/* ===================================================================== OUR APPROACH — page sheet for src/pages/our-approach.html No colour value appears here. Everything refers to css/tokens.css. SCOPE IS BY SELECTOR THROUGHOUT. Every rule is reached through `.a-hero`, `.a-cta`, `.jcol-lead`, or the two dark bands' own `.vm*` and `.tlw*`/`.beacon*` families. Nothing here is keyed to a position in the document or to a count of anything. THE FOUR PAGE-CHROME RULES WERE COPIED AS AUTHORED, not re-derived. `.a-hero .kick` and `.a-hero .body` are reached through the hero wrapper; `.a-cta` styles the hero's call-to-action paragraph; `.jcol-lead` styles the copy column of the join band. THE TWO BANDS' RULES ARRIVED BYTE-IDENTICAL and are governed by the blocks that precede each of them. **SINCE G3, 2026-08-16, ONLY ONE OF THEM IS DARK: `.vm-band` is --cream and `.tlw-band` is --ground-dark.** The `--accent`-on-dark surfaces in this sheet are `.beacon-badge` x6 in `.tlw-band` and NOTHING ELSE — `.vm-label` left that group when the band went sand and is now --accent on --cream. The token warning in the §2 block below is still load-bearing for this file, but it is now a warning about TWO DIFFERENT GROUNDS with two different figures. Read it before touching `--accent` anywhere, and read which ground each figure is on. `.jcol-lead` IS AUTHORED HERE RATHER THAN IN components/join-form.css BECAUSE THE ELEMENT IT STYLES IS NOT IN THAT PARTIAL. join-form.html carries no copy column. The paragraph lives in this page's own join wrapper, so the rule lives exactly as wide as the element does. A page whose sheet omits this rule inherits nothing for that paragraph — the shared sheet has never carried it. ===================================================================== */ /* ---- PAGE-LOCAL DEVIATION REGISTER ---------------------------------- Prefix OA-. Numbered from 1 and never renumbered. OA-1 · THE JOIN BAND'S COPY IS BORROWED, NOT WRITTEN FOR THIS PAGE. partials/components/join-our-approach.html carries the heading and lead paragraph of partials/components/join-about.html, character for character, on instruction and as a placeholder. It is NOT copy chosen for this page and no part of it was authored here. When the client supplies per-page join copy, that wrapper is rewritten and this entry is closed. Until then the two files say the same thing by borrowing, not by design — changing one does not change the other. OA-2 · THE FOUR HERO STRINGS ARE DIRECTOR-DRAFTED, NOT CLIENT-SUPPLIED. Eyebrow, h1, body and CTA in src/pages/our-approach.html were drafted and approved 2026-08-14 and are set exactly as given — no typographic correction, no re-wrapping, no smart quotes. Same status as OA-1: they stand until the client supplies copy written for this page, and replacing them is a copy change. Verified at authoring by codepoint: the body's dash is U+2014 and the CTA's arrow is U+2192 preceded by one U+0020. OA-3 · THE META DESCRIPTION IS BESPOKE AND DIRECTOR-DRAFTED. It was drafted and approved 2026-08-14 and is set exactly as given, with the dash as U+2014. It no longer duplicates the hero body — an earlier draft of this page repeated that string verbatim, which is the wrong shape for a description and roughly twice the useful length. Same status as OA-1 and OA-2: director-drafted, not client-supplied. AUTHORING WAS THE ONLY PATH, AND THAT WAS ESTABLISHED BY MEASUREMENT RATHER THAN ASSUMED. A mechanical derivation was looked for first. Across the seven pages that predate this one, NONE opens with its title's subject, NONE contains its own h1, and no two share an opening construction; lengths run 119 to 181 characters and sentence counts run one to three, and even the em dash appears in only some of them. Each is a bespoke editorial summary of its own page's argument. There is no pattern to follow, so no derivation was available and none was invented. A CLIENT-SUPPLIED DESCRIPTION REMAINS OPEN. Replacing this one changes a single attribute in src/pages/our-approach.html and this entry with it. OA-4 · TWO DARK SECTIONS ARRIVED HERE BYTE-IDENTICAL, 2026-08-14. Moved from the about page by director ruling: the Vision/Mission band landed as S2, The Longitude Way as S3, both immediately after the hero. Verified by sha taken FROM THIS FILE AND FROM src/pages/our-approach.html AFTER placement, never carried forward from the cut:
895 B sha 6f50dc73…
