GENERAL NOTES

The sprite decorates; the census governs. Every block is sheet 7's, unchanged: footprint 1.6·√sloc, height 3 px per authored file, spec annexes beside their buildings, gate severity in the same colours with the same uniform hatch including the cap. The Working Plant sprite (concept 3 of the sprite studies) adds four state channels as overlays. The design guard from the study is enforced: rust is a dotted speckle at partial opacity on the flanks only — never the cap, never a 45° line hatch — so a red-gated pristine plant (uniform hatch, cap included) and a rusting never-gating plant cannot be confused, in either theme.

Every channel is measured, and every threshold comes from a distribution. RUST is sheet 13's weathering census, read from www/atlas.lit-ui-router.dev/data/census-weather.json at origin/main @ 3f55e419: the member's median idle days — the very number sheet 13's schedule prints as "idle Nd" — stepped by a ladder this build cuts for itself. The rust ladder is not written down: it is re-cut from the idle distribution every cabinet, so a step means only what this plate’s own key says it means. The rule is the plate's stated method, mechanised: sort the distinct readings, measure the joints between them, and cut at the widest. The top step is reserved for a true outlier and opens only when the topmost joint is also the widest in the distribution; otherwise R4 is pinned above the highest reading and stands EMPTY. At this cabinet it opens: the topmost joint, 42d to 76d, is the widest in the distribution. The ladder cuts 0 ≤3d · R1 ≤10 · R2 ≤21 · R3 ≤42 · R4 >42, the deepest rust in the city is R4 (@tools/eslint-ts-parser), 4 of the 51 dated plants stand clean at 0, and @tools/eslint-ts-parser wears the cracked flanks the key draws. A step label therefore means nothing across cabinets — the cuts move with the distribution, which is why they are printed in the key rather than remembered. STEAM is distinct commits touching the member in a trailing 90-day window, read from www/atlas.lit-ui-router.dev/data/census-steam.json — window 2026-08-02..2026-10-30, counted at origin/main @ 3f55e419 (2026-10-31) — banded 0 puffs ≤2 · 1: 3–8 · 2: 9–15 · 3: ≥16. Those edges are editorial: 3, 9 and 16 are all occupied on this window, so they sit in traffic rather than in empty air. 19 plants steam at three puffs: lit-ui-router (56), docs (47), @tools/repo-checks (47), lit-ui-router-ssr (46), sample-app-lit-e2e (43), ui-router-server (37), sample-app-shared (36), @tools/release (35), examples (32), @tools/build_and_test (30), @tools/workers-builds (28), @tools/shared (25), lit-ui-router-effect (22), eslint-plugin-lit-ui-router (19), sample-app-lit-mobx (17), @tools/compat-guards (17), lit-ui-router-mobx (16), sample-app-lit-vanilla (16) and @tools/dts-backtest (16). LAMPS compress plate 7A's meter to one number — lit share = extent × line coverage — read from www/atlas.lit-ui-router.dev/data/census-shadow.json, metered at origin/main @ 3f55e419: three lamps at ≥90, two at ≥50, one above zero, and the accent lamp is 7A's honest category for light no meter reads. PIPES are the turbo run build graph from www/atlas.lit-ui-router.dev/data/census-plate.json: 47 real tasks in 187 nodes, last run green on 2026-08-17 (all cache hits — a replay of green, stated as such), so every pipe on the sheet connects and the key says so rather than inventing a broken one.

The channels disagree, which is the point. The channels are independent on purpose, and the city proves they must be: the flagship stands at rust R1 under full steam and every lamp, and the deepest rust in the city, R4, stands on @tools/eslint-ts-parser, none of them steaming. A single wreck-to-splendor axis would have to average these stories away. lit-ui-router is the oldest masonry in the city and its hottest steam and fully lamped — old and running. The typedoc plugin rusts at R2 on 2 files and still steams, because index.ts takes the commits while kind-indexes.ts, the other half of the plugin, idles its 21 days. examples steams at 3 puffs with 1 lamp (1% lit) at rust R2 — worked on, barely tested, barely aging — while docs steams at 3 puffs under a dim metered light in the city (54.9% lit). And @tools/happy-dom keeps plate 7A's oddest light: a plant whose spec annex is a canary pointed upstream, lighting only the workaround it guards (42.9% lit).

Every channel by import, and the ladder cuts itself. Placements, districts and gate tiers are imported from sheet 7's own placement table, and the masses from www/atlas.lit-ui-router.dev/data/census-city.json, so the two sheets cannot drift building for building. Rust, steam, lamps and pipes are each looked up by member directory in their own filed plate, and a member this sheet draws that a plate does not carry is a build error rather than a stale number. Rust was the last channel drawn by hand — an editorial step per member, keyed by badge — and it is not any more: the step is the ladder above applied to the weather plate's own medIdle, so 7B and sheet 13 cannot disagree about how long a member has stood idle, and a member with no dated source (@tools/wintercg-globals) carries no step rather than a guessed zero. The steam total — 763 member-touches from 801 window commits — double-counts commits touching several members, as any per-member count must, so the window commit count is printed beside it.