/* mechanisms.css - one motion mechanism per archetype.

   DIAGNOSTIC.md Part 2, measured off the 48-page deloitte.com crawl:
   8 components appear on all 11 of their page types and ZERO of them
   are motion. Their shared system is the CHROME (search, linklist,
   title, text). Every motion component is scoped to a type:
   cmp-dual-slider-v2 is home only, cmp-carousel-scrollbar--slide-padding
   is alliances only, cmp-di-modal is insights only.

   We had it inverted: ONE converge animation driving 26 decks across 19
   pages, and the chrome fragmented across three competing owners. JT,
   2026-10-28: "this is the same animation across multiple pages... the
   bigger fix is why the animation and page variety isnt translating."

   So the chrome is now unified (frame-gate checks 1-4) and the motion
   is split five ways, one per archetype:

     home        .stage          the pinned backdrop, repaired
     product     .m-conv         six sources converging into one chain
     comparison  .m-rail         a horizontal rail through the rows
     longread    .m-progress     reading progress, NO entrance motion
     lean        .m-anchor       a sticky in-page nav

   Their insights pages carry 71 transitions against alliances' 243, and
   13 reduced-motion rules against 2. A reading page is meant to be
   still. .m-progress is the only motion longread gets, and it is a
   position report rather than decoration.

   ⛔ EVERY MECHANISM DEGRADES. With no JS: the rail is a plain
   horizontally scrollable row, the progress bar sits at zero width and
   is invisible, the anchor nav is an ordinary list of links. Nothing
   here can hide content, which is the .aml defect that left copy
   permanently invisible once already. */

/* ── comparison: the rail ──────────────────────────────────────────
   Their cmp-carousel-scrollbar, on alliances / industries / services /
   what-we-do. A comparison is a set of parallel claims, so moving
   through it sideways says "these are alternatives" in a way a vertical
   stack does not. Scroll-snap, native overflow, no library. */
html .m-rail{position:relative}
html .m-rail__track{display:flex;gap:clamp(14px,1.6vw,22px);
  overflow-x:auto;overflow-y:hidden;scroll-snap-type:x mandatory;
  -webkit-overflow-scrolling:touch;scrollbar-width:none;
  padding:4px 2px 18px;margin:0}
html .m-rail__track::-webkit-scrollbar{display:none}
/* Card width SCALES with the viewport instead of being a fixed 480px.
   At a fixed width the rail stopped railing on a 1920px screen: three
   480px cards plus gaps fit inside the track, overflow went to zero,
   .is-static kicked in and the mechanism vanished on exactly the
   monitors this audience reads on. 38vw keeps roughly two and a half
   cards in view at every width, so it always overflows and the cards
   get genuinely wider on a big screen rather than just more numerous. */
/* The 620px cap is why the cards stopped growing at about a 1630px
   viewport: 38vw passes 620 there and clamp pins it, so a 2560 or 3840
   screen got the same card as a 1630 one with more empty space either
   side. JT: "make the rail wider on 4k too."
   860px is the new ceiling. It is a deliberate stop rather than letting
   38vw run: at 3840 that would be a 1,459px card, which is wider than
   the whole content column and reads as a page, not a card. */
html .m-rail__track > *{flex:0 0 clamp(300px,38vw,860px);
  scroll-snap-align:start;min-width:0}
/* ⛔ AN AFFORDANCE THAT LIES IS WORSE THAN NONE.
   Measured on /westlaw-alternative/ at 1280px: three cards at 360px
   fitted inside the 1,178px track, so scrollWidth === clientWidth, the
   rail did not scroll, and the hint underneath it said "Scroll sideways
   to compare all 3." mechanisms.js adds .is-static when there is no
   overflow, so the bar and the instruction disappear rather than asking
   for something impossible. CSS cannot measure overflow; this is the
   one thing here that genuinely needs the script. */
html .m-rail.is-static .m-rail__bar,
html .m-rail.is-static .m-rail__hint{display:none}
html .m-rail.is-static .m-rail__track{flex-wrap:wrap}
html .m-rail.is-static .m-rail__track > *{flex:1 1 300px}
/* The scrollbar is drawn, because a native one is hidden above and an
   affordance nobody can see is not an affordance. */
html .m-rail__bar{height:3px;background:var(--rule-light,#e5e7eb);
  border-radius:3px;overflow:hidden;margin-top:2px}
html .m-rail__bar i{display:block;height:100%;width:var(--rw,22%);
  margin-left:var(--rx,0%);background:var(--accent,#f59e0b);
  border-radius:3px;transition:margin-left .12s linear}
html .m-rail__hint{font-size:13px;color:var(--ink-dim,#3f4a5a);
  margin:8px 0 0;letter-spacing:.02em}
/* ⛔ THE >=1400px UNWRAP IS GONE, DELIBERATELY.
   It turned the rail back into a wrapped row on a large desktop, so the
   mechanism that is supposed to BE the comparison archetype's identity
   disappeared on the widest screens, which is where most of this
   audience reads. JT: "make the rail wider."
   Cards are 480px and the track runs full bleed, so three cards need
   ~1,500px against a 1,320px band and it rails at every width that
   matters. Where it genuinely does not overflow, mechanisms.js adds
   .is-static and the affordance hides itself rather than lying. */