2,663 B sha ff613b36… THE THIRTY RULES ARRIVED BYTE-IDENTICAL AT THE DECLARATION LEVEL — comments-stripped and whitespace-normalised, §2 1,639 B sha f4708aab… and §8 1,650 B sha 12b9d392…, both found verbatim in this file and absent from the origin sheet. Block-level shas could not survive, because the ruling required several comments inside those blocks to be rewritten; the declarations are the part that had to be exact and they are. The containment test was validated by mutation — an altered copy is NOT found. ALL FOUR FIGURES IN THIS ENTRY ARE AS AT 2026-08-14 AND ARE A RECORD OF WHAT ARRIVED, NOT A DESCRIPTION OF CURRENT DISK. **§2's declarations were rewritten by G3 on 2026-08-16, so §2 1,639 B / sha f4708aab… NO LONGER MATCHES THIS FILE AND IS NOT MEANT TO.** It is not corrected and must not be: re-deriving it against today's disk would destroy the only evidence that the section arrived unaltered, which is the single thing this entry exists to record. §8 / 1,650 B / sha 12b9d392… and both markup shas are untouched by G3 and still resolve. Same class as the 9ag lesson — an old checkpoint's figure drifts with every unit that lands after it and is still history, not drift. **"TWO DARK SECTIONS" likewise describes what arrived; since G3 only .tlw-band is dark.** THE DOTTED-DIVIDER BLOCK TRAVELLED WITH §2 DESPITE ITS OWN HEADING. It sits under "the dotted gold divider" and reads as a separate concern, but it styles `.vm-band .vm-item + .vm-item::before` and belongs to the Vision/Mission band. Leaving it behind would have stranded a rule with no element and removed the divider from the band that needs it. THE §53 --accent PAIR MOVED AS A PAIR, WHICH WAS THE NAMED HAZARD. Both halves — the markup half in the §2 client-header comment, the sheet half in the §2 dark-treatment block below — sat inside slices that move as units, so neither could land without the other. All four figures survive verbatim: #696FE8 / 3.2204 / #4E65DC / 2.6959. Four pointers were repointed at the new filenames and all four resolve. BOTH HALVES NOW CARRY THE HAZARD IN-FILE — "if this section ever moves again they move as a pair or not at all" — rather than only in this register, because a register is read when someone goes looking and a comment is read by whoever is already editing. THIS SHEET NOW HOLDS BOTH REMAINING --accent-ON-DARK SURFACES IN THE TREE: `.vm-label` x2 and `.beacon-badge` x6. **SUPERSEDED BY OA-5 ON 2026-08-16 — `.vm-label` LEFT THE DARK GROUP WHEN .vm-band WENT SAND. The --accent-on-dark set in this sheet is now `.beacon-badge` x6 alone.** OA-5 · THE VISION/MISSION BAND WENT SAND AND LOST ITS TEXTURE, 2026-08-16, unit G3. Director change. `.vm-band` moves --ground-dark -> --cream, the ::before texture layer is REMOVED rather than ported, and three foregrounds re-derive. `.tlw-band` DOES NOT CHANGE — it keeps its plum ground and its texture, and the two bands must not be reasoned about together any more. THE TEXTURE WAS REMOVED BECAUSE NO LIGHT-GROUND RECIPE EXISTS AND INVENTING ONE WAS NOT THIS UNIT. Every ::before texture in this tree is authored against a dark ground: the layer is --slate at opacity .10, which READS AS LIGHTENING on plum and would read as DARKENING on sand. Porting it would have been a new recipe, not a move. TWO DECLARATIONS DIED WITH IT AND BOTH WERE VERIFIED DEAD BEFORE REMOVAL, NOT ASSUMED. `.vm-band{position:relative}` existed only as the containing block for the layer's `inset:0`, and `.vm-band > .in{position:relative; z-index:1}` existed only to stack content above it. Removed all three together and measured at 1440 / 900 / 899 / 768 / 390 — BOTH SIDES OF THE DIVIDER'S OWN 900px BREAKPOINT: geometry identical to three decimals, document height unchanged, horizontal overflow 0. **`.vm-band .vm-item {position:relative}` IS NOT DEAD** — it is the containing block for the dotted divider and is independent of the texture. Do not remove it. THE ASSET IS NOT ORPHANED. `.tlw-band::before` in this same sheet still masks assets/longitude-texture.svg, and four other sheets reference it. Confirmed after the removal, not before. FOREGROUNDS, RE-DERIVED ON THE BARE SAND GROUND rgb(246,240,230). There is no composited ground here any more, so these are the ground's own figures and not an inked variant — which is itself the change from the block below: h2 --gold -> --accent 8.2161 -> 3.6759 LARGE 3.0 PASS h3.vm-label --accent, UNCHANGED 3.5594 -> 3.6759 LARGE 3.0 PASS p.vm-text --on-dark-body -> --on-page-body 8.6943 -> 7.5580 SMALL 4.5 PASS divider --gold -> --stone 8.2161 -> 1.8550 decorative **TWO OF THOSE FOUR ROWS WERE SUPERSEDED THE SAME DAY BY OA-6 AND THE FIGURES ABOVE ARE LEFT AS G3's RECORD, NOT CORRECTED.** The h2 is --ink at 13.0839 and the divider is --gold at 1.5925 since G4. The rows stay because this entry records what G3 changed and from what — re-deriving them would erase the only account of the ground move. **Read OA-6 for current values; read these for what happened.