/* ============================================================================
   home-refresh.css — homepage HERO only (2026-08-05)

   Scope, deliberately narrow: this file changes the hero section and nothing
   else. Every other section on the homepage — the results band, the program
   cards, the comparison table, the app showcase, the team, the craft band,
   the CTA band and the footer — keeps exactly the treatment styles.css gives
   it. An earlier version of this file also warmed those sections; that was
   out of scope and was reverted on 2026-08-05.

   The hero: the dim gym photo and the lean man under a loaded barbell are
   gone. In their place, a gradient that runs near-black at the left edge,
   where the copy sits, through deep purple and Human Purple, out to warm
   off-white on the right, where two members from the 07-23 shoot stand.

   The models are the untouched photographs with the studio wall removed
   (Higgsfield background removal) — the people are the real frames, only the
   wall is gone. They stand ON the section, feet cropped by its bottom edge,
   rather than inside a framed box.

   Chosen as "option D" from a five-way background comparison, 2026-08-05.

   The copy keeps the dark token set styles.css already gives .hero — white
   display type, purple accent, white-60 body — because its half of the
   gradient is still nearly black, exactly like the photo hero it replaces.

   Not one word of copy changed. Only the light.

   Loaded after styles.css, same specificity, so it wins on order.
   ========================================================================= */

.hero {
  position: relative;
  overflow: hidden;
  min-height: clamp(620px, 84vh, 840px);
  background:
    radial-gradient(760px 460px at 4% 8%, rgba(124, 86, 204, 0.20), transparent 66%),
    linear-gradient(99deg,
      #0B0A0F 0%,
      #100C1C 14%,
      #221145 27%,
      #3E2183 38%,
      #6B45BE 49%,
      #9E80DC 59%,
      #CFC0EA 69%,
      #EDE6DE 80%,
      #F7F2EA 100%);
}

/* Copy left on the dark end, models right on the pale end. styles.css lays
   .hero-inner out as a single centred flex row. */
.hero-inner {
  display: grid;
  grid-template-columns: 52% 48%;
  gap: 0;
  align-items: center;
  padding-top: 56px;
  padding-bottom: 72px;
}
.hero-copy {
  grid-column: 1;
  max-width: 600px;
}

/* Positioning layer only — no card, no background, no rounding. Full section
   height so the figures can stand on its bottom edge. */
.hero-photo {
  position: absolute;
  left: 50%;
  right: 0;
  top: 0;
  bottom: 0;
  z-index: 0;
  min-height: 0;
  pointer-events: none;
}

/* Sized by HEIGHT, not width, so both reach the bottom of the section and
   fill it whatever the viewport does. Both frames run head-to-mid-thigh, so
   bottom-aligning them puts their heads level; he is a little taller, as he
   is in life.

   No drop-shadow: a blur this large on an image this size gets clipped to
   the element box by the compositor and paints a visible rectangle over the
   gradient — worse than no shadow at all. */
.hero-cut {
  position: absolute;
  bottom: 0;
  width: auto;
  height: 92%;
}
/* ONE FIGURE, 2026-08-11. The second cut-out (.cut-1) came out of index.html;
   what is left is centred in .hero-photo and enlarged.

   Centred in .hero-photo, NOT in the viewport. .hero-photo is the right half
   of the section (left: 50%), and the left half is the headline — centring him
   on the page would run him straight through "Become More."

   THE OFFSET IS WHAT MAKES THE SIZE POSSIBLE. The cut-out has no transparent
   headroom: opaque pixels start around row 20 of 2400, so his hair is the top
   edge of the file. .hero clips with overflow: hidden, so simply raising the
   height crops his head. Pushing the bottom edge below the section by the same
   amount it grows keeps his head exactly where it was and spends the extra
   size on the crop at the thigh, where nobody misses it.

   44%, not 50%: dead-centre of .hero-photo left a heavier band of empty
   gradient between him and the copy than it did on his right, so the pair
   read as sitting right of centre. This walks him back toward the middle of
   the page without crossing into the headline.

   HEIGHT AND BOTTOM MOVE TOGETHER, and the sum is what fixes his head:
   bottom + height = 90%, so the top of the frame — which is his hair, the file
   has no transparent headroom — sits 90% of the way up the section no matter
   how large he gets. Grow him by changing BOTH by the same amount (112/-22 ->
   124/-34 on 2026-08-11) and he scales in place. Change only one and he slides
   vertically, which is the thing Alec has already corrected once.

