adopted reliability evaluation of the six: three satisfy it. Those records support the descriptive result “three of the six selected pilots meet the new criterion.” They do not support a reliability rate for all ten proposals, and the four absent outcomes are not four failures. The old cost criterion affected the population observed, not only the score assigned afterward.

The organization keeps all proposals and pilot observations, preserves the earlier cost judgements under their old use, and replaces the affected reliability judgements for the six. It withdraws an old all-project promise or resource priority if that claim depended on treating the selected pilots as representative. It then rebuilds the comparison needed for the next allocation.

If the four untried designs are serious alternatives, an appropriate new pilot or justified subject model supplies their missing consequences; if they are irrelevant to the present decision, it can retain a narrower answer without testing them all. If the new requirement concerns a longer service period than any pilot recorded, even the six lack the required observation. A revised score cannot extend observation time.

Human self-development. A person formerly chose familiar exercises because the self-assessment rewarded fluent completion. A new question asks whether the acquired way works on unfamiliar problems. Saved answers can reveal previously missed reasoning defects, but awarding them new marks supplies no performance on the excluded problem classes. Preserve what those answers show, retire a transfer claim based only on the familiar exercises, and obtain suitable new cases. Update the practice-selection rule that used the old success estimate. Neither erasing the person’s history nor averaging old fluency marks with new transfer results answers that question; the subsequent practice and its actual use remain separate from changing the evaluation.

Worked slice. A review evaluation rewards concise explanations but misses an omitted dependency that makes the instruction unusable. Adding the missing dependency improves the object under its existing usefulness requirement. Changing the evaluator so it notices that defect is a different proposal. Compare that evaluator with its predecessor on the defective text and a concise, fully usable alternative; reject a repair that makes every explanation longer without need. A new independently grounded case supports adoption beyond those development cases. Previously supported readability judgements remain scoped to their text and reading conditions; only conclusions depending on the missed dependency need reopening.

E.23:4.2 - Conditional names and kind settlement

Source and practitioner phrases such as “loop engineering”, “agent loop”, “harness loop”, “prompt loop”, and “workflow hardening loop” are entry phrases. Lower them into ObjectUnderImprovementRef, QualityEvaluationQuestionFrame, QualityEvaluationUseDeclaration, ImprovementAim, MethodFamilySelection, CostAndRiskAccount, and QualityImprovementLoopRecord, or else name the subject pattern for the live claim and leave E.23 closed.

Quick lowering map:

Entry cueE.23 useExit when this is the live claim
“Build a loop” or “loop engineering”Ask which object version is being improved and which evaluation will be rerun.If no object-version improvement claim is present, choose the subject pattern named by the live claim.
Agent retry, monitor, or escalation cycleUse E.23 only when the retry changes an object version and re-evaluation can show a changed result on declared coordinates.Performed execution and work plans use the A.15 family; gate passage uses A.21; transformation-flow cycle structure uses E.18.
Harness engineeringThe harness can be the object under improvement when its next version is evaluated against declared quality, cost, and risk conditions.Running the harness is work; comparing harness variants is G.9; retaining variants is C.18 or C.19; selected-set result declaration is G.5; for publication, use E.17 for a source-backed face and return to source and E.24.PUB for the occurrence, form, carrier, audience, bounded use, and availability.
Fast DPF seed hardeningA local DPF seed, pattern seed, relation record, or source pack can enter E.23 after the object version and evaluation are declared.Recover the needed source contribution through the smallest sufficient source route in E.4.DPF:4, using G.2 only when a broad, refreshable CG-Frame pack is needed. Construct a missing way through C.39 and explain the connected framework Methods through E.4.CM before pattern allocation when those contributions are needed. Carry the resulting contribution and its limits back to seed hardening. For source decay, edition change, and refresh, use G.11. For PFAD and PFR decisions, use E.4.PFAD and E.4.PFR. For first-entry publication, use E.11 only when publication is current.

The next table names the local values used after this routing choice.

Local nameKind and function
QualityImprovementLoopMethodRepeated improvement U.Method for one object version under one declared evaluation use.
ObjectUnderImprovementRefExact U.Entity version being changed, paired with its exact U.Kind.
QualityEvaluationQuestionFrameThe E.22 U.Episteme that binds one exact object version and use declaration to the selected characteristic space, predicate or comparator, ClaimScope, exact result-consuming work or decision, evaluation purpose, qualification window, and any grounded non-use boundary required by E.22. E.23 reuses that frame; it does not move the consuming-use position into the declaration.
QualityEvaluationUseDeclarationThe E.22 U.Episteme that keeps any question-changing evaluator condition or intended-evaluator identity separate from the actual evaluator, assignment and dated Work, and keeps the evaluation pattern, optional semantic Method, selected characteristic space, predicate or comparator, ClaimScope, quality-model descriptions, evidence basis, result form, and qualification window distinct. E.23 reuses it; it does not define a second evaluation ontology.
LoopEvaluationEvidenceBasis@ContextU.Episteme whose EntityOfConcern is the exact object version evaluated in one loop pass. It describes the evidence values checked and missing evidence positions found for that pass and is distinct from E.22’s expected evidence-basis description.
LoopEvaluationResultFormDescriptionU.Episteme describing the result-row form used for the current pass; normally the same form cited by the evaluation-use declaration.
ImprovementAimDesired evaluation-result change. It names the intended quality change, not a value established by the repair itself.
MethodFamilySelectionSelected method family for the current object and evaluation.
OperationFamilySelectionSetOptional operation-family set selected because its operations can change the evaluated result enough to justify cost.
ObjectUnderImprovementEvaluationWorkRefReference to one independently identified dated A.15.1 evaluation Work occurrence. The Work remains distinct from its application, returned value, result episteme, evidence, and judgment.
ObjectUnderImprovementEvaluationResultRefReference to one separately constituted result episteme whose claims state the evaluation result. The episteme is not the returned value; the exact A.6.1 result binding or direct evaluation-result relation remains separately identified.
ImprovementPassWorkRefReference to one independently identified dated A.15.1 Work occurrence that changes or attempts to change the object. Selection of a proposal supplies no such occurrence.
CostAndRiskAccountCost and risk account used to judge another pass or operation.
ImprovementLoopDecisionValueLocal closed value set `stop
QualityImprovementLoopRecordU.Episteme whose EntityOfConcern is the exact starting object version for one bounded improvement-loop application. Its ClaimGraph relates that version to one admitted unfolding structure, selected next-action proposals, independently identified evaluation and improvement Work, exact result bases and result epistemes, changed versions, evidence bases, trade-offs, cost and risk, and the selected continuation and boundaries. It describes those objects and relations; it is not the method, performer, Work occurrence, changed object, or structure.
QualitySideEvaluationChangeClaimControlled form for a claim within an episteme’s claim content; it compares before and after evaluation results for named object versions on declared Q coordinates. Each result uses the evaluation-use declaration bound to its own object version; keep the evaluation criterion, scope, scale and qualification conditions comparable, or state the evaluation change adopted under E.23:4.1a and which results remain comparable.
SourceComposedResultClaimControlled form for a claim within an episteme’s claim content; it relates one changed-object result claim to exact accepted source-use decisions and each source contribution. It is neither the changed object nor a source-use decision.
KindRestorationCheckConditionally present precision-repair check required by the selected restoration predicate and its evaluation result.

The claimGraph and referenceScheme positions in the episteme forms below carry, by value, the complete interpreted content and effective rule collection of that episteme. Their C.2.1:4.1.2 ranges are NonemptyClaimContent and EpistemeInterpretationScheme. The rules must interpret that form’s version designations, predicates and qualifications, including the evaluation scale used by a prediction. Interpreted negative claims or predictions whose truth is unresolved can satisfy the content range; their evaluation and warranted use remain separate questions.

LoopEvaluationEvidenceBasis@Context <: U.Episteme:
  entityOfConcernRef: U.EntityRef, referencing the exact object version evaluated in this loop pass
  entityOfConcernKindRef: U.KindRef, referencing the exact kind of that object version
  claimGraph: NonemptyClaimContent by value
  referenceScheme: EpistemeInterpretationScheme by value
  editionId
  qualityEvaluationUseDeclarationRef: U.EpistemeRef, referencing one QualityEvaluationUseDeclaration about that object version
  checkedEvidenceValueRefs[]: U.EntityRef, each referencing one evidence value checked
  checkedEvidenceValueKindRefs[]: U.KindRef, each referencing the exact kind of the paired evidence value
  checkedEvidenceRelationRefs[]: U.EntityRef, each referencing one governed evidence relation
  checkedEvidenceRelationKindRefs[]: U.KindRef, each referencing the exact kind of the paired evidence relation
  unfilledEvidencePositionDescriptionRefs[]: U.EpistemeRef, each referencing one description of an unfilled evidence position
  qualificationWindowDescriptionRef: U.EpistemeRef, referencing one EvaluationQualificationWindow description

QualityImprovementLoopRecord <: U.Episteme:
  entityOfConcernRef: U.EntityRef, referencing the exact starting object version for this bounded loop application
  entityOfConcernKindRef: U.KindRef, referencing the exact kind of that starting object version
  claimGraph: NonemptyClaimContent by value
  referenceScheme: EpistemeInterpretationScheme by value
  editionId
  improvementUnfoldingStructureRef: U.EntityRef, referencing one admitted A.22 constraint-governed unfolding structure; when transformation-flow membership is current, this same selected U.Structure also satisfies E.18/E.18.3 rather than designating a second structure
  qualityEvaluationUseDeclarationRef: U.EpistemeRef, referencing one QualityEvaluationUseDeclaration
  selectedNextActionProposalRefs[]: U.EpistemeRef, each referencing one CandidateImprovementProposalRow@Context; selection does not establish performance
  evaluationPassClaims[1..*]:
    evaluationWorkRef: U.EntityRef, constrained to U.Work
    evaluationApplicationRef?: U.EntityRef, referencing one exact A.6.1 application when that is the evaluation route
    evaluationResultBasisRef: U.EntityRef, referencing its A.6.1 result binding or the evaluation-result relation defined by the evaluation pattern
    evaluationResultEpistemeRef: U.EpistemeRef, referencing one separately constituted result episteme
    loopEvaluationEvidenceBasisRef: U.EpistemeRef, referencing one LoopEvaluationEvidenceBasis@Context
  improvementPassClaims[]:
    selectedNextActionProposalRef: U.EpistemeRef, referencing one still-propositional E.22 row
    improvementWorkRef?: U.EntityRef, constrained to U.Work and present only after improvement Work obtains
    changedObjectVersionRef?: U.EntityRef, present only when that changed version exists independently of this record
    changedObjectVersionKindRef?: U.KindRef, paired with changedObjectVersionRef
    workResultOrChangePredicateRefs[]?: U.EntityRef, each referencing a declared Work-to-result or Work-to-change predicate, or an A.6.1 result-binding predicate used by the basis
    workResultOrChangePatternLocators[]?: U.EntityRef, positionally paired with the predicate refs and each referencing the subject pattern that defines that predicate
    workResultOrChangeBasisRef?: U.EntityRef, referencing either the obtaining Work-to-result or Work-to-change relation, a filled local claim that names the Work, result or change, applicable conditions and facts, or an A.6.1 result binding
  tradeoffProtectionSet: TradeoffProtectionSet@Context by value
  costAndRiskAccountDescriptionRef: U.EpistemeRef, referencing one cost-and-risk-account description
  loopDecisionValue: ImprovementLoopDecisionValue
  selectedContinuationClaimRef?: U.EpistemeRef, referencing the current branch-selection claim without turning it into Work
  stopBoundaryRef: U.EntityRef, referencing one ImprovementLoopBoundaryCondition@Context
  reconsiderationBoundaryRefs[]: U.EntityRef, each referencing one ImprovementLoopBoundaryCondition@Context
  loopDecisionReasonDescriptionRef: U.EpistemeRef, referencing one loop-decision-reason description
QualitySideEvaluationChangeClaim within result or loop episteme claim content:
  beforeQualityEvaluationUseDeclarationRef and afterQualityEvaluationUseDeclarationRef
  beforeObjectVersionRef and afterObjectVersionRef
  beforeEvaluationResultRefs[] and afterEvaluationResultRefs[]
  evaluationCoordinateRefs[]
  qualificationWindowDescriptionRef