** Same handling as OA-4's arrival shas. --on-page-body RATHER THAN --ink-body, THOUGH THEY ARE ONE VALUE rgb(78,74,90). Both give 7.5580. This sheet's light-ground vocabulary is --on-page-body throughout (`.vm-text` base, `.jcol-lead`), and matching it keeps one name for one job on one ground. THE INHERITED FOREGROUND WAS A LATENT TRAP WITH NO CURRENT SYMPTOM, AND THAT IS WHY IT IS RECORDED RATHER THAN JUST FIXED. `.vm-band` carried `color:var(--on-dark)`, and **--on-dark IS --cream — one value.** On the new ground that inherits at 1.0000, invisible. NOTHING LOOKS BROKEN TODAY AND NOTHING WOULD HAVE: every descendant sets colour explicitly, so the band has no element that inherits. Enumerated to confirm it — section, .in, h2, .vm, .vm-item x2, .vm-label x2, .vm-text x2, and only the three text elements carry text, all three explicit. The trap is for whoever adds the fourth. Changed to --on-page in the same edit. THE h2 AND THE LABEL PASS FOR DIFFERENT REASONS AND MUST NOT BE REASONED ABOUT TOGETHER. The h2 is 46 / 34.952 / 28 / 28 px at 700 and is LARGE BY SIZE ALONE, clearing the 24px any-weight boundary by 4px at its floor. The label is clamp(21px, 1.4vw + 8px, 28px) — 28 / 22.392 / 21 / 21 — and is LARGE ONLY VIA 18.66px AT WEIGHT 700, with 2.34px of headroom at its floor. **The label's floor is still the accessibility decision it always was; only the ground under it moved.** Lower that floor below 18.66 and the threshold flips 3.0 -> 4.5 and 3.6759 becomes a failure, exactly as before. THE SEAM AT THE FOOT OF THIS BAND IS NOW A HARD LIGHT/DARK EDGE AND IT CARRIES NO RULE. .vm-band's bottom and .tlw-band's top coincide to a 0.001px subpixel — y=1042.531 at 1440, 1034.672 at 1028, 1227.609 at 768, 1511.859 at 390, gap 0.000 throughout, unmoved by this unit. The step goes 1.0000 (plum meeting plum, an invisible seam) to **13.0839**. Ruled adequate without a treatment, 2026-08-16: 13.0839 is the largest ground step this palette offers, and .tlw-band -> #join already runs plum->paper at 13.9920 with no rule immediately below it on this same page. What this splits is the 1,416px of continuous dark G1 measured — now 526px sand above 889px plum. THE HERO SEAM IS THE NEW SOFT ONE. hero -> .vm-band at y=516.266 goes 13.9920 -> 1.0694. ACCEPTED, director-ruled with the other four this unit creates. 9an routed a paper->cream step through plum; that precedent governed two argued sections carrying continuous prose and does not bind a hero handing off to a labelled band. OA-6 · TWO COLOUR CORRECTIONS AFTER SEEING 9bc RENDERED, 2026-08-16, unit G4. Director rulings taken on the built page, not on the mock. Colour only: no size, weight, spacing or geometry moved, and both changes measured metric-free at 1440 / 1028 / 768 / 390. .vm-band > .in > h2 --accent -> --ink 3.6759 -> 13.0839 LARGE 3.0 dotted divider --stone -> --gold 1.8550 -> 1.5925 decorative THE h2 GAINS HEADROOM AND LOSES NOTHING. --ink on --cream is the same pair the Governing Members h2 uses on our-team.html, measured there at 13.0839 in this same checkpoint. It is LARGE by size alone at 46/34.952/28/28 @700, so it was never resting on the 18.66/700 boundary and still is not. **--accent in this band is now `.vm-label` x2 and nothing else.** ############################################################ HAZARD (OA-6) — THE DIVIDER IS GOLD ON A LIGHT GROUND, AND THAT IS A KNOWN FLAGGED PATTERN — READ THIS BEFORE ACTING ON EITHER SIDE OF IT. Re-headed at C3, 2026-08-17. It read `OA-6 · …`, which MATCHES THE PATTERN THAT ENUMERATES ENTRIES IN THIS LOG, so the register counted 7 headings for 6 entries. Same defect as the two `TEAM-3` headings in our-team.css. A sub-heading inside an entry must not wear the entry shape. **THIS REVERSES G3's RULING 4 DELIBERATELY. G3 moved this divider --gold -> --stone specifically because gold falls to 1.5925 on sand; G4 moves it back, on a director ruling taken after seeing the rendered page.** It is not drift and it is not an oversight — it is a decision that the fainter gold reads better here than --stone's 1.8550, which was barely brighter anyway. **GOLD ON A LIGHT GROUND IS FLAGGED FOR RICK UNDER R4 AND THIS UNIT ADDS A SECOND DELIBERATE INSTANCE.** R4's flag is the client's gold emphasis on "Longitude is changing that." on why-longitude.html, which renders violet because --gold measures 1.7030 on --paper and fails even 3.0. That is a COPY-EMPHASIS decision on another page; this is a DECORATIVE RULE on this one, and the two are different nouns. **They share a cause: --gold has no contrast against any light ground in this palette.** **IF R4 RULES AGAINST GOLD ON LIGHT GROUNDS, THIS DIVIDER IS IN SCOPE FOR THAT RULING.** Whoever executes R4 must sweep for --gold on light grounds rather than only revisiting the why-longitude emphasis, and this entry is written to be findable from that side: the instances are `why-longitude.html`'s gold emphasis (1.7030 on paper) and `.vm-band .vm-item + .vm-item::before` here (1.5925 on cream). Reverting this one is a two-token edit in the two media queries below. **NO SC THRESHOLD BINDS EITHER FIGURE AND THAT IS NOT THE POINT.