Changelog

Addressed findings from an adversarial code review of the v1.51.0–v1.51.6 engine + calculator work

(0 critical, no XSS/secrets; these are the actionable correctness/consistency items):

• pue-calculator.html: the window.load handler re-ran the ENTIRE calculate() (a redundant full

model recompute) just to populate the additive panel after the deferred engine loaded. Now it caches the

inputs and calls only renderPueCurve(...) — matching the targeted-render pattern already used by

capex/opex. Verified the panel still renders (0 errors).

• tco-calculator.html: percentile accessors now clamp the index to [0, N-1] (the engine's pct()

convention) instead of a raw Math.floor(N·q) against the sample count — defensive against any future

case where filtered sample length differs from the nominal iteration count. MC values unchanged.

• rz-engine.js (doc-only; minified twin byte-identical, no cache-bust): documented that

sim.monteCarlo correlations are applied sequentially (safe for disjoint pairs; chained pairs are

transitively correlated); documented the opex.totalAnnual opts.climate constraint (don't combine with

a calibrated PUE — double-counts cooling); documented that charts.tornado labels must be trusted

(placed into SVG <text>, which doesn't execute HTML, but callers must not pass raw user input).

Changed — first batch of the full-site axe sweep (106 pages beyond the gated 8-page set;

11,334 critical/serious nodes found, heavily clustered):

• tools/build-changelog-html.py (~5,700 nodes, half the site total, one generated page):

light-mode was missing overrides for subheads/strong/code/tables/dates/nav — the

dark-default colors bled through. Light/dark .changelog-date values were reversed.

Tier badges/pills now derive a darker light-mode color from --tier-color via

color-mix. changelog.html now audits 0/0 both themes — permanent for all future releases.

• cdu-checklist.html / fire-checklist.html / geopolitics-1.html (~350 nodes): every

bare checklist checkbox gets an accessible name from its row text (tiny runtime; geopolitics

needed a MutationObserver — its kit checklist is JS-injected after DOMContentLoaded); chip

families (.src.*, .cad.*, .crit.*, status badges) darkened for light tints with dark

counterparts; table headers, print buttons, service-record inputs (aria-label), scrollable

table wrappers (tabindex), in-paragraph links underlined. All three pages now 0/0 both themes.

Remaining sweep clusters (article-15, dc-market-tracker, achievements, cockpit pages, ~25 pages)

tracked for follow-up batches.

panel)

Added

• Partial-Load PUE & Water panel on pue-calculator.html (free tier) — an additive readout powered by

RZEngine v2.0's previously-unconsumed pue.partialLoadPUE + pue.wue models. PUE is quoted at full IT

load, but data centers rarely run at 100% and fixed infrastructure overhead doesn't scale down — so the

panel shows the effective PUE across 20–100% utilization (e.g. a 1.46 design PUE degrades to ~1.72 at

50% load), plus a WUE (Water Use Effectiveness, L/kWh by cooling type) and annual water estimate, with

a small SVG curve. Additive only (does not touch the calculator's own PUE math), renders on window.load

after the deferred engine, hides gracefully if the engine is unavailable, dark-mode-safe. Placed in the

free results area (not the Pro-locked section). Verified 0 console errors both themes.

Added (engine)

• RZEngine.models.sim.monteCarlo gained correlated variables + discrete scenarios (engine

2.0.0 → 2.1.0, backward-compatible). New categorical distribution ({choices:[{value,weight}]})

for weighted discrete draws, and an optional opts.correlations:[{a,b,rho}] imposing pairwise

correlation between normal keys (z_b ← rho·z_a + √(1−rho²)·z_b). When opts is omitted the RNG

path is byte-identical to v2.0, so the capex/opex/roi panels are unaffected. tools/test-rz-engine.mjs

now 79 assertions (added categorical, correlation-widens-spread, no-opts-determinism). Min re-built

with terser (parity-checked); cache-bust 2026-07-05-v21.

Changed

• TCO Monte-Carlo consolidated onto the shared engine. Its simulation (correlated construction/power

at r=0.65 + 4 weighted macro scenarios: baseline / AI-boom / recession / power-crisis) now runs through

the seeded engine simulator, so the P5/P50/P95 are reproducible instead of jittering on every

recalculation. The cost model is unchanged — verified the engine path matches the prior inline model to

0.1% on P50 — and a full inline Math.random() fallback remains. This completes the sim-engine

unification (capex/opex/roi already on the engine; tco needed the correlation+scenario support added here).

Changed

• ROI Monte-Carlo consolidated onto RZEngine.models.sim.monteCarlo (Pro panel). The inline

simulation used Math.random(), so the P5/P50/P95 and probability-of-positive **jittered on every

recalculation** — an awkward wart for a financial tool. It now routes through the shared engine's

SEEDED simulator, so the percentiles are reproducible run-to-run. The uncertainty model is

unchanged (revenue ±20%, occupancy ±15%, opex ±15%); the histogram render is unchanged; a full

inline Math.random() fallback remains if the engine is unavailable. Verified: same-input reruns now

return identical P5/P50/P95, 0 console errors.

• TCO Monte-Carlo intentionally left as-is — its simulation uses correlated random variables

(construction/power r=0.65) and discrete macro-scenario selection (baseline / AI-boom / recession /

power-crisis) that the generic engine driver can't express; consolidating would have regressed the

model. Documented in the RZEngine↔calculator notes.

Added

• OPEX Uncertainty & Sensitivity panel on opex-calculator.html — same additive RZEngine v2.0

pattern as the CAPEX panel (v1.51.2). A 4,000-run seeded Monte-Carlo around the computed annual

OPEX, varying energy price (the dominant, volatile driver), PUE drift, and labor/maintenance

escalation → P10 / P50 / P90 + histogram + a sensitivity tornado (energy price ranks first,

as expected). Does NOT touch the calculator's own cost model/data; renders on window.load, hides

gracefully if the engine is absent, dark-mode-safe. Verified 0 console errors both themes.

Added

• Cost Uncertainty & Sensitivity panel on capex-calculator.html — an ADDITIVE, engine-powered

analysis that does NOT change the calculator's own (superior, city-level T&T/C&W 2025) cost data or

point estimate. It runs a 4,000-iteration Monte-Carlo (RZEngine.models.sim.monteCarlo, seeded/

reproducible) around the computed total, varying construction-bid spread, multi-year escalation, and

scope growth → P10 / P50 / P90 CAPEX range + a histogram, plus a sensitivity tornado

(RZEngine.models.sim.tornado) ranking which driver moves the total most. Charts are RZEngine v2.0

framework-free SVG builders. Renders on window.load (after the deferred engine is ready) and

degrades gracefully — the panel simply stays hidden if the engine is unavailable, never breaking the

calculator. Dark-mode-safe (built on the page's theme vars); verified 0 console errors both themes.

This is the owner-chosen approach ("leave the calculators' data, add new engine capabilities they

lack") for making RZEngine v2.0 deliver value to the calculators without regressing them.

Changed (docs/verification — no runtime change)

• A9 calculator audit: re-verified all 6 engine-consuming calculators (capex/opex/pue/roi/tco/cx)

on RZEngine v2.0.0 — 0 console errors; roi-calculator's IRR (now Newton-with-bracket honoring the

guess) computes correctly (NPV/IRR sane). Finding recorded: capex/opex/cx/tco are self-contained

(own defaults + math) and consume only the engine's shared MATH (pue/roi models) + auth/modal/pdf/ui

helpers — full adoption of the refreshed DATA values is a per-calculator enhancement, deliberately

NOT forced here to avoid regressing the hardened, working calculators.

• A10 dcmoc reconciliation: audited dcmoc/src/lib/capex-data.ts (per-city $/W, T&T/C&W/CBRE 2025)

and CarbonEngine.ts (IEA 2024-2025) against the refreshed RZEngine regional data. They are

consistent in magnitude and NOT divergent — dcmoc is intentionally finer-grained and equally

current, so its city-level numbers are NOT overwritten with RZEngine's coarser regional estimates

(that would degrade dcmoc). Coupling rule corrected in standarization/SUPER_ENGINE.md §Z; dcmoc

source unchanged (build state preserved).

Upgraded the shared Super Engine (rz-engine.js, window.RZEngine) and its database from

1.2.0 → 2.0.0. Staged, individually-verified, and fully backward-compatible: every existing

models.* signature is unchanged; all additions are new DATA keys or OPTIONAL params. Gated by a

new harness tools/test-rz-engine.mjs (76 assertions — worked examples + data invariants +

reachability + provenance). All 6 engine-consuming calculators re-verified: 0 console errors.

Added

• Schema + provenance (A1): DATA.meta (schemaVersion/engineVersion/asOf/lastReviewed/license),

a DATA.sources sidecar (every value → {source, asOf, unit?, method?}), and DATA.provenance[]

citations. Harness fails if any registered value lacks a source.

• Expansion tables (A3): country regions (ID/SG/JP/IN/MY), land, laborRates, carbon

(grid factors + carbon price + embodied), water (WUE + price), aiDensity, coolingTypes,

tiers (I–IV), extra roles + salaryRolesExt, discountDefaults (WACC), pueMatrix, refresh.

