/* Step Up AVL — "The Broadsheet", legacy article layer.
 *
 * The 42 articles converted in phase 3 keep the markup and the bespoke rules
 * (charts, timelines, card grids) they were hand-built with. This file does two
 * jobs so they land in the Broadsheet without being rewritten:
 *
 *   1. It redefines the old CSS variables (--paper, --rust, --sage, ...) in
 *      Broadsheet tokens, so every bespoke rule that reads var(--rust) retints
 *      itself.
 *   2. It overrides the article's shared core (masthead, act heads, prologue,
 *      stat row, aside, takeaway, sources footer) to the geometry the exemplar
 *      files/the-shortage-was-made-here.html sets by hand.
 *
 * LOAD ORDER MATTERS: broadsheet.css, then the article's own inline <style>,
 * then this file. Equal specificity, so the last one wins.
 *
 * "EQUAL SPECIFICITY" IS THE ASSUMPTION THAT BROKE. Load order only decides a
 * tie, and three of the rules below were never in one: see the "convergence
 * gaps" block at the foot of this file, added 2026-07-24 after measuring all 47
 * articles in a browser rather than reading the CSS. A convergence rule fails
 * silently in three ways, and this file had all three: it can lose on
 * specificity (`.pull`), it can target a class name the articles do not use
 * (`.stat .sn` where 12 articles write `.num`), or it can be missing for a
 * component nobody thought to converge (body links). None of those are visible
 * from inside this file, so a new rule here is not done until it has been
 * measured on a page that was supposed to be wrong.
 *
 * Spec: _docs/design/broadsheet-2026-07/README.md, sections 8-10.
 */

/* -------------------------------------------- old variables, new values */

/* THIS IS THE ONLY PLACE THE ARTICLE PALETTE IS DEFINED (as of 2026-07-25).
 *
 * It was not. Forty of the forty-seven articles carried their own :root copy of these
 * same 23 token names in their inline <style>, and four of the tokens had forked into
 * two values across the corpus: --ink #221d17 against #2b2724, --ink-soft, --rule, and
 * --paper-2 at #ece3d2 against #ece3d3, which is a single hex digit copied into ten
 * live pages. Every palette change would have had to find all forty.
 *
 * All forty were DEAD. This file loads after the inline block, so none of those values
 * had reached a page in a long time; the whole fork was invisible. Verified by reading
 * the computed tokens in a browser and confirmed by a whole-page render snapshot:
 * removing all forty moved nothing across 104,202 elements.
 *
 * So do not reintroduce a local copy. An article that needs a colour of its own should
 * name a new token here, or use a literal in the one rule that needs it. */
:root {
  --paper:        #F6F3EC;
  --paper-2:      #EAEFE9;
  --ink:          #1D2B2D;
  --ink-soft:     #5C6B66;
  --ink-faint:    #5C6B66;
  --rule:         #CFD7D0;
  --rust:         #AB422D;
  --rust-deep:    #A33D28;
  --rust-tintbg:  #F1EAD9;
  --rust-tintbd:  #C9A94E;
  /* --ochre carries small uppercase LABELS in several articles (.scene-kicker),
     where #B58A2A reads at 2.9:1. Darkened to the Broadsheet's --gold-text. */
  --ochre:        #7A5C14;
  --sage:         #2C7569;
  --sage-deep:    #223B33;
  --green:        #2C7569;
  --green-deep:   #216F57;
  --green-band:   #223B33;
  --green-tintbg: #EAEFE9;
  --green-tintbd: #DDE3DC;
  --orange:       #AB422D;
  --orange-num:   #AB422D;
  --result-bg:    #F1EAD9;
  --result-bd:    #C9A94E;
  --shadow:       rgba(0, 0, 0, 0.18);
}