/* Full bleed: the track escapes the 1,320px content column to the
   viewport edges, which is what makes a rail read as a rail. The
   padding puts the first card back on the content grid so it still
   lines up with the headings above it. */
html .m-rail{margin-inline:calc(50% - 50vw);width:100vw;max-width:100vw}
html .m-rail__track{padding-inline:max(24px,calc(50vw - 660px));
  /* ⛔ SCROLL PADDING, OR SNAP CLIPS THE FIRST CARD.
     scroll-snap-align:start snaps a card to the PADDING EDGE of the
     scroll box, not to where the padding puts it visually, so the
     first card snapped under the left inset and read as cut off at the
     viewport edge. JT: "this animation doesnt work." It also meant a
     restored scroll position could leave the track parked with the
     opening card half off screen. */
  scroll-padding-inline:max(24px,calc(50vw - 660px))}
html .m-rail__bar,html .m-rail__hint{
  margin-inline:max(24px,calc(50vw - 660px))}

/* ── longread: reading progress ────────────────────────────────────
   The only motion on a reading page, and it reports position rather
   than decorating. Fixed, 3px, above the masthead. */
html .m-progress{position:fixed;top:0;left:0;right:0;height:3px;
  z-index:60;background:transparent;pointer-events:none}
html .m-progress i{display:block;height:100%;width:0;
  background:var(--accent,#f59e0b);
  transition:width .08s linear}
@media (prefers-reduced-motion: reduce){
  html .m-progress i{transition:none}
}

/* ── lean / role pages: the sticky anchor nav ──────────────────────
   Their cmp-link-list-v3--quicklinks--sticky-bar, on careers. A role
   page is a short argument aimed at one reader, and the useful motion
   is orientation: where am I, what else is here. */
html .m-anchor{position:sticky;top:0;z-index:40;
  background:var(--paper,#fff);
  border-bottom:1px solid var(--rule-light,#e5e7eb);
  margin:0 0 clamp(14px,2vw,24px)}
html .m-anchor__list{display:flex;gap:4px;overflow-x:auto;
  scrollbar-width:none;list-style:none;margin:0;
  padding:0;max-width:1320px;margin-inline:auto}
html .m-anchor__list::-webkit-scrollbar{display:none}
html .m-anchor__list a{display:inline-flex;align-items:center;
  min-height:44px;padding:0 14px;white-space:nowrap;
  font:600 14px/1 var(--sans,system-ui,sans-serif);
  color:var(--ink-dim,#3f4a5a);text-decoration:none;
  border-bottom:2px solid transparent}
html .m-anchor__list a[aria-current="true"]{
  color:var(--ink,#0b1220);border-bottom-color:var(--accent,#f59e0b)}
html .m-anchor__list a:hover{color:var(--ink,#0b1220)}