• New models (A7): models.sim.monteCarlo/tornado/sensitivityGrid (seeded, reproducible),

models.carbon.*, models.water.*, and charts.* implemented as framework-free SVG builders

(histogram/tornado/sensitivity/roiLine/costStackedBar/hiringTrajectory).

• Model math (A6): cooling-aware capex + AI/GPU density + land/commissioning/permitting line items;

opex water/carbon/insurance/connectivity (opt-in) + PPA/TOU/demand + cooling-efficiency consumption;

roi npvAuto/discountedPayback + Newton-with-bracket IRR honoring guess; forecast R²/confidence

band + inflation wiring + scenario bands; tco lifecycleNPV (discounting + salvage); workforce

hiringPlan/attritionCostWeighted/cumulativeHiresCompounded.

Changed

• 2026 data refresh (A2): regional power $/kWh, salaries, capex/MW (full tier × cooling matrix),

PUE defaults, currency, inflation — all re-sourced to 2026 with asOf tags.

• No buried constants (A4): moved coolingClimate, contractCostBase, staffingLoadFactor,

opexDefaults, capexDefaults, refresh, workforceParams out of function bodies into DATA.

• Reachability (A5): liquid/immersion capex + PUE defaults are now actually consumed

(datacenterBuildCost(…,cooling), pue.defaultFor(cooling,tier)); inflation + attrition defaults wired.

• Fixed capex breakdown — it/mep/civil now split on the pre-contingency base (no contingency

double-count; civil ≥ 0); total magnitude unchanged.

Fixed / build (A8)

• Killed the version drift: DATA.version → 2.0.0, pdf.scriptTagsHTML() cache-bust and page ?v=

bumped together to 2026-07-04-v2. rz-engine.min.js regenerated reproducibly with terser

(parity-checked against source). tools/test-rz-engine.mjs added to the CLAUDE.md ship-gate.

See standarization/SUPER_ENGINE.md §Z.

Changed — cleared all 56 non-gating axe advisories from v1.50.41; audit-a11y now

reports zero findings of any severity on the 8-page set x both themes:

• Heading order: article-26 callout boxes (.pfas-insight/warning/danger-box) h4 → h3

with an !important re-assertion of the compact box-heading look (the shared editorial

.article-body h3 rule uses !important sizes); fire/cdu calculator input panels

h3 → h2 (they sat directly under the page h1); paper cards + footer column labels on

datacenter-solutions, articles, fire h4 → h3. Styling preserved (computed

font-size/color verified per page).

• Landmarks: the version stamp is now a labelled complementary landmark (shared

script.js/script.min.js — contentinfo would duplicate the footer landmark); the

TOC mobile drawer + sidebar get role="navigation" from shared js/rz-article-toc.js;

article-26 evidence block + series nav labelled; articles.html decorative particles

canvas aria-hidden and floating side cards a labelled complementary; the glossary

disclaimer moved inside <main>.

• Cache-busts: script.min.js (71 pages), rz-article-toc.js (32 pages).

Verification — audit-a11y CLEAN (0 critical/serious, 0 advisory); full suite green:

script-tags, js-syntax, version-stamp, mobile 0 fail, responsive-layout CLEAN (113pp),

dark-coverage CLEAN (114pp), charts 25/25, interactions CLEAN.

Fixed

• Second Brain "Wiki" pointed at the wrong content: the graph's wiki node and the

header Wiki link both opened standarization/repos/REPO_INSTALL_PLAN.md (a repo

install plan), not a knowledge wiki. Both now open the real vault landing

obsidian-knowledge-vault/00-Hub/Wiki.md.

Added

• obsidian-knowledge-vault/00-Hub/Wiki.md — a proper knowledge-wiki landing: what the

vault is, the 8 hubs (articles / calculators / apps / series / standards / comparisons /

reports / automation) with [[wikilinks]], and how to navigate by connections.

• repos graph node ("Repo Install Plans") — the repo/tooling install docs remain

reachable but no longer masquerade as the Wiki; edges wiki↔repos, rzstd↔repos,

asb↔repos. sync-graph.py: 0 dead URLs, graph renders 0 console errors.

Changed

• Portfolio consistency: extended the reference bottom-rising colored-glow card treatment

(already on the hero bento cards) to the rest of the homepage — .metric-card (Career

Achievements), .oe-card (Operational Excellence ×8), and .case-card (Case Studies ×3).

Each card now carries an always-visible bottom radial glow in a per-card accent (blue / green /

amber / violet / magenta / teal rotation via nth-child), content stacked above via

> * { z-index: 1 }, intensity scaled by mode (day .30 / dark .58 / rainbow .95). Index-only,

in styles-index.css per the 2-stylesheet architecture.

Removed

• Hero grid: dropped the gold 60px line-grid on .hero-background (background: transparent);

the soft radial washes + .rz-bg-gradient / aurora provide the backdrop — reference is a smooth

gradient, and this aligns with rejected-pattern #1 (no dot-grid/line-grid noise on the hero).

• Film-grain dots: disabled the sitewide [data-theme="dark"] body::before feTurbulence noise

overlay ON THE HOMEPAGE ONLY (index-only stylesheet — content: none; background: none), so the

dark + rainbow background stays a clean smooth gradient (removes the "titik2" speckle). Other

pages keep the grain via styles.css.

Added

• tools/audit-a11y.mjs — permanent accessibility render gate (in the ship suite +

CLAUDE.md): vendored axe-core 4.10.2 (tools/vendor/axe.min.js) over an 8-page

representative set (index, articles, article-13, article-26, fire/cdu calculators,

glossary, datacenter-solutions) x both themes, over local HTTP with the

localStorage+attribute theme method. Fails on any critical/serious violation;

moderate/minor (heading-order, region) are reported as non-gating advisories.

Changed — extended the WCAG-AA sweep from v1.50.40's 4-page audit set to the full

8-page gate set (341 flagged nodes → 0 critical/serious, both themes):

• article-13.html: brand company badges (AWS/Google/Microsoft/NVIDIA) darkened on

their tinted chips with dark-mode counterparts; 44 inline emerald + 39 amber/red/violet

table accents darkened with paired dark overrides; TOC section numbers + reference

numbers (#06b6d4 at 2.3:1) → #0e7490/#22d3ee; .article-body h4 dark override;

neon code-span colors (#67e8f9/#f472b6/#ff9900) darkened in light mode only.

• articles.html: card excerpts + category chips darkened (light) with dark pairs.
• datacenter-solutions.html: --text-muted both themes; open-buttons were near-white

on pale blue; neon instrument tokens darkened light-only; section/paper/FAQ headings

used --dark-blue (#1e3a5f) in dark mode → readable overrides.

• cdu-calculator.html: derived/warn/alarm chips, mode buttons, pro-gate overlay scrim

(.55 → .78), 2 remaining unlabeled inputs (aria-label), note links underlined.

• Shared styles.css: .calc-disclaimer .disc-fine used #64748b !important

(beat every page-level fix) → theme-correct colors + readable disclaimer links.

• Shared js/rz-article-editorial.js: focusable-scroller tagging now also covers

tables that scroll themselves (display:block tables), re-scans on window.load +

document.fonts.ready, and matches axe's any-overflow threshold.

• Cache-busts: styles.min.css (69 pages — was still on v=2026-05-18 after

v1.50.40's edit), styles-index.min.css, rz-article-editorial.js (37 pages).

Verification — audit-a11y --strict CLEAN (8pp x 2 themes, 0 critical/serious, 56

non-gating advisories); full suite green: script-tags, js-syntax, version-stamp 185/191,

mobile 0 fail, responsive-layout CLEAN (113pp), dark-coverage CLEAN (114pp), charts 25/25,

interactions CLEAN.

Added

• Form-label associations: 7 calculator inputs on fire-calculator.html and 13 on

cdu-calculator.html now carry explicit <label for="…"> associations (axe

label/select-name criticals — all resolved).

• Keyboard-reachable scrollable tables: js/rz-article-editorial.js now tags every

horizontally-scrollable table wrapper in the article body with tabindex="0" +

role="group" + an aria-label, so keyboard users can scroll wide tables

(axe scrollable-region-focusable).

• Cookie-banner landmark: the cookie notice on 78 pages is now a named region

(role="region" + aria-label="Cookie notice").

Changed

• WCAG-AA contrast sweep (~478 flagged nodes → 0 on the audit set, both themes):

darkened low-contrast light-mode colors (#94a3b8→#64748b, #f59e0b/#d97706→#92400e,

#22c55e→#15803d, #2563eb→#1d4ed8, #8b5cf6→#6d28d9, fire --fc-primary →

#b91c1c light-only) with paired dark-mode overrides where the swept colors sat on

dark panels (article-26 PFAS tables/cards, glossary term counter, fire preset buttons).

• Version stamp readable: the footer stamp no longer washes its text to 2.4:1 via

container opacity — text runs at full opacity with dedicated colors

(.rz-version-label / .rz-version-num, both themes); subtlety moved to the logo image.

• In-paragraph links distinguishable without color (WCAG 1.4.1): terms/privacy/footer

prose links are underlined (text-underline-offset: 2px).