SourceComposedResultClaim within result or loop episteme claim content:
  changedObjectVersionRef and changedObjectVersionKindRef
  resultClaimNodeRef?: C.2.1 ClaimAddress
  resultClaimEpistemeRef?: U.EpistemeRef, referencing the whole result episteme
  acceptedSourceUseDecisionRefs[1..*]
  sourceContributionDescriptionRefs[1..*]

The two named forms organize claims within the claim content of a result or loop episteme; a table row or serialization may publish them but does not become the claim.

A SourceComposedResultClaim uses exactly one of resultClaimNodeRef and resultClaimEpistemeRef. resultClaimNodeRef is the C.2.1:4.2.5 ClaimAddress pair: an exact episteme-edition reference and an intrinsic claim identity declared by that edition’s ClaimGraph. Resolve the edition and then require exactly one claim with that identity. An unresolved edition, absent or non-unique identity, or display-only node token leaves the address unresolved.

An edition change requires an explicit claim correspondence for any address migration. When no stable intrinsic claim identity exists, use resultClaimEpistemeRef if the receiving use can take the whole episteme’s content; otherwise constitute the required result claim as its own C.2.1 episteme and reference that whole episteme. A row number or rendering label does not supply the missing claim identity.

Checked evidence values stay paired with their kinds; evidence relations form a separate pair. Every actual Work reference points to an independently identified occurrence governed by the central §4 Work rule. A compact rendering may use only the omission allowed there.

For an improvement pass, the changed-version reference appears only after that version exists. The workResultOrChange... fields keep the declared predicate, its pattern locator, and the obtaining basis distinct. That basis must be an A.6.1 result binding, a direct Work-to-result or Work-to-change relation, or a local relation claim that names its participants, conditions, and obtaining facts.

If only the predicate or locator is known, keep the proposal, Work, changed object and Transformation separate. Recover the predicate’s applicable test and the facts it needs. An unavailable definition or condition is missing-governor; when occurrence evidence is unavailable, leave the relation unestablished and identify the missing basis. A supported negative or inapplicable result retains its own meaning. An A.15.PROD route points to its applicable local claim rather than to the pattern as a generic claim.

The two record epistemes follow C.2.1 identity: claim content, exact EntityOfConcern, and effective reference scheme determine each episteme edition. The listed loop fields contribute to claim content; editionId designates an already distinguished edition but does not constitute it. Empirical grounding, viewpoint membership, claim scope, model-use structure, applicability, qualification, evidence currentness, and source currentness remain separate relations or values defined elsewhere. A change in one of them changes a record episteme only when its claim content, EntityOfConcern, or reference scheme is revised; carrier and support serialization alone change neither episteme.

These records do not create quality values, project evidence, release state, selected-set result declaration, actual publication, parity, refresh, Work, Transformation, or proof of quality.

The @Context suffixes on support species such as LoopEvaluationEvidenceBasis@Context, CandidateImprovementProposalRow@Context, TradeoffProtectionSet@Context, and ImprovementLoopBoundaryCondition@Context are compatibility and retrieval spellings only. No suffix or context label supplies a container, participant, ClaimScope, applicability, or identity discriminator. Use the suffixless interface names QualityEvaluationQuestionFrame, QualityEvaluationUseDeclaration, and QualityImprovementLoopRecord.

E.23:4.2a - Conditional Improvement Unfolding Structure Block

Use this block when a named review or replay use relies on the improvement loop’s constraint-governed unfolding structure rather than only its method record. The block identifies the proposal epistemes, predicted evaluation-result changes, independently identified pass Work and results, guarded alternatives, decision value, information-basis hold, stop, and returns to neighboring work. A generic position in a structure does not identify which proposal, result or continuation occupies it.

ImprovementUnfoldingStructureBlock:
  unfoldingStructureRef: U.EntityRef, referencing one ImprovementLoopUnfoldingStructure
  objectVersionUnderImprovementRef: U.EntityRef
  objectVersionKindRef: U.KindRef
  evaluationFrameRef: U.EpistemeRef, referencing one QualityEvaluationQuestionFrame or equivalent exact frame
  qualityEvaluationUseDeclarationRef: U.EpistemeRef, referencing one QualityEvaluationUseDeclaration
  currentEvaluationResultRefs[]: U.EpistemeRef under that evaluation pattern
  candidateRepairProposalRefs[]: U.EpistemeRef, each referencing one CandidateImprovementProposalRow@Context under E.22
  tradeoffProtectionSet: TradeoffProtectionSet@Context by value
  expectedEvaluationResultChangeRefs[]: U.EpistemeRef, each referencing one ExpectedEvaluationResultChange@Context
  evaluationPassPositionRows[]:
    evaluationWorkRef: U.EntityRef, constrained to U.Work
    evaluationApplicationRef?: U.EntityRef, referencing one exact A.6.1 application when used
    evaluationResultBasisRef: U.EntityRef, referencing an A.6.1 result binding or evaluation-result relation
    evaluationResultEpistemeRef: U.EpistemeRef, referencing one separate result episteme under C.2.1
  improvementPassPositionRows[]:
    selectedNextActionProposalRef: U.EpistemeRef
    improvementWorkRef?: U.EntityRef, constrained to U.Work and present only after one dated U.Work occurrence obtains
    changedObjectVersionRef?: U.EntityRef, present only after that version exists
    workResultOrChangePredicateRefs[]?: U.EntityRef, each referencing a declared Work-to-result or Work-to-change predicate, or an A.6.1 result-binding predicate used by the basis
    workResultOrChangePatternLocators[]?: U.EntityRef, positionally paired with the predicate refs and each referencing the subject pattern that defines that predicate
    workResultOrChangeBasisRef?: U.EntityRef, referencing either the obtaining Work-to-result or Work-to-change relation, a filled local claim that names the Work, result or change, applicable conditions and facts, or an A.6.1 result binding
  guardedContinuationRows[1..*]:
    exactGuardOrConstraintClaimRef
    selectedObtainingRelationOccurrenceRefs[]
    admissibleContinuationDescription
  loopDecisionValue: ImprovementLoopDecisionValue
  selectedContinuationClaimRef?: U.EpistemeRef
  unfilledInformationBasisPositionDescriptionRefs[1..*]?: U.EpistemeRef
  informationBasisSufficiencyConditionRef?: U.EntityRef, referencing one ImprovementLoopBoundaryCondition@Context
  evidenceRelationRefs[]?: U.EntityRef, each referencing one exact evidence relation occurrence with its subject-pattern locator
  stopBoundaryRef: U.EntityRef, referencing one ImprovementLoopBoundaryCondition@Context
  reconsiderationBoundaryRefs[]: U.EntityRef, each referencing one ImprovementLoopBoundaryCondition@Context

ImprovementLoopUnfoldingStructure is a local A.22.CGUS U.Structure specialization whose improvement-loop membership predicate is defined here. Its constituents are the independently identified values named above; its selected obtaining relations and guard claims keep their exact predicates, occurrence-identity rules, and defining ClaimGraphs. A position row, adjacency, or selected continuation creates none of them.

When that selected structure additionally satisfies the transformation-flow membership and boundary conditions, E.18/E.18.3 recognizes the same U.Structure; do not manufacture a generic CGUS plus a second transformation-flow structure from reciprocal references. The organization is neither a root U-kind, enduring Work, context container, evidence, nor quality proof.

E.23 governs the coordinate-qualified prediction episteme:

ExpectedEvaluationResultChange@Context <: U.Episteme:
  entityOfConcernRef: U.EntityRef, referencing the exact object version whose later evaluation result is predicted
  entityOfConcernKindRef: U.KindRef, referencing the exact kind of that object version
  claimGraph: NonemptyClaimContent by value
  referenceScheme: EpistemeInterpretationScheme by value
  editionId
  qualityEvaluationUseDeclarationRef: U.EpistemeRef, referencing one QualityEvaluationUseDeclaration about that object version
  evaluationCoordinateRef: U.EpistemeRef, referencing one governed evaluation-coordinate description
  coordinateScaleRef: U.EpistemeRef, referencing one scale description that admits results for that coordinate
  currentEvaluationResultRef: U.EpistemeRef, referencing one current result episteme under the declared evaluation use
  changeExpressionKind: ExpectedEvaluationChangeExpressionKindValue
  expectedScaleValueRef?: U.EntityRef, referencing one value admitted by coordinateScaleRef
  expectedScaleValueKindRef?: U.KindRef, referencing the exact kind of that scale value
  expectedScaleRangeRef?: U.EpistemeRef, referencing one range description on coordinateScaleRef
  expectedScaleDirection?: EvaluationScaleDirectionValue
  candidateRepairProposalRefs[]: U.EpistemeRef, each referencing one CandidateImprovementProposalRow@Context
  predictionBasisRefs[]: U.EpistemeRef, each referencing one prediction-basis episteme
  tradeoffProtectionSet: TradeoffProtectionSet@Context by value

ExpectedEvaluationChangeExpressionKindValue is expectedValue | expectedRange | expectedDirection. Exactly one of value, range, or direction is present according to that kind. An expected value includes its exact kind and is admitted by coordinateScaleRef; an expected range belongs to that scale. EvaluationScaleDirectionValue is increaseOnScale | decreaseOnScale | preserveWithinRange | enterDeclaredRange | leaveDeclaredRange. Free direction prose does not close this episteme.

