/* ===================================================================== DEVIATION LOG — home page THE SECOND PAGE-LOCAL REGISTER OPENED UNDER THE DISTINCT-PREFIX CONVENTION, and naming that makes it a convention rather than two one-offs. Three registers already existed — why-longitude.css, utilities.css, privacy-policy.css — and the `D` prefix was overloaded by deploy unit names, so `D7` resolved three different ways. A COUNT THAT STOOD HERE WAS REMOVED RATHER THAN CORRECTED, B4 2026-08-17. This sentence used to say why-longitude.css ran "1-17, now 1-20". By B3 it was 1-22 plus WL-23..WL-26, so the figure had been wrong twice over. **IF YOU NEED THAT COUNT, COUNT THE REGISTER IN why-longitude.css** — the file where its convention lives. A cross-file count in a comment goes stale every time the other file grows and nothing warns you; a number that cannot go stale beats one corrected for the fifth time. Same class as §44 and as the four figures this session found already wrong. UTIL-1 was the first entry to take a distinct prefix in answer to that. HOME-1 is the second. A homepage decision filed in another page's log is filed where nobody looking for it would stand. HOME-1 · CLIENT-APPROVED COPY REUSED IN A NEW PLACEMENT. R2, 2026-08-13. Not a copy departure — the string is unaltered — but the same class of decision as R1's CTA labels: text approved for one position now appears in another, which is the director's call and not the build's. WHAT AND FROM WHERE. The visible link beneath the film band's video carries the video's title as it already ships in the hub page's second row. It was COPIED PROGRAMMATICALLY out of that file, not retyped, and verified after writing: byte-identical, 52 characters, one apostrophe at index 31 as U+0027 per the 9ah Insights-row ruling, no non-ASCII. THE DESTINATION IS THE HUB ROW'S OWN OUTBOUND URL, not the nocookie embed address the frame uses. The nocookie ruling governs the EMBED, which loads without the reader doing anything; a deliberate click is a different act, and the embed address is a bare player rather than a landing page. One destination for one video across both placements, and no new URL form enters the tree. CONTRAST, MEASURED, SO NOBODY RE-DERIVES IT. The link takes .cta-line unchanged, which is --accent on this band's --ground-dark: --accent #696FE8 on --ground-dark #292631 = 3.5594 .cta-line is 19px/700 (components/buttons.css:78-84), which clears the 18.66px bold boundary, SO ITS APPLICABLE THRESHOLD IS 3.0, NOT 4.5. It passes by 0.5594 and is used at its own threshold deliberately, rather than restyled to --on-dark (13.0839) — which would have authored a colour and made this one CTA look unlike every other in-body CTA. THIS IS A DISTINCT CASE FROM THE SITE-WIDE FORECLOSURE. That one bars --accent on LIGHT grounds, where it measures 3.9310 and 3.6759 against a 4.5 that does apply. Dark ground, larger type, different threshold. HOME-2 · A COMMISSIONED SECTION IS REMOVED, AND THE VIDEO IS RESIZED. CLIENT-REQUESTED, DIRECTOR-APPROVED (R3, 2026-08-13). Two changes, one unit, because they share an edit set. (A) THE FILM BAND'S CAP GOES 100vh -> 65vh. A sizing decision from the client — the video read too large. Figures and the 462px floor are recorded at the rule itself, further down this file, not repeated here. (B) THE STAT WHEEL SECTION IS DELETED IN FULL. `section.photo` and everything inside it: the pinned track and stage, the photograph, the connector arc, the five figure nodes and the headline bar. It was commissioned and built across 9aj, 9al and 9am — all three sealed THE SAME DAY it was removed. Their notes hold the mechanism; the complete pre-rework markup is preserved in _archive-checkpoint-9am-2026-08-13/. EVERY CLIENT-SUPPLIED STRING IT CARRIED, RECORDED HERE BECAUSE THIS IS NOW THE ONLY PLACE SOME OF THEM EXIST: bar headline A scaled platform, built on shared strength. (an wrapped "A scaled platform,") figure labels member systems / combined net patient revenue / hospitals / states / utilities full readings, from the