• Inline style="color:…" attributes (including JS-set el.style.color, which

serializes to rgb()) are neutralized in dark mode via per-value

[style*="…"] overrides on article-26.html + glossary.html.

Fixed

• tools/audit-dark-coverage.mjs false positive on index.html: pages that re-apply

the theme from localStorage on window.load (rainbow-mode init) undid the gate's bare

attribute flip — the gate now writes localStorage.theme together with the attribute,

mirroring the real toggle.

• tools/audit-mobile-responsive.py no longer walks into .claude/worktrees/ agent

checkouts (4 false FAILs from email-signature templates in a stale worktree).

• Version stamp injected on 5 internal pages that were missing it (plan-article-experience,

rz-index-mockup*, rz-index-polish, rz-index-redesign).

Verification — axe-core 4.10.2 sweep (index, article-26, fire-calculator, glossary ×

light+dark): 0 contrast / 0 link-in-text-block / 0 label violations. Full gate suite green:

script-tags, js-syntax, version-stamp 185/191, mobile-responsive 0 fail, responsive-layout

CLEAN (113pp), dark-coverage CLEAN both modes (114pp), article-charts 25/25, interactions CLEAN.

Fixed — a reported bug: after returning to the homepage while in Rainbow mode, the

bento cards rendered in light/day even though the rainbow background + toggle icon

were active. Root cause: the shared script.min.js runs applyTheme(getPreferredTheme())

at DOMContentLoaded from the theme key; if that key had drifted to light (e.g. the

2-way toggle was used on another page) while rzRainbow=1, it overwrote data-theme

to light while the index-only .rz-rainbow class stayed → light cards on a rainbow

background. The index theme controller wasn't re-asserting afterward.

• The controller now re-asserts the stored mode on load (enforce()) and on

bfcache restore (pageshow, browser back/forward), and normalizes the theme

key back to dark whenever rainbow is active.

• Added a MutationObserver on data-theme: while rzRainbow=1, if anything flips it

off dark, it's restored immediately (catches the clobber with minimal flash).

• Verified headless across all states (clobber rzRainbow=1+theme=light → corrected

to dark cards; normal rainbow/dark/day consistent) + visual screenshot of the

previously-broken state now correct; 0 console errors; js-syntax + script-tags CLEAN.

Index-only inline JS (no CSS change).

Added

• article-25 (PJM Grid Crisis) — a live grid schematic: GENERATION FLEET → PJM GRID → LOAD. Click

"Fast-forward: 2027–28" and ~40 GW of thermal retires (warn), the grid's capacity gap hits 6 GW

(alarm), generation flow slows — with 65M people downstream and data centers driving 40% of load growth.

All figures as stated in the article body.

• article-16 (SE Asia bubble) — a scrollytelling of Johor's 5.8 GW pipeline, stage by stage: 487 MW

live → +422 MW building → +1.4 GW committed → +3.4 GW on paper, with a cumulative MW counter (487 → 5,709)

and proportional bars that make the bear case visible (most of the headline is the red "planning" bar).

Values from the already-verified data/article-16/johor-pipeline.csv (PUBLISHED chip).

• Both verified interactively (scenario alarms/flows; scrolly 0→3 + reverse; 0 errors). The site now has

3 living diagrams + 2 scrollytelling stories, all provenance-gated (25 configs CLEAN).

Added

• rz-tracker events (guarded, no-op when the tracker is absent) so the admin dashboard can measure the new

systems: rz_search_open / rz_search_go (destination + query) / rz_command (palette),

rz_diagram_scenario (living diagrams), rz_scrolly_complete (scrollytelling reached the final step, fired

once). Verified recording into rz_user_events; interaction gate CLEAN. Modules cache-busted to ?v=1.50.37.

Changed

• The article TOC is now a shared module — 32 pages each carried a ~3.6KB duplicate inline copy of the

TOC/scrollspy script; all excised (marker- and signature-bounded, with sanity asserts) and replaced by

js/rz-article-toc.js (idempotent: skips pages without TOC markup or where a list was already built;

strips the heading-anchor glyph from labels). ~115KB of duplicated inline JS removed site-wide; one

implementation to maintain. Scrollspy verified live (active-section highlight on scroll).

• The Inter/JetBrains Google-Fonts stylesheet is now non-blocking (media="print" onload + <noscript>)

on 34 article/FF/geopolitics pages — the second render-blocking font request removed (the editorial

Fraunces set went async in v1.50.24). Interaction + dark-coverage gates CLEAN.

Fixed — the SEO audit's only REQUIRED-tag errors: three pages

(all-in-one-dashboard.html, network-compare.html, network-visualization-hub.html)

had no Open Graph tags at all, so they rendered as bare links when shared on

social/messaging. Added the full block per page — og:type/url/title/description/image

(+ image dims/alt where a dedicated OG card exists; assets/profile-photo.jpg fallback

for the dashboard), twitter:card/title/description/image, and a WebPage/CollectionPage

JSON-LD. audit-seo.py REQUIRED errors: 3 → 0 (61+ clean pages). JSON-LD validated,

js-syntax + script-tags CLEAN.

Added

• Section-level deep search in the command palette: new search-sections.json (345 entries — every article

h2 with a static id, generated by tools/build-search-sections.py --apply) is lazy-loaded alongside the main

index; queries ≥3 chars surface up to 3 "§ Section ·

" hits that deep-link straight to the heading

(e.g. "workforce cliff" → article-27.html#section-1, verified to land the H2 in view). Regenerate the file

when adding/renaming article sections (noted in CLAUDE.md tooling + interaction gate stays green).

Added

• tools/audit-interactions.mjs — a permanent ship-gate that exercises the interactive systems with REAL

keyboard/mouse/scroll in a headless browser (spins its own static HTTP server so search-index.json fetches

work): command palette (Ctrl+K / query / Esc / "/" on a migrated + a previously-dead page), living-diagram

scenario injection (article-13/9), scrollytelling step advance + reverse with counter targets (article-23),

and the reading-polish features (anchors, min-left chip, exactly one progress bar). Counter checks poll until

settled. Added to the CLAUDE.md audit suite. First run: CLEAN.

Fixed — the previous portfolio pass made the amber gradient apply to BOTH primary

CTAs (Download Resume + Get Started, which share .btn-bento-primary). Per the

reference, only Download Resume is amber; **Get Started is now a distinct violet-tinted,

violet-bordered button** (.bento-cta-row .btn-bento-primary), with a stronger violet

edge in rainbow mode. Contact Us stays the plain outline. Verified across day/dark/

rainbow (amber vs violet confirmed, identity card matches reference), 0 console errors;

gates CLEAN (dark both modes, responsive). Index-only CSS, re-minified.

Changed — reworked the homepage portfolio (bento) section to match the owner's

reference across day / dark / rainbow, closing the "only partial elements checked" gap.

• Per-card colored glow. Each experience card now has a pronounced gradient glow

rising from its bottom edge in its own accent — ZTE blue, Astra green, Cargill amber,

AWS violet, Pure Data magenta (recolored AWS orange→violet + Pure Data purple→magenta

to match the reference). The bottom feature cards get the same treatment (governance

teal, SAP green, Explore amber). Previously the accent was a faint hover-only corner

circle (~0.10 alpha); now it's an always-visible bottom radial with content stacked

above it (.bento-exp-card > * { z-index:1 }).

• Intensity scales by mode — day subtle pastel (opacity .34), dark medium (.66),

rainbow full-strength (1.0) — so rainbow reads as the vibrant reference while day

stays clean.

• Download Resume CTA → amber-gold gradient (#E8B563→#f59e0b, a brand accent),

matching the reference (was emerald).

• Verified visually against the reference in all three modes (screenshots), 0 console

errors; gates CLEAN (dark both modes, responsive, version-stamp, js-syntax, script-tags).

Index-only CSS in styles-index.css (re-minified, cache-bust bumped).

Changed — rolled the shared RZCalc hardening (input validation with error states

• CSV export) across the remaining engine-backed calculators, completing the program.

• capex-calculator (validate IT load + fuel autonomy), opex-calculator (IT load),

roi-calculator (10 financial inputs), tco-calculator (4 core inputs; the

4.8k-line flagship, integrated non-invasively — bespoke theme + Pro machinery

untouched), carbon-footprint (4 editable inputs; readonly/derived excluded), and

cx-calculator (validation on its button-triggered cxCalculate(); its existing

Gantt CSV kept). Each gets an Export CSV button (except cx, which already had one).

• All use the shared js/rz-calc-utils.js (window.RZCalc) — no per-page duplication.

Validation is non-gating on these (they read inputs defensively with ||default

/ getVal(id,default)), so out-of-range inputs are flagged (.rz-invalid + inline

summary) without altering the compute.

• Every engine-backed calculator on the site is now hardened: cdu, fire, pue, capex,

opex, roi, tco, carbon, cx. Tracker: standarization/CALC_HARDENING_ROLLOUT.md.

• Verified headless — all 6: util loaded, calc renders results, out-of-range → summary +

field marked, restore → cleared, 0 code-level console errors (only pre-existing

external geo-IP requests). Audits CLEAN (js-syntax, script-tags, version-stamp,

dark-coverage both modes, responsive).