/* ------------------------------------------------------ surround and sheet */

/* The old articles painted the charcoal surround on <body> and padded it.
 * The surround is now a band inside the shell. */
/* The page's own `body { line-height: 1.62 }` inherits into the shared shell and
 * made the sticky compact masthead taller on an article than on every other page.
 * Reset it here (this file loads last) and hand the reading line-height to the
 * band, where it belongs. */
body {
  background: var(--paper);
  color: var(--body-text);
  font-family: var(--serif);
  line-height: normal;
  padding: 0;
}
.article-band { line-height: 1.5; }

.article-band .sheet {
  max-width: 780px;
  margin: 0 auto;
  background: var(--paper);
  background-image: none;
  color: var(--ink);
  padding: 48px 56px 52px;
  box-shadow: 0 34px 80px -34px rgba(0, 0, 0, 0.7);
  overflow: visible;
}
.article-band .pad { padding: 0; }
.article-band section { padding: 0; }
.article-band .act-band { display: none; }

/* ------------------------------------------------------------ article head */

.article-band .sheet > header {
  padding: 0;
  border-bottom: 0;
}
.article-band .sheet > header .eyebrow {
  display: block;
  font-family: var(--sans);
  font-size: 11px;
  font-weight: 600;
  letter-spacing: 0.16em;
  text-transform: uppercase;
  color: var(--avl-green);
  margin: 0 0 14px;
}
.article-band .sheet > header .eyebrow::after { content: none; }

/* Second head shape (11 articles): header.masthead > .kicker > .label x2. */
.article-band .sheet > header.masthead,
.article-band .sheet > .mast {
  padding: 0;
  border-bottom: 0;
  position: static;
}
.article-band .sheet > header.masthead::after { content: none; }
.article-band .sheet > header.masthead .kicker {
  display: flex;
  justify-content: space-between;
  flex-wrap: wrap;
  gap: 8px;
  margin: 0 0 14px;
}
.article-band .sheet > header.masthead .kicker .label {
  font-family: var(--sans);
  font-size: 11px;
  font-weight: 600;
  letter-spacing: 0.16em;
  text-transform: uppercase;
  color: var(--avl-green);
}
.article-band .sheet > header.masthead .kicker .label:last-child {
  color: var(--text-secondary);
}
.article-band .sheet > .mast .eyebrow {
  display: block;
  font-family: var(--sans);
  font-size: 11px;
  font-weight: 600;
  letter-spacing: 0.16em;
  text-transform: uppercase;
  color: var(--avl-green);
  margin: 0 0 14px;
}
.article-band .sheet > .mast .eyebrow::after { content: none; }

.article-band .sheet h1 .hl {
  font-style: italic;
  font-weight: 500;
  color: var(--article-red);
}

.article-band .sheet h1 {
  font-family: var(--serif);
  font-weight: 600;
  font-size: 46px;
  line-height: 1.05;
  letter-spacing: -0.015em;
  color: var(--ink);
  margin: 0 0 18px;
}
.article-band .sheet h1 em {
  font-style: italic;
  font-weight: 500;
  color: var(--article-red);
}
/* No `h1 br { display: none }` here: hiding the authored line break also swallowed
 * the only space between the red accent word and the word after it. The <br>s were
 * stripped from the markup at conversion time instead; the 46px title wraps on
 * its own. */

.article-band .dek {
  font-family: var(--serif);
  font-size: 20px;
  line-height: 1.5;
  font-style: italic;
  font-weight: 400;
  color: var(--body-text);
  max-width: none;
  margin: 0 0 20px;
}
.article-band .dek strong { font-weight: 500; color: var(--ink); }

.article-band .byline,
.article-band p.byline {
  border-top: 1px solid var(--ink);
  border-bottom: 1px solid var(--hairline);
  padding: 9px 0;
  margin: 0 0 28px;
  font-family: var(--sans);
  font-size: 12px;
  font-weight: 600;
  letter-spacing: 0.08em;
  text-transform: uppercase;
  color: var(--text-tertiary);
}