The episteme predicts a later re-evaluation result. Its listed prediction fields contribute to claim content; a new claim content, EntityOfConcern, or effective reference scheme yields another C.2.1 episteme edition. A changed grounding, viewpoint, applicability, qualification, source-currentness, carrier, or rendering relation does not by itself change the prediction episteme; revise its claims when that change alters the prediction. It is not an operation, move, transition, work occurrence, or proof of improvement.

ImprovementLoopDecisionValue is stop | continue | switchMethodFamily | openNewFrame | holdUntilInformationBasisSufficient. The hold value has non-empty unfilledInformationBasisPositionDescriptionRefs[] and an informationBasisSufficiencyConditionRef; other values leave both absent. Each description says which information-basis position is unfilled without referencing a nonexistent entity. The sufficiency condition says what information would make continuation admissible. A decision value or selected-continuation claim neither authorizes nor performs the next action.

ImprovementLoopBoundaryCondition@Context carries boundaryConditionKind = stop | subjectAssertionReconsideration | informationBasisSufficiency, a condition description, the affected object-version ref and exact kind, the unresolved assertion ref, and an optional non-semantic candidateSubjectPatternLocator. Source currentness, selected-set result declaration, actual publication, Work, evidence, and assurance remain distinct subject assertions under their exact predicates. A reconsideration boundary ends or redirects this E.23 use; it makes no later Work, decision, or relation obtain.

A visible cycle such as “draft -> evaluate -> repair -> re-evaluate” may be useful before execution. While any constituent, obtaining relation, guard, expected result change, protected trade-off, selected continuation, decision value, stop, or return needed for the wider improvement CGUS remains unresolved, keep that presentation as a ProvisionalUnfoldingDemonstrationDescription@Context about the object version and proposed continuation set. It may guide slot discovery, but it is not yet a structure or a slice.

Admit the wider ImprovementLoopUnfoldingStructure first. Only then may a separate DemonstrativeUnfoldingSlice@Context select one traversal through that admitted structure and name it as EntityOfConcern. Neither episteme is a QualityImprovementLoopRecord, performed Work, actual Transformation, or proof of improvement.

E.23:4.2b - How the conditional detail supports the loop

The names and schemas in 4.2 and the unfolding-structure block in 4.2a support the ordinary steps above; they do not define a second loop. Use them only when a receiving use must inspect exact record identity, evidence positions, independently admitted Work and result relations, guarded alternatives, or replayable structure. Otherwise keep the ordinary account and do not manufacture a record or structure merely to complete the schema.

For a structured use, the complete result of the selected evaluation belongs to step 4 (including the complete E.21 result when E.21 is selected), proposal and Work separation to steps 5–7, before/after comparison to step 8, guarded continuation to step 9, and the replayable record to step 10. The central Work rule, direct result or change relation, and precision-restoration checks remain the same in both forms.

E.23:4.3 - Stop, continue, and reopen

Stop when the current object version meets the declared floor or improvement aim and no feasible non-dominated proposal remains worth its cost under the current use, comparison set, source state, and protected trade-offs. If the remaining proposal mainly makes a value easier to argue while adding apparatus or worsening use, affordability, locality, source preservation, or ecology, reject that proposal; continue searching for a substantive content improvement if the improvement aim is still open, and stop only with a by-value no-proposal disposition.

Continue only when at least one predicted evaluation-result change states a scale-qualified change worth its cost and risk. Use ExpectedEvaluationResultChange@Context when the conditional account in §4.2a is needed.

Switch method when the current method family is not changing the evaluated result, is too costly, or no longer fits the evaluation. Use holdUntilInformationBasisSufficient only with non-empty unfilled-position descriptions and the sufficiency condition that would make continuation admissible.

A needed observation can prepare that continuation without itself changing the object under improvement. Identify the unresolved question, how the observation could change the next decision, and what information would be enough. For example, §4.4.1 can return to current-process observation before selecting an intervention. Keep that preparatory result distinct from an improvement pass and from evidence that quality improved.

An all-5, all-exceptional, current-front-reaching, or current-front-improving result is neither necessary nor sufficient to stop. Use the practical gain, protected trade-offs and attainable cost in the preceding rule. A stopped loop can reopen when a new use, Q component, source anchor, SoTA front, comparison set, affordability boundary or worthwhile proposal changes that judgement. If the object still fails its required use and no repair is feasible, choose replacement, withdrawal or another admissible continuation; stopping work does not make that object adequate.

Treat the five decision values as current continuation dispositions, not as Work states. When the admitted A.22 structure is used, a branch is usable only when its guarded continuation cites the exact current guard or constraint claim and the already-obtaining relation occurrences that make that alternative admissible. A stop or subject-assertion reconsideration is a boundary until an exact stronger predicate and current facts establish another relation. Naming A.15, E.22, G.11, G.5, or another subject pattern as a locator neither performs Work nor creates an object described there.

E.23:4.4 - Method-family selection

Method familyUse when
PDSAorPDCAFamilyLearning quality, baseline comparison, measuring instruments, or standardize-then-repeat action matter for the improvement loop.
Improvement Kata for work processes (§4.4.1)A desired direction is meaningful, the next operating condition can be bounded, and the effective way to reach it needs to be learned from the process. Add coached practice through HCD.9:4.10 when developing the people applying this way is also a goal.
POOGIFamilyThe evaluation problem is throughput-shaped or constraint-shaped.
OODAFamilyOrientation quality and feedback under changing conditions affect the evaluation.
RalphLikeGeneralAdaptiveFamilyA broadly capable agent can improve the object through repeated specification, feedback, memory, and verification under C.19.1 cost and risk discipline.
FixedPerformerObjectVersionUnderImprovementOptimizationFamilyThe performer or harness stays fixed while the object version is edited and re-evaluated.
NQDQualitySideImprovementFamilyThe evaluation supplies the Q side for a declared NQD and OEE comparison and loop changes seek a non-dominated change in evaluated Q coordinates.
SoTAReachAndMaintainFamilyReaching or maintaining an externally assigned front depends on composing several accepted source or practice anchors.
SpecializedObjectFamilyCycleA specialized method family fits a declared characteristic space and is BLP-compatible.

The selected family is justified by characteristic-space fit, the declared scale-qualified predicted evaluation-result changes, cost and risk, and protected trade-offs. Familiarity, automation, or current popularity is not enough.

For PDSAorPDCAFamily, use the selected formulation. Deming’s PDSA emphasizes study of a prediction; ASQ’s PDCA also includes analysis and learning in Check. Distinguish maintaining work under an adequate standard, testing a change to it and organizing sustained use of the change. The relevant method supplies the operations needed by the improvement question.

For POOGIFamily, distinguish selecting the system constraint, responding to a current buffer condition, changing buffer settings and organizing a transformation. The Cox–Boyd account connects Buffer Management, Five Focusing Steps and Change Question Sequence as distinct contributions. Their use does not impose a fixed number of control levels. Current buffer regulation remains a neighbouring operational method, not a compulsory step of every quality-improvement pass.

Where recurring departures or changed conditions require more than current regulation, connect the selected inquiry to the change it can justify and to later performance that can test the change. A local improvement may require an agreement across organizations; tactical configuration may require reconsidering strategic assumptions; a proposed remedy may require inquiry into feedback and delayed effects. Library.Reference:6.5.3.2 explains these distinct extensions and their source conditions. Select the contribution needed by the improvement question, preserving the evaluation basis and the rights of affected participants.

For OODAFamily, recover orientation, implicit guidance, explicit hypotheses and their tests through action. Common understanding and intent can support coordinated initiative under changing conditions. Preserve Boyd’s own distinction between command/control and appreciation/leadership when relating these contributions to the general control account in C.43.

If choosing the family requires clarification of the controlled subject, a changed goal, development of the performer or the organization needed for performance, use Library.Reference:6.5. It connects these family conditions to the subject methods and control arrangements that supply the missing contribution. An already adequate cycle can be used directly.

E.23:4.4.1 - Improve a work process through target conditions and experiments

Use this branch when a team must improve an actual process but cannot responsibly prescribe the whole path to the desired state. The working question is: What should the process do next, what currently prevents that, and which observation or experiment can justify the next move? Rother’s Improvement Kata provides a way to answer it repeatedly. A useful local result can be a bounded target condition and an informative next step; it need not be a complete solution to the longer-term challenge.

Choose an observable process and an attainable period of work. When people are learning the way, a small process with frequent, visible repetitions makes the connection between action and consequence easier to examine. Later applications can address harder or slower work. In a long-cycle process, an offline trial or simulation can provide earlier information, while the conditions that only real operation can establish remain to be checked there. Use an already adequate way directly when the receiving task needs no inquiry or practice.

Establish direction and study the current process. State the challenge that makes this improvement worthwhile, the desired outcome and the qualities that must be protected. A direction may come from a larger organizational target, but the team still needs to understand its own process. A long-term direction is not a predetermined list of intermediate solutions.

Observe where the work happens. Follow the material, information or customer request through the relevant steps; identify how work arrives, how an output is completed and where waiting, rework or variation occurs. Compare recent outcome observations with the required performance. Then examine the operating pattern that could explain that difference. For repeated output, observations at a consistent completion point distinguish the interval between outputs from the time spent on an individual task. Retain unusual observations and the circumstances that may explain them. An average can conceal the interruption or variant that should change the next action.

Use the analysis that the process requires. Where demand and available time govern output, derive the needed rate from those values. Where equipment is involved, distinguish unattended running, handling that delays the next cycle, and work that can occur in parallel. Where staffing capacity is the question, examine repeatable work content and variation before treating a simplified capacity estimate as available capacity. For office or project work, trace actual cases, their information dependencies and delays. These distinctions select a useful next question; they do not impose one manufacturing worksheet on every process. Initial observations can be sufficient to choose a step even while other facts remain unknown. Refresh them after intervening changes.

Construct the next target condition. Combine a near achieve-by date, the outcome sought within that scope, and the operating attributes expected to produce it. For example, a target might concern how complete inputs become available, how often an output is completed, or how work passes between two operations. Keep safety, correctness and other protected qualities explicit. The target describes a desired operating state; buying software or instructing operators describes a possible means of reaching it.

Derive this target from the current observations and the direction. Choose enough stretch to require useful learning, with a horizon and scope in which the team can act and see consequences. A wider ambition may first require a smaller target. Keep the operating attributes already adequate and leave genuinely unknown details unresolved rather than filling a form with invented knowledge. When a missing fact prevents even a meaningful target, obtain that fact first. The next target after this one will depend on what is learned; it need not be fixed now.