Added — a reusable production-hardening utility for the engine-backed calculators

(M-303: one shared engine, not per-page hardcoded), plus its first rollout.

• js/rz-calc-utils.js (window.RZCalc) — self-contained, zero-dependency ES5

helper that injects its own theme-aware CSS and provides: validateNumbers()

(checks numeric inputs vs their min/max, marks .rz-invalid + aria-invalid,

returns {ok,errors}), showErrors() (toggles a .rz-calc-validation summary),

and downloadCSV() / csvEscape().

• pue-calculator.html now uses it: out-of-range inputs (e.g. IT load beyond

50–100000) surface an inline summary + field marking (non-gating — the calc is

already defensive with ||default), and a new Export CSV button downloads all

inputs + the computed results (PUE, DCiE, total power, cooling load, UPS loss,

annual energy/cost, rating). Verified headless: calc renders; out-of-range → summary

• .rz-invalid; restore → cleared; CSV → 16 rows; 0 console errors.
• standarization/CALC_HARDENING_ROLLOUT.md — tracker + integration pattern for the

remaining calculators (capex/opex/roi/tco/carbon/cx), so the rollout is consistent.

Added / Changed — same production-readiness pass as v1.50.28, now on

fire-calculator.html (Program Phase 3 — calculator-suite consistency).

• Input validation + error states. Numeric inputs (room volume/area/temp, battery

energy) are validated against their min/max before the engine runs; the design

concentration stays optional (blank = default Class-A). Invalid fields get

.invalid + aria-invalid and a clear inline summary ("Room volume is required",

"Design concentration must be 1–60"); results show "Awaiting valid inputs" instead of

NaN.

• CSV export. New "Export CSV" button downloads all inputs plus the computed KPIs

(agent quantity, design concentration, occupant-safety margin, GWP/residual-O₂,

detectors, discharge time, and the Li-ion runaway/off-gas panel when present) —

in-browser, no server transmission.

• Verified headless: valid → 9 KPIs + 9 CSV rows; cleared field + out-of-range → summary
• marking; blank optional concentration → no error; 0 JS errors; audits clean.

This completes the production-readiness hardening across the site's two bespoke

engine-backed calculators (CDU + Fire). The Finance-Terminal gateway (Phase 1) is built

• tested and awaits the owner's one-time Cloudflare deploy (cf-worker/DEPLOY.md).

Added / Changed — production-readiness pass on cdu-calculator.html (Program Phase 2).

• Input validation + error states. Every numeric input is validated against its

min/max before the engine runs. Invalid or empty fields are marked (.invalid +

aria-invalid) and a clear inline summary appears (e.g. "ΔT must be 3–20", "IT load

is required"); the results panel shows an "Awaiting valid inputs" prompt instead of

feeding NaN to the engine. Fixes silent/NaN output when a field is cleared or out of range.

• CSV export. New "Export CSV" button downloads the full run — all inputs plus the

8 computed KPIs (flow, velocity, ΔP, HX approach, NPSH margin, dew-point margin, pump

power, N+1 count) — built entirely in the browser (no server transmission), matching

the existing PDF export.

• Verified headless: valid state renders 8 KPIs + 8 CSV rows; cleared/out-of-range

fields trigger the summary + field marking; recovery restores results; 0 JS errors;

script-tag + js-syntax audits clean.

Added — the homepage (index.html) now has a full-page gradient backdrop and a

third, index-only Rainbow theme, per owner reference designs.

• Gradient backdrop — a fixed, GPU-composited .rz-bg-gradient layer behind all

content (no background-attachment:fixed, so it never repaints on scroll — safe on

low-power tablets). Three tasteful variants driven by the theme:

• Day — soft pastel radial wash (blue / lavender / peach, 0.10–0.18 alpha).
• Dark — subtle teal / blue / violet aurora on the near-black body.
• Rainbow — vibrant multi-hue aurora (violet → blue → teal → magenta → amber),

a slow 26s hue-drift (reduced-motion-safe), brightened hero aurora-orbs, and a

faint iridescent bento-card edge.

• Index-only 3-way theme toggle (day → dark → rainbow → day) with a dedicated

rainbow glyph. Architecture keeps it isolated: Rainbow = data-theme="dark" + an

additive html.rz-rainbow class, so every dark component style still applies and

stays readable, and the shared theme key stays light/dark — **the other 66

pages are completely unaffected** (they never see Rainbow; an index-only rzRainbow

localStorage key drives it). A pre-paint FOUC guard applies the theme before first paint.

• Verified headless: 3-way cycle + rainbow persistence across reloads + rainbow icon;

other pages correctly stay dark (not rainbow); 0 console errors; no mobile overflow;

dark-coverage CLEAN (both modes); hero-text contrast over the rainbow wash 6.28:1 (AA);

uiux-reviewer APPROVED (no blockers). styles-index.min.css re-minified + cache-bust bumped.

CSS lives ONLY in styles-index.css (index-only feature; 2-stylesheet architecture).

Changed

• The remaining 29 pages with per-page inline Fuse.js search (index, articles, insights,

datacenter-solutions, articles 1–19, geopolitics + geopolitics-1/2/3, FF-1, future-forward) migrated to the

shared js/rz-command-palette.js: each page's dedicated "Feature 21: Global Search" inline script block

removed and replaced by the module. Zero inline search copies remain — one implementation site-wide, and

every page now also gets the "/" shortcut + Commands group (theme toggle, quick nav). Verified over HTTP on 7

representative migrated pages: Ctrl+K opens, queries return results, commands present, 0 page errors.

(article-19 had a marker variant + an eager fuse.js CDN tag — both removed; fuse.js is now lazy-loaded

everywhere.) This completes the search unification begun in v1.50.23.

Added

• js/rz-scrolly.js — a zero-dep scroll-driven narrative system: a pinned canvas (position:sticky)

evolves as step cards cross the viewport middle (IntersectionObserver band). Steps set data-step on the

container (CSS reveals canvas layers) and animate <text data-cv> counters (rAF easeOutCubic; instant under

prefers-reduced-motion; no-IO browsers see the final state). Works forwards AND backwards.

• Flagship: article-23 ("xAI Colossus: 150 MW in 122 Days") — right after "xAI did it in 122 days," the

reader scrolls through the build: (0) an empty Electrolux plant, construction 19 days after conception →

(1) unpermitted gas turbines + Tesla Megapack BESS power up, MW counter climbs to 150 → (2) rack rows fill

the hall, GPU counter spins to 100,000 → (3) DAY 122 — LIVE, with the Huang quote. Every milestone as

stated in the article body (ILLUSTRATIVE-chipped schematic). Canvas reuses the living-diagram SVG tokens +

dash-flow. Verified by real scroll-through: steps 0→3 advance with correct counters/layers and fully reverse,

0 page errors.

• Scrollytelling authoring recipe added to ARTICLE_DATAVIZ_STANDARD.md. This completes the 4-phase

"jauh lebih keren" program (v1.50.23 palette → .24 polish → .25 living diagrams → .26 scrollytelling).

Added

• js/rz-article-diagram.js — the cockpit animated-schematic language (CSS dash-flow on SVG pipes/busbars +

a live 1s ticker + fault-injection scenarios, the cdu-mini-bms/chiller-plant idiom) packaged as a reusable

in-article widget: [data-rz-diagram] figure = authored SVG + JSON config (baselines with min/max/exact

clamps, flows, scenarios with deltas/alarms/msg). Instrument scenario buttons ("Normal" first,

aria-pressed), aria-live narration, ticker pauses offscreen (IntersectionObserver), prefers-reduced-motion

disables the dash animation, theme-aware --dg-* tokens, caption + basis chip (ILLUSTRATIVE) reusing the chart

styles. Conduit stroke = 4px base + 2px dashed overlay at ~32 px/s (matches the cockpit spec).

• Reference pair:
• article-13 (Power Distribution) — live 2N single-line diagram: UTILITY → TX → UPS-A/UPS-B → PDU →

GPU RACK (NVL72 ≈ 120 kW, 480 V AC per the article). "Inject: UPS-A failure" kills the A-path, pins UPS-A at

exactly 0%, UPS-B carries 100% — the rack never notices. The 2N story, animated.

• article-9 (HVAC Shock) — live warm-water DLC loop: DRY COOLER ↔ CDU ↔ GPU RACK cold plates

(supply ≈45°C per NVIDIA, per the article). "Tropical: Jakarta 35°C" collapses the dry-cooler approach and

pushes supply past the design point (the article's thesis, animated); "CDU pump failover" slows the loop and

recovers.

• Provenance gate extended: tools/audit-article-charts.mjs now validates rz-diagram-cfg blocks too

(source + basisTag mandatory) — 24 configs clean. New "Living diagrams" section in ARTICLE_DATAVIZ_STANDARD.md.

• uiux-reviewer: APPROVED, no blockers; its MEDIUM patches (exact-value scenarios, aria-pressed,

aria-live, basis-chip mapping) shipped in this release. Conduit-stroke idiom ruled conformant.

Changed

• tools/audit-dark-coverage.mjs + tools/audit-responsive-layout.mjs now relaunch the browser when the

Chromium process dies mid-run (ConnectionClosedError under resource pressure) instead of aborting the audit.