/* ---------------------------------------------------------------- act heads */

/* The legacy .act-head is an inline kicker + title: `div.act-head > .act-num +
 * .act-title`. The Broadsheet history variant stacks them under a double rule,
 * so restack rather than restyle.
 *
 * Scoped to `div` on purpose. Eleven articles use `.act-head` for a DIFFERENT
 * shape: `section.act > .label + h2.act-head`, where the double rule already
 * comes from `.act` in article.css. Unscoped, this rule fired there too and
 * drew a second rule between the label and the heading. */
.article-band div.act-head {
  display: block;
  border-top: 3px double var(--ink);
  margin: 32px 0 0;
  padding-top: 20px;
}
.article-band .act-head .act-num {
  display: block;
  font-family: var(--sans);
  font-size: 11px;
  font-weight: 700;
  letter-spacing: 0.3em;
  text-transform: uppercase;
  color: var(--article-red);
  margin-bottom: 6px;
}
.article-band .act-head .act-title {
  display: block;
  font-family: var(--serif);
  font-size: 30px;
  font-weight: 600;
  line-height: 1.15;
  letter-spacing: -0.01em;
  color: var(--ink);
}
.article-band .act-head .act-title em {
  font-style: italic;
  color: var(--article-red);
}
.article-band .act-sub {
  font-family: var(--serif);
  font-size: 17.5px;
  line-height: 1.65;
  color: var(--text-tertiary);
  max-width: none;
  margin: 12px 0 18px;
}

/* --------------------------------------------- act heads: one geometry, four shapes
 *
 * The corpus grew four markup shapes for the SAME element (a small uppercase
 * red kicker over a serif head, under a double rule). `.act-num` reads like a
 * numbering scheme but is not: only 12 of 123 are literally "ACT I".
 *
 *   .act > .al + h2                        article.css canonical  (4 articles)
 *   section.act > .label + h2.act-head     11 articles
 *   div.act-head > .act-num + .act-title   26 articles
 *
 * The canonical geometry is `.act` / `.act .al` / `.act h2` in article.css. The
 * legacy `div.act-head` shape above already matches it. The strays are pulled
 * into line HERE rather than by rewriting 36 articles' markup, which is the job
 * this file exists to do (see the header). Change the geometry in article.css;
 * these rules only re-point the other shapes at it. */

/* `.article-band section { padding: 0 }` further up is more specific than
 * `.act { padding-top: 20px }` in article.css, so the section.act shape lost the
 * gap between its rule and its kicker. The inline page styles also set
 * `.act { margin: 2.6rem 0 0 }` (41.6px) against the canonical 32px. */
.article-band section.act {
  margin-top: 32px;
  padding-top: 20px;
}

/* The kicker: the 11 section.act articles style a bare `.label` in their own
 * page <style> (500 weight, .16em, .72rem). Re-point it at `.act .al`. Direct
 * child on purpose: `.label` is also the kicker inside .chart-card, .series,
 * .takeaway and the masthead .kicker, which keep their own treatment. */
.article-band section.act > .label {
  display: block;
  font-family: var(--sans);
  font-size: 11px;
  font-weight: 700;
  letter-spacing: 0.3em;
  text-transform: uppercase;
  color: var(--article-red);
  margin-bottom: 6px;
}

/* `.act h2` sets `margin: 0 0 12px`; the legacy .act-title carried none. */
.article-band .act-head .act-title {
  margin: 0 0 12px;
}

/* --------------------------------------------------------------- body copy */

.article-band .body-copy {
  font-size: 17.5px;
  line-height: 1.65;
  color: var(--body-text);
}
.article-band .body-copy p {
  font-size: 17.5px;
  max-width: none;
  margin: 0 0 16px;
}
.article-band .body-copy p b,
.article-band .body-copy p strong { font-weight: 600; color: var(--ink); }