Select an obstacle and the boundary of present knowledge. Ask what observed condition prevents this target. A current queue, an inconsistent input or an unexplained variation can be an obstacle. “We have not installed our preferred solution” has not yet identified one. Describe what that solution would change and check whether the corresponding problem is present. An acknowledged lack of knowledge can itself determine the first step.

Keep a provisional list when several possible impediments would otherwise be lost, and select one relevant, manageable obstacle for the next investigation. The list is not a commitment to remove every item. Later observations can show that an item was mistaken, has disappeared or no longer matters to the target. A beginner may learn from a tractable obstacle; selecting the largest one is not a universal prerequisite.

Choose the next step from what is not yet known. Three useful choices have different results:

  • Observe the process without changing it when the current condition or the supposed obstacle is insufficiently known. State what the observation should distinguish, where to look and what facts to bring back. An expectation about what will be seen helps expose assumptions; it need not be a fabricated numerical effect prediction.
  • Try a bounded exploratory change when a small, permissible perturbation can reveal how the process responds. Predict the response as far as the present knowledge permits and identify what remains uncertain. Protect the people, product and continuing work if the prediction fails.
  • Test a particular explanation when the proposed causal relationship is clear enough to discriminate. State the predicted effect, the observations and comparison that could support or challenge it, and the conditions needed for that inference.

Match the comparison to the claim. To learn how a component affects performance, arrange variation that distinguishes its contribution from other changes. When components may interact, compare the relevant combinations: the effect of one change may depend on whether another is present. A qualified factorial design can support those distinctions.

When the decision concerns adopting a complete new way of working, compare that way with the available alternative under comparable receiving conditions; the warranted conclusion concerns their overall difference. A comparison of whole ways does not by itself identify their components’ separate effects. The subject’s experimental method determines allocation, replication, measurement and analysis. Carry changes in inputs, assistance or operating conditions into that design or restrict the resulting claim.

Choose the size and timing of the next contribution. Select the least burdensome sufficient action or coordinated set of actions for the decision, including observation, analysis or preparation where needed. One obstacle may require several factors or several experiments to be considered together. When the experimental arrangement remains unsettled, compare feasible alternatives, such as a single trial, a planned batch or an adaptive sequence, by the information they can supply, their full preparation and execution burden, delay and risk.

With long feedback, a batch can examine useful combinations in one waiting period; an adaptive sequence can justify additional waiting when an earlier result can usefully change a later experiment. A planned sequence may also need preparation whose value appears only in the later comparison. C.40:4.7 supplies the related choice of how to search and compare candidate ways; the relevant domain method supplies the experiment itself.

Before acting, agree the intended action or set, the expected result or learning, how it will be observed, the limits on harm and the point at which useful information should be available. A short trial is useful only if the process has had time and opportunity to produce the relevant observation. Involve the people who perform and understand the work. A coach can help the learner design this contribution while leaving the learner’s needed judgement observable.

Compare and decide what follows. After the step, distinguish the observed facts from the explanation drawn from them. Compare them with the expectation stated before acting. The result may support the explanation, contradict it, partly fit it or leave the question unresolved. Use the difference to update the current condition and choose what now needs to be learned or changed. A successful prediction can also provide useful knowledge. Rother uses experimentation broadly for this learning sequence; an observation or exploratory trial does not acquire the warrant of a controlled causal test by that name.

Retain a useful change only within the conditions its evidence supports. Before sustained use, check the protected qualities, the affected process as a whole and the operating conditions under which the change must continue to work. A successful trial and a stable working arrangement are different results. Return to §4.3 for the justified improvement continuation or stop; observation undertaken to fill a missing basis retains its preparatory status. Keep an adequate current arrangement when another change has no worthwhile contribution.

When the target is achieved or its agreed date arrives, compare the result with the original target and review what the process and the participants have learned. A missed target remains a missed target for that interval. Decide the next target from the resulting current condition and direction. If demand, safety, scope or the larger purpose has materially changed, reconsider the target and its grounds; do not preserve an obsolete target merely to keep a routine going or rewrite the old result as success. HCD.9:4.10 supplies the separate review of the learner’s reasoning and the coach’s help.

Worked example — a contract-entry bottleneck. This constructed example adapts the reasoning difficulty in the Practice Guide’s sales case; its times and subsequent trials are illustrative, not measurements reported by Rother.

A team wants an ordinary contract completed within 20 minutes. Older observations suggest 31–39 minutes overall, including 10–14 minutes entering data. The learner proposes automating entry. The coach asks when those observations were made and what now consumes the entry time. Because the available observations predate a change to the input form, the next useful step is to observe current contracts, separate typing from lookup and correction, and bring back the sequence and times. The team does not yet have a justified automation proposal.

Current observations of the selected ordinary variant show 11–13 minutes of entry, including three to five minutes repeatedly searching for approved item codes. The team agrees a near target: by Friday, entry for this variant should take at most eight minutes, with verified inputs available when entry starts, no invalid codes and no added delay elsewhere. The 20-minute overall challenge remains open. The desired input condition does not prescribe whether its means should be software, a changed handover or another arrangement.

The learner now predicts that a view of the existing approved code source, available at entry, will remove repeated lookup without adding a separate preparation task. A safe offline comparison uses matched saved contracts, keeps their content and correctness checks comparable, and alternates which procedure is tried first. This helps distinguish the changed lookup procedure from simple familiarity with a particular contract. Both code correctness and time are observed. In the constructed result, entry takes seven to eight minutes and the codes remain correct for the tested variant. That is a reason to examine a bounded supervised application and the whole-process consequences. It is not proof that every contract can now be completed within 20 minutes. Representative live work and maintained access still determine whether the change can become the ordinary arrangement.

The same contract work can pose a different experimental question. Suppose the proposed route changes both the input handover and the code view. If the team needs to understand whether the changes work separately or together, it needs observations that distinguish the four combinations, with the assignment of contracts, repeated observations and analysis chosen to support that inference. If it only needs to choose between the complete new route and the current route, compare those two arrangements for correctness, completion time and total work. That second comparison can support the route choice without identifying which component caused the difference.

A condition that changes the next move. A special contract needs a customer choice before an item code can be approved. A faster lookup cannot supply that missing decision. If this contract type is now required, the team first establishes when and by whom the choice can be obtained, and revisits the input handover or target scope. It preserves the demonstrated benefit for the ordinary variant without claiming the broader target. If the customer response time makes Friday’s broader target infeasible, that fact changes the target decision; it is not a reason to keep running identical lookup experiments.

The process and coaching sequence is developed from Mike Rother’s The Toyota Kata Practice Guide, the 2018 electronic edition, Chapters 5–9 and 11–12. It supplies current-condition analysis, target-condition construction and learning at the boundary of knowledge. The experimental distinctions use the NIST/SEMATECH handbook’s interaction, factorial-design and randomized-comparison explanations. Treating a complete way as the compared alternative is the application made here. Huan, Jagalur and Marzouk, revised 2026, §§5 and 6.3 distinguish batch and adaptive experimental design and its modeling and computational demands; model-based experimental-design optimisation requires a suitable model and enough expected benefit to justify its computation. The contract examples are constructed applications. Rother supplies a practice prescription and teaching cases; the experimental-design sources justify comparisons under their stated model conditions.

E.23:4.5 - Operation-family selection

An operation family is selected only when the loop account names:

  1. one scale-qualified predicted evaluation-result change;
  2. failure mode addressed;
  3. cost or risk reason;
  4. protected trade-offs;
  5. stop or removal condition.

Typical operation families are specification articulation, task decomposition, context refresh with carry-forward evidence, failure-context retry, verification against specification, memory or distillation, external critic or co-regulation, proposal portfolio use, search breadth or variants, bounded object-change budget, held-out evaluation, rejected-change memory, optimizer-memory separation, source-anchor contribution assignment, agent-tool-interface hardening, and task-family adaptation signature. They remain selectable only for the loop that justifies them.

E.23:4.6 - Cost and BLP discipline

C.19.1 governs the preference for broad, scale-amenable methods when safety, admissibility, and practical fitness are comparable. E.23 uses that preference but does not assume that accepted-work cost is one number. Compare material resources, tools and instruments, adaptation attempts, skilled attention, rework or delay, risk exposure, and avoided loss on their admitted scales. Keep the components separate, reject a dominated option, and use the declared project policy to choose or hold when no option dominates.

Net-cost arithmetic is permitted only after every term has been converted to one declared unit through an admissible conversion whose basis, uncertainty, and scope remain visible. Until then, avoided loss is a separate project estimate rather than a quantity subtracted from concrete burden. A justified avoided loss can still make an expensive loop preferable. For a simple object, a direct edit or adjustment, small repair, lower-cost performer, specialized cycle, or one-shot evaluation can remain the better option.

Choose harness improvement when a named defect in framing, tools, memory, verification or stopping causes avoidable retry and repairing it compares favourably with direct object repair or another available method. Apply C.11.DUA to preparation, recurring use, transition, independent judgement, maintenance and displaced useful work. Leave unavailable cost evidence unknown; a larger harness is not itself a gain.

E.23:4.7 - Source-composed, OEE, and NQD improvement

Accepted SoTA is the working external front only when assigned by the object-under-improvement evaluation, accepted source-use decision, or declared comparison set. E.23 can govern a loop that reaches, maintains, or improves relative to that front; it does not self-assign SoTA.

When an evaluation-result change depends on source use, source currentness, or a dated external front, the loop record cites the exact accepted source contribution and bounded use, including a G.2 pack result when that route was used, or the relevant G.11 currentness result, with the edition or date needed for replay. E.23 carries that reference; it does not make the source-use or currentness decision.

When several source anchors are used, retain each accepted source-use decision and contribution. If the improvement depends on their combined use, construct the needed connection under the subject Method; for framework Methods, use C.39 and E.4.CM. Preserve independently usable contributions when no composite Method is needed. Relate the changed result claim to those decisions and contributions and re-evaluate the changed object version. Use SourceComposedResultClaim within a separately constituted result episteme when the receiving use needs that account.

For NQD and OEE, use E.23 to change one object version or candidate and re-evaluate it on declared Q coordinates. Use C.17 for novelty, diversity, descriptors, and distances, C.18 for archive and front insertion, C.19 for pool policy, G.5 for selected-set result declaration, G.9 for parity, and G.11 for currentness and refresh. When audience availability is current, use E.17 for a source-backed publication face and return to source and E.24.PUB for the publication occurrence, form, carrier, audience, bounded use, and availability.

E.23:5 - Archetypal Grounding

Working rule. Name the object version and evaluation, make one bounded change, and re-evaluate the changed object before claiming improvement. Identify dated improvement Work only when that occurrence is asserted. Keep proposals, performed Work, the changed object or Transformation, the later evaluation, and its returned result distinct.