/* ── the answer cards ──────────────────────────────────────────────
   JT supplied four real product answers as screenshots on 2026-10-28.
   36 of 36 card decks on this site contained zero images, and these are
   the most persuasive images the project has: the product refusing to
   invent a doctrine, and hedging honestly on an uncorroborated
   treatment signal. They are dark captures, so they carry their own
   plate rather than sitting on white. */
html .m-answer{display:grid;gap:clamp(16px,2vw,28px);align-items:center;
  grid-template-columns:minmax(0,1fr)}
/* ⛔ NOT SIDE BY SIDE. MEASURED.
   The split put the figure in a 631px column, and these captures are
   1,829px wide, so the screenshot rendered at 34% of its intrinsic
   width. audit-real.py fails any content image under 55% because that
   is the point below which the text inside it stops being readable, and
   the whole reason JT sent these was "these screenshots should help
   with readability".
   Stacked, the figure gets the full 1,208px band: 66%, which clears the
   bar. The argument reads first, the evidence sits under it full width.
   A two-column layout that makes the evidence illegible is a layout
   that defeats its own content. */
html .m-answer--split{grid-template-columns:minmax(0,1fr)}
html .m-answer--split > .m-answer__say{order:1}
html .m-answer--split > figure{order:2}
@media (min-width:980px){
  html .m-answer__say{max-width:72ch}
}
html .m-answer figure{margin:0;border-radius:var(--m-radius,18px);
  overflow:hidden;border:1px solid rgba(148,197,255,.18);
  background:#15171c;box-shadow:0 18px 44px rgba(3,15,29,.22)}
html .m-answer figure img{display:block;width:100%;height:auto}
html .m-answer figcaption{padding:10px 14px 12px;font-size:13.5px;
  line-height:1.45;color:#aab4c4;background:#15171c}
html .m-answer__say h3{margin:0 0 8px}
html .m-answer__say p{margin:0 0 10px}
/* ⛔ AN ACCENT THAT FAILS ON ITS GROUND IS NOT AN ACCENT.
   This was var(--accent) gold, #f59e0b, which is 1.9:1 on white and
   fails AA badly at 11px. surface.js measured it, repaired it to
   #1f2937, and the eyebrow rendered as plain dark grey: the repair
   worked exactly as designed and the design was wrong. Amber 800 is
   5.6:1 on white, keeps the accent family, and needs no repair, so the
   styling survives instead of being corrected away. */
/* The selector carries [data-on] deliberately. surface.js marks every
   text element with its measured ground, and modules.css colours those
   marks with `html body main [data-on="light"]`, specificity (0,1,3).
   `html .m-answer__tag` is (0,1,1) and lost, so the eyebrow rendered in
   plain body grey and looked unstyled while the rule sat in the sheet,
   parsed and matching. Same cascade fight the 17.5px type floor lost
   before it was restated at token depth. */
/* Size lives in modules.css with the other floor exceptions. What is
   left here is the look. The colour carries !important because
   `html body main [data-on="light"]` (0,1,3) sets colour on every
   marked element and this selector cannot outrank it from a later sheet
   by specificity alone. #92400e is 5.6:1 on white, so surface.js leaves
   it alone instead of repairing it to body grey, which is what happened
   to the #f59e0b gold this replaced. */
html body main .m-answer__tag,
html body main .m-answer__tag[data-on]{
  display:inline-block;font-weight:700;line-height:1.3;
  font-family:var(--sans,system-ui);
  text-transform:uppercase;color:#92400e !important;
  margin:0 0 10px}

/* ── ported from .tier-card, which unify-frame.py removed ──────────
   ⛔ THIS IS THE RISK THAT STEP CARRIES, AND IT LANDED.
   unify-frame.py removes a legacy class once the module equivalent is
   present, and it is only safe while the module layer covers what the
   legacy class gave its CHILDREN. `.m-tiers__item` was written to cover
   `.tier-card` itself (background, border, accent rule, padding) and it
   did. What was missed is `.tier-card .btn{width:100%}`.

