/* ============================================================================ HEARTH pairing P2A — "Warm paper, cool signals" · editorial · quiet APPLIED TO THE PUBLIC SITE ---------------------------------------------------------------------------- Owner call 2026-08-30bi (tracker §13 row 85 · study record §6a-xxvii). Source of every value below: ui_kits/Helix-Counter-Brand-And-Surfaces-Handoff/pairings/tokens/P2A-warm-paper-cool-signals.css Values are TRANSCRIBED from that sheet, not re-derived. Do not "improve" one here — change the study sheet and re-transcribe, or the two silently diverge. ✅ THE TYPE HALF IS NOW APPLIED (2026-08-30bq, §13 row 88 leg ① — owner: "go ahead and vendor the fonts"). **Newsreader** and **Hanken Grotesk** are vendored as SIL OFL 1.1 latin-subset variable woff2 in `fonts/`, declared in the root `fonts.css`, so this sheet sets `--font-display` and `--font-sans` from the study spec. ⛔ The Google-Fonts `@import` in the study sheet is STILL not copied and never will be — the repo bans a remote font outright (check.sh guard 1); vendoring is what made the token settable. ⚠ `--font-mono` is deliberately NOT set: P2A names JetBrains Mono, which IS the canon's mono already, and the canon's fallback stack is richer than the study sheet's. Setting it would narrow the stack for no gain. ⚠ `--font-display-weight: 400` from the study sheet is also NOT set — measured, that token has ZERO consumers anywhere in the tree, so it would be a declaration nothing reads. ⛔ THE 9 `--signal-*` TOKENS ARE DELIBERATELY OMITTED. Measured 2026-08-30bi: ZERO consumers across ui_kits/public-site, and the canon does not define them. The site already carries that layer as `--sev-*` / `--status-*` — 48 consumer references here — which P2A correctly leaves alone ("severity and status keep their own hues in every direction"). Their values live in the study sheet if that layer is ever wired. ⚠ NOTE THE ONE REAL LOSS: P2A's stated character is that cool signals "free clay to mean failure", and its `--signal-failure: #A94E2E` IS clay. That re-assignment lands only when the signal layer is wired; today failure still reads from `--sev-*`. ⚠ GROUND CHANGE, CHECKED: canon canvas #F6F1E7 → #FAF8F4 (a cooler, lighter paper). All eight severity/status inks re-measured against the new grounds and all pass — 5.01 → 7.41 on canvas, 5.31 → 7.41 on surface, none below AA 4.5:1. The site's own text inks as APPLIED: 16.87 (text) / 6.97 (secondary) on canvas, and the canon's muted #6D6355 at 5.55 — see the muted note in the block below, the one token that deviates. Guarded from here on — this file is in `hearth-contrast-audit.py`'s APPLIED_PAIRINGS set, which re-checks every canon severity ink against the grounds this sheet moved, so a future edit that breaks a pair fails check.sh instead of shipping. ⚠ DARK THEME IS NOT ADDRESSED AND DOES NOT NEED TO BE. All 17 public-site pages are hard-coded `` (site.js:54 says so explicitly), so the canon's `:root[data-theme="dark"]` block and its `prefers-color-scheme` twin (guarded `:root:not([data-theme="light"])`) never match. If the site ever gains a dark mode, `:root[data-theme="dark"]` outranks this file's plain `:root` and it falls back to the canon dark — deliberately. P2A's own dark counterpart is **P2C · Ink base, signals only**, in the same study directory. ⚠ NOT LINKED FROM `motion-lab.html` — that page deliberately links no external stylesheet and is self-contained by design. 16 of the 17 pages carry this. LOAD ORDER MATTERS: must come AFTER colors_and_type.css, which redefines 14 of these token names. The pages link it after hearth-components.css; site.css / site-v2.css / site-motion.css define no tokens of their own (measured: 0), so nothing downstream re-overrides these. ============================================================================ */ :root { /* ---- ground + ink: warm paper ---------------------------------------- */ --hearth-canvas: #FAF8F4; --hearth-surface: #FFFFFF; --hearth-raised: #FFFFFF; --hearth-sunken: #F1EEE8; --hearth-border: #E7E3DB; --hearth-border-strong: color-mix(in srgb, #191713 14%, #FAF8F4); /* = #DAD8D4 */ --hearth-border-subtle: color-mix(in srgb, #191713 6%, #FAF8F4); /* = #ECEAE6 */ --hearth-text: #191713; /* 16.87:1 on canvas */ --hearth-text-secondary: #5A554D; /* 6.97:1 on canvas */ /* ⛔ P2A's OWN muted ink #6E6A61 IS DELIBERATELY NOT APPLIED — the canon's #6D6355 stands. Measured 2026-08-30bi by `hearth-contrast-audit.py`'s APPLIED_PAIRINGS leg: #6E6A61 on the canon severity/status CHIP backgrounds bottoms out at 4.27:1, below AA 4.5 — this repo already holds a text ink to EVERY ground it can land on, and those chip fills are grounds. The canon ink clears all of them (worst 4.66) and reads 5.55:1 on this direction's #FAF8F4 canvas, so the deviation costs a barely-perceptible shift in one de-emphasised grey and buys back 4 failing pairs. ⚠ THIS IS THE ONE TOKEN OF THE 19 THAT DEVIATES from the study sheet — every other value here is transcribed verbatim. Revisit it the day the severity chip fills are restated for this paper. */ /* NON-TEXT ONLY — hairlines, dashed borders, dots. Never a text carrier (post-2026-08-28l canon; check.sh guard 26 enforces it). = #A3A099 */ --hearth-text-faint: color-mix(in srgb, #6E6A61 62%, #FAF8F4); --hearth-text-on-accent: #FFFFFF; /* ---- accent: ink holds the action, not a brand hue -------------------- */ /* The direction's whole argument: keep the paper warm, take the colour OUT of the chrome, and let a coloured thing on screen always be carrying meaning. */ --accent: #191713; --accent-hover: #191713; --accent-active: #000000; --accent-solid-bg: #191713; --accent-subtle-bg: #F1EEE8; --accent-text: #191713; --accent-border: #E0DCD4; /* ---- type: the study spec, now that both faces are self-hosted ------- */ /* Fallbacks are the study sheet's own. Georgia/system-ui are what render in the swap window, so they are the shape the page briefly wears, not dead text. */ --font-display: 'Newsreader', Georgia, serif; --font-sans: 'Hanken Grotesk', system-ui, sans-serif; color-scheme: light; }