aria-labels, which carried the figures the animated counters rendered visually: 5 member systems $100B+ combined net patient revenue 250+ hospitals 20 states 5 utilities list label Longitude Health at a glance photograph alt Clinicians from different health systems in consultation THE ALT TEXT IS THE ONLY DESCRIPTION OF THAT PHOTOGRAPH ANYWHERE IN THE TREE and it leaves with the markup. The image file itself is retained in src/assets; only its built copy is dropped, because nothing references it any more. THREE FILES ARE NOW ORPHANED AND ARE DELIBERATELY LEFT ON DISK, UNLINKED: components/stat-arc.css (0 of 45 rules match any element on any page now), js/stat-wheel.js, and js/counter.js — which lost its last consumer with .fnum and .pbub, the only ones in the tree. Their tags were dropped from index.html in this unit; shipping a 42,686 B sheet that matches nothing is a defect, not a tidiness question. The FILES stay: the client removed a section commissioned hours earlier and the rework is still running page by page, the tree already retains rollbacks as standing practice, and unlinked files cost nothing at runtime. Deleting them is REMOVED 6 (src+dist x3) and its own unit. js/hero-sphere.js STAYS LINKED — #herosphere is in the hero. MEASURED BEFORE REMOVING, so no one re-derives it: only ONE for stat-arc.css ever existed, on index.html. The "14 pages" figure carried in the handoff was a grep over source text matching comments — see the §25 amendment. File reach and behaviour reach were both 1. GROUND SEQUENCE, LOGGED AND NOT RESTYLED AROUND. This band was dark and gave a hard break on BOTH sides. Without it the page runs cream -> paper -> slate with no dark band after #film, and #members -> #utilities is a soft tonal step of 5/8/12 per channel. That is the same class of seam R1 created after the hero (checkpoint-9an) and R2 closed with the film band (checkpoint-9ao) — read those three entries together; the ground-sequence history is one thread across R1, R2 and R3. HOME-3 · THE QUOTE-BAND TREATMENT LEAVES THIS SHEET. R6, 2026-08-14. Not a copy and not a deletion — a relocation, recorded here because this file is where anyone looking for it will start. The five `.band-slate.quote-feature` rules, and both comment blocks that documented them, moved to components/quote-block.css. NOTHING ABOUT THE QUOTE BAND REMAINS IN THIS FILE — verified by count: zero occurrences of "quote" and zero of "band-slate". WHY THEY WENT. They lived here because the homepage was the only consumer, and living in a one-page sheet gave them a second, accidental barrier. why-longitude.html now carries the same band from the same partial, and both pages already linked quote-block.css — so the modifier now sits beside the component it modifies, in the file both consumers load. Keeping it here would have meant either a duplicate or a page sheet reaching a second page. KEYED TO THE SELECTOR, NOT TO A PAGE. In its new home the scope note reads: applies wherever `.band-slate.quote-feature` appears, which is wherever the partial that sets the class is included. The old headers said "HOMEPAGE ONLY" and went stale the day a second page adopted it. NO ROSTER WAS WRITTEN IN ITS PLACE — a roster is stale the moment the next page adopts it. THE COMPOUND IS WHAT MADE THE MOVE SAFE. In this sheet the rules loaded after quote-block.css; in it they load before their own base rules. They still win because `.band-slate.quote-feature` out-specifies `.band-slate` — specificity, not order. MEASURED: 54 computed values on index.html identical before and after, at 1440, 760 and 390. HOME-4 · THE QUOTE PARTIAL IS FORKED AND THE HOMEPAGE GETS ITS OWN. 2026-08-15. index.html now includes partials/components/quote-home.html; why-longitude.html keeps quote-block.html, unchanged, still carrying Pete McCanna. ONE PARTIAL SERVED TWO PAGES while its own header claimed "HOMEPAGE ONLY" — the fork removes that mis-scoping rather than re-documenting it. Only the homepage changes; the other page's rendered output is byte-for-byte what it