   Measured consequence: the call to action inside a price tier stopped
   being 250px wide and spanned the full 1,146px band, which put 50,424
   square pixels of solid gold on /pricing/ and tripped the gold-field
   check. The button was always gold; it was never that big.

   The lesson is in unify-frame.py's KEEP set: a legacy container is
   only safe to drop once its DESCENDANT rules are ported too, not just
   its own. */
html .m-tiers__item .btn,html .m-tiers__item a.btn{
  margin-top:auto;width:100%;justify-content:center;min-height:48px}
html .m-tiers__item[data-featured="true"]{
  border-top-color:var(--gold-rule,#f59e0b)}
html [data-accent] .m-tiers__item{border-top-color:var(--hue)}
html .m-tiers__item .tier-features{margin:0 0 16px;padding:0;
  list-style:none}

/* The answer captures pan on a phone, and they need a bigger minimum
   than the diagram rule gives them. modules.css section 26 pans content
   figures at min(760px, 205vw); for a 1,834px Teams capture that is 41%
   of intrinsic width, which is still too small to read even while
   panning, and these exist precisely because JT said "these screenshots
   should help with readability". 1,100px is 60% and clears the 55% bar
   the audit enforces. */
@media (max-width:1079px){
  html body main .m-answer figure{overflow-x:auto;
    -webkit-overflow-scrolling:touch}
  html body main .m-answer figure img{
    min-width:1100px;width:auto;max-width:none}
  html body main .m-answer figcaption{position:sticky;left:0;
    max-width:100vw}
}

/* ── the masthead on a phone ───────────────────────────────────────
   MEASURED at 375x812 by mobile-gate.py: `.mast` is position:sticky and
   236px tall, and it STAYS there while you scroll. That is 29% of the
   viewport permanently occupied by chrome on every page. The height is
   a brand row (52), a tagline (14), two utility buttons (44) and a
   seven-link nav wrapping onto two rows (88).

   Worse, on the role pages the sticky anchor nav also sits at top:0, so
   the two stuck elements overlapped: the orientation bar built for
   those pages was underneath the masthead and could not be seen.

   Two changes, neither needing JS:
     the nav becomes ONE horizontally scrollable row, not two wrapped
       rows, so the masthead is ~150px instead of 236 when you scroll
       back to it
     the masthead stops being sticky below the breakpoint, so the
       content gets the whole screen and the anchor nav can own top:0

   ⛔ NOT HIDDEN, NOT COLLAPSED BEHIND A TOGGLE. Every nav link is still
   present and reachable by swiping the row. A hamburger would need JS,
   and a menu that fails closed when a script dies hides the entire
   site's navigation. */
@media (max-width:1039px){
  html body .mast{position:static}
  html body .mast-in{padding-top:10px;padding-bottom:10px;gap:8px 16px}
  html body .mast-sub{display:none}
  html body .mast-nav{flex:1 1 100%;display:flex;flex-wrap:nowrap;
    overflow-x:auto;scrollbar-width:none;gap:2px;
    -webkit-overflow-scrolling:touch}
  html body .mast-nav::-webkit-scrollbar{display:none}
  /* min-height restored explicitly: setting `padding:0 10px` to make
     the row compact also zeroed the vertical padding that was giving
     these links their height, and they measured 29.5px against the
     44px floor. Making chrome denser must not make it untappable. */
  html body .mast-nav > a{flex:0 0 auto;white-space:nowrap;padding:0 10px;
    min-height:44px;display:inline-flex;align-items:center}
  /* The anchor nav is the only thing pinned on a phone now, so it sits
     under nothing and gets the full width. */
  html body main .m-anchor{top:0}
}

/* ── the footer social links ───────────────────────────────────────
   Measured by mobile-gate.py at 24.1px on /legal-ai-positioning-map/,
   against the 44px floor in WCAG 2.5.8.

