ASSAY component gallery

Live components — the demonstrator's actual code, rendered over Meridian vignette fixture data (SPEC-04). Components, not product: no service, no compute, no plan exists behind this page. Every assessed value is banded with provenance welded on (ASSAY-DEC-9); the Meridian scenario is engineered fiction (ASSAY-DEC-8).

The knowledge base, K1–K14

Minimal S1 knowledge table (SPEC-05)

The knowledge service's four discipline moments, rendered by the actual firewall over the same fixtures: the refused K10 save (banner), the contested K12 pair (compile-blocking flag), K8's waiver and single-source markings, and confidence-lint cautions. Banded honesty holds — no bare assessed scalar appears (G2).

Compile — knowledge → CompiledWorld (SPEC-06)

The Stage-2 demo moment, run by the actual compile service. Contested never compiles (G5): with the K12 mine-stock pair disputed, the compile refuses rather than average an irreconcilable band into a channel —

Resolve the contest, and the recompile builds the world. Channels are stored sparse — a per-channel default plus named region overrides (the MCOO idiom; no dense 1.2M-cell surface is ever stored or hashed). Every channel value walks back to the named knowledge it came from, with its owner (G3). stamp 4fa981bd2093ed28… · 9 knowledge objects consumed · sparse channels (no dense cell stored)

Score & organise — the honest matrix (SPEC-07 · SPEC-08)

The Stage-3 demo moment, run end-to-end by the actual machinery: /plan/handful (SPEC-08) fans out over Meridian's four axes — approach, suppression, causeway, extraction — scores every candidate with the SPEC-07 scorer, and organises by banded non-domination into the genuinely distinct plans below. No author picked the set: a plan is kept only when no other conservatively beats it on every commitment (overlapping bands are honestly incomparable and both survive), and dropped only when strictly beaten everywhere — never on a scalar total (DEC-19). Each plan's reason is derived from the organiser, not captioned:

The same handful, as the S2 planner matrix. Verdicts are a four-stop scale — robust / marginal / tight / violated, one colour language, no decimals anywhere; the banded margin rides on hover, never a headline number (G2). Each verdict propagates its metric by interval arithmetic — worst-and-best-case on pure bands, no midpoint (DEC-15) — validated by the vignette §9 oracle cases (O-1–O-3 exact, O-4 containment). Pick a verdict, walk it to the assessment and owner it rests on: every cell was written with a scored_from edge back to the world, whose channels trace to named knowledge above. handful stamp 99ca45289c4582da… · 4 genuinely distinct plans, organised not authored · same stamp + seed ⇒ identical handful (G1) · every verdict carries a scored_from edge to the world (G3)

Relax — least-worst, never silence (SPEC-09)

The Stage-4 demo moment, run end-to-end by the actual machinery. Against the R3m world (both approaches mined, the causeway dropped) no plan satisfies C2, C3 and C4 together — by construction (vignette §6). /relax does not pick one compromise and hide the trade: it returns every inclusion-minimal way forward, each naming what it gives up in command language. The sacrifices are computed — a commitment is sacrificed only when the SPEC-07 scorer returns violated for it — not authored. Ranked least-worst first by the commander's ordinal tiers (must / should / prefer, no numeric weight — DEC-19): the two should-sacrifices lead, the must-sacrifice (C2, "the strait opens two days late") is ranked last but present — never dropped, never silent (G4). Plans that give up more — {C2,C4}, {C3,C4} — are inclusion-dominated and never shown as separate options. relax stamp 69fd05662fc5b616… · 3 least-worst candidates over R3m, computed not authored · never empty, never a silent drop (G4) · tie-break stated (DEC-19) · each candidate cited_in the report back to named knowledge (G3)

Scenario robustness — don't plan on most-likely (SPEC-10)

The Stage-5 demo moment (thesis C; JP 2-01.3: "don't plan on most-likely"). The generated handful scored across BASE, R1, R2, R3 — the adversary COA set from the vignette. Each cell is a four-stop verdict chip; a ▼ marks a verdict that dropped from its BASE value under that scenario ("collapse"). The worst column is the minimax: the plan's worst verdict for that commitment across all scenarios — a real verdict on a real scenario, not a weighted blend (DEC-15/19). Toggle R2 (Strait Denial) and watch the strait-early plan die on C1/C2 while robust alternatives hold. robustness stamp 704866e7a49d8d33… · 4 scenarios × 4 plans × 6 commitments · worst-case (minimax) verdict per plan×commitment · stamps_compatible=true

Sensitivity — which beliefs bear load (SPEC-11)

The Stage-6 sensitivity demo moment (thesis E). Band-edge perturbation of each consumed knowledge object against the first handful plan: push each answer to its lo and hi edges, re-score, and count how many commitment verdicts change. K8 tops the ranking with single_source: true — its band-edge perturbation changes C4 verdicts via the battery fire-control threshold. The single_source flag is carried through from provenance, never used in the ranking arithmetic (DEC-19) — shown alongside so the commander sees the collection risk without it being laundered into a scalar. sensitivity stamp 6216affdcb7cef23… · 9 knowledge objects perturbed · ranked by verdict-change count · single_source carried, not used in ranking (DEC-19)

Discrimination — which questions separate COAs (SPEC-12)

The Stage-6 discrimination demo moment (thesis D). COA-pair band separation over open questions' expected-answer miniatures (DEC-18): for each open question with expected_answers, measure how well the answer bands separate each pair of adversary COAs. K11 ranks above K13 despite higher collection cost — the R1/R2 expected-answer bands are disjoint (mines loaded vs not), while K13's radio-traffic bands overlap across all COAs. Cost is shown alongside, never collapsed with value (DEC-19): the commander sees both and decides. No Shannon entropy, no VOI, no scenario weights (research note 08-analysis.md §3). discrimination stamp 4a01e2f5832ecd4f… · 2 open questions ranked · COA-pair separation over expected-answer bands (DEC-18) · cost alongside, never collapsed (DEC-19)

Staleness — what goes stale when knowledge changes (SPEC-13)

The Stage-6 staleness demo moment (thesis F). Transitive forward trace walk from K9 (the storm-window knowledge that superseded K5): follow every downstream edge — consumed_by, scored_from, compiled_into — using the EDGE_ORIENTATION map from traceView.ts. The walk flags exactly the dependent verdicts and nothing else. It does not recompute — the constitution says flags, then humans decide. staleness stamp 0d19d6fa5868c966… · 98 verdicts + 1 worlds invalidated · transitive walk from K9 · flags only, no recompute (constitution)

Generated from fixtures/ by npm run gallery · identifiers frozen per assay-vignette.md §8 · compile per research note docs/research/02-compile.md · score per docs/research/03-score-plan.md · relax per docs/research/04-relaxation.md · robustness per docs/research/06-robustness.md · analysis per docs/research/08-analysis.md