Fixed

• Editorial articles ran TWO read-progress bars (an older per-page inline #scrollProgress + the editorial

runtime's .rz-read-prog). The inline bar is now retired wherever the editorial register is active (one CSS

rule in css/rz-article-dark.css) — exactly one bar remains.

• View Transitions coverage: the @view-transition cross-page fade/slide block existed only in styles.css

— mirrored into styles-index.css per the 2-stylesheet rule, so the homepage participates too.

Added

• Heading anchor links — hover any article h2 for a # affordance; click copies the deep link

(clipboard + ✓ confirmation) and sets the URL hash. Auto-slugs ids where missing. (js/rz-article-editorial.js)

• Live "≈N min left" chip in the related-rail head — prose-only word count (excludes embedded calculator/

widget text) ÷ 220 wpm, counts down as you scroll.

• Non-blocking editorial fonts — the Fraunces/Plex Google-Fonts stylesheet now loads async

(media="print" onload) with a <noscript> fallback on 68 pages; system-font fallback shows during load

(display=swap behavior), removing a render-blocking request.

• Note: the article TOC was found already present on every article (33 inline copies with scrollspy) — the

planned shared-TOC extraction was skipped as pure churn risk with no user-visible gain.

Fixed

• Search was completely dead on 9 pages (FF-2, FF-3, articles 20–25, 27): the navbar search button +

"Ctrl+K" tooltip + modal markup rendered, but zero JS was wired — clicking/typing did nothing. And

article-26 ran a degraded inline search that filtered a non-existent tags field (the index has

keywords), so keyword search silently never matched. All 10 now run the shared module below; article-26's

broken inline block removed.

Added

• js/rz-command-palette.js — the shared site search + command palette (standard going forward, see

UI_FEATURES_STANDARD.md). Ports the canonical Fuse.js modal (lazy CDN, search-index.json, recents, category

chips, match highlighting, hover preview, Ctrl/Cmd+K + arrows + Enter + Esc) and adds: a "/" shortcut

(outside inputs), a Commands group (theme toggle + quick navigation to Home / Articles / DC Solutions /

Glossary / Insights) shown when the query is empty and substring-matched while typing, and self-injecting

modal markup so it works on any page. Guarded by window.__rzPalette; the 29 pages with a working inline copy

are untouched this batch (migrate opportunistically).

• Verified by real keyboard interaction over HTTP: Ctrl+K opens, "fire" → 8 results, Esc closes, "/" reopens,

theme command toggles; article-26 + FF-2 same; index.html inline search regression-free.

Fixed / Changed — the Commodities tab rendered empty KPIs/table and a flat line

chart with no candlesticks.

• KPIs + table: removed the Enter API key gate; loadCommodities() now uses

batchQuotes(...,{yahooFirst:true}) (keyless Yahoo spark first, no fhLim stall —

same fix as B-009/B-010/B-011). 14 commodity ETFs render in ~0.4s.

• Real candlesticks: loadCmdChartLegacy() now renders through the in-house

renderCandles() (lightweight-charts: candle + volume histogram + SMA20 + crosshair

OHLC) using keyless yahooCandles data — the same TradingView-grade renderer the V2

gateway path uses, previously V2-only. New yCandleRows() adapts the candle arrays.

The Chart.js line remains as a graceful fallback if lightweight-charts is unavailable.

• Added a crosshair OHLC readout (#cmdOhlc) to the Commodity Chart card header.
• Verified headless (no key): 14 rows + drawn candlestick canvas; timeframe switch

(1W–1Y), commodity selection, and light/dark re-render cleanly; 0 console errors.

Tracker: B-006 candlestick chart → SOLVED (the analytics-panel + related-news half of

B-006 is the broader R-008 scope, gateway-backed — deferred).

Fixed — the Sectors, Economy and Futures tabs rendered empty (they gated on a

Finnhub API key, or stalled behind the shared 55/min Finnhub rate-limit budget that

the overview/watchlist tabs spend on page load).

• New keyless Yahoo quote fallback in the data layer: yahooQuote() +

yahooBatchQuotes() derive a Finnhub-shaped {c,dp,d,v} quote from Yahoo's

CORS-proxy-reachable chart/spark endpoints (the same proxy race that already

powers the candlestick charts). The multi-symbol spark endpoint resolves a whole

tab's ETF list in one request (~400ms) instead of N proxied requests that trip

public-proxy rate limits.

• batchQuotes(syms,{yahooFirst:true}) — Sectors / Economy / Futures now resolve via

Yahoo first (fast, keyless, no fhLim contention); Finnhub only fills any gaps.

Removed the Enter API key early-returns on those tabs.

• loadSectors() now renders the data table before its charts and guards each

new Chart() in try/catch, so a charting exception can no longer blank the tab

(matches the table-before-chart order the Economy/Futures tabs already use).

• Verified headless (no API key): all three tabs render real data in ~0.4s, 0 console

errors; the default batchQuotes path (overview/watchlist) is byte-identical

(opts defaults to {}).

Tracker: B-009, B-010, B-011 → SOLVED. (B-006 commodity candlesticks / B-008 news

remain — News is Finnhub-only; the broader fix is the gateway Worker, ROOT item.)

Added — the CDU↔FMECA integration (Ships 1–3) is now reflected in the Obsidian

second-brain vault + the web knowledge graph, closing the 4-ship CDU-integration plan.

• Apps/second brain/obsidian-knowledge-vault/05-Standards/CDU-Suite.md (new) — living

note documenting the 6-page CDU toolkit, the FMECA fault-mode integration (F11.1–F11.5,

component C-LQC-001..007, with S·O·D + RPN table), the 6 sourced charts, and a

standards-compliance status table. [[wikilinks]] to Standards-Hub / Calculators-Hub /

Comparisons-Hub / Dark-Mode-Rollout.

• Standards-Hub.md — added a "CDU Liquid-Cooling Suite" section (6-page table) linking to [[CDU-Suite]].
• Apps/second brain/index.html (web graph) — added 8 nodes (cduh, cducal, cdusel, cducmp,

cduchk, cdubms, aim, cspr) + their edge cluster (hub → 5 sub-tools; checklist/mini-BMS → AI

maintenance FMECA engine; calculator ↔ spares; cooling-pillar + Air-vs-Liquid compare + DC

Solutions). sync-graph.py: 0 dead URLs; node + edge arrays parse (146 nodes / 296 edges,

0 dangling endpoints); page renders headless with 0 console errors.

See standarization/CONTENT_LINKAGE_PLAYBOOK.md §2.6.

Added

• Interactive sourced charts (js/rz-article-chart.js + Chart.js, theme-aware, crosshair) on the CDU

tools — 6 charts, each data-rz-chart figure carrying a source + basisTag, all passing

audit-article-charts --strict. Datasets under data/cdu/chart-*.csv (each with source/basis_tag):

• cdu-checklist: PM-cadence by interval (DERIVED) · spare-parts cost × criticality (ILLUSTRATIVE) ·

FMECA fault RPN (DERIVED, from sod_rpn.csv).

• cdu-comparison: max single-unit capacity by CDU type (VENDOR, from cdu-models.csv).
• cdu-selection-guide: secondary flow vs capacity, ~0.85–0.93 LPM/kW (VENDOR).
• cdu-calculator: pressure drop vs flow, dP ∝ Q² Darcy-Weisbach (DERIVED).
• Chart CSS is supplied inline per page (un-scoped, theme-aware via each page's --cdu-/--ck-/--cp-

tokens) since the editorial chart CSS in css/rz-article-dark.css is register-scoped + parallel-owned.

Verified: 6 charts render + redraw on theme toggle, source caption + basis chip on each, 0 console errors;

script-tags / js-syntax / version-stamp / mobile / responsive-layout / dark-coverage / article-charts all

pass for the 4 pages. Ship 3 of the 4-ship CDU integration.

Added

• cdu-checklist.html §10 — FMECA fault reference (liquid cooling, F11.x): 5 expandable fault-mode

cards sourced from the FMECA-KG (docs/research/csv/), ordered by RPN (Severity × Occurrence ×

Detectability, IEC 60812): F11.1 chemistry-drift 140 · F11.4 fluid-degradation 120 · F11.3

manifold/hose-leak 108 · F11.2 CDU-pump-fail 60 · F11.5 filter-clog 60. Each card carries

component (C-LQC-00x), mechanism, effect chain, detection, corrective + preventive actions, S·O·D and a

source/confidence tag (ASME/OCP = STD · Vertiv/3M = VENDOR). Notes that RPN understates the high-severity

F11.2 (S=10). Links to the Mini-BMS + the FMECA knowledge base. Printable form renumbered §10 → §11.

Verified: 5 cards render + expand, nav + cross-links resolve, 0 console errors; script-tags / js-syntax /

version-stamp / mobile / responsive-layout / dark-coverage all pass. Ship 2 of the 4-ship CDU integration.