was. THE HOMEPAGE QUOTE IS NOW VISHAL AGRAWAL. Attribution follows the existing `Name · Title, Org` cite pattern and takes the §6 LEADERSHIP title form from our-team.html — "Vishal Agrawal, MD" / "Chief Executive Officer, Longitude Health". NOT the §4 board form: "(Ex-Officio)" is a governance qualifier and does not belong on a quote attribution. LOGGED COPY DEVIATION — ENCLOSING QUOTATION MARKS REMOVED. The supplied string opened U+201C and closed U+201D. Both were STRIPPED under Ruling F, because `blockquote::before` already renders a U+201C ornament — 160px gold at this size — and marks plus ornament render the quotation TWICE. This is the same handling as our-team.html §4 and the inverse of the memorial in our-origins.html, which keeps its typed marks precisely because it has no ornament. **IF THE ORNAMENT IS EVER REMOVED FROM THIS BAND, THE MARKS COME BACK IN THE SAME EDIT.** The two curly apostrophes inside the string, in "isn't" and "It's", are U+2019 and were NOT touched — only the two enclosing double marks were removed. THE PORTRAIT IS BELOW THE 2x STANDARD AND THAT IS ACCEPTED, NOT MISSED. assets/leadership/Vishal.jpg is 400x400 into a 300px slot: 1.333x at 1x and 0.667x at DPR 2 — below native. Accepted by ruling as a known deviation. THE REMEDY IS NAMED: a higher-resolution original has been requested from the client, and the swap will be ONE FILE with no markup change, because the slot is fixed-basis and the image is object-fit:cover. DO NOT UPSCALE TO CLOSE IT. The file this replaced, lh-portrait-mccanna.jpg at 500x500, already failed the same gate at 0.833x — the deviation is inherited, not introduced. OPEN, LOGGED, NOT ACTED ON: quote-block.html is now why-longitude's page-specific partial while carrying a generic name, beside quote-home.html and quote-utilities.html which are named for their pages. A generic name on a page-specific file is exactly how the "HOMEPAGE ONLY" claim went stale. Renaming it to quote-why-longitude.html needs why-longitude.html edited and is ITS OWN UNIT. HOME-5 · THREE CTA DESTINATIONS AND FIVE LOGO WIDTHS. CLIENT-REQUESTED (BugHerd #1 and #2), DIRECTOR-RULED (B4, 2026-08-17). Two tickets, one entry, because they land in the same unit on the same page. ── #1, AND THE FIRST OF THE THREE IS A CORRECTION, NOT A REDIRECTION. The hero CTA read `href="#problem"`. **`id="problem"` EXISTS ON ZERO PAGES** — measured across all 20 built pages, control `id="utilities"` returning index.html and utilities.html. R1 moved index's `#problem` section to why-longitude.html and **the id did not survive anywhere, on either page.** The hero's primary call to action has therefore pointed at a non-existent fragment since R1, silently: a dead same-page anchor does not error, it just does nothing. **Repointing it to why-longitude.html fixes a live defect. Do not read this row as a preference.** section.hero .btn "See the model →" #problem -> why-longitude.html section#members .cta-line "See how it works →" -> our-approach.html relabelled "Learn about Longitude →" section#utilities .cta-line "Explore what we're building →" NO-OP, untouched THE THREE STRINGS IN THE TICKET ARE SECTION HEADINGS, NOT CTA LABELS, and reading them as labels is what made the third row look like a change. They are the h1/h2 of the three sections — "Solving together what no health system can solve alone.", "Built by health systems. For health systems.", "A growing ecosystem of health system-built utilities." — and each identifies which section its instruction applies to. **The third row already pointed at utilities.html, so it was reported as a NO-OP and NOT edited.** No change was manufactured to match the shape of the other two. TWO CTAs ARE DELIBERATELY EXCLUDED AND ARE NAMED SO THE OMISSION IS NOT READ AS AN OVERSIGHT: `.btn` "Join Us" -> #join, and `.cta-line` "Solving What Health Systems Can't Solve On Their Own" -> the off-site youtu.be link. Five CTA-class elements render on this page and the ruling addressed three. THE .html FORM IS THE SITE CONVENTION, COUNTED NOT ASSUMED: **26 `.html` page hrefs across src/pages and src/partials, 9 distinct targets, and ZERO extensionless page hrefs.