WM-12: Make horizon and hierarchy describe the decision being made

First authored design, 2026-09-15. This elaborates exactly WM-12. It is a reviewable design, not implementation, independent acceptance or machine completion. The existing build loop continues under claude-2; these documents do not dispatch new jobs or override recorded decisions.

All WM cascades · Checklist item · Machine-readable cascade

The problem and existing evidence

WM-12 governs the comparison semantics shared by model, G and execution. Changing pattern count must not secretly change the number of scored observations. Shared-horizon mathematics is useful but does not establish the machine’s anticipation source, hierarchy or interpretation correspondence. The existing effective-horizon and FPI-consumer obligations remain open.

Governing checklist snapshot

WM-12 — Horizon, interaction, hierarchy and interpretation mean what the specification says. OPEN: legitimate current anticipation, effective horizon >=2 and FPI consumer evidence; fair common observation schedule; order/overlap and nested interpretation correspondence. No policy-length penalty from raw sums over different horizons. Declared extrapolations and constructed arithmetic are not established world dynamics. Preserve pending paired fairness/composition, nonfiring-versus-progressing and concrete compounding-misspecification controls and evidence-bound interpretation admission. Shared-T proofs alone do not settle the later Alexander formulation; universal worsening and metadata-only ordering are unproved proposals, not adopted facts. [S,H,T] Built evidence: Shared-horizon mathematics and scoped composition/independence examples exist.

The preceding evidence is the checklist’s dated account. It has not been re-audited as a current deployment census for this design.

Pattern applications

WM-12-C1 — Separate three clocks

Pattern: wm-pattern-design/separate-design-from-use.

Declare construction stage, prediction/observation time and the grain of each nested decision. A new authoring step is not automatically an extra scored consequence. Output a schedule declaration tied to the actual comparison, not a universal episode-step convention.

WM-12-C2 — Ground anticipation and comparison context

Pattern: aif/declare-the-conditioning.

Identify actual current anticipation inputs, their timestamps, effective horizon and the consumer of that configuration. Comparable candidates need a justified shared observation/preference opportunity. Record horizon >=2 and FPI consumer evidence where the checklist requires them; a flag or proposed setting is not an effective execution record.

WM-12-C3 — Align nested execution semantics

Pattern: wm-connected-learning/history-and-guard-information-are-state.

Declare history, available guard information, precedence, repetition and termination at each level. A lower-level episode and a higher-level step need an explicit relation; hidden-state guard evaluation remains a stated approximation unless correspondence is established. Do not reset board state between steps merely to simplify an example.

WM-12-C4 — Test fairness and interaction together

Pattern: wm-pattern-design/composition-changes-the-problem.

Compare order/overlap effects at the declared schedule. No-interaction controls require the relevant per-step observable and ambiguity invariance; commuting transitions or terminal equality alone do not prove equal cumulative G. Non-firing outcomes must preserve an unfinished target rather than earn an automatic reward for inactivity.

WM-12-C5 — Preserve progressive evaluation

Pattern: wm-pattern-design/partial-design-without-imaginary-completion.

Require initial, intermediate and completed designs to have their own grounded prediction semantics under WM-10. Retain which continuations are modeled, what stopping means, and what search coverage was examined. Equal or worse proposals do not justify extension; local stopping is not enactment authorization.

WM-12-C6 — Retain the effective semantic configuration

Pattern: wm-qualifying-run/certificate-and-world-must-meet.

Bind the enacted schedule, hierarchy, model and actual consumer to its receipt. Include controls for concrete compounding misspecification, cancellation/saturation and metadata-only attestation twins. Reject universal inequalities unsupported by the modeled dynamics. Mathematical examples and runtime configuration observations remain different evidence.

Composition and revision

These are design input/constraint relationships; they do not assert automatic runtime firing. Each contribution constrains the next actual artifact.

Build interfaces

Grain/time declaration → schedule. Anticipation reader → current inputs and effective configuration. Hierarchy adapter → nested state/observation relation. Comparison validator → fair context or explicit incompatibility. Execution receipt → observed consumer settings and semantic identities. No new episode, fuel or yield law is selected here.

Each envelope must retain its schema, content identity, source/model revisions, task/occurrence and information boundary. Typed missing inputs remain missing; do not replace them with success-shaped defaults. Implementers may reuse existing compatible components, but must identify the real producer and consumer before crediting integration.

Worked situation

Adding a pattern that lets the receiver interpret an existing evidence bundle may improve the handoff without lengthening the observation window. Counting an extra adverse observation simply because the design now has another node would manufacture a length penalty. Conversely a genuinely longer observation horizon can change predicted outcomes; fairness must not erase that real difference.

This situation is illustrative, not a measured production result or an adopted probability table.