Added

• cdu-hub.html — new "Maintenance intelligence" card row linking the CDU toolkit to

ai-engineering-maintenance.html (FMECA + Knowledge-Graph + ML-advisor; calls out the CDU

liquid-cooling fault modes F11.1–F11.5) and spares-readiness-calculator.html (CDU critical-spares

readiness — plain link; the spares calc reads only financial query params).

• cdu-mini-bms.html — the educational note now maps each injected fault to its FMECA fault mode

(leak → F11.3, pump-fail → F11.2, filter-clog → F11.5) and links to the FMECA knowledge base.

• standarization/CONTENT_LINKAGE_PLAYBOOK.md §2.6 — new handoff rule for maintenance tools / CDU↔FMECA

integration (cross-link AI-maintenance + spares, tag fault IDs, carry sourced charts, update the second-brain).

Cross-links verified (ai-engineering-maintenance + spares-readiness-calculator resolve 200); 0 console errors;

script-tags / js-syntax / version-stamp / mobile / dark-coverage pass for the touched pages. Part of the

4-ship CDU↔maintenance integration. (ai-engineering-maintenance.html is parallel-session-owned — linked TO, not edited.)

Fixed

• B-004 completion (Apps/finance-terminal/index.html): the V2-gate removal (v1.50.9) made

wireTable() work in all modes, but the non-V2 render paths for the Sector and Futures tables

(loadSectors / loadFutures) never called it (the calls lived only in the V2 paths). So with a

Finnhub key in the default non-V2 mode, those tables populated but stayed un-sortable/un-filterable.

Added wireTable('sectorTable','sectorFilter') / wireTable('futuresTable','futuresFilter') to the

non-V2 paths. Both tables have proper <thead>s; page parses clean.

Fixed

• B-005 — "Market Dominance" cards render empty (Apps/finance-terminal/index.html): the working

horizontal-bar renderer (renderDominanceCards) was only wired into the V2 (gateway) crypto path; the

non-V2 loadDominance() still drew a Chart.js doughnut on #dominanceChart that rendered empty

(canvas sizing). Routed the non-V2 path through the same bar renderer (renderDominanceCards(g.data),

since CoinGecko /global wraps the payload in {data:{…}}). Verified: 6 dominance bars

(BTC/ETH/USDT/BNB/…) render with live data, empty canvas gone. (Same V2-gating root pattern as B-004.)

Fixed

• B-002 — "Error loading forex data" (Apps/finance-terminal/index.html): the forex API

(CFG.FK) pointed at api.frankfurter.app, which now **301-redirects to api.frankfurter.dev

without CORS headers on the redirect** — so the browser blocks the cross-origin redirect and the

fetch throws. Pointed CFG.FK straight at https://api.frankfurter.dev/v1 (HTTP 200 + ACAO:*,

identical {rates} shape; /latest, /{date}, /{start}..{end} all verified). Forex now loads

(29 live rates returned in-browser).

• B-003 — slow data load: yahooCandles() tried the 3 CORS proxies sequentially with an 8s

timeout each, so a single hung proxy stalled a chart up to 8s before the next was tried (compounding

across symbols → the reported multi-minute load). Now races all proxies in parallel

(Promise.any, 7s) — the fastest valid responder wins, a hung proxy can't block. Verified a chart

fetch returns 22 points in ~350ms.

Fixed

• Finance Terminal (Apps/finance-terminal/index.html) — table column sorting + live filtering

(B-004) were wired through wireTable(), which early-returned on if(!CFG.V2). So whenever the V2

gateway flag was off (the default, and the state when the gateway is down), every data table was

un-sortable and un-filterable — the reported symptom. Sorting/filtering are pure client-side DOM

operations (reorder the rendered rows / hide by text) with no dependency on the data source, so the

V2 gate was removed. Verified: header click sorts asc, re-click toggles desc, and the filter input

hides non-matching rows (crypto / screener / sector / futures tables). Self-contained app — does not

use rz-version.js; recorded here for the changelog only.

Changed

• Hardened the Contact-box email copy-to-clipboard (index.html) into a 3-tier fallback:

navigator.clipboard.writeText → document.execCommand('copy') → select the visible email text so

the user can press Ctrl+C. Verified the execCommand tier succeeds even with the Clipboard API forced

unavailable (shows "Copied!"); the final tier selects the address. js-syntax + script-tags clean.

Fixed

• The v1.50.0 changelog entry documenting the WhatsApp-number removal had itself printed the full

number into the public /changelog.html — which defeats the privacy intent. Scrubbed the literal

number from the CHANGELOG entry and regenerated changelog.html. (The number remains only in an

internal, noindex, non-sitemapped security-audit report that records the original finding.)

Changed

• Callout AI-slop, eliminated site-wide. The earlier de-slop covered only a handful of callout class names;

the uiux-reviewer flagged it as a systemic gap, and a scan confirmed nearly every article had bespoke

callout namespaces (formula-box, conclusion-box, warning-box, insight-box, danger-box, plus

per-article prefixes aif-* / pjm-* / col-* / dcj-* / a21-* …) still carrying saturated gradient

fills + 3–4px borders. Added attribute-selector rules in css/rz-article-dark.css that catch the whole

-box / -note callout family at once and apply the editorial language (flat color-mix tint +

background-image:none + hairline + 2px semantic accent rail + no heavy shadow), with a --cl accent driven by

the class-name hint (warning→amber, danger→red, success/positive→green). **Cards / panels / grids are

intentionally NOT swept** (structural, not text callouts). Cache-bust on all 38 editorial pages bumped to

?v=1.50.19 so the shared CSS reaches live visitors. Dark-coverage gate CLEAN (114 pages, both modes).

Added

• article-22.html ("NVIDIA's $4B Photonics Play") — free interconnect-power bar: **pluggable optics ~16W

vs co-packaged optics (CPO) ~9W per 800G port** (~44% cut; ≈350 kW saved across a 50,000-GPU factory). New

data/article-22/optics-power.csv. A better, on-thesis angle than the flat $2B+$2B investment split.

Rendered by js/rz-article-chart.js. (article-3 / article-9 checked and skipped — calculator-generated /

scattered-prose data, no defensible free static series.)

Added

• article-14.html ("The $64 Billion Rebellion") — free opposition-impact bar: **$18B blocked + $46B

delayed = $64B** of US data-center projects stalled by community opposition (24 states; cancellations

quadrupled 6→25, 2024→2025). New data/article-14/opposition.csv. Rendered by js/rz-article-chart.js.

15 interactive sourced charts now live (gate clean). (article-4 MTTR data is calculator-generated/dynamic →

not charted, to keep the free-flow series defensible.)

Added

• article-7.html ("From Reliability to Resilience") — free Uptime Tier downtime bar: max annual downtime

28.8 h (Tier I) → 0.4 h (Tier IV) (99.671% → 99.995% availability), red→green severity ramp. New

data/article-7/tier-availability.csv (Uptime Institute Tier Standard).

• article-5.html ("Technical Debt Is Operational Risk") — free cost-premium floating bar: vendor-lock-in

premiums integration 30–50% / parts 50–200% / service 20–40%. New data/article-5/tech-debt-premiums.csv

(Schneider WP37). 14 interactive sourced charts now live (gate clean).

Changed

• datacenter-solutions.html — the DC Solutions landing now adopts the editorial skin:

data-rz-register="editorial" + Fraunces serif hero. The .hero h1 rule is page-INLINE (scoped to this page,

overriding the global .hero), so the gradient-clipped title keeps its gradient and only the global .hero

on other pages is untouched. **This completes the editorial serif-hero rollout across every content landing

on the site** (articles · series landings · CDU landings · pillars · compares · reports · infographics ·

glossary · DC solutions). Cockpits keep the instrument register.

Changed

• asean-dc-report-2026, carbon-footprint, glossary, and the 3 infographic-* pages now adopt the

editorial landing skin: data-rz-register="editorial" + Fraunces serif hero (was bold Inter sans).

Inline-serif approach. This completes the editorial serif-hero rollout across the site's content landings

(articles, series landings, CDU landings, pillars, compares, reports, infographics, glossary). Cockpits

(calculators / mini-BMS / EPMS / grid monitors) intentionally keep the instrument register.

Changed

• All 10 compare-* pages now adopt the editorial landing skin: data-rz-register="editorial" + Fraunces

serif hero (covers both hero-class variants, .compare-hero and .cmp-hero; was bold Inter sans).

Inline-serif approach — each page keeps its own palette/Inter body. Continues the landing-skin rollout

(articles, series landings, CDU landings, pillars).

Changed

• The 5 pillar pages (pillar-cooling/power/fire-safety/standards/sustainability) now adopt the editorial

landing skin: data-rz-register="editorial" + Fraunces serif hero (.pillar-hero h1 was bold Inter sans).

Inline-serif approach — the rest of each page keeps its existing palette/Inter body. Brings the topic-hub

landings in line with the CDU landings + series landings (insights / geopolitics / future-forward).

Changed

• cdu-selection-guide.html — the last CDU landing/guide page without the editorial skin now matches its

siblings: data-rz-register="editorial" + Fraunces serif hero (.cdu-hero h1 was generic bold sans). Page

keeps its own palette (inline-serif approach, like cdu-checklist / cdu-comparison). All four CDU landing pages

(hub / selection-guide / checklist / comparison) now share the editorial serif-hero treatment.