** The 11 extensionless hrefs are newsroom article DIRECTORY links carrying a trailing slash, a different pattern. Both repoints take `.html`, matching the untouched third CTA and the header's nine. **An extensionless page link here would have been the site's first.** `cleanUrls` may well redirect `.html` to extensionless at the edge — firebase.json is ⌘4's and was not opened — but **what the edge does to a URL does not change what the source should say.** FRAGMENTS AFTER THE REPOINTS, REPORTED AND DELETED NEITHER. `#problem` is now wholly unreferenced and its target never existed to begin with, so there is nothing to remove. `#utilities` is now unreferenced from this page **but its target element remains** — it is the section's own id and utilities.html carries one too. Leaving an unreferenced id is not dead code: it is an addressable landmark, and deleting it would break any external or bookmarked link. ── #2, AND THE WIDTH CHANGE HAS THREE DEPENDENTS, NOT ONE. All five member logos go +10%: 157 -> 172.7, 121 -> 133.1, 110 -> 121, 124 -> 136.4, 113 -> 124.3. Every one stays far under its ~500px intrinsic width; the binding headroom is Baylor at 2.896x, and 1.448x at DPR 2, still above native. No image file was opened, moved or re-encoded. `--cx` SCALES x1.1, AND THE REASON IS THE NOUN. The first measurement asked whether the dots MOVE when the logos grow. They do not — `.mem` is `flex:1 1 168px` under `space-between`, so the box is sized by the flex algorithm and not by the logo, and the box widths held at 196.8px to two decimals. **That measurement is true and answers the wrong question.** `--cx` corrects each dot onto its logo's OPTICAL CENTROID, and the note at the declaration records that Novant and Providence carry their ink weight ~8px left of bbox centre. Scale the artwork and that offset scales with it: at +10% the dot is left 1.08px behind on Novant and 1.03px on Providence. **The dot not moving is not the dot still being aligned.** Recomputed at the file's existing 2-decimal precision: -0.04, -0.43, -11.85, -11.28, -0.19. THE ANIMATION DELAYS ARE THE THIRD DEPENDENT AND THE SHEET SAYS SO. The delay for each logo is its dot's fraction of the drawn line x the line's duration, and the note there instructs re-measurement if `--cx`, the widths or the row's justification change. All three triggers fired, so the fractions were re-measured against the FINAL 2-dp `--cx` values rather than derived. The figures and which delays moved are recorded beside the delays themselves, where whoever edits them will be standing. THE 84px ZONE IS THE CEILING AND +10% IS INSIDE IT. `.mem .lg` is a fixed 84px band so no tall mark can push its name below the short ones — every name baseline lands on one line. Michigan is tallest at 70.50px, reaching **77.55px at +10% with 6.45px to spare. The zone is reached at about +19.1%**, so a future "make them bigger again" of +20% would overflow it and break that shared baseline. **Measure the zone before granting a second increase.** our-team.html DOES NOT MOVE AND WAS NOT EDITED. Its logo row is a COPY of this component under a different selector family — `.gm-strip`/`.gm-mem.g-*` against `.mstrip`/`.mem.m-*` — with no shared selector between them. The `--cx` values were identical only because one was copied from the other, and our-team.css's own banner warns the copy does not follow the original. **They are now deliberately different, and aligning them is not a tidy-up.** ── THIS UNIT WROTE INSIDE ANOTHER SHELL'S LIVE GRANT. `src/pages/index.html`, `src/css/pages/home.css` and `src/css/components/member-strip.css` are all on the **HOMEPAGE SHELL's** enumerated grant — the first two on the base list, the third by the 2026-08-07 amendment. **That delegation has never been recorded as closed** and this unit did NOT close it: closure would assert the homepage change list is finished and no record establishes that. Written on the `8w` precedent, where the BUILD SHELL — which issued the delegation and holds the write lock — wrote the same two paths. **The authority is not in question; the collision risk is, and a declaration cannot detect a collision (§35).