Build order and acceptance

1. Resolve the first two applications against the real work and exact authority. Preserve the unresolved decisions below; bell design questions before choosing dependent behavior.
2. Implement the middle producer/checker interfaces with explicit source and numeric contracts. Retain a meaningful failing example before repair and independent review after it.
3. Connect the final consumer and collect the particular evidence described above. Unit tests, source acceptance and ordinary use receive separate credit.

Distinguishing controls: Reject T set mechanically to construction length, invented zero-cost padding, hidden-state guards unavailable to the executor and stale anticipation. Compare aligned dynamics, interacting orders, stalled versus progressing designs and cancellation of errors. Verify actual effective horizon/FPI consumption separately from settings files or unit stubs.

Do not rerun the entire machine merely to validate a local edit. Select checks for the changed invariant; retain commits, commands and results. A future qualifying run still requires every required node and connection.

Dependencies and open decisions

WM-02 · WM-03 · WM-05 · WM-06 · WM-09 · WM-10 · WM-11 · WM-13

Open: nested episode interpretation, fuel units/objective, mid-run board changes and yield pricing where still unresolved. The current Stage-0 source replaces the earlier global formula fork. The design preserves the no-structural-length-penalty ruling without asserting universal length neutrality or universal worsening under composition.

The dependency links express interface constraints, not a topological build order. A producer and consumer can jointly constrain an interface without requiring that all other implementations be finished first. Use a bounded packet once its relevant semantics are settled.

Sources and review boundary

The companion JSON pins the governing checklist, September rulings, current cascade-policy semantics and every applied pattern. Where the semantics source distinguishes a ruling, extrapolation or open proposal, that distinction controls. No generic finite witness is promoted to a machine instance. Existing patterns are reused within the application stated above; their other examples, thresholds and historical implementation claims are not automatically adopted.

Existing AIF reference snapshots

Retained source text, not new patterns or adopted historical claims.

@flexiarg aif/declare-the-conditioning
@title Declare the Conditioning (Score the Fit, Not the Form)
@keywords conditioning, situation, context, fit, EFE, curried, referent
@audience aif implementers, pattern authors, futon developers
@tone technical
@style pattern
@references [aif/currency-before-merge aif/valuation-reads-the-paperwork]

@how aif/structured-observation-vector
@see-also aif/term-to-channel-traceability
@why no matching problem dossier node (@holds-at absent; not among the twenty dossier nodes R1-R17, R20, TRACE). Negative claim per the committed check futon2 holes labs library-loop runs l3_no_dossier_check.py, receipt runs L3-no-dossier-check.edn (enumerates every aif pattern's @holds-at against the dossier node set extracted from PROBLEMS-*.md headings). Rationale not invented (L3 rationale-backfill pilot, zai-1, 2026-09-05)
@how own mechanism: Every valuation surface must declare what slice of the situation it consumed; G(candidate) is a lambda — only G(candidate | situation) is a value. (source: this pattern's ! conclusion line; L3 rationale-backfill pilot, zai-1, 2026-09-05)
! conclusion: Every valuation surface must declare what slice of the situation it consumed; G(candidate) is a lambda — only G(candidate | situation) is a value.

  + context: A stack grows multiple valuation surfaces (blend terms, wholeness, ΔG legs, field diagrams), each labelled with the same vocabulary ("EFE", "value", "score").

  + IF:
    You want valuations to be comparable, auditable, and honestly labelled.

  + HOWEVER:
    Different surfaces silently consume different slices of the world — one eats the live observation vector, another a 160-char prose scrap, a third nothing at all — so identical labels name incomparable objects, and cross-surface comparisons are apples-to-oranges without anyone having lied.

  + THEN:
    Give every valuation surface a three-axis declaration: SENSE (which formal object it claims), SELECTION-SEMANTICS (extremum / amortized sample / bound — over what menu), and CONDITIONING (what situation-slice it consumed). Render the declaration where the number is rendered. In pattern terms: Context → (Problem → Solution) — application is modus ponens; the world supplies the antecedent; score the fit, never the form alone.

  + BECAUSE:
    Canonical expected free energy is G(π; Q(s), C) — conditioned on beliefs and preferences by definition; a policy has no value in vacuo, and Alexander says the same of form and context. Surfaces that skip the declaration re-invent the mislabeling one level up.

    + evidence: M-evaluate-policies §9.6–9.7 + exhibit "conditions on" column (futon2, 2026-07-03): the atomic-action lane conditioned on the live observation vector while the cascade lane ate a title+status scrap and two native components consumed no situation at all — one arena, incomparable conditioning. E5 measured the conditioning channel: situation-enriched ψ rebuilt 11/13 cascades. The fold contract already typed circumstance as an argument; the plumbing had starved it.

  + NEXT-STEPS:
    - Add the sense × selection-semantics × conditioning triple to each EFE-labelled surface's declaration (extends the efe-referent-declaration shape).
    - Route all lanes toward ONE situation source (the structured observation vector) where feasible.