** The divider is decorative — it separates two already-labelled blocks and carries no information the labels do not — so 1.5925 is legal and always was. The flag is a legibility and brand-consistency question, not a compliance one, and it stays open under R4 either way. ############################################################ --------------------------------------------------------------------- OA-7 · A TRUNCATED BRAND NAME IS COMPLETED IN CLIENT-APPROVED COPY. DIRECTOR-DIRECTED, C3, 2026-08-17. One word appended to the sixth beacon description, in `pages/our-approach.html`: before "...share success, and move forward as One" after "...share success, and move forward as One Longitude" **+10 bytes. No terminal period added — neither string carries one, and that remains deliberate.** The other five descriptions end in U+002E; this one does not, and OA-7 does not change that. ── PROVENANCE: THIS IS APPROVED COPY, NOT AUTHORED-IN-BUILD. The section's copy-provenance block records the six value DESCRIPTIONS as **director-approved new copy that SUPERSEDES T04**, and the sixth as "deliberate transcription from the approved source". **So this is a directed change to approved copy and is logged as a deviation, not as a plain edit.** The five FROZEN paragraphs and the six titles are untouched. ── ⚠ THE OLD COMMENT DEFENDED THE RIGHT CHARACTER FOR THE WRONG REASON. `our-approach.html` carried a standing instruction: the bare capital O in "as One" is deliberate, **do not lowercase it.** That instruction was correct and is now explained rather than merely asserted — **"One" was the first word of the brand name "One Longitude", and the transcription was TRUNCATED.** The capital was never a style choice; it was the start of a proper noun that got cut. **A rule that is right for an unknown reason survives only until someone tidies it.** Had a later editor "corrected" `One` to `one` on the grounds that a sentence-internal capital looked like a typo, the truncation would have become invisible. The comment has been rewritten in place to say what the capital is FOR. **The no-terminal-period half of that instruction is untouched and still binding.** ── THE SIX ITEMS ARE SIX LITERAL STRINGS, NOT A REPEATED PARTIAL. `ul.beacon` holds six inline `
  • ` blocks, badges B/E/A/C/O/N spelling BEACON. **There is no partial to fork and no loop to edit once.** Changing one description leaves the other five untouched; a future editor asked to change "the beacon copy" must edit each literal. This edit landed in exactly one of the six. ── GEOMETRY: NOTHING MOVED, INCLUDING INSIDE THE ITEM. Ten characters were expected to risk a reflow at narrow widths. **Measured at 1440 / 1028 / 768 / 390 before writing: dH 0.000 on `.beacon-text`, `.beacon-item`, `ul.beacon`, the footer and the body at every width.** The added word fits the existing line count throughout. Controlled — appending 36 characters instead moved the item by **+26.391**, one line-height — so the zero is a measurement, not a blind probe. ── NOT IN SCOPE, AND DELIBERATELY SO. `span.beacon-badge` x6 sits in this same list and regressed to **2.5821** on plum under C1. **Accepted there as decorative and aria-hidden; C3 does not revisit it.** Nothing gold on this page was touched either (see WL-28 for the `em.g` hook on why-longitude). Both were in the director's screenshot alongside this string and both stay as they are. OA-8 · ONE PERIOD IS APPENDED TO THE SIXTH BEACON DESCRIPTION. CLIENT-REQUESTED (BugHerd #24), N4, 2026-08-18. before "…share success, and move forward as One Longitude" after "…share success, and move forward as One Longitude." **+1 B. One character, one sentence, one page.** All six descriptions now end in U+002E; the sixth was the only exception and no longer is. ── ⚠ THIS REVERSES OA-7's RULING, AND OA-7 IS NOT WRONG. C3 recorded, at this same sentence: *"The no-terminal-period rule is unaffected and still deliberate."* **The client has now asked for the period.** OA-7 is left standing — **it did not add or remove punctuation; it COMPLETED A TRUNCATED BRAND NAME**, and its finding about the capital O is untouched and still correct. Only its punctuation half is superseded. The markup comment that carried the no-period instruction has been rewritten in place rather than left reading as live. ── A SIXTH DIRECTION FOR TERMINAL PERIODS, AND THE SECOND "ADD". index.html:167 client dropped it, the build leaves it dropped WL-24 client supplied none, the build adds none UTIL-3 client supplied one, the build removes it (deviation) C3 / OA-7 the string was TRUNCATED; no period was added or removed TEAM-4 client REPLACES a terminated string with an unterminated one OA-8 client asks that a period be ADDED to a string that had none **Six directions, six records. None generalises to another.