Changed

• cdu-hub.html ("Liquid-Cooling CDU Toolkit" landing) now adopts the editorial register like the other

landing/hub pages (insights / geopolitics / future-forward): data/rz-register="editorial" + Fraunces serif

hero title + css/rz-article-dark.css. The .hub-hero h1/p selectors were added to the shared editorial hero

rules (dark + light). The hero title renders in Fraunces serif (was generic sans); body + cards unchanged.

Added

• article-19.html ("Singapore vs Batam: Why Cost Alone Doesn't Win") — free industrial power-cost

comparison: Singapore USD 0.17–0.22/kWh vs Batam USD 0.07–0.09/kWh (SG 2–3× more; a 50 MW Batam

facility saves >US$25M/yr) — floating bars that set up the article's "cost alone doesn't win" thesis. New

data/article-19/power-cost.csv. Rendered by js/rz-article-chart.js. 12 interactive sourced charts now live

(gate clean).

Added

• article-12.html ("How AI Data Centers Fund $57B in Grid Modernization") — free renewable-contracts bar:

Amazon 34 GW vs Microsoft 23.2 GW vs Malaysia's entire 35 GW grid (reference), making the article's "a

single company contracts as much renewable capacity as a whole country" point visible. New

data/article-12/renewable-contracts.csv (basis distinction noted: contracted PPAs vs total grid). Rendered by

js/rz-article-chart.js. 11 interactive sourced charts now live (gate clean).

Added

• article-18.html ("AI Factories vs Traditional Data Centers") — free rack power-density evolution line:

5 kW (2015 x86) → ~600 kW (Rubin Ultra NVL576, 2027+) across NVIDIA generations (A100, H100, GB300, Vera

Rubin), conveying the AI density explosion. New data/article-18/rack-density.csv (vendor generation specs).

Rendered by js/rz-article-chart.js. 10 interactive sourced charts now live (gate clean).

Added

• article-16.html ("Southeast Asia DC bubble?") — free Johor pipeline-by-stage bar: **487 MW operational /

422 MW under construction / 1.4 GW committed / 3.4 GW early-stage planning** (5.8 GW total), making the

"mostly unbuilt / speculative" bear case visible. New data/article-16/johor-pipeline.csv.

• article-11.html ("AI Data Centers vs Citizen Bills") — free electricity-cost-index bar: Bloomberg's

267% higher costs near major data-center activity vs 5 years ago (index 100 → 367). New

data/article-11/electricity-cost.csv.

• Both via js/rz-article-chart.js. 9 interactive sourced charts now live (gate clean).

Added

• article-21.html ("Nuclear SMRs for AI") — free SMR-capacity bar: Oklo Aurora 75 / NuScale VOYGR 77 /

X-energy Xe-100 80 / GE-Hitachi BWRX-300 300 / TerraPower Natrium 345–500 MWe (vendor design ratings).

New data/article-21/smr-capacity.csv.

• article-23.html ("xAI Colossus: 150 MW in 122 Days") — free build-pace bar: Colossus ~4 months

(122 days, 100,000 GPUs / 150 MW) vs typical hyperscale 18–24 months. New data/article-23/build-pace.csv.

• Both rendered by js/rz-article-chart.js (theme-aware, Source caption + basis chip), using the floating

(min–max) bar support added in v1.50.4. 7 interactive sourced charts now live (gate clean).

Added

• article-24.html ("Data Center Manpower Shortage") gains a free interactive salary-ladder chart — a

min–max floating bar per role/level: Technician $38–57K → Operations mid $68–84K → senior

$105–142K → Liquid-cooling specialist $90–160K → AI-infrastructure specialist $140–200K (no

four-year degree). New validated data/article-24/salary-by-role.csv (Glassdoor / PayScale / SalaryExpert +

BLS). Per ARTICLE_DATAVIZ_STANDARD.md.

Changed

• js/rz-article-chart.js — bar charts now support floating (min–max) bars; the tooltip renders the range

($min–$max).

Added

• article-25.html ("PJM Grid Crisis") gains a free interactive chart: PJM's Base Residual Auction clearing

price jumped ~9× in a single cycle — $28.92 → $269.92/MW-day (July 2024, 2025/26 delivery; total

procurement $2.2B → $14.7B). New validated data/article-25/auction-prices.csv (source + basis_tag per row),

rendered by js/rz-article-chart.js (theme-aware tokens, Source caption + PUBLISHED chip). Per

ARTICLE_DATAVIZ_STANDARD.md.

Added

• article-20.html ("AI Data Center Water Use: Altman vs the Data") now carries a free, interactive,

finding-titled chart in the reading flow — U.S. data-center water use **2023 measured (17B gal, Joule,

peer-reviewed) → 2028 projected (68B gal, Global Water Intelligence), ≈4×** — directly grounding the

fact-check. Driven by a new validated data/article-20/water-usage.csv (source + basis_tag per row) via

js/rz-article-chart.js; theme-aware brand tokens, Source caption + PUBLISHED chip. Per

ARTICLE_DATAVIZ_STANDARD.md. (First Phase-2 rollout increment after the v1.50.1 article-26/27 reference pair.)

Phase 1 (shared system + article-26/article-27 reference pair). The shared CSS/JS lands on all 27 editorial articles; the bespoke calculator/chart work is the reference pair, with the rest to roll out in batches.

Fixed

• Article body alignment ("nggak lurus"). Paragraphs were individually centered with a ch-based measure, so the 1.14rem lead paragraph and 1rem body paragraphs computed different widths and landed on different left edges. Replaced with a single left-aligned, justified reading column at a fixed rem measure (css/rz-article-dark.css) — every prose block (incl. the drop-cap lead), the hero figure, and tables now share one left+right edge.
• Callout AI-design-slop. .info-box(+variants), .ws-insight-box, .ws-engineer-note, .ws-evidence-block carried saturated gradient fills + 3–4px borders. All adopt the editorial language: flat color-mix tint + 1px hairline + a 2px semantic accent rail + one restrained shadow + capped radius — scoped to the editorial register so it lands site-wide. Inner callout text re-aligned to the heading edge.
• Article number badges washed out (articles.html): white text on saturated category gradients → one legible dark-glass instrument chip (hairline + mono tabular number) that reads on any thumbnail in both themes.

Added

• Section dividers — a tier-3 hairline above each <h2>, aligned with the hero/reading column.
• Related-articles rail — a compact sticky right rail built by js/rz-article-editorial.js from each page's existing "Continue Reading" cards (thumbnail derived from the href). 2-col on wide screens, collapses to the bottom grid < 1024px. Zero per-article markup edits.
• Calculator tooltip compliance (article-27.html .ws-calc-wrap): all 12 inputs + 9 KPI cards now carry .tip help affordances (formula / denominator / source) per TOOLTIP_STANDARD.
• Interactive, validated-data charts — new js/rz-article-chart.js renders on-brand Chart.js charts (CNBC-style hover crosshair + tooltip, instrument tokens, theme-aware, finding-titled) from an inline sourced config, with a visible Source caption + basis chip. Reference charts: article-26 dual-axis fluid-loss line (from the verified data/article-26/worked-model-scenarios.csv) and article-27 aging-workforce bar (new data/article-27/workforce-stats.csv, AFCOM 2024). New gate tools/audit-article-charts.mjs blocks any chart without a source+basisTag. See standarization/ARTICLE_DATAVIZ_STANDARD.md.

Changed

• tools/audit-dark-coverage.mjs + tools/audit-responsive-layout.mjs hardened (--disable-dev-shm-usage, resilient page close) to survive headless browser crashes when cycling many pages.

Fixed

• Hover wobble (index.html cards) — the company-logo hover used a springy easing

cubic-bezier(0.34, 1.56, 0.64, 1) (Y=1.56 overshoots) that made the logo bounce/oscillate. Replaced

with a smooth cubic-bezier(0.4, 0, 0.2, 1) + smaller scale(1.02) → a subtle micro-movement.

• Hover blink/disappear (index.html cards) — the global hover rule set animation: none !important

while the entrance animations (swingIn on .oe-card, bentoRise on .bento-exp-card) stayed bound

to the cards, so every mouse-leave re-applied and replayed the entrance from opacity:0. Several

reveal observers also kept re-adding .visible. Fix: the IntersectionObserver now unobserves after the

first reveal, an observer-independent init-time handler freezes each card's entrance animation

(animation:none) once it ends so it can never replay, and the animation:none was removed from the

hover rule. Verified by instrumented hover-testing: 0 entrance-animation restarts on hover.

Removed

• Personal WhatsApp number removed from index.html — the visible contact card

and all four structured-data spots (Person telephone+sameAs, ProfessionalService telephone,

Organization sameAs+contactPoint). The generic "Share on WhatsApp" share-bar button (shares the page

URL, not the number) is kept.

Changed

• Contact box enhanced (index.html .contact-info): availability line with a pulsing status dot +

a primary "Email me" CTA; the methods are grouped under labels (Direct contact · *Find me

elsewhere · Based in*); each email has a copy-to-clipboard button with "Copied!" feedback; refined

card styling with matching dark-mode overrides. Re-minified styles-index.min.css + script.min.js,

cache-busts bumped.