   112/-22 -> 124/-34 -> 134/-44 -> 128/-38, and 50% -> 44% -> 40% -> 37%.
   Size and horizontal position trade against the same budget: his left edge
   has to stay clear of .hero-photo's own left edge, which is where the copy
   column ends. Coming down 6 points bought back the room that moving 3 points
   further left spent, so this pass is roughly clearance-neutral. */
.cut-2 { left: 37%; transform: translateX(-50%); height: 128%; bottom: -38%; z-index: 1; }

/* Both of these small-caps lines are brand purple (#9B6CFF), which worked on
   the near-black photo hero. Here they run rightward into the purple middle
   of the gradient and lose contrast halfway along, so they lift to a pale
   lavender that holds from black through to mid-purple. */
.hero-badge,
.hero-trust-line { color: #DCCFFF; }

/* Phone: the desktop composition ROTATED, not replaced. On desktop the eye
   travels left to right — dark end with the copy on it, then the ramp
   through purple, then the off-white end with the two members standing on
   it. On a phone that same journey runs top to bottom: the copy sits on the
   dark top in the same white type, the gradient ramps through purple, and
   the models stand on the off-white bottom, cropped by the section's edge
   exactly as they are on desktop.

   The earlier mobile version stacked the models above the copy and stayed
   purple all the way down — it never reached the off-white end, which is why
   it read as a different design rather than the same one on a narrow screen.

   The figures are height-sized, so the band they stand in scales with the
   viewport and is capped and centred; at 390px they come out 186 and 208
   wide, spanning the screen edge to edge instead of floating as two small
   cutouts.

   styles.css also darkens the wash under 900px for the old photo hero,
   which has to go with it. */
@media (max-width: 900px) {
  .hero {
    min-height: 0;
    background:
      radial-gradient(620px 380px at 82% 2%, rgba(124, 86, 204, 0.26), transparent 70%),
      linear-gradient(176deg,
        #0B0A0F 0%,
        #140E24 12%,
        #241348 26%,
        #3E2183 40%,
        #6B45BE 54%,
        #9E80DC 66%,
        #CFC0EA 78%,
        #EDE6DE 89%,
        #F7F2EA 100%);
  }
  /* Copy first on the dark top; the bottom padding clears the band of
     models below it, and tracks the band height rather than being a fixed
     number so their heads never collide with the trust line. */
  .hero-inner {
    grid-template-columns: 1fr;
    padding-top: 44px;
    padding-bottom: calc(clamp(380px, 112vw, 470px) + 24px);
  }
  .hero-copy { grid-column: 1; }
  /* The band is a centred flex row rather than two absolutely-placed
     figures. Percentage lefts had to be re-tuned every time a cutout's
     proportions changed, and any drift showed up as the pair sitting left of
     centre. Here the browser centres the group at every width, and the
     overlap is a negative margin. */
  .hero-photo {
    left: 50%;
    right: auto;
    transform: translateX(-50%);
    width: min(100%, 460px);
    top: auto;
    bottom: 0;
    height: clamp(380px, 112vw, 470px);
    display: flex;
    align-items: flex-end;
    justify-content: center;
  }
  /* position: relative, not static — they still need z-index to overlap. */
  .hero-cut { position: relative; left: auto; bottom: auto; }
  /* Same trade as the desktop rule above: grow, then push down by what he
     grew, so his hair stays clear of the copy. The band is already a centred
     flex row, so with only one figure left the browser centres him for free —
     no left/margin tuning needed. `bottom` works here because .hero-cut is
     position: relative on this breakpoint.

     THE TRANSFORM MUST BE RESTATED HERE, not left to inherit. The desktop rule
     above centres him with translateX(-50%), and a transform is not undone by
     `left: auto` — carried into this breakpoint it shifted him half his own
     width left and clipped his arm off the screen edge. The flex row already
     centres him, so this value is only the small nudge off centre Alec asked
     for: +5% of his own width, about 12px right at 375px. */
  .cut-2 { height: 112%; bottom: -16%; transform: translateX(5%); z-index: 1; }
  /* Avatars and the line beside them share a row that has no width for both
     on a phone: the sentence ends up in a six-word-tall column. */
  .hero-trust-row { flex-direction: column; align-items: flex-start; gap: 12px; }
  .hero-trust-sub { max-width: 36ch; }
}

/* ============================================================================
   PHONE HERO — the model, bigger and set right (Mishy, 2026-08-16)
   ============================================================================
   Alec's ask: he should stand a lot larger, and move right far enough that the
   "Book a consultation" button — which is pinned to the left gutter — meets the
   left side of his head rather than sitting in empty gradient above him.

