FPF Reference

Onboarding Glossary (NQD & E/E‑LOG)

Pattern A.0 · Stable Part A - Kernel Architecture Cluster

One‑screen purpose (manager‑first). This pattern gives newcomers a plain‑language starter kit for FPF’s generative engine so they can run a lawful problem‑solving / search loop on day one. It explains the few terms you must publish when you generate, select, and ship portfolios (not single “winners”), and points to the formal anchors you’ll use later. (OEE is a Pillar; NQD/E/E‑LOG are the engine parts.)

Builds on. E.2 (P‑10 Open‑Ended Evolution; P‑2 Didactic Primacy), A.5, C.17–C.19 - Coordinates with. E.7, E.8, E.10; F.17 (UTS); G.5, G.9–G.12 - Constrains. Any pattern/UTS row that describes a generator, selector, or portfolio.

Keywords & queries. novelty, quality‑diversity (NQD), explore/exploit (E/E‑LOG), portfolio (set), illumination map (report‑only telemetry), parity run, comparability, ReferencePlane, CL^plane, ParetoOnly default

Engineer‑managers meeting FPF for the first time need a plain, on‑ramp vocabulary for the framework’s generative engine so they can run an informed problem‑solving/search loop on day one—before formal specifications. Without that, Part G and Part F read as assurance/alignment only, and teams default to single “best” options. This undercuts P‑10 Open‑Ended Evolution and weakens adoption.

Keywords

  • novelty
  • quality-diversity (NQD)
  • explore/exploit (E/E-LOG)
  • portfolio (set)
  • illumination map (report-only telemetry)
  • parity run
  • comparability
  • ReferencePlane
  • CL^plane
  • ParetoOnly default.

Relations

A.0builds onThe Eleven Pillars
A.0builds onOpen-Ended Kernel & Extension Layering
A.0builds onCreativity‑CHR — Characterising Generative Novelty & Value
A.0builds onExplore–Exploit Governor (E/E‑LOG)
A.0coordinates withArchetypal Grounding Principle
A.0coordinates withFPF Authoring Conventions & Style Guide
A.0coordinates withLEX-BUNDLE: Unified Lexical Rules for FPF
A.0coordinates withUnified Term Sheet (UTS)
A.0coordinates withMulti‑Method Dispatcher & MethodFamily Registry
A.0coordinates withParity / Benchmark Harness
A.0coordinates withDHC Dashboards — Discipline-Health Time-Series (lawful telemetry, generation-first)
A.0route stepCG-Spec — Frame Standard & Comparability Governance
A.0explicit referenceThe Eleven Pillars
A.0explicit referenceOpen-Ended Kernel & Extension Layering
A.0explicit referenceCreativity‑CHR — Characterising Generative Novelty & Value
A.0explicit referenceExplore–Exploit Governor (E/E‑LOG)
A.0explicit referenceArchetypal Grounding Principle
A.0explicit referenceFPF Authoring Conventions & Style Guide
A.0explicit referenceLEX-BUNDLE: Unified Lexical Rules for FPF
A.0explicit referenceUnified Term Sheet (UTS)
A.0explicit referenceMulti‑Method Dispatcher & MethodFamily Registry
A.0explicit referenceParity / Benchmark Harness
A.0explicit referenceDHC Dashboards — Discipline-Health Time-Series (lawful telemetry, generation-first)

Content

1) Problem frame

Engineer‑managers meeting FPF for the first time need a plain, on‑ramp vocabulary for the framework’s generative engine so they can run an informed problem‑solving/search loop on day one—before formal specifications. Without that, Part G and Part F read as assurance/alignment only, and teams default to single “best” options. This undercuts P‑10 Open‑Ended Evolution and weakens adoption.

2) Problem

In current practice:

  • Single‑winner bias. Teams look for “the best” option and publish a leaderboard, suppressing coverage & diversity signals essential to search.
  • Metric confusion. “Novelty” and “quality” are used informally; units/scales are omitted; ordinal values are averaged, breaking comparability.
  • Hidden policies. Explore/exploit budgets and governor rules are implicit; results are irreproducible and refresh‑unsafe (no edition/policy pins).
  • Tool lock‑in. Implementation terms (pipelines, file formats) leak into the Core, violating Guard‑Rails.