/* The drop cap lives in article.css now, as ONE rule covering every act-head
 * shape (see ".dropcap" there). The copy that used to sit here fired per
 * .body-copy rather than per section, so a section holding prose + a chart +
 * more prose took two caps. Do not re-add a copy: two sources of truth for the
 * cap is what let the corpus drift to 10-of-41 in the first place. */

/* ------------------------------------------------- prologue -> gold callout */

.article-band .prologue {
  background: var(--callout-tint);
  border-left: 4px solid #B58A2A;
  padding: 20px 24px 22px;
  margin: 0 0 26px;
}
.article-band .prologue p {
  font-size: 17.5px;
  line-height: 1.65;
  max-width: none;
  margin: 0 0 14px;
}
.article-band .prologue p:last-child { margin: 0; }
.article-band .prologue .act-num {
  display: inline-block;
  font-family: var(--sans);
  font-size: 10.5px;
  font-weight: 600;
  letter-spacing: 0.22em;
  text-transform: uppercase;
  color: var(--gold-text);
  border: 1px solid #C9A94E;
  padding: 4px 10px;
  margin-bottom: 4px;
}

/* --------------------------------------------------------------- pull quote */

.article-band .pull {
  font-family: var(--serif);
  font-style: italic;
  font-weight: 600;
  font-size: 24px;
  line-height: 1.3;
  color: var(--ink);
  margin: 18px 0 4px;
}

/* --------------------------------------------------- stat row -> stat strip */

.article-band .stat-row {
  display: grid;
  grid-template-columns: 1fr 1fr 1fr;
  gap: 0;
  border-top: 1px solid var(--ink);
  border-bottom: 1px solid var(--ink);
  margin: 26px 0;
}
.article-band .stat-row .stat,
.article-band .stat-row .stat.sage {
  background: none;
  border-top: 0;
  border-right: 1px solid var(--panel-rule);
  padding: 18px 16px 16px;
}
.article-band .stat-row .stat:last-child { border-right: 0; }
.article-band .stat .sn {
  font-family: var(--serif);
  font-weight: 700;
  font-size: 34px;
  line-height: 1;
  color: var(--article-red);
}
.article-band .stat .sl {
  font-family: var(--sans);
  font-size: 11.5px;
  line-height: 1.45;
  color: var(--text-secondary);
  margin-top: 8px;
}
.article-band .stat .src {
  font-family: var(--sans);
  font-size: 10px;
  letter-spacing: 0.08em;
  text-transform: uppercase;
  color: var(--text-secondary);
  margin-top: 8px;
  /* Removed with its twin in article.css:373 -- see the note there. This sheet is
     still live: before-the-door.html and corrections.html load it. */
}

/* -------------------------------------------------------- aside -> callout */

.article-band .aside {
  background: var(--panel-bg);
  border-left: 4px solid var(--avl-green);
  padding: 20px 24px 22px;
  margin: 26px 0;
}
.article-band .aside p {
  font-size: 17.5px;
  line-height: 1.65;
  max-width: none;
  margin: 0 0 14px;
}
.article-band .aside p:last-child { margin: 0; }
.article-band .aside .lead {
  font-family: var(--serif);
  font-style: italic;
  font-weight: 600;
  font-size: 21px;
  line-height: 1.32;
  color: var(--ink);
  margin-bottom: 12px;
}