** THE PROTOCOL, EXECUTED IN ORDER: hash, byte count, mtime and delimiter baseline recorded for ALL NINE grant paths before any edit, the JS members enumerated by measurement rather than by name; `src` confirmed quiescent against the 9bg seal with the freshness probe controlled against 9bf; hashes re-compared IMMEDIATELY BEFORE THE BUILD, not only at the start; and every untouched grant path verified byte-unchanged at exit. QUIESCENCE EVIDENCE, MEASURED NOT ASSUMED: index.html unchanged at `435858e8e96f76e1` across NINE checkpoints, 9ay through 9bg, mtime 2026-08-15 10:49:04. member-strip.css untouched since 2026-08-07 16:44:16 — **the day the grant was issued.** Nothing had written either file in ten days. **If the HOMEPAGE SHELL is still live it must re-read all three before writing any of them; its copies are now stale.** ===================================================================== HOME-6 · THE #film CTA IS RE-GROUNDED TO GOLD. CONSEQUENCE OF THE --accent TOKEN MOVE. C2, 2026-08-17. `#film .film-cta .cta-line` takes `--gold` instead of the inherited `--accent`. **Colour only — one added declaration, no copy, no geometry.** before --accent #2B61C9 on --plum 2.5821 19px/700 LARGE, thr 3.0 FAIL after --gold #F0B83F on --plum 8.2161 19px/700 LARGE, thr 3.0 PASS ── WHY THIS IS A CONSEQUENCE AND NOT A NEW DECISION. C1 moved `--accent` #696FE8 -> #2B61C9 at client request. On light grounds every instance improved; on `--plum` the same move took this link from 3.5594 to **2.5821, below its 3.0 threshold.** **See TOK-1 in tokens.css for the 55-instance census — that entry and this one are ONE decision recorded in two places**, and C1 shipped this failure knowingly rather than widen its own manifest. C2 closes it. ── `--gold` IS NOT A BESPOKE VALUE. IT IS ALREADY THE DARK-GROUND ACCENT. `h2.tlw-h2` on our-approach measures **8.2161 on plum — the identical figure**, because it is the identical pair of colours. This sheet is adopting an established treatment, not inventing one. Rejected alternatives, measured: --ink-on-dark 8.6943 clears, but is the dark-ground BODY colour --cream 13.0839 clears, but is the dark-ground BODY colour --slate 3.0421 clears by 0.0421 — not a margin --pill-hover 2.8600 fails **The two that clear most easily are the colours body text already uses on dark grounds, so a CTA wearing them stops reading as a link.** Legibility is not the only thing being measured here. ── ⚠ THIS DOES NOT GRANT R4, AND MUST NOT BE READ AS A PRECEDENT FOR IT. **R4 is the client asking for gold on the `h1 em`, which sits on `--paper`. Gold on paper measures 1.7030 and fails even the 3.0 large-text threshold. That request stands REFUSED.** This is gold on `--plum` at 8.2161. **Same colour, different ground, opposite verdict** — a 4.8x spread between the two figures. A future reader finding "gold approved, 2026-08-17" in this log is reading about a DARK ground only. The ground is the variable, not the colour. ── SCOPING, BECAUSE `.cta-line` IS A SHARED CLASS. Nine `a.cta-line` render site-wide and **seven sit on light grounds, passing at 5.0671 or 5.4187 — they must not move.** The base colour is declared once, at `components/buttons.css:78`, as `color:var(--accent)`; **neither page sheet declared it, so this is an ADDED scoped rule, not a swap.** Section-scoped to `#film`, which contains exactly one `.cta-line`. §46. The cascade was enumerated rather than assumed (§65): the only rules setting `color` that match this element are `a{color:inherit}` and `.cta-line{color:var(--accent)}`. **No `:hover`, `:focus` or `:visited` rule for `.cta-line` exists in any sheet**, so there is no second state to carry. The new rule is (1,1,0) against (0,1,0) and wins on specificity alone. HOME-7 · THE FILM EMBED MOVES FROM YOUTUBE TO WISTIA. CLIENT-DIRECTED, V1, 2026-08-18. `