FPF needs a short, normative glossary that names the generative primitives in Plain register and ties each to its formal anchor—so portfolios, not single scores, become the default publication.

3) Forces

ForceTension
Readability vs RigorOne‑liners for managers ↔ lawful definitions with editions and scale types.
Creativity vs AssuranceOpen‑ended search (OEE/QD) ↔ conformance, parity, and publication discipline.
Comparability vs LocalityShared N‑U‑C‑D terms ↔ context‑local CG‑frames and bridges with CL.
Tool‑agnostic CoreConceptual publication in UTS ↔ engineering teams’ urge to cite specific tools.

4) Solution — Normative onboarding glossary and publication hooks

4.1 Plain one‑liners (normative on‑ramp; formal anchors in C.17–C.19)

TermPlain definition (on‑ramp)See
Novelty (N)*How unlike the known set in your declared CharacteristicSpace. Compute lawfully (declared DescriptorMapRef + DistanceDefRef; no ad‑hoc normalisation).C.17, C.18
Use‑Value (U / ValueGain)*What it helps you achieve now under your CG‑Frame; tie to acceptance/tests; publish units, scale kind, polarity, ReferencePlane.C.17, C.18
Constraint‑Fit (C)Satisfies must‑constraints (Resource/Risk/Ethics); legality via CG‑Spec; unknowns propagate (never coerce to zero).C.18, G.4
Diversity_P (portfolio)*Adds a new niche to the portfolio; measured against the active archive/grid, not a single list; declare ReferencePlane for each head.C.17, C.18
E/E‑LOGNamed, versioned explore↔exploit policy; governs when to widen space vs refine candidates; policy‑id is published.C.19
ReferencePlaneWhere a value lives: world (system), concept (definition), episteme (about a claim). Plane‑crossings add CL^plane (penalties to R only); cite policy‑id.F.9, G.6
Scale Variables (S)The monotone knobs along which improvement is expected (e.g., parameterisation breadth, data exposure, iteration budget, resolution). Declare S for any generator/selector claimed to scale.C.18.1
Scale Elasticity (χ)Qualitative class of improvement when moving along S (e.g., rising, knee, flat in the declared window). Used as a selection lens; numeric laws live in domain contexts.C.18.1
BLP (Bitter‑Lesson Preference)Default policy that prefers general, scale‑amenable methods over domain‑specific heuristics, unless forbidden by deontics or overturned by a scale‑probe.C.19.1, C.24
Iso‑Scale ParityFair comparison across candidates at equalised scale budgets along S; may also include scale‑probes (two points) to test elasticity.G.9, C.18.1

(Registers & forbidden forms per LEX‑BUNDLE; avoid “axis/dimension/validity/process” for measurement and scope.)

4.2 Publication & telemetry duties (where these terms show up)

  1. UTS surface (Part F). When a UTS row describes a generator, selector, or portfolio, it MUST surface N, U, C, Diversity_P, E/E‑LOG policy‑id, ReferencePlane, with units/scale/polarity typed under MM‑CHR / CG‑Spec, and lawful references to DescriptorMapRef/DistanceDefRef. (Row schema: F.17; shipping via G.10.)
  2. Parity & edition pins (Part G). When QD/OEE is in scope, pin DescriptorMapRef.edition and DistanceDefRef.edition (and, where applicable, CharacteristicSpaceRef.edition, TransferRulesRef.edition) and record policy‑id + PathSliceId. Treat illumination/coverage as report‑only telemetry; publish an Illumination Map where G‑kit mandates parity artefacts. Declare S (Scale Variables) and run at least one scale‑probe (two points along S) when claiming scale‑amenability. Dominance policy defaults to ParetoOnly; including illumination in dominance MUST cite a CAL policy‑id.
  3. Tell‑Show‑Show (E.7/E.8). Any arhitectural pattern that claims generative behaviour MUST embed both a U.System and a U.Episteme illustration using this glossary (manager‑first didactics).

4.3 Minimal recipe (run this on day one)

  1. Declare CG‑Frame (what “quality” means; lawful units/scales) and ReferencePlane.
  2. Pick 2–4 Q components + a simple DescriptorMap (≥2 dims) for N/D; publish editions.
  3. Choose an E/E‑LOG policy (explore↔exploit budget); record policy‑id.
  4. Call the selector under G.5 with parity pins; return a set (Pareto/Archive), not a single score.
  5. Publish to UTS + PathIds/PathSliceId; Illumination Map is report‑only telemetry by default.