/* --------------------------------------------------- takeaway -> the point */

.article-band .takeaway {
  background: var(--spruce);
  color: #F1EBDD;
  padding: 36px 40px 38px;
  margin: 34px 0 30px;
}
.article-band .takeaway .eyebrow {
  display: flex;
  align-items: center;
  gap: 16px;
  font-family: var(--sans);
  font-size: 12px;
  letter-spacing: 0.3em;
  text-transform: uppercase;
  color: #DCE5DF;
  margin-bottom: 20px;
}
.article-band .takeaway .eyebrow::after {
  content: "";
  flex: 1;
  height: 1px;
  background: #4A6156;
}
.article-band .takeaway h2 {
  font-family: var(--serif);
  font-weight: 700;
  font-size: 38px;
  line-height: 1.12;
  letter-spacing: -0.01em;
  color: #F1EBDD;
  margin: 0 0 22px;
}
.article-band .takeaway h2 em {
  font-style: italic;
  font-weight: 600;
  color: var(--gold);
}
.article-band .takeaway p {
  font-family: var(--serif);
  font-size: 17.5px;
  line-height: 1.7;
  color: #EDE7D8;
  max-width: none;
  margin: 0 0 18px;
}
.article-band .takeaway p:last-child { margin: 0; }

/* ------------------------------------------------ sources footer -> notebox */

.article-band .sheet > footer {
  background: var(--panel-bg);
  border-top: 3px solid var(--avl-green);
  padding: 16px 20px;
  margin: 0 0 22px;
  font-size: 13.5px;
  color: var(--body-text);
}
/* The citation block's label. h2 is the element everywhere as of 2026-07-25 (audit B3:
 * it was h4 on 26 articles, h2 on 5, a run-in <p> or <b> on 13, and absent on 3, which
 * is why the block was not addressable as a section and why the outline skipped h2->h4
 * on ~30 pages). h4 stays matched so a page converted later, or a draft built from an
 * older shell, still gets the label rather than a bare browser heading. */
.article-band .sheet > footer h2,
.article-band .sheet > footer h4 {
  font-family: var(--sans);
  font-size: 11px;
  font-weight: 600;
  letter-spacing: 0.14em;
  text-transform: uppercase;
  color: var(--avl-green);
  margin: 0 0 7px;
}
.article-band .sheet > footer p {
  font-size: 13.5px;
  line-height: 1.6;
  max-width: none;
  margin: 0 0 8px;
}
.article-band .sheet > footer a {
  color: var(--avl-green);
  border-bottom: 0;
  font-weight: 600;
}
.article-band .sheet > footer .meta {
  font-family: var(--sans);
  font-size: 11px;
  letter-spacing: 0.06em;
  text-transform: uppercase;
  color: var(--text-secondary);
  margin-top: 14px;
}

/* ------------------------------------------------------------- responsive */

@media (max-width: 760px) {
  /* No trailing paper band before the footer on mobile (see article.css). */
  .article-band .sheet { padding: 32px 20px 0; max-width: none; }
  .article-band .sheet h1 { font-size: 33px; }
  /* Mobile body copy: see the matching block in article.css. Same intent,
   * different selector, because the legacy sheet wraps prose in .body-copy. */
  .article-band .dek { font-size: 18px; color: var(--ink-deep); }
  .article-band .body-copy,
  .article-band .body-copy p {
    font-size: 18px;
    line-height: 1.7;
    color: var(--ink-deep);
    font-optical-sizing: none;
    font-variation-settings: 'opsz' 16;
  }
  /* Four narrative articles (between-jobs, by-the-time-you-see-me,
   * rent-or-eat-reframed, what-the-belt-measures) set their prose in .story p
   * instead of .body-copy, and clamp it to ~16.5px at phone widths. Two class
   * names of specificity so this outranks their own inline .story p rule. */
  .article-band .story p {
    font-size: 18px;
    line-height: 1.7;
    color: var(--ink-deep);
    font-optical-sizing: none;
    font-variation-settings: 'opsz' 16;
  }
  /* The 48px mobile drop cap is in article.css, on the same selector list as
   * the desktop rule. Both must move together. */
  .article-band .act-head .act-title { font-size: 26px; }
  .article-band .pull { font-size: 20px; }
  .article-band .aside .lead { font-size: 18px; }
  .article-band .stat-row { grid-template-columns: 1fr; }
  .article-band .stat-row .stat { border-right: 0; border-bottom: 1px solid var(--panel-rule); }
  .article-band .stat-row .stat:last-child { border-bottom: 0; }
  .article-band .takeaway {
    margin-left: -20px;
    margin-right: -20px;
    padding: 28px 20px 30px;
  }
  .article-band .takeaway h2 { font-size: 28px; }
}

