Canonical Evolution Loop
About this pattern
This is a generated FPF pattern page projected from the published FPF source. It is canonical FPF content for this ID; it is not a FPF Reference product feature page.
How to use this pattern
Read the ID, status, type, and normativity first. Use the content for exact wording, the relations for adjacent concepts, and citations to keep active work grounded without pasting the whole specification.
Status: Stable Type: Pattern
Use this when. Use this pattern when repeated adaptation must keep one exact subject's current identity, an observed basis, a proposed or actual successor, the Systems and dated Work that make the change, and renewed use connected. Name the subject kind first: System, episteme, Method, MethodDescription, or a sequence of distinct Work occurrences.
What goes wrong if missed. Teams treat drift, learning, release, and improvement as unrelated events. Specifications become stale, operational surprises lose their evidence relation, and changes appear without a clear predecessor, successor, performing System, or dated Work. At the opposite extreme, one generic loop is imposed on every subject: a description edit is called a Method change, completed Work is said to be revised, or an internal adaptation is rejected because its acting System is not external.
What this buys. A compact, reviewable adaptation cycle whose identity rule comes from the subject pattern. It keeps observed basis, design-time change, run-time use, acting Systems, dated Work, evidence, publication, acceptance, and responsibility distinct while connecting only the facts needed by the receiving use.
Not this pattern when. Not this pattern when one direct subject-pattern claim answers the change question without a repeated adaptation cycle. Use B.3.5 for relation grounding, B.4.1 for early cue stabilization and routing, B.5.2.0 for abductive hypothesis work, C.27 for temporal status, or A.15 for method/work alignment without an adaptation-loop claim.
The FPF is built on the Principle of Open-Ended Evolution (P-10): continued use can reveal reasons to adapt a System, an episteme, a Method, or a description. The useful commonality is a repeated move from use through an observed basis and an explicit change back to use. The identity question is not common. The same subject may continue through a change, a successor may be identified, or later dated Work may be a distinct occurrence rather than a revision of earlier Work. B.4 therefore supplies a shared cycle only after the relevant subject pattern has supplied that distinction.
Keywords
- evolution loop
- DesignRunTag feedback
- observe-notice-stabilize-route
- drift repair
- knowledge refinement
- method refinement
- open-ended evolution.
Relations
B.4.xContent
Problem Frame
The FPF is built on the Principle of Open-Ended Evolution (P-10): continued use can reveal reasons to adapt a System, an episteme, a Method, or a description. The useful commonality is a repeated move from use through an observed basis and an explicit change back to use. The identity question is not common. The same subject may continue through a change, a successor may be identified, or later dated Work may be a distinct occurrence rather than a revision of earlier Work. B.4 therefore supplies a shared cycle only after the relevant subject pattern has supplied that distinction.
Problem
Without a canonical, shared model for evolution, projects fall into predictable and costly failure modes:
- Design-Reality Divergence (The "Drift"): The run-time subject in use slowly diverges from its design-time account. Formal models become elegant fictions, assurance cases become irrelevant, and the project loses the ability to reason reliably about what it uses.
- Learning Stagnation (The "Ivory Tower"): Observation produces valuable findings, but no explicit change path carries them into a revised design or renewed use. "Lessons learned" remain static documents.
- Chaotic Change (The "Whack-a-Mole"): Reactive patches have no stated observed basis, identity decision, or return-to-use condition. Hidden dependencies and unintended consequences accumulate.
Forces
Solution
Use the Canonical Evolution Loop as a coordinating cycle, not as a universal identity rule. First recover the exact subject and what continuity means for it. Then name the actual Systems and dated Work, traverse only the phases that occurred, and connect the result to renewed use.
Name the subject and continuity question
Not every holon is an adaptation-loop subject, and the five branches are not interchangeable. If the continuity or successor relation is still open, keep that question explicit rather than hiding it behind the word evolution.