** The MD chip "Snehal Patel, MD., Executive in Residence" is untouched by all six — **and it is not on this page**: it lives in our-team.html and utilities.html, confirmed by measurement rather than assumed. ── ⚠ THE BLAST RADIUS OF A NAIVE REPLACE-ALL, ON RECORD. The word "Longitude" occurs **351 times in CODE and 54 times in COMMENTS across src/, 405 raw, in 31 files.** This edit touched **one**. **Any sweep, replace-all or pattern match on that word is forbidden here** — the target is a single sentence, matched on its full text, asserted unique before writing. **N0 reported 401 and that figure has since moved.** It was a raw count taken before N1, N2 and N3, and the register prose those units added carries the word too. **A blast-radius figure ages the moment anyone writes about the thing it counts** — re-derive it, and split code from comments, because a replace-all hits both and only one of them renders. ── SCOPE AND PROVENANCE. The values band lives **directly in our-approach.html — zero beacon markup in any partial**, so there is no propagation. The second textual match at `:201` is C3's own comment prose, not an occurrence (§25). **No T-document sources this string**: it is director-approved copy that supersedes T04 for this section, per the provenance block above. */ .a-cta{margin-top:32px} .a-hero .body{max-width:60ch} .jcol-lead{ margin-top:22px; font-size:17.5px;line-height:1.6; color:var(--on-page-body); max-width:46ch; } .a-hero .kick{font-size:14px} /* ---- §2 vision / mission -------------------------------------------- Vision: and Mission: are the CLIENT DOCUMENT'S OWN labels and they render, unlike Header:, Body: and CTA:, which name a slot rather than a thing and do not. Without these two labels §2 is two unlabelled paragraphs of similar length with no way to tell which is which. THE TRAILING COLON IS BACK, AND THAT REVERSES THE LOGGED DEVIATION THIS COMMENT USED TO RECORD. It was dropped when §2 was built; the 2026-08-11 mock restores it, and the client document has it. It is restored as .vm-band .vm-label::after{content:":"} rather than in the markup, because the copy string is frozen for this unit — the h3 text node still reads exactly "Vision" and "Mission". A CSS-generated colon is not in the accessibility tree the same way as authored text; that is acceptable here because the colon carries no meaning the label does not already carry, but it is the reason this is a note and not a silent one-liner. If the colon is ever wanted as real content, it belongs in the markup and this rule should go. */ .vm{margin-top:36px;display:grid;gap:40px} @media(min-width:900px){.vm{grid-template-columns:1fr 1fr;gap:56px}} /* min-width:0 is not cosmetic. A GRID ITEM'S DEFAULT min-width IS auto, so a long unbroken token in the prose can push the track wider than its 1fr share and overflow the column. Any fixed-fraction grid child holding arbitrary prose needs this; it is not specific to these two items. This rule was added because the class-resolution probe reported .vm-item as a class with no rule behind it — the check exists to catch a hook that styles nothing, and it caught one of mine. */ .vm-item{min-width:0} .vm-label{ font-family:var(--mono);font-size:11px;letter-spacing:.26em; text-transform:uppercase;color:var(--on-page-soft); margin-bottom:14px;font-weight:400; } .vm-text{font-size:17.5px;color:var(--on-page-body);max-width:52ch} /* ---- §2 SAND TREATMENT — director change, 2026-08-16, unit G3 --------- Visual only. Not one character of the vision or mission copy moves. This block REPLACES the dark treatment of 2026-08-11; that treatment's own figures are gone with the ground they described, and OA-5 records what was removed and why. THE GROUND IS STILL BUILT HERE, NOT BORROWED, AND STILL NOT .band-cream. .vm-band restates its own ground for the same reason it always did — a bands.css class would bring `.band-cream .in` layout with it, and bands.css is barred to this session. It is also a §46 decision: the .band-cream RULE lives in exactly one sheet, components/bands.css:19, and the class renders on 11 dist pages, so reusing it would inherit every future change to it on a page-local surface. THERE IS NO TEXTURE HERE AND NO COMPOSITED GROUND. Every ratio below is measured against the bare token, rgb(246,240,230), luminance 0.87626 — NOT against an inked variant. The dark treatment's figures were inked ones and none of them carries over. The ::before layer was removed rather than ported because --slate at opacity .10 lightens plum and would darken sand; there is no light-ground recipe in this tree. --ink h2 46/34.952/28/28 @700 LARGE 3.0 13.0839 --accent .vm-label 28/22.392/21/21 @700 LARGE 3.0 3.6759 --on-page-body .vm-text 19px/400 SMALL 4.5 7.5580 --gold divider 2px dotted, decorative 1.5925 THE h2 AND THE DIVIDER TOOK THEIR CURRENT VALUES IN G4, NOT G3 — see OA-6. G3 shipped the h2 as --accent at 3.6759 and the divider as --stone at 1.8550; both were corrected the same day on a ruling taken from the rendered page. **--accent in this band is now `.vm-label` x2 and nothing else.