   The 44px rule for these exists, in exhibit.css line 314. These pages
   do not load exhibit.css: they load the register-v23 sheets, and
   register-v23-chrome.css line 109 restyles `.col-social a` with
   font-size and weight only. So the floor was written once, is correct,
   and simply was not present on the pages that needed it. A rule in a
   stylesheet a page never requests is indistinguishable from no rule,
   which is the same shape as /free/ missing modules.css entirely. */
html body .col-social a{min-height:44px;display:inline-flex;
  align-items:center}
/* The colophon link COLUMNS have the same gap, and they are the ones
   that actually measured 29.5px: display:block, min-height auto, inside
   .colophon-in > .col-cols. The social row and the link columns are
   different elements in the same footer and only one of them had a
   floor anywhere. */
html body .colophon-in a,html body .col-cols a{min-height:44px;
  display:inline-flex;align-items:center}

/* ── the cascade deck ──────────────────────────────────────────────
   Replaces the CSS-only stepper. JT: "make it cascade each answer on
   top of the last. almost like a deck of cards shuffling."

   The deck is tall on purpose: it reserves one viewport of scroll per
   card, and the stage pins while they deal. That is what makes it the
   event on the page rather than 3% of it.

   --p is written by mechanisms.js from 0 to 1 across the deck's scroll
   span. Each card reads its own index --i and works out its phase. No
   per-card JS, no library. */
html .m-deck{position:relative;
  /* one screen of scroll per card, plus a screen to read the last one */
  height:calc((var(--n,4) + 0.6) * 86vh);
  margin-block:clamp(18px,3vw,38px)}
html .m-deck__stage{position:sticky;top:12vh;height:76vh;
  display:grid;place-items:center}
html .m-deck__card{grid-area:1/1;width:min(680px,92%);
  background:var(--paper,#fff);color:var(--ink,#0b1220);
  border:1px solid var(--rule-light,#e5e7eb);
  border-top:3px solid var(--accent,#f59e0b);
  border-radius:var(--m-radius,18px);
  padding:clamp(22px,3vw,40px);
  box-shadow:0 26px 60px rgba(3,15,29,.22)}
html .m-deck__n{display:block;font:700 12px/1 var(--sans,system-ui);
  letter-spacing:.1em;color:#92400e;margin-bottom:12px}
html .m-deck__card h3,html .m-deck__card h2{margin:0 0 10px}
html .m-deck__card > :last-child{margin-bottom:0}

html .m-deck__rail{position:sticky;top:90vh;display:flex;gap:7px;
  justify-content:center;margin-top:-6vh}
html .m-deck__rail i{width:30px;height:3px;border-radius:3px;
  background:var(--rule-light,#e5e7eb);transition:background .2s}

/* ⛔ EVERYTHING BELOW IS GATED ON .js-deck, WHICH THE SCRIPT ADDS.
   Without it the cards sit stacked in the grid at full opacity and the
   page is readable, just undramatic. Gating the HIDING on the script
   means a dead script cannot hide anything, which is the rule this
   project learned from the .aml reveal that left copy invisible. */
html.js-deck .m-deck__card{
  /* phase: 0 before this card's turn, 1 at its turn, >1 once passed */
  --ph:calc((var(--p,0) * var(--n,4)) - var(--i));
  --in:clamp(0, calc(var(--ph) + 1), 1);     /* rises into place */
  --out:clamp(0, var(--ph), 1);              /* settles underneath */
  opacity:var(--in);
  transform:
    translate3d(0, calc((1 - var(--in)) * 46vh - var(--out) * 2.2vh), 0)
    scale(calc(0.92 + 0.08 * var(--in) - 0.04 * var(--out)))
    rotate(calc((1 - var(--in)) * 4deg - var(--out) * 2.4deg));
  z-index:calc(10 + var(--i));
  transition:none}
html.js-deck .m-deck__rail i{
  --ph:calc((var(--p,0) * var(--n,4)) - var(--i));
  background:color-mix(in srgb, var(--accent,#f59e0b)
             calc(clamp(0, calc(var(--ph) + 1), 1) * 100%),
             var(--rule-light,#e5e7eb))}