Separate the changed subject from the acting side
The subject does not observe, refine, or deploy itself by grammatical convenience. A System performs each actual piece of Work. The changed subject, performing System, dated Work, Method enacted by that Work, and result remain distinct. Practitioner prose can still say "the engineer refined the design" or "the controller adjusted the valve" when that recognizable actor and action are enough. If that ordinary sentence is all the receiver needs, do not open a technical Work account. If B.4 identifies one particular dated U.Work occurrence, first recover every actual performer's A.13 core and independently admit the Work under A.15.1 from its performance history, enacted Method, temporal extent, and containing System. Add F.6 only when the receiving claim also needs precise assignment-bound attribution. A short B.4 account may omit an unused assignment identifier or classification only when every relation it consumes remains recoverable.
The performing System need not be external to the larger holon. For internal adaptation, apply the A.12 reflexive split: identify the changed subsystem or part and the acting subsystem or part as exact, distinct participants, and establish their parthood in the containing holon independently. Use an external System when that is what the case actually has. For any particular dated U.Work, recover every performer's A.13 core and independently admit the occurrence under A.15.1; add F.6 afterward only when precise assignment-bound attribution is current. Name an assignment in the short B.4 account only when the receiving claim uses its identity. State authority, responsibility, permission, acceptance, or admission through its own direct predicate, actual participants, and applicability basis; neither a phase label nor Work supplies them.
Keep the four phases non-overlapping
Use explicit transition conditions:
- Operate -> Observe: a named cue, question, monitoring result, or review need requires interpretation.
- Observe -> Refine: the receiving use has an observed basis or a routed cue from which a change question can be formed.
- Refine -> Deploy: one candidate is selected and its subject-specific identity question or intended treatment is explicit. State an obtaining continuity, successor, substitution, refinement, or edition relation here only when both endpoints already exist; otherwise keep the question open until the relevant change or production has occurred.
- Deploy -> Operate: actual renewed operation or use begins; availability alone does not close this transition.
Evidence is not a fifth phase. Evidence relations warrant the observed basis, candidate choice, transition, or renewed-use claim when a receiver relies on them. Evidence can be produced or used during several phases without duplicating those phases.
Connect neighbouring cycles without collapsing them
The Canonical Reasoning Cycle (B.5) can supply reasoning Work within Observe and Refine. The B.5.1 development states and B.4 phases coordinate, but they are not a one-to-one implementation: a finding may reopen Exploration or Shaping, evidence use can support Evidence, and renewed use can enter or return to Operation.
When Observe finds only a weakly articulated cue, use the optional B.4.1 sequence Notice -> Stabilize -> Route. Its routed result can return to Refine or enter another subject pattern. That sequence does not replace the four B.4 phases.
Keep the account proportional. A local repair can name only the current subject, observed basis, actual Work, resulting identity relation, and next use. Expand to a complete cycle trace when a named relying decision, assurance case, audit, or later replay needs it. Never invent phases merely to make the record look complete.
Didactic Note: four practical questions
- Operate: What exact subject is operating or being used now?
- Observe: What has actual use shown, and through whose Work?
- Refine: What change is being considered, and would it continue or replace the subject?
- Deploy: What deployment Work occurred, and what separately established fact now supports the next use?
The gain is a readable connection from an observed basis to a real change and back to use, without losing subject identity, performers, Work, or evidence.
Archetypal Grounding
The phase names can be shared, but each subject branch keeps its own identity and return-to-use rule.
-
B.4.1 - Observe -> Notice -> Stabilize -> Route (optional pre-abductive route):
- Context: A fleet of autonomous delivery drones (
U.System) is in operation, and operators begin to notice that winter deliveries feel "off" before a clean anomaly statement exists. - Loop Example:
- Operate: The drones perform deliveries.
- Observe: The monitoring service and named operators perform observation Work and find recurring cold-weather battery strain, but the cue still has low articulation.