** THE h2 AND THE LABEL NO LONGER SHARE A TOKEN, AND ONLY ONE OF THEM IS FRAGILE. Since G4 the h2 is --ink at 13.0839 — LARGE by size alone (its 28px floor clears the 24px any-weight boundary by 4px) and carrying so much headroom that its classification stopped mattering. **The label is the fragile one and always was:** --accent at 3.6759, LARGE ONLY via 18.66px at weight 700, with 2.34px of headroom at its 21px floor. **That floor is an accessibility decision and remains one.** Lower it below 18.66 and the threshold flips 3.0 -> 4.5 with no visible symptom: the ratio will not move, and 3.6759 will simply stop being enough. **Until G4 both elements sat at the same 3.6759 and it was easy to reason about them as a pair; they were never one, and now they do not even look like one.** THE LABEL'S RATIO WENT UP, NOT DOWN, WHEN THE GROUND WENT LIGHT. 3.5594 on bare plum -> 3.6759 on sand. Its margin over 3.0 goes +0.5594 -> +0.6759. The 2026-08-12 resize to 28/700 that closed this label's SC 1.4.3 failure is therefore still doing its work and must not be read as made redundant by the ground change. THE TOKEN WARNING SURVIVES THE GROUND CHANGE BUT ITS FIGURES DO NOT, AND THE DIFFERENCE MATTERS. Stated on its own terms for THIS ground: --accent is #696FE8 and this label measures 3.6759 on --cream against the 3.0 it needs at 21/700. **THE PROPOSED --accent #4E65DC MEASURES 4.3910 HERE — a LARGE margin of +1.3910, where on the old plum ground it was 2.9797 and FAILED.** Moving this band to sand did not merely preserve the token exposure, it substantially reduced it FOR THIS ELEMENT. Two cautions: 4.3910 still MISSES the SMALL threshold by 0.1090, so the 18.66/700 classification is what carries it either way; and **this says nothing about `.beacon-badge`**, which is still --accent on the plum .tlw-band and is where this sheet's real token exposure now lives. Read the §8 block for that one — do not carry this figure to it. THE --accent-ON-DARK SET IN THIS SHEET IS `.beacon-badge` x6 AND `.vm-label` IS NOT IN IT. The claim this block used to carry — that .vm-label was the only --accent-on-dark instance in the tree — was wrong in both directions before G3 even touched it: the tree held ten such instances, not one, and two of them were never on this page. It is now wrong in a third way, because this label is not on a dark ground at all. NO ROSTER IS RESTATED IN ITS PLACE. Count them when you need the number; a written count of this has gone stale four times in nine days. THIS NOTE HAS A TWIN in the §2 client-header comment of src/pages/our-approach.html, which carries the same warning and the same figures. R7 corrected one and found the other only by sweeping for it; A4 moved both together; G3 rewrote both together. The two must say the same thing — if you correct one, correct the other, and if this section ever moves again they move as a pair or not at all. THE COLON after each label is CSS-generated (.vm-label::after) so the copy string stays frozen; the h3 text nodes still read exactly "Vision" and "Mission". Unaffected by this unit. TRACKING IS `normal`, AND THAT IS A MEASURED CHOICE unaffected by the ground. The base `.vm-label` rule carries `letter-spacing:.26em`, mono display tracking for 11px caps; at 28px that resolves to 7.28px per character and Charter's caps do not want it. Measured at 28/700, `MISSION:` renders 127.28px at `normal` against 185.52px at .26em, in a column 252px at its narrowest. `normal` is the literal removal and invents no constant — .02em / .04em measure 131.77 / 136.25. --------------------------------------------------------------------- */ /* NO `position:relative` AND NO ::before. Both were removed with the texture in G3 and both were verified dead first, not assumed — see OA-5. `.vm-band .vm-item{position:relative}` further down IS still needed; it is the divider's containing block and has nothing to do with the layer that left. */ .vm-band{background:var(--cream);color:var(--on-page)} .vm-band > .in > h2{color:var(--ink);text-align:center} /* THE FLOOR IS THE ACCESSIBILITY DECISION, NOT THE CEILING. 