PLOT
What the public package proves.
These are checks on Skua Strait, a Dead Index reference theater. They are not Navy validation. Every row is computed from the same engine the desk uses, on the frozen packs, with n = 2000 and seed = 20260911 for the Monte Carlo rows.
21 / 21 pass
| Measure | Value | Note |
|---|---|---|
| Reference systemP1 | Skua Straitpass | Synthetic Dead Index reference theater. Not operational Navy data. |
| Ingest packs · CAPP2 | 3 / 3 reconstruct · one hash 8fa9fba14365pass | A-GRA XML, near-A-GRA log, platform JSON reconstruct SKUA-CAP-01 to the same mission hash as the frozen catalog. |
| Ingest packs · StrikeP9 | 3 / 3 reconstruct · one hash 55ec8c59a829pass | Same three shapes reconstruct SKUA-STK-01 to one hash. |
| Normalization scoreN1 | agra 1.00 · near-agra 0.74 · platform 0.82pass | (present + mapped + ½ guessed) / 19 canonical fields. Same per shape on both stories. |
| Field mapN2 | agra 18/1/0/0 · near-agra 8/5/2/4 · platform 1/13/3/2pass | present / mapped / guessed / missing. Missing fields are filled from the frozen catalog and say so. |
| Conversions and re-based timestampsN3 | agra 0 conv · 0 rebased · near-agra 190 conv · 97 rebased · platform 192 conv · 0 rebasedpass | Feet to km, metres to km, milliseconds to seconds, epoch to mission time. Every conversion is a row in the report. |
| Blue remaining at recoverP3 | 2pass | Four launched. Two do not recover. |
| Red readings per storyR1 | 2 · 2 · all < 1pass | Neither reading is drawn as fact. Two readings on CAP, two on Strike. |
| Traced confidenceR2 | H-A (6 + 2) / (6 + 3 + 4) = 0.62 · H-B (5 + 2) / (5 + 8 + 4) = 0.41pass | (F + 2) / (F + A + 4) over observations that hold on the pack. Recomputed equals displayed equals gold, both stories. |
| Observations evaluatedR3 | 22 / 22 hold · 13 / 13 holdpass | Each observation is a predicate over track ids, events, and times, with its numbers shown. |
| Timeline what-ifP5 | Blue move updates redpass | A station change at 08:00 recomputes both sides. Batteries stay. |
| Courses of actionP6 | 2 immediate + 1 later · both storiespass | Preferred pairing plus a backup that assumes pairing is blocked. Retrain is labeled later and scored as no change to the next launch. |
| Monte Carlo determinismM1 | n 2000 · seed 20260911 · e68f3e4d98aa · 5c7579e15467pass | Two runs from the frozen packs produce the same mcHash on both stories, and the hash equals the gold. |
| Median orderingM2 | COA-1 > COA-2 > COA-3 · COA-1 > COA-2 > COA-3pass | Courses ordered by median utility match the gold ranking on both stories. |
| Probability of bestM3 | COA-1 0.997 · 0.947 · sums 1.000000 · 1.000000pass | Share of sampled worlds in which each course has the highest utility. Sums to 1 within 1e-9. |
| Survivability interval · CAP COA-1M4 | p10 0.55 · p50 0.62 · p90 0.70pass | Interval overlaps COA-2, COA-3; the desk flags it. |
| Baseline geometryP7 | surv 0.59 · effect 0.42pass | Deterministic from tracks and class cards at the mark. Not a model API. |
| Export brief + A-GRA-shaped packX1 | deterministic · 1fd699308f10pass | Same approved desk → same brief and pack bytes. Pack carries COA and HYP messages. |
| Algorithm selection matrixA1 | 4 areas · 19 families · 34 works · 0 danglingpass | One selected family per area, scored 0–2 on five criteria. Every citation resolves to a listed work. |
| Ingest to Approve · controlled runT1 | 00:05 · Chrome 153.0.8010.36 headless, Darwin arm64pass | 2026-09-11 on localhost:3100, 1440×900, SKUA-CAP-01 from agra: load pack, jump to the loss, read both readings, pick a course, Approve. Deck clock 00:00:04. |
| StandaloneP8 | In-browserpass | No inference URL. Packs ship with the page. Monte Carlo runs in the tab. |