   Three values do it, and they are coupled, so change them together:

     .hero-photo height   the band he stands in
     .cut-2 height/bottom how far he overshoots that band, and how far he is
                          pushed back down so his hair clears the copy
     .hero-inner padding-bottom
                          how much room the copy reserves above the band — which
                          is what sets the button's height on the page

   The overlap is the point. Before this, padding-bottom was the band height
   plus 24px, so the copy always cleared him completely and the button could
   never reach his head. Here it is deliberately LESS than his full height, so
   the top of the figure rises past the bottom of the copy.

   Phones only (640px, not the 900px block above): on a tablet the band is at
   its clamp ceiling and the copy column is wide, so the same overlap puts the
   button across his chest instead of beside his head.
   ========================================================================= */
@media (max-width: 640px) {
  .hero-photo { height: clamp(430px, 128vw, 540px); }
  .hero-inner { padding-bottom: calc(clamp(430px, 128vw, 540px) * 1.06); }
  /* He is clipped at the right edge by .hero's overflow: hidden, which is fine
     and intended — he is a cutout standing on the section, not a framed photo.
     Do not push the translate much past this or the clip starts taking his arm
     rather than the empty margin beside it. */
  .cut-2 { height: 132%; bottom: -17%; transform: translateX(21%); z-index: 1; }
}

/* ============================================================================
   THE GLASS PILL beside "Book a consultation" (Alec, 2026-08-16)
   ============================================================================
   Lifted from the header tab, not approximated: same rgba(13,13,13,.62) fill and
   the same blur(22px) saturate(170%). The saturate is the point — it amplifies
   whatever is behind the pill rather than merely blurring it, so the button takes
   a tint from the hero ramp it is sitting on. There is a longer note on
   .nav.is-tab .nav-inner in v9-header-hero.css about why that must not be
   flattened; it applies here too.
   ========================================================================= */
.btn-glass {
  /* FROSTED, to Alec's reference (2026-08-16). Three passes to get here and the
     differences matter:

       .62 black   the header tab's fill — reads as a dark pill, not as glass
       .10 white   more of the ramp came through, but a white tint over a coral
                   gradient goes orange, and it did
       .44 neutral where it landed: a dark, desaturated frost that dims what is
                   behind it instead of tinting it. .30 was the first try and it
                   sat too light over skin — the reference reads clearly darker
                   than its background, and over a bright passage .30 did not

     The saturate comes DOWN to 110% here, unlike the header tab which runs 170% on
     purpose so it picks up the colour of whatever it flies over. This button is not
     supposed to do that — the reference is neutral against a warm background, and at
     170% it took the hero's coral and turned amber. The heavier blur is what makes it
     frost rather than tint. The hairline is faint and there is no inset highlight:
     both were doing the work the fill does now. */
  background: rgba(24, 22, 28, 0.44);
  -webkit-backdrop-filter: blur(30px) saturate(105%);
  backdrop-filter: blur(30px) saturate(105%);
  color: var(--warm-white);
  border: 1px solid rgba(255, 255, 255, 0.12);
  box-shadow: 0 10px 30px rgba(16, 14, 20, 0.22);
  transition: background 0.25s ease, border-color 0.25s ease, transform 0.25s ease;
}
.btn-glass:hover {
  background: rgba(24, 22, 28, 0.56);
  border-color: rgba(255, 255, 255, 0.22);
  transform: translateY(-1px);
}

@media (max-width: 640px) {
  /* ONE SOURCE FOR THE BAND. The hero's bottom padding and the CTA row's offset are
     the same distance measured from opposite ends, so they read it from here. Change
     this and both follow; hard-code either one and they drift apart. */
  .hero { --hero-pad: calc(clamp(430px, 128vw, 540px) * 0.95); }
  .hero-inner { padding-bottom: var(--hero-pad); }

  /* THE CTA ROW MOVES TO THE FOOT OF THE HERO (Alec, 2026-08-16). It used to sit in
     the copy flow directly under the paragraph, which put it in the middle of the
     section with the member below it.