21px is not a taste value: --accent measures 3.6759 on this ground since G3 took it to --cream (3.2204 on the inked plum before it), which clears 3.0 and misses 4.5, so the label is legal ONLY while it classifies LARGE — i.e. while it stays at or above 18.66px AT WEIGHT 700. The clamp floor is what guarantees that at every width, and 21px leaves 2.34px of headroom above the boundary. ANY EDIT THAT LOWERS THIS FLOOR BELOW 18.66px SILENTLY REINTRODUCES AN SC 1.4.3 FAILURE — the ratio will not move and nothing will look broken; only the threshold flips 3.0 -> 4.5. The ceiling and slope are the typographic half. h2 is clamp(28px,3.4vw,46px) and BOTTOMS OUT AT 28px, so a flat 28px label was exactly the h2's size at 768 and below. 1.4vw+8px reaches the 28px ceiling at 1440 and sits at the 21px floor from ~930px down, which keeps the label subordinate everywhere the h2 is on its floor. */ .vm-band .vm-label{ font-family:var(--serif); font-size:clamp(21px, 1.4vw + 8px, 28px); font-weight:700; letter-spacing:normal; color:var(--accent); text-align:center; } .vm-band .vm-label::after{content:":"} /* 19px, NOT 20. At 20px the 52ch cap resolves to 578.24px and exceeds the 562px track at 1440, so the COLUMN starts governing the measure and the cap that was tuned for this copy stops binding. 19px resolves to 549.33px and stays inside the track. Line counts are unchanged at 1440 either way — a ch-based cap widens with the font (§42) — so the 20px cost would have been invisible at the width it was chosen for and paid at the others. */ .vm-band .vm-text{font-size:19px;color:var(--on-page-body);text-align:center;margin-inline:auto} /* ---- the dotted divider --------------------------------------------- ONE MECHANISM, ROTATED, SO THE STACKED STATE IS NOT A SECOND DESIGN. .vm is two columns only at >=900px (the breakpoint is .vm's own, 40px of gap below it and 56px above). Below 900 the items stack and a VERTICAL rule has nothing to divide, so the same absolutely-positioned pseudo-element is laid horizontally into the row gap instead. Each offset is HALF the gap that applies at that width, so the rule lands in the centre of the gutter and NO CONTENT MOVES in either state: -28px of 56, -20px of 40. The divider is drawn on the SECOND item, so a future third .vm-item would get one too rather than the rule being pinned to a count. Decorative: it separates two already-labelled blocks and carries no information the labels do not. Reported regardless. --gold, AND IT WENT --gold -> --stone -> --gold IN ONE DAY. It was gold against the dark ground, where it measured 8.2161. G3 moved it to --stone when the band went sand, because gold falls to 1.5925 there. **G4 moved it back on a director ruling taken from the rendered page — the fainter gold reads better here, and --stone's 1.8550 was barely brighter anyway.** Both swaps are recorded rather than one overwriting the other, so the next reader sees a decision that was made twice and not a value that drifted. No threshold binds either figure — this rule is decorative and carries no information the labels do not — so 1.5925 is legal, and so was 1.8550. The requirement is a hairline, not a rule that competes with the two labels either side of it. For the record, the alternatives measured on --cream: --ink-soft 4.6361 and --accent 3.6759, both of which would draw a divider heavier than anything else in the band, and --stone 1.8550. **GOLD ON A LIGHT GROUND IS FLAGGED FOR RICK UNDER R4 AND THIS IS A SECOND DELIBERATE INSTANCE OF IT. If R4 rules against gold on light grounds, this divider is in scope for that ruling** — reverting it is a two-token edit in the two media queries directly below. OA-6 carries the full entry and names both instances so the sweep can be run from the R4 side. Do not read this swap as closing that flag; it widens it. */ .vm-band .vm-item{position:relative} @media(min-width:900px){ .vm-band .vm-item + .vm-item::before{ content:"";position:absolute;pointer-events:none; top:0;bottom:0;left:-28px; border-left:2px dotted var(--gold); } } @media(max-width:899.98px){ .vm-band .vm-item + .vm-item::before{ content:"";position:absolute;pointer-events:none; left:0;right:0;top:-20px; border-top:2px dotted var(--gold); } } /* ---- §8 the longitude way ------------------------------------------- The row of asterisks between the six values and the closing three paragraphs is a MANUSCRIPT DIVIDER in the source and is NOT RENDERED — not as text, not as an