Agent harness improvement from a loop-engineering request. A user asks to improve a local DPF seed. The record names that seed version as the object under improvement, selects E.4.DPF.DA or E.21 for evaluation, and states the aim: make the seed usable for local first entry without public-Core claims. The loop may change only that seed or another declared evaluation or harness slice. Prompts, adversarial examples, and harness checks enter only when the record states the expected evaluation change and their removal or stop condition. Selection makes them proposals. Each actual harness run or seed edit is separate Work governed by the central §4 rule; returned values, durable results, changed versions, and Transformations stay separate and use their declared bindings or relations.

Use the smallest sufficient source route in E.4.DPF:4 for the seed’s needed source contribution, using G.2 only for a broad, refreshable CG-Frame pack. When the seed lacks the needed way or its usable explanation, develop that contribution through C.39 and E.4.CM before allocating pattern text. Carry the resulting contribution and limits back to the seed. Use G.11 for refresh, G.9 for parity, C.18 or C.19 for retained variants, and G.5 for selected-set results. Use E.17 and E.24.PUB for publication, and E.4.PFAD or E.4.PFR for their respective claims. A change outside the declared slice opens that neighboring work; it does not enlarge one E.23 loop without a new boundary.

Affordable floor evaluation. E.22 resolves the evaluation question if needed; the evaluator applies E.21 to every coordinate and returns the result through the declared application binding or result relation. If that evaluation is recorded as dated Work, the central §4 rule applies. If the result is admissible and no improvement aim was requested, E.23 stays closed. Any admission, refresh, landing, or release claim still uses its own E.19 or release gate; the E.21 result is quality evidence, not the gate.

Pattern exceptional improvement. A pattern already passes the declared floor. An additional worked case could extend its coverage, or a clearer account of dependent uses could make source refresh more replayable. Use E.22 when this further improvement aim needs framing. Selecting a proposal does not perform it. Add the useful worked case or refresh the source-bound rule, identify the repair pass as Work only if that dated occurrence is asserted, and re-evaluate the changed pattern through a later E.21 pass with its own result. Check what became worse. Stop at 4 when no worthwhile content improvement remains under the declared use; do not add apparatus merely to defend all-5.

Physical prototype improvement. The object is PumpAssembly@Prototype-3. Its evaluation declaration states any evaluator condition that changes result admissibility and keeps the vibration evaluation pattern and Method, Q-Bundle and characteristic-space descriptions, expected calibrated evidence, and result form distinct. The question frame binds those values to the engineering decision that will consume the result.

A test-bench measures Prototype-3 vibration. If this evaluation is asserted as dated Work, the central §4 rule applies. Its returned vibration value uses the evaluation’s binding or result relation; any durable result claim is a separate episteme. E.22 then proposes an impeller-geometry change while protecting efficiency and manufacturability. The proposal is not performance.

Machining and assembly produce Prototype-4 through separately identified Work. Link them to the changed version or Transformation only through an obtaining A.6.1 binding, Work-to-result relation, Work-to-change relation, or applicable A.15.PROD local claim. A later vibration evaluation uses the same characteristic space and evidence basis before any measured-improvement claim obtains; each asserted Work occurrence follows the central §4 rule.

Three proposals remain three evaluated alternatives. Under the same evaluation use, quality model, and expected evidence basis, E.22 can return three CandidateImprovementProposalRow@Context values: change impeller geometry, change bearing-support stiffness, and add vibration isolation. The QualityImprovementLoopRecord cites all three separately, without treating any as performed. Each has its own ExpectedEvaluationResultChange@Context for the pump-assembly version and keeps protected trade-offs such as efficiency, mass, manufacturability, and service access separate. If comparable operating-point measurements are missing, the actual LoopEvaluationEvidenceBasis@Context names that gap and holdUntilInformationBasisSufficient states the comparability condition. A later pass may select only proposals still worth their cost and risk; actual improvement begins with separately identified Work and an obtaining result or change basis.

DRR improvement. A DRR needs drafting adequacy for authoring across several selected pattern hosts. Use the coordinates supplied by E.9.DA, return row-atomic proposals, repair the decision, and re-evaluate the changed DRR through a separate E.9.DA pass and result. Identify each repair or evaluation as dated Work only when that occurrence is asserted, using the central §4 rule. The improved object is still a decision record, not prewritten pattern prose.

NQD quality-side improvement. A generated candidate has declared Q components and a comparison set. An E.22 application returns proposal rows. Use E.23 to organize candidate changes and later re-evaluation of Q; apply the central §4 Work rule only to performed passes. Use the direct definitions and tests for archive or front insertion, selected-set result declaration, publication, parity, and refresh. None is a quality-loop decision.

E.23:5.1 - Improve the way further improvements are obtained

The following constructed cases concern a way of improving work. They show how to obtain and use a changed way; they are not evidence that a particular training intervention succeeds.

A training organization develops its preparation of facilitators. Its participants can repeat a worked solution but keep using it after a decisive condition changes. The current preparation gives facilitators examples to explain. A proposed preparation instead teaches them to construct pairs of cases: retain the apparent task, change one condition that determines the appropriate action, and explain why the answer must change. The object being improved is this preparation method and its usable description. The facilitators, methodologist and tools are its actual means of application; stages in their reasoning are not thereby parts of a person under A.12.

Begin with one consequential failure. A solution is acceptable only when its entire error interval lies inside a tolerance. An exercise that compares only the point estimate teaches the wrong decision. The first facilitator repairs the exercise and the rule for constructing exercises: recover the condition that makes the action appropriate, retain its basis in the explanation, and build a contrast in which losing that condition changes the answer. C.40:4.7 supplies the comparison of obtaining ways, and C.40:4.11 supplies their further development from retained material. Preserve the previous preparation, its useful cases and the reasons for the change; a less successful variant may still contain a useful operation without becoming the selected preparation.

Try the old and proposed preparation on suitably comparable work under a declared allowance. Include case construction, tools, the methodologist’s help, practice, assessment, correction and subsequent use in the cost. Judge the exercises’ validity and what participants do on a fresh case, including earlier valid uses that must survive. If repeated grading itself changes which answers pass, apply §4.1.1 to the unchanged answers before attributing those transitions to learning. The comparison of human learning still needs its own basis. Use enough independent judgement for the decision; extra generated exercises do not create additional assessment capacity.

A better exercise is the first result. A different result occurs when the second facilitator uses the changed preparation to obtain the next improvement in that preparation. For example, the second facilitator applies its condition-and-contrast construction to a new failure, discovers that it omitted how to obtain the condition’s basis, and changes the preparation so the next facilitator can do that work. Preserve the received construction, the new case, the reasoning behind the change and the resulting preparation. An instruction to make such a change is only a proposal. If the original methodologist still does all the construction, the result demonstrates a supported arrangement, not the successor’s claimed independent operation. Assistance can remain part of the selected arrangement when that is the useful result.

A changed condition opens a different question. The next project depends on who has authority to make a decision, rather than on a numerical error bound. Substituting the word authority for a number is insufficient. Obtain the applicable basis for the decision right, determine the action permitted when that basis holds and what to do when it is absent, then construct cases that require the participant to use that distinction. Training does not create a missing right. C.40.CD develops this problem and the way of solving it together, retaining the original need to prepare useful facilitators.

Before calling this an expansion of possibilities, identify what the earlier way could obtain. If it already supported constructing and justifying cases about any relevant condition, the new example may be ordinary adaptation. If obtaining and using a non-numerical decision basis required a previously unavailable operation, demonstrate the work that the changed way now makes feasible. Use C.18:4.3a for that stronger comparison. A new project, a novel exercise or a higher score alone does not establish an expanded space. Better means may enable harder work that takes longer; acceleration is not required.

Make continuation possible for the next participant. Use C.36.RP:4.3–4.6 to decide whether the recipient needs to execute, adapt or develop the preparation. For development, give access to the construction, its reasons and conditions, relevant cases and the means or assistance for obtaining a missing contribution. Observe the recipient using these means on changed conditions and making the resulting construction usable in another preparation episode. Merely circulating the exercise set leaves this result untested. If a methodologist leaves, a surviving document may no longer be sufficient: recover the missing support or revise the arrangement before relying on the old success.

An independent participant outside the organization can continue this work without becoming its component. If the question now concerns transmission, retention or change across a cultural population, use C.36; count actual receiving and developing episodes for the declared question rather than treating popularity as the result. Changing those transmission conditions is itself a proposal to test.

A person improves a way of learning. An engineer has 90 minutes of personal practice and 15 minutes of mentor help each week. Familiar solutions are available, but changed constraints are missed. Instead of another rereading, the engineer retains a first unaided answer, compares it with the governing condition and a contrast, practices the missing operation, and tests a fresh case. To examine improvement of the learning method, follow its next use: can the engineer derive a useful learning question from a new work episode, select suitable practice and use the result for the following choice? Include help and the whole practice cycle in the allowance. One correct subject answer does not settle this question. If restored access to current data removes the difficulty without training, use the capability-access distinction in E.23.CAE and retain the adequate learning method. Developing a person’s capability, when separately selected, uses E.23.CDI and the relevant human-development practice; the person cannot be reset to an earlier state for comparison.

An AI system provides the same distinction with different means. A fixed external author may repeatedly improve an agent’s code without using the resulting agent as the author of the next improvement. To test that latter claim, place the candidate in the improving function and observe the next bounded development work under comparable resources and evaluation. A copied software version can preserve a starting basis; it does not justify an analogous rollback claim about a human or organization.

Return the observations to the selected question: what work improved, what changed in the way of improving, which changed means were reused, and what subsequent development became better or newly feasible at what cost. Keep each supported result within its conditions. Open a new E.23 frame when the object or evaluation question changes; use §4.1a when revising the evaluation. Stop additional method development when the available way is sufficient and a further change is not worth its burden.

E.23:6 - Bias-Annotation

This pattern biases FPF toward adaptive improvement with explicit re-evaluation. The bias is useful because many objects improve only through feedback and revision.

The bias is bounded. One direct evaluation can close without a loop. Repetition is justified only by a scale-qualified predicted evaluation-result change and acceptable cost and risk.

Scope: limited. The pattern covers repeated improvement of one declared object version under one rerunnable evaluation. It is not a universal account of change, learning, capability development, cultural evolution, publication, release, or project authorization; use the subject pattern for those claims.