/* ----------------------------------------------------------------- print */

@media print {
  .trust, .masthead, .navrule-thin, .site-foot { display: none; }
  .article-band { width: auto; margin: 0; background: #fff; padding: 0; }
  .article-band .sheet { max-width: none; box-shadow: none; padding: 0; }
}

/* ------------------------------------------------- mobile legibility floor */

@media (max-width: 760px) {
  .article-band .sheet > header .eyebrow,
  .article-band .sheet > .mast .eyebrow,
  .article-band .sheet > header.masthead .kicker .label { font-size: 12px; letter-spacing: 0.1em; }
  .article-band .byline,
  .article-band p.byline { font-size: 12.5px; letter-spacing: 0.06em; }
  /* The `.act > .al` shape takes this same floor from article.css, which every
   * article loads (four of them do not load this file at all). `section.act >
   * .label` needs its copy HERE: its desktop rule above is more specific than
   * anything article.css can say, and this file loads last. */
  .article-band .act-head .act-num,
  .article-band section.act > .label { font-size: 12px; letter-spacing: 0.2em; }
  .article-band .prologue .act-num { font-size: 11.5px; letter-spacing: 0.16em; }
  /* `.stat .cap` belongs here too, but it CANNOT live here: its desktop rule is
   * in the convergence block below, at the same specificity and later in the
   * file, so a copy here is dead code (it was, from 2026-07-24 to 07-25). The
   * mobile counterpart sits at the very foot of this file instead, after every
   * rule it has to beat. */
  .article-band .stat .sl { font-size: 12.5px; }
  .article-band .stat .src { font-size: 11px; letter-spacing: 0.04em; }
  .article-band .sheet > footer,
  .article-band .sheet > footer p { font-size: 14.5px; }
  .article-band .sheet > footer h2,
  .article-band .sheet > footer h4 { font-size: 12px; letter-spacing: 0.1em; }
  .article-band .sheet > footer .meta { font-size: 12px; letter-spacing: 0.03em; }
}

/* ===========================================================================
   CONVERGENCE GAPS (added 2026-07-24)

   Found by loading all 47 articles in a browser and grouping every shared
   component by its COMPUTED value, not by reading this file. Each rule below
   restates a value that already exists above or in article.css; none of them
   invents a new one. They exist because the earlier rule never reached the
   page. Method and the full before/after:
   _docs/reviews/article-shell-drift-before-after-2026-07-24.html
   =========================================================================== */

/* --- 1. Body links -------------------------------------------------------
 * broadsheet.css line 91 makes AVL green THE link colour and article.css sets
 * `.art-body a` to 600. Nothing converged the legacy body wrappers, so the
 * inline <style> in each article won by default and an in-text link rendered
 * three ways across the corpus: rust with a rust hairline underline (16
 * articles), green at 400 with no underline (4), green with a grey hairline
 * (3). The SNAP series alone shipped all three. Two class names of specificity
 * to outrank the inline `.body-copy a` (0,1,1) rules.
 *
 * This rule could not finish the job on its own: 20 links in 11 articles also
 * carried a hardcoded `style="color:#C04A33;..."` ATTRIBUTE, which no selector
 * outranks. Those attributes were removed from the markup on 2026-07-24 rather
 * than answered with `!important` here, because a hex baked into a served page
 * is invisible to the token system and would have drifted again at the next
 * palette change. One gold link survives on purpose: it sits inside the dark
 * takeaway panel in the-day-manna-lost-everything, where green would fail
 * contrast. If a link ever needs its own colour again, give it a class. */
.article-band .body-copy a,
.article-band .story a,
.article-band .prologue a,
.article-band .aside a {
  color: var(--avl-green);
  font-weight: 600;
  border-bottom: 0;
  text-decoration: none;
}
.article-band .body-copy a:hover,
.article-band .story a:hover,
.article-band .prologue a:hover,
.article-band .aside a:hover { color: var(--deep-green); }