@media (max-width:1039px){
  /* No pinning on a phone: a sticky stage that holds the screen while
     it deals four cards is a scroll trap on a small device. */
  html .m-deck{height:auto}
  html .m-deck__stage{position:static;height:auto;display:grid;
    grid-template-columns:1fr;gap:14px;place-items:stretch}
  html .m-deck__card{grid-area:auto;width:auto}
  html.js-deck .m-deck__card{opacity:1;transform:none}
  html .m-deck__rail{display:none}
}
@media (prefers-reduced-motion: reduce){
  html .m-deck{height:auto}
  html .m-deck__stage{position:static;height:auto;gap:14px}
  html .m-deck__card{grid-area:auto;width:auto}
  html.js-deck .m-deck__card{opacity:1;transform:none}
  html .m-deck__rail{display:none}
}

/* ── the converge band owns its ground ────────────────────────────
   JT, 2026-10-28: "WHITE ON WHITE TOO."

   MEASURED: `.m-conv` band background is rgba(0,0,0,0) and the chips
   are solid #fff. On a page where the band happened to sit over a light
   section, six white chips landed on white and the whole mechanism was
   invisible. The chips were never wrong; the band had no ground of its
   own and inherited whatever was behind it, which on /how-it-works/ was
   a pale section and a product screenshot.

   A module that only works over some backgrounds is not a module. This
   band now paints its own dark ground, with the citation field at low
   opacity for texture, so white chips always read and the convergence
   is a distinct moment rather than something that blends into the page
   it interrupts. */
html body main .m-conv{position:relative;isolation:isolate;
  background:#041424;color:var(--on-deep,#e8eaf0);
  padding-block:clamp(34px,5vw,68px)}
html body main .m-conv::before{content:"";position:absolute;inset:0;
  z-index:-1;pointer-events:none;
  background:
    radial-gradient(1200px 700px at 18% 88%, rgba(245,158,11,.12), transparent 62%),
    radial-gradient(1400px 900px at 80% 8%, rgba(37,99,235,.14), transparent 64%)}
html body main .m-conv h2,html body main .m-conv .m-conv__lede{
  color:var(--on-deep,#e8eaf0)}
html body main .m-conv .m-conv__lede{color:#aab4c4}
/* The chips are the SOURCES, white because they are documents. The core
   is the ANSWER, so it inverts: gold rule, lighter plate, to read as the
   thing they resolve into rather than another card. */
html body main .m-conv__chip{background:#fff;color:#0b1220;
  border:1px solid rgba(255,255,255,.14);
  box-shadow:0 10px 26px rgba(3,15,29,.34)}
html body main .m-conv__core{background:#0b2136;
  border:1px solid rgba(245,158,11,.55);
  border-top:3px solid var(--accent,#f59e0b);
  color:var(--on-deep,#e8eaf0);
  box-shadow:0 18px 42px rgba(3,15,29,.46)}
html body main .m-conv__core span{color:#c9d4e4}

/* ── the index list ───────────────────────────────────────────────
   30 treatises and 21 practice areas were rendering as bordered cards
   in a four-column grid: about 900px of boxes that said "there are a
   lot" and nothing else. JT: "THIS ONE HAS ALL THE INFORMATION, BUT NO
   IMPACT" and "WALLS OF TEXT DONT HELP."