- Optional B.4.1 route inside Observe: A named team performs stabilization Work. Under
A.16.1, aU.PreArticulationCuePackpreserves the cue nucleus, primary witness traces, and current language-state position without pretending that a final anomaly or action record exists; when the pack is made available for this use, name the separate publication occurrence underE.24.PUB. The same or another team performs routing Work. UnderB.4.1, aRoutedCueSetkeeps multiple continuations visible—for example, battery-chemistry investigation or route-planning adjustment; again, name its publication occurrence underE.24.PUBwhen availability matters. - Continue the loop: The selected route enters Refine or another fitting subject pattern. Only a selected and tested change proceeds to Deploy and renewed drone operation.
- Context: A fleet of autonomous delivery drones (
-
Knowledge-instantiation slice (theory refinement loop):
- Context: A scientific theory of protein folding (
U.Episteme) is used to predict structures. - Loop Example:
- Operate: Named researchers perform theory-application Work using the current theory episteme.
- Observe: A research lab performs observation Work. A separately identified
C.2.1finding episteme states that the current theory fails to predict the structure of a protein class; name separately anyE.24.PUBpublication occurrence that makes this finding available. - Refine: A research team performs revision and testing Work. A later theory episteme, identified under
C.2.1from its changed claim content, includes a term for the new protein class. Assert an edition relation between the two theory epistemes only if that relation obtains. - Deploy: The team performs publication Work for the later theory. The publication occurrence, journal acceptance, admission into a configured knowledge base, and later community use are separate relations. Note. If a chart or CG-frame readings are derived from this episteme, they MUST cite the
MethodDescriptionof the measurement protocol actually used (per A.19.CN CC-A19.D1-3) to keep comparability auditable.
Adaptive-specialization note. When the knowledge-instantiation slice carries a bounded-specialization claim for one declared task family, that claim SHALL name the prior basis being refined from, the named work-measure threshold being pursued, the adaptation budget being spent, and the freshness or provenance basis for claiming the specialization is reusable. If the refinement is claimed as one specialization step, it SHALL also cite the declared
TaskFamilyorTaskSignatureanchor consumed byC.22.1,G.5, andG.9. This keeps the refinement legible as contextual task-family specialization rather than vague general capability growth. - Context: A scientific theory of protein folding (
-
Method-instantiation slice (adaptive method loop):
- Context: A field-maintenance organization uses a declared inspection-and-repair Method (
U.Method) described by one currentU.MethodDescription. - Loop Example:
- Operate: Maintenance teams perform dated maintenance Work that enacts the current Method.
- Observe: A reviewer performs review Work and records that the time from fault detection to safe restoration repeatedly exceeds the allowed window.
- Refine: Method maintainers perform revision and testing Work. A wording clarification can yield a later MethodDescription while the same Method remains current. Adding an earlier isolation action or changing a classification checkpoint can instead change identity-bearing Method semantics; decide under
A.3.1whether the result is a refinement, substitute, or distinct successor Method, and useB.1.5if its composition changes. Then identify the MethodDescription episteme that describes the chosen Method. - Deploy: A named publishing team performs publication or release Work for the later MethodDescription and, where needed, configuration or training Work for renewed Method use. Decision results, authority, acceptance, admission, and later Work that enacts the Method remain separate. Completed maintenance Work is never revised.
Adaptive-specialization note. When the method-instantiation slice carries a bounded-specialization claim for one declared task family, that claim SHALL name the narrower higher-fit specialist method or specialist portfolio being activated, the refinement budget being spent, the escalation or commit checkpoints, and the fallback when that method fails. If the method update is being used as evidence of specialization, the note SHALL keep the bearer of that specialization explicit: the holder, dyad, team, or scoped portfolio carries the claim; the method is only one selected vehicle. This keeps method evolution reviewable as bounded specialist acquisition rather than as hidden budget inflation.
- Context: A field-maintenance organization uses a declared inspection-and-repair Method (
Bias-Annotation
Conformance Checklist
- CC-B4.1 (Proportional loop integrity): Record the actual path from observed basis through change to renewed use. A complete phase-by-phase trace is required only when a named relying decision, assurance case, audit, or replay needs it. Do not invent a phase or claim completion merely to fill a loop.