/* The wrapper list above is an enumeration, and this file already knows how that
 * ends: "there is always one more shape". Three articles put prose links inside a
 * bespoke component none of those four names reach (a timeline node, a path row),
 * so they rendered at weight 400 against 600 everywhere else. Scoping to any
 * paragraph in the sheet says what is meant instead of listing the wrappers that
 * happen to exist today.
 *
 * WEIGHT ONLY, deliberately. Setting colour here at (0,3,1) would also capture the
 * one gold link that is kept on purpose, in the-day-manna-lost-everything's dark
 * takeaway panel, where green fails contrast. Those links are already green from
 * their own components; weight was the only property out of step. Measured: this
 * moves 10 elements on 3 pages, all 400 -> 600, and nothing else in the corpus. */
.article-band .sheet p a { font-weight: 600; }

/* --- 2. Pull quotes ------------------------------------------------------
 * `.article-band .pull` further up this file is the right value and the wrong
 * selector. Every legacy pull quote is a `<p class="pull">` inside a wrapper
 * that sets its own paragraph size, and each of those beats it: `.prologue p`,
 * `.body-copy p` and `.aside p` are all (0,2,2) against `.article-band .pull`
 * at (0,2,0), so the pull quote was being reset to 17.5px body size in 22
 * articles. A pull quote set at body size is not a pull quote, and the rule
 * meant to prevent that has been losing quietly since the conversion.
 * Scoped through `.sheet` to reach (0,3,1), which clears every wrapper at once
 * rather than enumerating them and missing the next one. */
.article-band .sheet p.pull {
  font-size: 24px;
  line-height: 1.3;
  /* colour is NOT optional here, for the same reason the size is not. The panel
   * rule in block 4 below (`.article-band .prologue p`) is (0,2,2) and outranks
   * `.article-band .pull` (0,2,0), so without this line the pull quote takes
   * body-text grey instead of ink. That is exactly what happened between
   * 2026-07-24 and 07-25: fixing the panel colour silently recoloured 22 pull
   * quotes, because this rule set the property being fixed and not the ones the
   * fix could reach. */
  color: var(--ink);
}

/* --- 3. Stat rows, the other half of the corpus --------------------------
 * The stat-row block above converges `.stat .sn` / `.stat .sl`, which is the
 * markup 17 articles use. The other 12 write `.num` / `.cap` (76 and 120
 * occurrences), so they were never reached and kept their inline 40px numerals
 * and uppercase captions against the shell's 34px and sentence case. Scoped
 * through `.stat-row .stat` on purpose: 48 of those `.cap` elements are chart
 * and figure captions elsewhere on the page and must not be restyled. */
.article-band .stat-row .stat .num {
  font-family: var(--serif);
  font-weight: 700;
  font-size: 34px;
  line-height: 1;
  color: var(--article-red);
}
.article-band .stat-row .stat .cap {
  font-family: var(--sans);
  font-size: 11.5px;
  line-height: 1.45;
  letter-spacing: normal;
  text-transform: none;
  color: var(--text-secondary);
  margin-top: 8px;
}

/* --- 4. Panel body colour ------------------------------------------------
 * `.aside` and the sources footer are converged structurally above, but those
 * rules set size and leading and never colour, so each article's own
 * `color: var(--ink-soft)` survived. The shell's equivalents (`.callout`,
 * `.notebox`) inherit body text. Same panel, two greys, decided by which file
 * happened to mention colour. */
.article-band .aside p,
.article-band .prologue p,
.article-band .sheet > footer p { color: var(--body-text); }