LensDeclared bias and check
GovFavors an explicit evaluation, protected trade-offs, and a local stop or switch condition. Keep evaluation evidence separate from the decision, gate, publication, or release that may later use it.
ArchFavors one bounded object-under-improvement loop with named exits to specialized Methods and neighboring patterns. Do not let the loop absorb capability development, cultural evolution, DPF authoring, archive, selection, parity, or refresh architecture.
Onto-EpistFavors keeping a proposal, selected continuation, performed Work, changed object or Transformation, later evaluation, evidence, and result episteme distinct. Completion or a better score alone establishes none of the neighboring claims.
PragFavors rerunnable evidence and non-dominated improvement under cost, risk, and protected qualities. For a cheap one-pass question, a direct evaluation is preferable to maintaining a loop.
DidFavors an ordinary first move and unlike worked cases before formal loop records. Loop language can invite readers to mistake a visible cycle for enduring Work or context, so the grounding and anti-patterns show the distinctions in use.

E.23:7 - Conformance Checklist

CheckPassing condition
CC-E23-1Name the exact object version, its evaluation and the settled evaluation question before claiming a changed evaluation result. Use QualityEvaluationQuestionFrame and QualityEvaluationUseDeclaration when the receiving use needs their declarations.
CC-E23-2Reuse a question only while its object version, selected characteristic space, predicate or comparator, ClaimScope, result-consuming work or decision, purpose, qualification window, and any grounded non-use boundary remain applicable. Resolve missing or changed bindings through E.22; a separate frame is conditional on its receiving use, not on repetition alone.
CC-E23-3Represent returned repair possibilities as row-atomic E.22 findings or proposal rows with closure tests; pair proposals selected for the next pass with scale-qualified predicted evaluation-result changes, using ExpectedEvaluationResultChange@Context when the conditional §4.2a account is needed. A grouped memory summary does not discharge skipped rows, and proposal selection does not establish performance.
CC-E23-4For each evaluation or improvement pass asserted as U.Work, first identify every actual performer through A.13, then use A.15.1 to identify the occurrence, time, Method, and containing System independently. The obtaining A.13 assignment remains part of actual-performer admission. Add separate A.2.1 assignment-occurrence and F.6 references when the account expressly represents precise assignment-bound attribution; missing or failed attribution alone leaves independently admitted Work intact, subject to the premise-recheck rule in section 4. Then name the evaluation application and result binding or direct result or change relation, plus any separate result episteme. Re-evaluate the changed object before claiming coordinate, status, Q, or front-relation change.
CC-E23-5Record what became worse and protected trade-offs.
CC-E23-6Continue only when a scale-qualified expected evaluation-result change and the cost and risk account support another pass.
CC-E23-7Decide a local stop from the achieved aim, protected trade-offs and absence of a worthwhile feasible improvement under §4.3. All-5, exceptional or front-reaching results neither require nor suffice for that decision.
CC-E23-7aDo not treat 5, all-5, or 5-defensible as a repair target. Repair below-floor results first. Exceptional-improvement work proceeds through non-dominated proposal rows that name the expected substantive content change, protected trade-offs, and cost and risk. A no-proposal or stay-at-current-value disposition is admitted only when it names the checked evaluation evidence basis, using LoopEvaluationEvidenceBasis@Context when the conditional structured account is needed, and explains why every plausible content improvement is dominated, unavailable, or outside the declared scope. Reject changes that add guards, relation catalogues, evidence theatre, or quality proof while reducing use, affordability, locality, or ecology.
CC-E23-8When a neighboring claim appears during a loop, name the live claim and its subject pattern before continuing. That subject pattern still governs the claim when its object version becomes the object under improvement. E.23 organizes the improvement; it does not acquire the subject pattern’s authority.
CC-E23-8aFor a precision-restoration defect, apply F.19 as guidance and open a subject pattern only for unresolved FPF-specific meaning; claim Method or MethodDescription only after A.3.1 and A.3.2 admit it. Apply CC-E23-4 only when actual repair Work is asserted. Consume E.21’s compact PrecisionRestorationProfile when that evaluation is active. Require one bounded KindRestorationCheck when the changed expression can alter the object, kind, relation, slot or use position, claim kind, admissible use, or scope; otherwise F.19’s local revalidation completes the ordinary repair.
CC-E23-9Apply F.19 to wording introduced or repaired by the loop; use E.10 cues and a subject pattern only for an unresolved FPF-specific name, kind, claim or use.
CC-E23-10Preserve the named evaluation’s evidence basis, result-row shape, short-rationale rule, required result summaries, and coordinate-specific payloads in every re-evaluation. For E.21, consume the compact PrecisionRestorationProfile under E.21:4.3a.
CC-E23-11If a practitioner entry phrase such as “loop engineering”, “agent loop”, or “harness loop” appears, lower it to object version plus object-under-improvement evaluation before opening E.23, or name the direct neighboring subject pattern and stop the E.23 overread.
CC-E23-12In agent or harness cases, state which slice the loop may change: the target object version, the evaluation, or the harness object. Any other slice becomes neighboring work under its own subject pattern, not implicit E.23 scope.
CC-E23-13Keep the selected proposal, actual improvement Work governed by CC-E23-4, its result or change relation, changed object or Transformation, later evaluation pass, and result episteme distinct. When a required relation has no governor, retain those objects and the blocker; do not mint a generic Work-result relation.
CC-E23-14When §4.2a’s receiving-use condition holds, represent current alternatives, exact guards, selected obtaining relations, selected continuation, stop, and subject-assertion reconsideration conditions in one admitted A.22 unfolding structure. When transformation-flow membership is current, E.18/E.18.3 recognizes that same selected U.Structure; do not mint a parallel loop object. A visible cycle, record, structure, decision value, or branch is not enduring Work, context, authorization, or performance.
CC-E23-15A low value, finding, floor miss, or improvement aim does not establish an actual Problem. Any actual Problem used by the loop resolves to one current C.22.PFR occurrence with its direct participants and temporal identity.

E.23:8 - Common Anti-Patterns and How to Avoid Them

Anti-patternRepair
Checklist closed, quality improved. Discharge count replaces re-evaluation.Re-evaluate the changed object and apply CC-E23-4 when dated Work is asserted.
Loop result without evaluation form. The loop says the object improved but retains no evidence in the declared evaluation form.Restore that result form and evidence basis, then apply CC-E23-4 to any dated Work claim.
Agentic retry as method law. Repetition continues without a scale-qualified predicted evaluation-result change.State a scale-qualified predicted evaluation-result change, cost and risk, trade-offs, and a stop or switch condition.
Operation-family creep. Verification, memory, supervision, or search is added everywhere.Keep only operations that can change the evaluation result enough to justify cost.
Goodharted pass. Visible values rise while protected qualities worsen, or a non-5 value is treated as a defect to be fixed by more apparatus.Use trade-off inspection; apply E.13 when the visible value is replacing the intended value; reject, delete, split, relocate, or hold dominated changes; continue searching for substantive content improvement when the improvement aim is still open; record stay at current value only when the checked evaluation evidence basis shows that no non-dominated content improvement remains.
Lexical substitution closure. A trigger word disappears, but the replacement narrows, widens, or changes the object kind; for example a graph-shaped method or workflow cue becomes a work sequence without a selected ontology decision.Reopen the row, compare the pre-repair and post-repair meaning through F.19, using E.10, F.18 or the subject pattern for an unresolved kind or name, and leave the repair blocking if the kind cannot be preserved or explicitly changed by accepted decision.
Maturity-ceiling stop. All-5 is treated as an automatic stop or permanent end of development.Apply the gain, protected-quality and cost test in §4.3; state the reason for stopping and what could reopen it.
SoTA citation as self-assignment. Sources are cited as proof of frontier quality.State source contributions and re-evaluate the composed result.
Loop engineering as ontology. A fashionable source phrase is treated as a new Core kind or as proof that all repeated activity is one improvement loop.Use the phrase only as an entry cue; recover object version and evaluation, or use its subject pattern for the live claim. Common exits are work, gates, evolutionary retention and publication, source use, refresh, transformation-flow, and DPF subject patterns.
Proposal as performance. A selected proposal or continue decision is treated as if the repair happened.Apply CC-E23-13: keep selection epistemic until separate Work and an obtaining result or change relation exist.
Cycle as Work or context. A record, dashboard, retry label, or visible arrow cycle is treated as enduring Work or ambient context.Apply CC-E23-14: recover the conditional structure only when needed, and identify each asserted performed pass separately.
Finding as actual Problem. A low coordinate, floor miss, or loop-entry need is treated as a Problem occurrence.Keep the finding epistemic; cite C.22.PFR only when one actual condition and one criterion-applicability occurrence make the temporally identified ProblematicForRelation obtain.

E.23:9 - Consequences

ConsequenceBenefitCost
Repeated improvement follows one explicit improvement method and, when §4.2a is needed, one current unfolding structure, while each performed pass retains its own dated Work identity and attribution under CC-E23-4.FPF no longer relies on hidden authoring habits or one fictitious enduring loop occurrence.A complete loop record names its object, evaluation, structure, independently identified Work and result routes, and boundaries.
Row discharge is separated from evaluated quality change.Improvement claims become replayable.The claim remains inadmissible until a separate evaluation of the changed object supplies its result; identify dated Work only when the account asserts it.
General and specialized loops are comparable.BLP can be applied without craft folklore.Comparison is admitted with explicit cost, risk, and characteristic-space fit.
Exceptional stop remains local.All-5 or front-reaching closure no longer freezes future development.The closure record includes its reopen conditions.

E.23:10 - Rationale

The shared method is to select a proposed improvement, perform it, connect any returned value or changed object through its direct relation or A.6.1 binding, then re-evaluate through a separate pass and result. CC-E23-4 supplies the dated-Work account whenever either performed pass is asserted as Work. Check trade-offs and cost, then stop, continue, switch method, open a new frame, or hold.

When §4.2a is needed, A.22 carries the guarded alternatives, selected continuation, stop, and returns; when transformation-flow membership is independently current, E.18 and E.18.3 recognize that selected structure rather than a second loop object. Classical improvement cycles, agentic loops, fixed-performer optimization, MCDA, Goodhart, and OEE and NQD lines contribute useful operations and boundaries, but they do not replace this method or turn the cycle into enduring Work or context.

E.23:11 - SoTA-Echoing

SoTA here means the current best-known problem-solving practice for the stated question, not the newest, most official, or most familiar source. Comparisons are qualified as of 2026-08-19 unless a row states its own qualification date. Lineage, bounded current research, internal governing dependencies, and rejected transfers are named as such.