- CC-B4.2 (Subject identity): Name the exact subject kind and apply its direct continuity rule: System continuity or transformation; episteme predecessor, successor, or edition; Method identity, refinement, substitution, or successor; MethodDescription edition; or separate dated Work occurrences.
- CC-B4.3 (Acting-side distinction): Work that observes or changes a subject MUST have an identified performing System distinct from the changed subject in that Work account. Internal adaptation is permitted when
A.12establishes exact distinct subsystems or parts and their independently obtaining parthood. Every particular datedU.Workreuses each performer's A.13 core and is independently admitted under A.15.1; F.6 is added afterward only for precise assignment-bound attribution. Its short B.4 account may omit an assignment identifier unused by the receiver only when every consumed relation remains recoverable. - CC-B4.4 (Adaptive-specialization anchoring): When the knowledge-instantiation or method-instantiation slice carries a bounded-specialization claim, that claim MUST name the declared
TaskFamilyorTaskSignature, the work-measure threshold target, the adaptation budget, and the freshness or provenance basis for reuse. - CC-B4.5 (Adaptive-specialization boundary): The knowledge-instantiation and method-instantiation slices SHALL NOT silently re-govern selector or parity semantics. If transfer, retention, downstream exploitation efficiency, corridor entry, or downside cost are comparison-relevant, the pattern-local note MUST leave those fields recoverable by the downstream
C.22.1,G.5, andG.9patterns. - CC-B4.6 (Phase and transition separation): Operate, Observe, Refine, and Deploy MUST have the distinct outputs and transition conditions stated in B.4:4.3. Evidence use and optional B.4.1 cue routing do not become duplicate phases.
- CC-B4.7 (No success inference): Deployment or publication establishes neither acceptance nor successful renewed use. Record failure, reopening, fallback, or another iteration when that is what occurred.
Common Anti-Patterns and How to Avoid Them
Consequences
Rationale
This pattern operationalizes the Open-Ended Evolution Principle (P-10) by connecting use, observation, explicit change, and renewed use. It does not supply one generic ontology of evolution. Subject patterns decide identity and continuity; Systems perform dated Work; evidence supports relied-on claims; and B.4 makes their repeated coordination inspectable.
SoTA-Echoing
The phase rhythm has historical lineage in iterative cycles such as Plan-Do-Check-Act and Observe-Orient-Decide-Act. B.4 adapts that lineage rather than treating those labels as an ontology: it distinguishes design-time accounts from run-time use, applies a kind-specific identity rule, identifies actual Systems and dated Work, admits internal acting-side splits, and keeps evidence, deployment, publication, acceptance, and renewed use separate.
The result is a practical review language for repeated adaptation. It avoids both agentless "self-evolution" stories and the opposite mistake of requiring every acting System to be external. Canonical means the four questions recur; it does not mean every subject or project follows one identical history.
Relations
- Operationalizes:
P-10 Open-Ended Evolution. - Uses:
A.4 Temporal Dualityfor design-time/run-time distinctions;A.12for external or reflexively split acting sides;A.15.1for dated Work; and the direct subject patterns named in B.4:4.1 for identity and continuity. - Coordinates with:
B.5 Canonical Reasoning Cycle,B.5.1development states, andB.3 Trust & Assurance Calculus. B.4 does not implement the B.5.1 states one-for-one, and evidence is not a B.4 phase. - Is detailed by:
B.4.1 Observe -> Notice -> Stabilize -> Routefor optional early cue routing, together with B.4.x instantiation patterns for specific subject families.
Pre-abductive seam compatibility
For early language-state routing, Observe does not have to jump directly into anomaly or hypothesis forms. Observe may publish a U.PreArticulationCuePack and a RoutedCueSet through B.4.1; a selected route then enters Refine or another fitting pattern. A downstream loop consumes the routed cue publication directly or a later typed publication such as U.AbductivePrompt, as appropriate.
B.4:End
Last Updated: 2026-09-10 — upstream FPF commit a87d0ef4 (github.com/ailev/FPF)