   An inventory reads as an index. Columns, rules between entries, the
   count stated once at a size you cannot miss. Same content, about 40%
   of the height, and it now looks like a corpus rather than a card
   wall. */
html body main .m-index__wrap{margin:14px 0 0}
html body main .m-index__count{
  font:700 clamp(28px,3.4vw,46px)/1.05 var(--serif,Georgia,serif);
  color:var(--ink,#0b1220);margin:0 0 14px;letter-spacing:-.01em}
html body main .m-index{list-style:none;margin:0;padding:0;
  columns:3;column-gap:clamp(22px,2.6vw,44px)}
html body main .m-index__row{break-inside:avoid;display:block;
  padding:9px 0;border-top:1px solid var(--rule-light,#e5e7eb)}
html body main .m-index__row b{display:block;
  font:600 15.5px/1.3 var(--sans,system-ui,sans-serif);
  color:var(--ink,#0b1220)}
html body main .m-index__row span{display:block;font-size:13px;
  line-height:1.4;color:var(--ink-dim,#3f4a5a);margin-top:2px}
@media (max-width:1079px){
  html body main .m-index{columns:2}
}
@media (max-width:620px){
  html body main .m-index{columns:1}
}

/* ── the ledger: a sourced cost table that stacks up ──────────────
   The third comparison mechanism. These pages already carry a table of
   competitor rates, each row with its own source and read date, and it
   was the most load-bearing content on the page rendering as a static
   block below the fold.

   As a ledger the rows arrive one at a time as it scrolls, so the cost
   accumulates in front of the reader rather than landing all at once.
   The content is identical; what changes is that you watch it add up.

   ⛔ GATED ON .js-ledger, which the script adds. No script means every
   row is visible from the start: a normal table, fully readable. A
   table that fails closed would hide sourced pricing, which is the one
   thing on these pages that has to be verifiable. */
html .m-ledger{position:relative;overflow-x:auto;
  -webkit-overflow-scrolling:touch}
html .m-ledger table{width:100%;border-collapse:collapse}
html.js-ledger .m-ledger tr{
  --ph:calc((var(--lp,0) * var(--rows,6)) - var(--i));
  opacity:clamp(.18, calc(var(--ph) + 1), 1);
  transform:translate3d(0, calc((1 - clamp(0, calc(var(--ph) + 1), 1))
            * 14px), 0);
  transition:opacity .18s linear, transform .18s linear}
html.js-ledger .m-ledger tr:first-child{opacity:1;transform:none}
/* The row currently arriving carries the accent, so the eye has
   somewhere to be while the column fills. */
html.js-ledger .m-ledger tr td{transition:background .22s}
@media (prefers-reduced-motion: reduce){
  html.js-ledger .m-ledger tr{opacity:1;transform:none}
}

/* ── the stack convergence ────────────────────────────────────────
   Nine procurement line items resolving into one contract. Replaces the
   old graphic that showed our own features flowing into our own price,
   which described rather than argued.