Practice questionExact source and statusSelected payload and limitSource-use decision, receiving locus, qualification, and reopen
What must a repeated improvement pass expose so that it produces learning rather than merely naming a cycle?Gerald Langley et al., The Improvement Guide, 2nd ed. (2009), retained Model-for-Improvement lineage; Michael Taylor et al., Systematic review of the application of the plan-do-study-act method to improve quality in healthcare, BMJ Quality & Safety 23 (2014), DOI 10.1136/bmjqs-2013-001862; Julie Reed and Alan Card, The problem with Plan-Do-Study-Act cycles, BMJ Quality & Safety 25 (2016), DOI 10.1136/bmjqs-2015-005076; D. Royce Sadler, Formative assessment and the design of instructional systems, Instructional Science 18 (1989), DOI 10.1007/BF00117714; John Hattie and Helen Timperley, The Power of Feedback, Review of Educational Research 77(1) (2007), DOI 10.3102/003465430298487. The last two are retained formative-feedback lineage.The sources contribute aim, explicit measures, prediction or proposal, tested change, comparison with the prior result, learning, and the connection among desired condition, current condition, and next move. Their healthcare and education evidence establishes neither a universal lifecycle nor FPF ontology or improvement.Adapt as lineage — reason: the common learning structure changes the E.23 action, while the named domain methods do not dominate current cross-domain practice. Receiving loci: E.23:4.1 steps 1, 3, 7–10; Affordable floor evaluation; Pattern exceptional improvement. Qualification/currentness: retained for this structure, not as present-front authority. Reopen: comparative evidence overturns the learning structure, or E.23 changes its proposal, re-evaluation, or stop action.
When does a specialized improvement-cycle family fit better than a general adaptive loop?Theory of Constraints Institute, Five Focusing Steps (https://www.tocinstitute.org/five-focusing-steps.html), living institutional explanation of POOGI; John R. Boyd, The Essence of Winning and Losing (1996 briefing), retained OODA lineage.POOGI contributes constraint selection, throughput-shaped improvement, and attention to inertia after a constraint shifts. OODA contributes orientation and feedback under changing external conditions. Neither source makes every quality problem a constraint or makes loop speed, cadence, or action volume a quality result.Adapt as lineage — reason: each branch supplies a useful selection discriminator without supplying a universal method. Receiving loci: POOGIFamily and OODAFamily in E.23:4.4. Qualification/currentness: living explanation and historical lineage, not current SoTA for all improvement. Reopen: either family is used outside its stated discriminator, or current comparative practice supplies a better branch at comparable effort.
Which repeated-agent and harness mechanisms warrant bounded use rather than one generic “agent loop”?Thoughtworks Technology Radar Vol. 34, Ralph loop (2026-04-15, Assess), current external-technique signal; Ralph CLI, Ralph loop (https://ralph-cli.dev/docs/core-concepts/ralph-loop/), and Wiggum.dev, The Loop (https://wiggum.dev/concepts/the-loop/), implementation/rationale sources; Noah Shinn et al., Reflexion (arXiv:2303.11366), Aman Madaan et al., Self-Refine (arXiv:2303.17651), Shunyu Yao et al., ReAct (arXiv:2210.03629), Andy Zhou et al., LATS (arXiv:2310.04406), and John Yang et al., SWE-agent (arXiv:2405.15793), retained stepping stones; Boyuan Wang et al., Harnesses for Inference-Time Alignment over Execution Trajectories (arXiv:2605.21516), Wenze Wang et al., A Physical Agentic Loop for Language-Guided Grasping with Execution-State Monitoring (arXiv:2604.07395), and Roxana Geambasu et al., Engineering Robustness into Personal Agents with the AI Workflow Store (arXiv:2605.10907), current 2026 preprints in distinct settings.The combined branch contributes fresh-context work against a specification, feedback memory, action–observation coupling, decomposition versus guided execution, partial-harness and over-structuring limits, bounded physical monitoring/retry/escalation, finite termination, and the flexibility–robustness trade-off of hardened workflows. It does not establish convergence by repetition, a general FPF loop kind, or that more harness is better.Adapt and combine — reason: Thoughtworks supplies a current practice signal, the 2026 papers supply distinct mechanisms and failure limits, and the older papers remain mechanism lineage. Reject as load-bearing current evidence: Ralph CLI and Wiggum documentation remain implementation/rationale only. Receiving loci: RalphLikeGeneralAdaptiveFamily, E.23:4.5, and Agent harness improvement. Qualification/currentness: primary official and arXiv sources checked 2026-08-19; evidence remains setting- and benchmark-bound. Reopen: the Radar posture or cited papers materially change, or comparative evidence changes the mechanism, cost, or stop choice.
When do an extra supervisor or accumulated search memory improve the loop rather than add apparatus?Zeda Xu, Nikolas Martelaro, and Christopher McComb, Supervising Ralph Wiggum / CRDAL (arXiv:2603.24768v2, 2026-05-07), current engineering-design research signal; Yanlong Wang et al., FactorMiner (arXiv:2602.14670v1, 2026-02-16), current financial-alpha-search research signal.CRDAL contributes metacognitive co-regulation when fixation, underexploration, or expensive design mistakes are live. FactorMiner contributes retrieve/generate/evaluate/distill, modular skills, experience memory, and reduced redundant search in a large comparable search space. Neither supports a universal supervisor, financial-alpha objectives, or automatic transfer to every object.Adapt as two domain-bounded branches — reason: the operation cues are useful only when E.23 can name the corresponding risk or comparable search. Receiving loci: E.23:4.5 operation-family selection, E.23:4.6 cost/risk discipline, and Agent harness improvement. Qualification/currentness: current preprints for engineering design and financial alpha, not general improvement evidence. Reopen: either paper changes materially, a comparative source dominates its cue, or E.23 begins using the domain objective as a general quality value.
When may a fixed performer improve by changing one external method-description object?Yifan Yang et al., SkillOpt: Executive Strategy for Self-Evolving Agent Skills (arXiv:2605.23904v2, 2026-05-25), current preprint.SkillOpt keeps the target model fixed while a separate optimizer makes bounded add/delete/replace edits to one external skill document, accepts only held-out improvement, and keeps rejected-edit and optimizer memory separate. Its benchmark results do not transfer automatically to arbitrary Methods, physical systems, or work-facing system-role kinds and assignments.Adapt — reason: the fixed-performer, mutable-object, bounded-edit, held-out-acceptance distinction sharpens an E.23 family without importing the optimizer as FPF method law. Receiving loci: FixedPerformerObjectVersionUnderImprovementOptimizationFamily, BoundedObjectChangeBudget, held-out re-evaluation, and Agent harness improvement. Qualification/currentness: current preprint, limited to its models, benchmarks, skill documents, and harnesses. Reopen: the paper changes materially or stronger comparable evidence changes the held-out acceptance or bounded-edit rule.
How should several quality coordinates and OEE/NQD alternatives be compared without turning one score or algorithm into authority?Xi Lin et al., Quality-Diversity Optimization as Multi-Objective Optimization (arXiv:2602.00478, 2026), current preprint; Haoxiang Qin et al., A survey on Quality-Diversity optimization: Approaches, applications, and challenges, Swarm and Evolutionary Computation 100:102240 (2026), DOI 10.1016/j.swevo.2025.102240, current survey; Rick Kazman, Mark Klein, and Paul Clements, ATAM (CMU/SEI-2000-TR-004, 2000), retained software-architecture lineage.QD/MOO contributes set-valued multi-coordinate comparison and explicit behavior or descriptor spaces. ATAM contributes scenario-based exposure of quality-attribute trade-offs. These sources do not supply one universal scalar, an FPF archive/front policy, or authority over candidate generation, retention, selection, parity, refresh, or publication.Adapt — reason: set-valued and scenario-visible trade-offs support non-dominated choice while direct neighbours retain their own decisions. Receiving loci: E.23:4.1 steps 8–9, Physical prototype improvement, Three proposals, and NQD quality-side improvement; C.17–C.19/G.5/G.9/G.11 remain the exits. Qualification/currentness: QD sources are current for QD/MOO; ATAM is domain lineage. Reopen: a current QD overview dominates these payloads, or E.23 starts defining a neighbour-owned archive, front, pool, selection, parity, or refresh rule.
When does optimization of a visible measure cease to improve the intended value?Jacek Karwowski et al., Goodhart’s Law in Reinforcement Learning (ICLR 2024), and Thomas Kwa, Drake Thomas, and Adrià Garriga-Alonso, Catastrophic Goodhart: regularizing RLHF with KL divergence does not mitigate heavy-tailed reward misspecification (NeurIPS 2024), current narrow proxy-risk anchors; Charles Goodhart, Problems of Monetary Management: The U.K. Experience (1975); Donald T. Campbell, Assessing the Impact of Planned Social Change, Occasional Paper 8 (1976); David Manheim and Scott Garrabrant, Categorizing Variants of Goodhart’s Law (arXiv:1803.04585, 2018); and Jongwoon Choi, Gary Hecht, and William Tayler, Lost in Translation: The Effects of Incentive Compensation on Strategy Surrogation, The Accounting Review 87(4) (2012), retained monetary, social-indicator, taxonomy, and strategy-surrogation lineage or evidence.The sources distinguish proxy misspecification, optimization pressure, behavior change, heavy-tailed reward error, and strategy surrogation. They do not forbid measurement, predict every failure, or make RL/RLHF results a general project-value model.Adapt and combine — reason: the 2024 papers provide current domain-bounded mechanisms while the older sources preserve distinct cross-domain failure explanations. Receiving loci: protected trade-offs, E.23:4.3 stop, Goodharted pass, and the E.13 exit. Qualification/currentness: the current anchors are narrow RL/RLHF results; older sources remain lineage or bounded experimental evidence. Reopen: stronger current proxy-risk evidence changes a used mechanism, or E.13’s intended-value boundary changes.
When can the evaluation change without silently moving the improvement target or retaining a dependent selection bias?Iacob et al., The Red Queen Gödel Machine: Co-Evolving Agents and Their Evaluators, v3, §3.2 and Appendices F.4–F.5, bounded current research; Kim et al., Rethinking Reward Model Evaluation Through the Lens of Reward Overoptimization (2025), reward-evaluation research.RQGM separates fixed within-epoch criteria from changes between epochs; criterion dependencies identify utility records to remove from current estimates and derived selection statistics to rebuild. Preserved artifacts can be reevaluated on demand. Kim et al. shows that optimizing an evaluator proxy need not optimize downstream performance.Adapt: §4.1a uses A.19.ECS:4.5 for adoption, then traces dependencies in evidence acquisition and rebuilds the affected decision basis. This does not require destroying history or globally rescoring every artifact. Qualification: the additional v3 mechanism was checked 2026-10-10; fixed-epoch assumptions and computational experiments do not establish unconditional guarantees for adaptive search or human and organizational improvement. Reopen: changed populations, missing observations, leakage, harmful repairs or full cost invalidate the current basis.
How should a source-composed improvement claim expose what each source contributes?Joanne McKenzie and Sue Brennan, Cochrane Handbook for Systematic Reviews of Interventions, version 6.5 (2024), Chapter 12, current evidence-synthesis reference; FPF G.2 for a broad, refreshable SoTA pack and G.11 for source currentness, each an internal governing dependency only when that contribution is needed.Chapter 12 contributes disclosure of the selected synthesis method and limitations. FPF supplies the cross-domain decision and currentness rules. Cochrane’s healthcare evidence hierarchy and statistical methods are not imported, and the broad synthesis of cycle, agent, trade-off, proxy, and OEE/NQD lines remains rationale rather than independent current evidence.Adapt Chapter 12 for transparent contribution and limitation reporting; retain conditionally G.2 for broad SoTA-pack work and G.11 for currentness or refresh; reject healthcare-specific apparatus and citation-count front claims. Receiving loci: SourceComposedResultClaim, E.23:4.7, and the source-use/currentness exits. Qualification/currentness: current reference plus internal rules, not external cross-domain SoTA. Reopen: Chapter 12 materially changes the used reporting rule, G.2/G.11 changes, or a composed claim no longer names its admitted sources and limits.
How can an improving method be compared without confusing better downstream work with actual use of the changed improver?STOP: Self-Taught Optimizer (2023), retained research anchor; Srikanth et al., Recursive self-improvement of AI research agents (2026), bounded current AIDE² research.STOP compares improvers through downstream code improvement under a budget. AIDE²’s main run keeps the outer author fixed while changing the agent being edited; its separate promotion experiment is inconclusive.Adapt and distinguish: §1 and §5.1 compare subsequent work and require actual reuse for that claim. Qualification: primary scopes checked 2026-10-10; neither result establishes universal acceleration or effectiveness for humans and organizations. Reopen: the compared function, resources, evaluation or claimed reuse changes.
Could the evaluation manufacture the gain being reported?Xu et al., Phantom Gains: Auditing Self-Improvement Against a Measured Null (2026), bounded current measurement study.An unchanged model passed through the same pipeline can produce apparent transitions; each statistic needs its appropriate no-change comparison and uncertainty. A threshold fitted to one comparison need not survive further controls. The study concerns a particular model family and short LoRA training, not all self-development.Adapt: §4.1.1 develops the improvement-specific connection to measurement construction and resolving power. Applicable analytical models remain usable; a measured control is not compulsory when an adequate basis already exists. Qualification: checked 2026-10-10. Reopen: changed recording, selection, dependence or tuning invalidates the comparison.
Can improved means be used to improve the work of improving?Douglas Engelbart, Bootstrapping, retained strategic lineage, §§3–4 and 8–10.Apply developing means in improvement work and develop human and technical means together. The proposed strategy and pilots are not proof of a particular organizational intervention.Adapt as lineage: §5.1 connects actual later application, changed preparation and available support. Qualification: relevant primary passages checked 2026-10-10. Reopen: a changed arrangement cannot support the claimed later application or places omitted costs on its recipients.
How does further development relate to cultural transmission and open-endedness?Borg et al., arXiv:2203.13050v2, theoretical cultural-evolution synthesis; Hughes et al., Open-Endedness is Essential for Artificial Superhuman Intelligence (2024), formal and strategic proposal.Borg treats conditions of transmission, retention and variation as themselves capable of change. Hughes relates novelty and learnability to an observer; this predictive-model learnability is not practical acquisition by a person. Neither supplies an empirically established recipe for indefinite development.Adapt with limits: §5.1 distinguishes a next participant’s development, newly feasible work and the stronger open-endedness claim, using C.36.RP, C.40.CD and C.18. Qualification: relevant primary passages checked 2026-10-10. Reopen: changed claims about the observer, receiving competence, transmission or reachable work require a different construction.
Which relationships and assumptions distinguish compact cycle formulations?Deming Institute, PDSA, IHI, Model for Improvement and implementation, and ASQ, PDCA, current institutional formulations; Cox III and Lynn H. Boyd, System Design, §3, a particular BM/5FS/CQS treatment; John R. Boyd, original briefings in Air University’s A Discourse on Winning and Losing, retained lineage, Organic Design pp.218–253 and the complete Essence of Winning and Losing.Preserve the selected PDCA/PDSA formulation, testing versus sustained implementation, distinct POOGI contributions, and Boyd’s orientation, implicit guidance and coordinated initiative. Boyd’s command/control and appreciation/leadership retain their own meanings. These sources establish neither a fixed level count nor one universal TOC structure or comparative superiority across cycle families.Adapt for family selection: E.23:4.4; Library.Reference:6.5 supplies the connected explanation. Buffer regulation remains a neighbouring operational use. Qualification: named passages and formulations qualified 2026-10-30; current institutional availability and retained lineage are distinct from a best-known-practice conclusion. Reopen: the selected formulation, operating conditions or comparative claim requires a different source basis.
When does a current flow response need a change in operating conditions, coordination or reasoning?Cox III, PASS execution, §5, online 2019, issue 2021; Cox III and Mabin, healthcare supply-chain proposal, §§4–5.3, 2025; Demand Driven Institute, DDAE and linked component descriptions, living methodology; Mabin and Cavana, TOC and qualitative SD, 2024 methodological study.Distinguish immediate response, recurring-cause inquiry, changed conditions, interorganizational coordination, bidirectional tactical/strategic adaptation and mutually revising causal/dynamic accounts. Cox–Mabin’s decomposition and reconnection stages were not fully implemented. Mabin–Cavana’s illustrative trajectories are hypothetical and their teaching exercise did not demonstrate implementation.Adapt for conditional extension: E.23:4.4 and Library.Reference:6.5.3.2/6.5.4. Qualification: relevant primary passages and live DDI method texts consulted 2026-10-30; selected source contributions, not a complete TOC survey or comparative advantage claim. Reopen: the required implementation, transfer or comparative claim exceeds those passages or changes their conditions.