5) Archetypal Grounding

Informative; manager‑first (E.7/E.8 Tell‑Show‑Show).

Show‑A - SRE capacity plan (selector returns a set). Frame. We must raise service commitment headroom for Q4 without breaking latency SLOs. Portfolio. {cache‑expansion, read‑replicas, query‑shaping, circuit‑breaker tuning, schema‑denorm}. Glossary in action. U = latency@p95 & error‑rate, C = budget ≤ $X, risk ≤ R, N = dissimilarity to current playbook, Diversity_P = adds a previously empty niche in our archive (e.g., “shifts load to edge”). E/E‑LOG starts Explore‑heavy, flips Exploit‑heavy once ≥ K distinct niches are lit. (Publish UTS row + parity pins; illumination stays report‑only telemetry.)

Show‑B - Policy search with QD archive (MAP‑Elites‑class). Frame. Robotics team explores gaits that trade stability vs energy use. Glossary in action. CharacteristicSpace = {step‑frequency, lateral‑stability}, ArchiveConfig = CVT grid, N from descriptor distance, U = task reward, Diversity_P = coverage gain; PortfolioMode=Archive. Families include MAP‑Elites (2015), CMA‑ME/MAE (2020–), Differentiable QD/MEGA (2022–), QDax (2024); publish editions and policy‑ids; treat illumination as report‑only telemetry.

(Optional) Show‑C - OEE parity (POET/Enhanced‑POET). Co‑evolve {environment, method} portfolios; publish coverage/regret as telemetry metrics; pin TransferRulesRef.edition; return sets, not a single winner.

Show‑Epi - Evidence synthesis (U.Episteme). Frame. A living review compares rival causal identification methods (e.g., IV vs. DiD vs. RCT‑adjacent surrogates) across policy domains. Glossary in action. U = external‑validity gain @ F/G‑declared lanes, C = ethics & data‑licence constraints, N = dissimilarity in **ClaimGraph** transformations, D_P = coverage of identification niches in the archive. ReferencePlane = episteme. Illumination/coverage stays report‑only telemetry; selection returns a portfolio of methods per niche. (Publish UTS rows; cite Bridges + CL for cross‑domain reuse; edition‑pin Descriptor/Distance defs where QD applies.)

6) Bias‑Annotation

Scope. Trans‑disciplinary; glossary applies to both System and Episteme work. Known risks & mitigations. Over‑aggregation: forbid mixed‑scale sums; use CG‑frame and MM‑CHR. Terminology drift: enforce LEX‑BUNDLE registers; ban tool jargon in Core. Optimization monoculture: require portfolio publication where G‑kit mandates parity; illumination stays report‑only telemetry unless a CAL policy promotes it (policy‑id cited).

7) Conformance Checklist (SCR/RSCR stubs)