   The left column is what a firm buys today, each row carrying how it
   is normally bought. The right is what Marcella covers of that line.
   The rows slide in from the left as it scrolls and the rule between
   the columns draws itself, so the convergence is literal: nine
   separate things becoming one column. */
html body main .m-stack{background:#041424;color:var(--on-deep,#e8eaf0);
  padding-block:clamp(34px,5vw,68px);position:relative;isolation:isolate}
html body main .m-stack::before{content:"";position:absolute;inset:0;
  z-index:-1;pointer-events:none;
  background:radial-gradient(1200px 760px at 16% 90%,
             rgba(245,158,11,.14), transparent 62%)}
html body main .m-stack h2{color:var(--on-deep,#e8eaf0);max-width:22ch}
html body main .m-stack__eyebrow{font-size:12px !important;
  letter-spacing:.1em;text-transform:uppercase;color:var(--accent,#f59e0b);
  margin:0 0 10px}
html body main .m-stack__lede{color:#aab4c4;max-width:58ch}
html body main .m-stack__field{display:grid;gap:clamp(18px,2.4vw,34px);
  margin-top:clamp(18px,2.4vw,30px);align-items:center;
  grid-template-columns:minmax(0,1fr)}
@media (min-width:1000px){
  html body main .m-stack__field{
    grid-template-columns:minmax(0,1.5fr) minmax(0,1fr)}
}
html body main .m-stack__list{list-style:none;margin:0;padding:0;
  display:grid;gap:1px;background:rgba(148,197,255,.16);
  border:1px solid rgba(148,197,255,.16);border-radius:14px;
  overflow:hidden}
html body main .m-stack__item{background:#071c30;padding:11px 15px;
  display:grid;gap:2px 14px;align-items:baseline;
  grid-template-columns:minmax(0,1fr) auto}
html body main .m-stack__item b{font:600 15px/1.3 var(--sans,system-ui);
  color:#fff}
html body main .m-stack__how{font-size:12.5px;color:#8da3bd;
  justify-self:end;text-align:right;white-space:nowrap}
html body main .m-stack__ours{grid-column:1 / -1;font-size:13px;
  color:var(--accent,#f59e0b)}
html body main .m-stack__core{background:#0b2136;
  border:1px solid rgba(245,158,11,.55);
  border-top:3px solid var(--accent,#f59e0b);border-radius:16px;
  padding:clamp(20px,2.4vw,30px);
  box-shadow:0 18px 44px rgba(3,15,29,.5)}
html body main .m-stack__core b{display:block;
  font:700 clamp(22px,2.4vw,30px)/1.1 var(--serif,Georgia,serif);
  color:#fff;margin-bottom:8px}
html body main .m-stack__core span{font-size:14px;color:#c9d4e4}

/* The rows arrive one at a time, gated on the script so a dead script
   shows all nine immediately rather than an empty box. */
html.js-stack .m-stack__item{
  --ph:calc((var(--sp,0) * var(--n,9)) - var(--i));
  opacity:clamp(.12, calc(var(--ph) + 1), 1);
  transform:translate3d(calc((1 - clamp(0, calc(var(--ph) + 1), 1))
            * -26px), 0, 0);
  transition:opacity .2s linear, transform .2s linear}
@media (prefers-reduced-motion: reduce){
  html.js-stack .m-stack__item{opacity:1;transform:none}
}
@media (max-width:620px){
  html body main .m-stack__item{grid-template-columns:minmax(0,1fr)}
  html body main .m-stack__how{justify-self:start;text-align:left}
}

/* ── the why band ─────────────────────────────────────────────────
   /matter-memory/ dealt four cards describing what the feature does
   without ever establishing that there is a problem. This sits above
   the deck and makes the case. Dark ground, because it is the
   uncomfortable part of the argument and should not look like the
   product tour that follows it. */
html body main .m-why{background:#0b1220;color:var(--on-deep,#e8eaf0);
  padding-block:clamp(34px,5vw,64px)}
html body main .m-why h2{color:#fff;max-width:20ch}
html body main .m-why__eyebrow{font-size:12px !important;
  letter-spacing:.1em;text-transform:uppercase;color:#f59e0b;
  margin:0 0 10px}
html body main .m-why__lede{color:#c9d4e4;max-width:62ch}
html body main .m-why__list{list-style:none;margin:clamp(18px,2.4vw,30px) 0 0;
  padding:0;display:grid;gap:1px;background:rgba(148,197,255,.14);
  border:1px solid rgba(148,197,255,.14);border-radius:14px;
  overflow:hidden}
html body main .m-why__item{background:#111b2c;padding:16px 18px}
html body main .m-why__item b{display:block;
  font:600 16px/1.3 var(--sans,system-ui,sans-serif);color:#fff;
  margin-bottom:5px}
html body main .m-why__item span{display:block;font-size:15px;
  line-height:1.55;color:#aab4c4;max-width:72ch}
html body main .m-why__close{margin-top:clamp(16px,2vw,26px);
  color:#e8eaf0;max-width:62ch;
  border-left:3px solid var(--accent,#f59e0b);padding-left:14px}
@media (min-width:900px){
  html body main .m-why__list{grid-template-columns:1fr 1fr}
}