     IT IS POSITIONED AGAINST .hero-copy, NOT .hero — v9-header-hero.css makes
     .hero-copy position: relative so it can carry z-index over the member, and that
     makes it the containing block for anything absolute inside it. A plain
     `bottom: 26px` therefore lands at the foot of the COPY, which is halfway up the
     section. Offsetting by the padding below the copy is what reaches the section's
     own foot. Moving this out of .hero-copy in the markup would be the other fix, but
     it costs the desktop layout, where the row belongs in the copy column. */
  .hero .hero-ctas {
    position: absolute;
    /* PAST THE COPY'S GUTTERS (Mishy, 2026-08-16). At left/right 0 the row was only as
       wide as the container's content box — 342px at 390, 327px at 375 — and that was
       the whole reason the glass pill had to be squeezed to 16px a side to keep the pair
       on one line. Pulling the row out by the gutter gives it the section's full width
       and buys 48px, which is what pays for the wider pill below.

       The buttons stay clear of the screen edge because they are centred and never fill
       this width: the pair measures ~341px inside 390. */
    left: calc(var(--gutter) * -1);
    right: calc(var(--gutter) * -1);
    bottom: calc(26px - var(--hero-pad));
    z-index: 4;
    display: flex;
    flex-direction: row;
    align-items: center;
    /* CENTRED (Mishy, 2026-08-16). The row ran from the left edge, which left the pair
       hanging off the copy's margin with the member's leg in the gap on the right. It is
       absolute with left/right 0 inside .hero-copy, so this centres the pair across the
       full width of the section. */
    justify-content: center;
    /* 8 -> 16 -> 24px (Mishy, 2026-08-17, twice: "move them apart", then "pull them
       apart a little bit more").
       PAID FOR, NOT ADDED, BOTH TIMES. The pair has a hard width budget — one line at
       360px — so gap taken from thin air would have pushed the row past 350 and wrapped
       it. Each round the two buttons gave up the same amount of side padding as the gap
       gained: 17/29 -> 15/21 -> 13/19 -> 11/17. The row has measured 342px through all
       of it, so the edge gaps are untouched and only the space BETWEEN the pills has
       moved.

       24 IS ABOUT THE END OF THAT TRADE. At 11px a side the primary's label is close to
       its own pill; the next 8px of gap would have to come from the type, which is the
       thing the previous round was asked to make bigger. If more separation is wanted
       after this, the honest answer is stacking them, which costs the hero ~50px of
       height. */
    gap: 24px;
    margin: 0;
  }
  /* width: auto is load-bearing — styles.css gives .hero-ctas .btn width: 100% for the
     stacked phone layout, and at 100% these two wrap into a column however the flex
     direction is set. */
  /* SIZE, IN TWO PASSES (Mishy, 2026-08-16). Type went 12.5 -> 13.5px and stayed there.
     Vertical padding went 12 -> 15px and then back to 10px: at 15 the pills stood 44px
     tall and read as slabs — "too thick" — and once the row had been widened they did
     not need the height to feel substantial. 10px puts them at 34px, slimmer than the
     38px they were before any of this, which is what suits the wider pill.

     Width, though, is the dimension with no room, and the side padding is set by it:
     the row must stay on ONE LINE, and the narrowest phone sets the budget. Since the
     row was pulled out past the gutters it has the section's full width — 375px on an
     iPhone SE, 390 on a 14/15/16 — rather than the 327/342 it had inside them.

     Measured with an at-390 layout probe, the pair is 353px, which clears 375 with 11px
     of margin either side. It went to 361px first, at 19/31px padding, and that left only
     7px at 375 — fine at 390 and uncomfortably close to the bezel on an SE. An early attempt at 14px type with 18/20px side padding
     measured 347px against the old 342 budget and wrapped into a column on every phone,
     which grows the hero's foot by a whole button. Measure the rendered widths at 375,
     not just 390, before raising any of these numbers again. */
  /* BIGGER, WITHOUT GETTING WIDER (Mishy, 2026-08-17). "The 2 bottom tabs at the hero
     section seem quite small" — so type goes 13.5 -> 14.5px and the vertical padding
     11 -> 14px, which stands the pills at 41px instead of 34.