E.23:12 - Relations

PatternRelation
A.19.ECSConstructs or repairs an object-under-improvement evaluation when none exists.
C.40:4.7, C.40:4.11Construct and compare ways of obtaining results through their later work, including changing the means of further development and retaining useful material.
C.40.CDDevelops a next problem and its solving means together when an acquired operation opens work that the earlier question did not settle.
C.36.RP, C.36Supply the next participant’s acquisition, adaptation and further development, and the separately current cultural-population question.
C.16.MR, C.16.IR, C.16.RMConstruct the measurement relation, determine what it resolves and repair the arrangement. E.23:4.1.1 connects these contributions to the reported improvement claim and its no-change comparison.
E.22Resolves a missing or disputed evaluation question, preserves adequate direct requests, and supplies the conditional frame and use declaration plus finding/proposal forms.
E.21Defines the evaluation used to obtain pattern-quality values in pattern-improvement loops.
E.9.DADefines the decision-adequacy evaluation used in DRR loops.
E.2.DADefines the FPF Pillar-adequacy evaluation used in corpus-level loops.
A.22.CGUSDefines constraint-governed unfolding structures. E.23 specifies their improvement-loop specialization: exact constituents, already-obtaining relations, guards, admissible alternatives, selected continuation, stop, and subject-assertion reconsideration conditions. The structure, visible cycle, record, and slice perform no Work.
A.13, A.15.1, A.2, A.2.1, F.6, A.6.1, C.2.1, A.3.4, A.15.PRODDefine or constrain actual performers, independently admitted evaluation or improvement Work, optional local classification and precise assignment-bound attribution, application and result binding, result episteme, actual Transformation, and any separately current production branch. E.23 mints no generic Work-result or Work-to-change relation.
C.22.PFRGoverns an actual Problem occurrence when one is used by an improvement claim; a finding, floor miss, evaluation need, or loop record does not establish its actuality or temporal identity.
E.13Governs pragmatic utility and proxy-to-value alignment when loop targets, quality values, metrics, or review results become substitutes for the intended value.
G.2Supplies a broad, refreshable CG-Frame pack when that source-synthesis contribution is needed. Carry its adopted contributions and limits into the affected improvement claim; other source routes keep their own bounded basis. For a DPF seed, select the smallest sufficient source route through E.4.DPF:4; any evidence claim remains under its applicable evidence rule.
F.18Supplies durable-name evaluation for naming loops.
C.25, C.16.QGovern engineering quality bundles and quality-word precision repair.
C.19.1Governs BLP and cost and risk comparison for method-family choice.
C.22.1, C.24Govern durable task-family adaptation and tool-call planning when the loop makes those claims.
C.17, C.18, C.19, G.5, G.9, G.11Govern OEE and NQD candidate characteristics, archive, front, pool, selected set, parity, and refresh.
E.18, E.18.3When the selected A.22 improvement structure also satisfies transformation-flow membership and boundary conditions, recognize that same U.Structure as the current transformation-flow unfolding structure. A visible loop, method, record, or series of Work occurrences supplies neither membership nor a second TFS by shape.
E.18.1Carries accepted problem-side records or generated seed records toward the next FPF relation, including DPF seed-to-hardening routes before a quality-improvement loop is ready.
E.4.DPFGoverns DPF authoring routes and publication carriers when a fast local framework seed is the object being carried toward use or admission.
E.4.PFAD, E.4.PFRGovern framework architecture decisions and framework relation records; E.23 may improve a declared artifact version but does not decide those framework slots.
A.21Governs the profile application, check-result mapping, gate decision and bounded action consequence. Monitoring, retry, escalation or a green harness state does not supply that decision. E.17 and E.24.PUB govern its publication when needed.
C.32.P2SUses improvement-loop results only when they reopen architecture problem-to-structure carry-through; E.23 still governs the loop record and re-evaluation.
C.11, A.10, B.3, A.15, A.20, A.21Govern decision, evidence, assurance, work, gate, and release claims when a loop result is reused beyond quality improvement.
E.10, E.10.ROLE, A.3.1, A.3.2, A.6.P, C.2.P, F.18, F.19Repair load-bearing wording and names introduced by loop records. E.10.ROLE resolves ambiguous source role before any system-role-kind, assignment, participant, or non-system use is asserted. A.3.1 supplies the Method admission test; A.3.2 supplies the membership test for a qualifying U.MethodDescription. Selected pattern content otherwise defines, constrains, tests, or guides without becoming the Method or its description.
E.23.CAESupplies the separate observation-and-disposition probe when apparent capability loss or failed transfer leaves the object to improve ambiguous. It returns candidate routes, not an improvement-object choice or loop decision.
E.23.CDIDescribes the separate Method for developing one admitted holder System’s capability for a named Work family and checking transfer in representative Work after development is separately selected. E.23 routes to it but does not absorb its action sequence; member distributions and cultural propagation keep their own results.
C.43, Library.Reference:6.5Explain how a compact cycle connects observation, choice, action and learning, what its realization requires, and which changed relationship calls for a different method or arrangement. Use this comparison when family selection depends on those conditions; E.23 retains object-quality improvement and its evaluation discipline.

E.23:End

Referenced in the corpus

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