IDRequirementPurpose
CC‑A0‑1If a pattern/UTS row describes a generator, selector, or portfolio, it MUST surface N, U, C, Diversity_P, ReferencePlane, and E/E‑LOG policy‑id; units/scale/polarity MUST be declared.Makes generative claims comparable and auditable (UTS as publication surface).
CC‑A0‑2When QD/OEE is in scope, pin editions: DescriptorMapRef.edition, DistanceDefRef.edition (and, where applicable, CharacteristicSpaceRef.edition, TransferRulesRef.edition); log PathSliceId and policy‑ids.Enables lawful parity/refresh; edition‑aware telemetry.
CC‑A0‑3No mixed‑scale roll‑ups; ordinal data SHALL NOT be averaged; any roll‑up MUST live under a declared CG‑frame.Prevents illegal scoring; keeps comparisons lawful.
CC‑A0‑4Where the G‑kit requires parity, publish an Illumination Map (coverage per niche); single‑number leaderboards are non‑conformant on the Core surface when a ParityReport is required.Portfolio‑first publication; avoids single‑winner bias.
CC‑A0‑5Keep illumination/coverage as report‑only telemetry; dominance policy defaults to ParetoOnly; any change is CAL‑authorised and cited by policy‑id.Separates fit from exploration; preserves auditability.
CC‑A0‑6Apply E.7/E.8: include a U.System and a U.Episteme illustration when claiming generative behaviour; obey E.10 register hygiene; use the exact subsection title “Archetypal Grounding.”Locks didactic primacy; prevents jargon drift.
CC-A0-7ReferencePlane declared for every N/U/C/Diversity_P head and CL^plane penalties route to R only; Φ_plane policy-id published when planes differ.Prevents plane/stance category errors; aligns with Bridge/GateCrossing visibility guards (Bridge+UTS+CL/Φ_plane).
CC‑A0‑8Diversity_P ≠ Illumination. Diversity_P may enter dominance; Illumination remains report‑only telemetry unless explicitly promoted by CAL policy‑id.Matches QD triad semantics and parity defaults.
CC‑A0‑9If a generator/selector is claimed scale‑amenable, declare S (Scale Variables) and an E/E‑LOG scale policy‑id; otherwise mark S = N/A.Makes scale assumptions explicit and comparable across contexts.
CC‑A0‑10For scale‑amenable claims, execute a scale‑probe (≥ 2 points along S) and report a Scale Elasticity class (rising/knee/flat) in the UTS row.Forces early strategy‑relevant evidence without over‑specifying numerics.
CC‑A0‑11Apply Iso‑Scale Parity in parity runs when S is declared; where infeasible, state the loss notes and treat results as non‑parity with an explicit penalty in R.Keeps comparisons fair and auditable under scale constraints.
CC‑A0‑12BLP default. If a domain‑specific heuristic is selected over a general, scale‑amenable method, record a BLP‑waiver reason: deontic, scale‑probe overturn, or context‑specific.Prevents silent violations of the Bitter Lesson; improves selector transparency.

8) Consequences

Benefits. • Immediate usability for engineer‑managers (plain one‑liners) with formal anchors for auditors. • Portfolio‑first culture (sets & illumination) instead of brittle leaderboards. • Edition‑aware comparability; parity/refresh is routine, not ad‑hoc.

Trade‑offs & mitigations. • Slightly longer UTS rows → mitigated by consistent schema and copy‑paste snippets. • Requires discipline on units/scales → mitigated by CG‑frame templates.

9) Rationale

This pattern instantiates P‑10 Open‑Ended Evolution by making generation‑selection‑publication operational at the on‑ramp: readers get just enough shared vocabulary to run search as standard practice. It aligns with Didactic Primacy (P‑2) and LEX‑BUNDLE (E.10) by keeping definitions plain‑first and scale‑lawful, and with Patterns Layering (P‑5) by pointing to C.17–C.19 for formal anchors without tool lock‑in. The post‑2015 line (MAP‑Elites → CMA‑ME/MAE → Differentiable QD/MEGA → QDax; POET/Enhanced‑POET/Darwinian Goedel Machine) normalised quality‑diversity and open‑endedness as first‑class search objectives; this glossary surfaces those ideas as publication standards, not tool recipes.

10) Relations

Builds on. E.2 Pillars (P-10, P-2, P-6), A.5 (Open-Ended Kernel), B.5/B.5.2.1 (Abductive loops + NQD integration), C.17–C.19 (Creativity-CHR, NQD-CAL, E/E-LOG).

Coordinates with. E.7/E.8 (Archetypal Grounding; Authoring template), E.10 (LEX‑BUNDLE), F.17 (UTS), G.5/G.9–G.12 (set‑returning selectors, iso‑scale parity, shipping & refresh). Constrains. Any generator/selector/portfolio publication on the Core surface: N‑U‑C‑Diversity_P + policy‑ids; S/Scale‑probe where applicable; parity pins; lawful scales; portfolio‑first where mandated. (Ties into UTS rows and parity artefacts.) Editor’s cross‑reference. For agentic orchestration of scalable tool‑calls under BLP/SLL, see C.24 (Agent‑Tools‑CAL).

Editor’s note (implementation hint)

This pattern is an on‑ramp: it does not replace C.17–C.19. It binds Plain definitions to publication/telemetry expectations so newcomers can use NQD/E/E‑LOG immediately while experts follow the formal trails.

A.0:End