     THE WIDTH BUDGET IS UNCHANGED AND THAT IS DELIBERATE. Everything above still holds:
     the pair has to stay on one line at 360px, and the moment it wraps the hero grows by
     a whole button. Bigger type is wider type, and then two rounds of widening the gap
     between the pills wanted room of their own — the side padding paid for all three:
     17 -> 15 -> 13 -> 11 here and 29 -> 21 -> 19 -> 17 on the glass pill below. Measured
     at 345/360/375/390/414 after every step, the pair has held at 342px against the 341
     it was before any of this, so the edge gaps are still the ~9/17/25px the note below
     was written about, to within a pixel.

     14 AND NOT 14.5, which was tried first and measured. At 14.5 the pair came out at
     353px — 3.5px of margin either side at 360, which is the bezel problem the note
     below warns about, and it wrapped outright below 350. Taking the type back half a
     point returned 11px of that and costs nothing visible: the height is doing the work
     in this change, not the type.

     Height was the dimension with room, which is why it took most of the increase. The
     "too thick" 44px slab noted above was 15px of padding on 13.5px type in a 34px-tall
     row of pills; 14px on 14.5px type is a pill of the same proportion, scaled. */
  .hero .hero-ctas .btn {
    flex: 0 1 auto;
    width: auto;
    padding: 14px 11px;
    font-size: 14px;
    white-space: nowrap;
  }
  /* THE GLASS PILL, UNSQUEEZED (Mishy, 2026-08-16). It went to 16px a side when the row
     was still boxed in by the container's gutters, and at that value the label sat hard
     against both ends — Mishy read it as squeezed, which it was.

     With the row now spanning the full section width there is room to give it back and
     more: 16 -> 26px a side. The pill goes 143 -> 163px, so it is finally the wider of
     the two (the primary is 170px including its arrow), which is the "longer than the
     one beside it" Alec asked for.

     26 AND NOT 30, which was tried first. The ceiling here is not whether the pair fits
     but how close it comes to the screen edge once it does: at 30 the pair measured
     349px and left only 5px of margin either side on a 360px Android. At 26 it is 341px,
     which keeps ~10px at 360, 17px at 375 and 25px at 390. Measure the edge gap, not
     just the fit, before raising this again. */
  /* 29 -> 17 (Mishy, 2026-08-17), and this is NOT the pill being squeezed again — it is
     the type getting bigger and the pill holding the same outside width. See the note on
     .btn above: at 14px the label itself is ~6px wider, so 17px a side lands the glass
     pill at ~149px. It has given up 12px a side in total — some to the bigger type, the
     rest to the gap between the pair (see the gap note above). It is still the wider of
     the two relative to its label, which is the property Alec asked for. */
  .hero .hero-ctas .btn-glass { padding-left: 17px; padding-right: 17px; }

  /* HIGHER, AND CUT AT THE THIGH. Two numbers, and they move together — their
     DIFFERENCE sets where his head lands and `bottom` alone sets how much of him falls
     below the section edge.

       height 132 -> 142   he grows a little
       bottom -17 -> -24 -> -50 -> -45 -> -41   and is pushed down by more than he grew

     `bottom` went to -50% on 2026-08-16 because at -24% his face sat 115px INTO the
     paragraph — measured, not guessed: the copy's foot is 474px above the section's
     bottom edge and the top of his hair was 589px above it. The arithmetic is

       model top above the section's foot = (height - |bottom|) x band

     with the band at 499px on a 390px screen. -50% put his hair 459px up, 15px CLEAR of
     the copy, which Alec found too low; -45% put his hair 484px up, about 10px into the
     last line; -41% gives (1.42 - 0.41) x 499 = 504px, so roughly 30px of hair now sits
     behind the copy. His eyes are ~90px below the crown in this file, so at this value
     they are still well clear of the text — that is the number to watch if it is raised
     again. If the copy or the band ever changes, redo the sum rather than nudging by eye.

     translateX 21% -> 17% at the same time, at Alec's ask; that one is independent of
     the vertical pair.

     -41% -> -36% (Mishy, 2026-08-16), "a little bit higher". By the sum above that is
     (1.42 - 0.36) x 499 = 529px, so he rises 25px and about 55px of hair now sits behind
     the copy. His eyes are ~90px below the crown, which puts them at 439px — still 35px
     BELOW the copy's foot at 474px, so the face stays clear of the text. That 90px is the
     number that runs out first: past about -31% his eyes reach the last line. */
  .cut-2 { height: 142%; bottom: -36%; transform: translateX(17%); z-index: 1; }
}