/* --- 4b. Sources block typeface ------------------------------------------
 * Eleven articles set `.foot p` in Archivo in their inline <style>, so the
 * Sources & notes block reads as SANS on those and serif on the other 35. The
 * shell's `.notebox p` sets size, leading and colour and never family, so it
 * inherits the sheet's serif; that is the canonical value, and the fact that
 * `.notebox h2` names `var(--sans)` explicitly for the label is the tell that
 * the body underneath it is meant to be the other thing.
 *
 * Missed until 2026-07-25 because every earlier sweep measured this component
 * on size, leading and colour only. A component is not verified by the
 * properties you thought to look at.
 *
 * Container AND descendants, and it took two tries to get that right. Naming
 * `p` and `a` missed housing-funding-comparison, which writes its sources into
 * `div.src`; naming only the container missed the twelve articles whose inline
 * `.foot p` sets Archivo directly, because inheritance loses to any rule that
 * matches the element itself. Enumerating shapes is the `.stat .sn` vs `.num`
 * mistake over again: there is always one more shape. So say what is actually
 * meant, which is that everything in this block is serif unless something says
 * otherwise. The two children that ARE deliberately sans say so above:
 * `> footer h4` (0,2,2) and `> footer .meta` (0,3,1) both name var(--sans) and
 * both outrank the `*` at (0,2,1), so the label and the meta line keep their
 * face and their tracking. Verified, not assumed. */
.article-band .sheet > footer,
.article-band .sheet > footer * {
  font-family: var(--serif);
  letter-spacing: normal;
}
/* Size, not family: three articles use a bare <footer> whose inline rules leave
 * links unsized, so they fell through the mobile floor's `footer, footer p` pair
 * and rendered a step small on a phone. */
.article-band .sheet > footer a { font-size: inherit; }

/* --- 5. Mobile counterparts to blocks 1-4 --------------------------------
 * LAST BLOCK IN THE FILE, AND IT HAS TO STAY LAST. Every rule above is scoped
 * high enough to beat a wrapper, which means it also beats the phone rules it
 * was never thinking about: `.article-band .sheet p.pull` (0,3,1) outranks
 * article.css's `@media .pull { font-size: 20px }` (0,1,0), and the desktop
 * `.stat .cap` rule outranks the identical-specificity copy in the mobile
 * block further up. Both of those shipped broken on 2026-07-24 and are the
 * reason this block exists: a convergence rule is not finished until its
 * breakpoint counterpart is written at a specificity that can win, in a place
 * the cascade reaches last.
 *
 * Values are not new. 20px is article.css's mobile `.pull`; 12.5px is the
 * mobile floor `.stat .sl` already gets. Verified at 375px and 1280px across
 * all 47 articles, comparing EVERY property of each component, not the one
 * being changed. */
@media (max-width: 760px) {
  .article-band .sheet p.pull { font-size: 20px; line-height: 1.3; }
  .article-band .stat-row .stat .cap { font-size: 12.5px; }

  /* The prologue never got the mobile legibility bump. `.body-copy p` goes to
   * 18px/1.7 in the darkest ink on a phone; `.prologue p` sits at 17.5px/1.65 with
   * no breakpoint rule of its own, so on 38 articles the FIRST thing a phone
   * reader reads is a step smaller than everything under it. Nothing says that was
   * intended, and the current shell bumps its equivalent (`.callout .lead`) to
   * 18px, so this converges rather than inventing a value.
   *
   * It has to live here rather than in the mobile block further up, because
   * block 4's `.article-band .prologue p { color: var(--body-text) }` is the same
   * specificity and sits later in the file: written up there the size would land
   * and the colour would not, which is how the prologue would end up 18px in one
   * grey while the copy under it is 18px in another. Same trap as the two rules
   * above. Values copied from `.body-copy p`, not chosen. */
  .article-band .prologue p {
    font-size: 18px;
    line-height: 1.7;
    color: var(--ink-deep);
    font-optical-sizing: none;
    font-variation-settings: 'opsz' 16;
  }
}