    , not as a decorative rule. There is deliberately no rule here to style one. */ /* ---- §8 DARK TWO-COLUMN + BEACON CONNECTOR — 2026-08-12 -------------- THE GROUND IS BUILT HERE, AND THIS IS THE FOURTH LOCAL RESTATEMENT OF A bands.css GROUND ON THIS PROJECT. `.band-plum` would have imported `.band-plum .in{display:flex;gap:5vw;align-items:center}`, and this section's `.in` holds one block-level grid child — a lone flex item at `flex:0 1 auto` sizes to content and can exceed the track. The ground is restated locally instead. **RECORDED, NOT SILENT: this is one of four places on the project where a ground is re-declared locally rather than reached from a component sheet, each for the same reason — reaching the component would have crossed a write grant or imported a layout contract. The other three are elsewhere in the tree; the count is a consolidation backlog scheduled post-launch, and it is deliberately NOT restated as a roster here, because a roster of files goes stale the moment one of them moves.** The two on THIS page are the band above and this one. ################################################################ FLAGGED FOR RICK — THE PLUM-ON-PLUM SEAM. This band and the one above it are BOTH --ground-dark and they abut: measured on the built page, the upper band's bottom edge and this band's top edge are both at y=1043, and the join is 220px of continuous dark (110 + 110 of section padding, flat at every width). NOTHING MARKS THE BOUNDARY BUT THIS BAND'S GOLD h2 — no rule, no border, no ground change. AND THE TEXTURE DOES NOT RUN CONTINUOUSLY ACROSS THE JOIN. Each band paints its own ::before at inset:0 with mask-position 0% 0%, so the 18px tile grid restarts at each band's own top edge. The upper band measured 526px tall = 29.22 tiles, so the phase RESETS mid-surface rather than coincidentally aligning. A continuous-texture treatment across stacked bands is a BRAND decision, not a build one. Shipped as-is by ruling. ################################################################ TEXTURE IS #fragmentation's RECIPE, unchanged: --slate masked by longitude-texture.svg at 18px, opacity .10. Every ratio in this block was measured against the COMPOSITED ground rgb(46.3,45.8,57.3), never bare plum — the texture LIGHTENS plum, so the inked end binds for every foreground here. EVERY COLOUR IS AUTHORED. A new dark section inherits NONE of the four hand-rolled `.band-plum .body{color:var(--on-dark-body)}` overrides, and `--ink-body` on this ground is **1.5663**. Nothing here is left to inherit its colour: the h2, the five paragraphs, the six titles, the six descriptions and the badge letters each carry an explicit rule. THE BADGE FILL IS --accent AND THAT ADDS SIX INSTANCES TO A HELD EXPOSURE GROUP. --accent on this ground measures 3.2204. The badge is a non-text UI shape judged at 3.0, so it clears by 0.2204 — the same thin margin the other seven --accent-on-plum sites carry. Under the deferred --accent #4E65DC the fill would fall to 2.6959 and drop BELOW 3.0. Counted and reported; not acted on. THE LETTER-ON-FILL RATIO IS A SEPARATE MEASUREMENT from the fill-on-ground ratio, and it is the one that governs the glyph. White on --accent is measured in the report; the letter is 24px so it classifies LARGE whatever its weight. NO MOTION. The utilities connector animates its nodes and rule on a `.on` class; this one does not, so there is no selector for a prefers-reduced-motion block to restate and none is authored. Adding a transition later would require restating EVERY selector in full. --------------------------------------------------------------------- */ .tlw-band{background:var(--ground-dark);color:var(--on-dark);position:relative} .tlw-band::before{ content:"";position:absolute;inset:0;z-index:0;pointer-events:none; background:var(--slate); -webkit-mask-image:url(../../assets/longitude-texture.svg); mask-image:url(../../assets/longitude-texture.svg); -webkit-mask-size:18px 18px; mask-size:18px 18px; -webkit-mask-repeat:repeat; mask-repeat:repeat; -webkit-mask-position:top left; mask-position:top left; opacity:.10; } .tlw-band > .in{position:relative;z-index:1} .tlw{display:grid;gap:48px} @media(min-width:1000px){ .tlw{grid-template-columns:minmax(0,1fr) minmax(0,1fr);column-gap:64px;align-items:start} } .tlw-main{min-width:0} .tlw-band .tlw-h2{color:var(--gold);margin:0} .tlw-band .tlw-main p{ margin-top:20px; font-size:16.5px;line-height:1.7; color:var(--on-dark-body); max-width:62ch; } .beacon{list-style:none;padding:0;margin:0;min-width:0} .beacon-item{position:relative;padding:0 0 40px 76px} .beacon-item:last-child{padding-bottom:0} .beacon-badge{ position:absolute;left:0;top:0; width:52px;height:52px;border-radius:50%; background:var(--accent); display:flex;align-items:center;justify-content:center; font-family:var(--serif);font-size:24px;font-weight:700;line-height:1; color:var(--on-pill); } /* The rule runs badge-to-badge and stops at the last node, so there is no dangling tail below the sixth. */ .beacon-item:not(:last-child)::after{ content:"";position:absolute;left:25px;top:52px; width:2px;height:calc(100% - 52px); background:var(--gold); } .beacon-title{ margin:0; font-family:var(--serif);font-size:21px;line-height:1.25;font-weight:700; color:var(--on-dark); } .beacon-text{ margin:8px 0 0; font-size:16.5px;line-height:1.6; color:var(--on-dark-body); max-width:46ch; }