Changed

• cdu-calculator.html KPI value displays (.kpi-val, .cc-pp-kpi .v) now use

font-variant-numeric:tabular-nums slashed-zero, matching the instrument-grade numeric style

already used on fire-calculator.html and the checklist tables. Closes the last item of the

CDU-hub uiux back-port (theme-key was already unified in v1.49.5; dark chips already present).

Fixed

• Added a multi-resolution root favicon.ico (16/32/48 px, generated from assets/Favicon.png).

The site declared a PNG favicon via <link rel="icon">, but browsers still auto-request

/favicon.ico from the web root — which 404'd on every page (a console error visible on all

100+ pages). A health probe across the CDU + Fire pages confirmed this was the only remaining

console artifact; with the file in place, those pages now report 0 console errors and 0 4xx.

Fixed

• Articles no longer sprawl or sit left-stuck on wide screens. Root cause: there was no site-wide

reading-width rule on .article-content / .article-body (only inside @media print), so the body grew

with the viewport while the 68ch paragraph cap held text left-aligned and tables filled the full body width

— at a 2400px viewport article-27's body reached 1520px with a 1520px table beside a 930px text column

("table lebih lebar dari text, berantakan"). Added a shared reading-layout system in styles.css:

.article-body is capped to 1180px and centered, and its direct prose children (`p, h2–h4, ul, ol,

dl, blockquote, figure, pre, table) share one centered 760px` reading column. Because only direct prose

children are capped, the interactive widgets embedded in .article-body (calculators, strategy grids,

gantt charts) keep full width — verified article-27's Workforce calculator still renders at 1116px.

• Tables scroll instead of overflowing on phones. .article-body table becomes an overflow-x:auto

scroll container ≤900px. Fixes article-1 (4 data tables, was +210px horizontal page scroll) and

tia-942-checklist (.gap-table, was +38px).

• EPMS_Telemetry no longer scrolls sideways on mobile. The position:absolute .ui overlay held a

fixed-height toolbar wider than a phone; the page's overflow-x:hidden guard was defeated by the

overflow-x:hidden+default-overflow-y:visible scroll-promotion quirk. Switched the guard to

overflow-x:clip and constrained the toolbar to 100vw with internal horizontal scroll (cockpit stays a

desktop dashboard).

• Editorial skin (css/rz-article-dark.css) reconciled: the 68ch measure is now centered

(margin-inline:auto) and extends to direct-child tables, so editorial articles match the new column.

Added

• tools/audit-responsive-layout.mjs — new ship-gate. Renders every content page at 390 / 768 / 2400px

and FAILS on real user-facing horizontal scroll (measured via actual scrollX, not the scrollWidth

artifact) or an article prose table wider than the reading column. CLEAN across 113 pages. Added to the

CLAUDE.md audit suite. See standarization/RESPONSIVE_STANDARD.md "Article reading column".

Added

• pillar-cooling.html now includes a Liquid-Cooling CDU Toolkit resource card (→ cdu-hub.html),

placed next to the Air-vs-Liquid comparison. The pillar's own intro promises it "links every

cooling-related resource," but the CDU suite — the biggest recent cooling addition — was missing.

Found via a cross-linkage evaluation (search-index coverage for all 9 new pages is complete; the

fire pillar↔tools links were already in place; this was the one real pillar→tool gap).

Added

• datacenter-solutions.html Engineering & Compliance Tools now lists the **Fire Suppression

Calculator (fire-calculator.html) and Fire Safety Checklist** (fire-checklist.html) alongside

the CDU Toolkit — previously only the Fire-Safety pillar was linked, not the interactive tools

(the same discoverability gap just fixed on tools.html). Tool-count badge 9 → 11.

Fixed

• Corrected a stale <!-- coming soon --> code comment on the PLN Sumatra card (the card itself has

been published/linked since v1.46.6).

Fixed

• Cross-page theme persistence. 19 pages stored the theme under non-standard localStorage keys

(rz_theme on the cdu-, compare-, pln-java-grid-*, dc-market-tracker, tier-advisor pages;

rfs_theme on rfs-readiness-workbench) while the rest of the site + script.js + the FOUC guard

use 'theme'. Result: setting dark on one page didn't carry over when you navigated to one of

these — it reverted to its own key's default. Unified all of them to 'theme' so the theme

persists across the whole site (the admin panel's separate rz_admin_theme is intentionally left).

Verified: the affected pages now honour the shared key on load and write it on toggle; dark-coverage

gate CLEAN in both modes; JS audits CLEAN.

Known (pre-existing, separate)

• tier-advisor.html has a chartjs-plugin-annotation console error (chart-lifecycle, unrelated to

the theme-key change) — tracked for a separate fix.

Changed

• cdu-mini-bms.html P&ID schematic now colours each tagged instrument by its live alarm state

(amber WARN / red ALARM), mirroring the tiles and the active-alarm banner — so a fault reads

consistently across all three views instead of the schematic only showing the leak marker. Example:

on Filter clog, PDT-01 (filter ΔP) and PDT-02 (loop ΔP) go red and FT-01 (flow) goes amber. The

LT-01 reservoir-level instrument is now also link-interactive (click ↔ its tile) and state-coloured.

The both-mode audit's structural check (body must match the active theme) caught the

inverse of the v1.48.1 bug: pages with a light palette that stay dark in light mode.

Fixed

• changelog.html stuck dark in light mode — it had [data-theme="light"] rules for the

content cards but no [data-theme="light"] body, so light mode showed a light navbar over a

dark body (a broken middle state). Added the light body/surface rule to the generator template

(tools/build-changelog-html.py) and regenerated; the hero band stays dark intentionally.

(achievements.html is intentionally dark-only — no light palette — so it's correct as-is.)

Changed — enforcement now covers BOTH modes

• tools/audit-dark-coverage.mjs now also fails STUCK-DARK-IN-LIGHT: a page declaring a

light palette ([data-theme="light"] / :root:not([data-theme="dark"])) must render a light

body in light mode. Pages with no light palette are dark-only (cockpits, dark trophy pages) and

skip the light check. Gate CLEAN across 114 pages in both themes.

Added

• tools.html now lists the recently-shipped tools that were missing from the hub:

Fire Suppression Calculator (fire-calculator.html) and Fire Safety Checklist

(fire-checklist.html) in Compliance & Standards Tools, and the **PLN Sumatra Interconnected

Grid** (pln-sumatra-grid.html) in Market & Grid Monitors. (All three pages already existed and

were in sitemap/search-index/llms — only the hub listing was stale.)

Fixed

• tools.html tool count corrected 18 → 23 across the hero copy, hero badge, meta/OG/Twitter

descriptions, and the JSON-LD CollectionPage ItemList — which was also completed to 23 entries

(it had been missing the CDU toolkit and Spares Readiness calculator in addition to the new tools).

Removed

• index.html Insights navbar dropdown — removed the "Tools & Calculators" sub-item (it mixed a

tools link into a reading/insights menu). The dropdown now reads Engineering Journal · Global Analysis

· Future Forward · Glossary · — · Second Brain · All Insights, matching the Insights dropdown already

used on the rest of the site. tools.html stays reachable via the global search and page footers;

the calculators remain in the DC Solutions dropdown.

Added

• cdu-checklist.html §07 — "Spares planning" subsection turning the parts register into an

actionable stocking plan: per-item min stock on site, typical service life (real replacement

cadence, not just the inspection interval), lead time (the real driver for holding a critical

spare), and estimated annual spend per CDU. Closed with a budget roll-up: recurring

consumables ≈ $1,000–4,000/yr/CDU, one-time on-site critical-spare kit ≈ $1,500–6,000, shared

fleet spare pump ≈ $600–3,500 — all EST planning bands (not a quote), noting filters + fluid as

the dominant cost drivers.

• cdu-mini-bms.html tile reading-guide — a compact "how to read a tile" strip above the live

tiles explaining the anatomy (instrument tag matching the P&ID · value+unit · range bar with normal

band + live marker · NORMAL/WARN/ALARM status · click for trend), so the redesigned board is

self-explanatory to learn.

Audited all 114 pages in both light and dark (228 renders) with a refined,

gradient-aware text-contrast probe, then visually verified the worst-flagged pages.

Fixed

• .cmp-badge-a/-b low contrast on the comparison pages — the A/B badges set

accent-coloured text on a light tint of the same accent (color:var(--cmp-accent) on

background:var(--cmp-accent-light)), ≈2.5 contrast (sub-WCAG) in both themes. Changed to a

solid accent background with white text (high contrast, theme-independent). 5 compare pages.

Audit result

• The rest of the contrast flags were false positives confirmed by screenshot: hero photo /

overlay backgrounds, auth-gate dimming (gated labs/cockpits), gradient card backgrounds, and

intentionally-dim mono labels/captions. Body text, headings, lists, and tables render readable

in both modes across the site. The genuine white-body-in-dark class was already fixed + gated

in v1.48.1 (audit-dark-coverage.mjs, still CLEAN across 114 pages).

Owner report: many content pages render broken in dark mode (only nav + title dark,

article body stays white) and the standard isn't enforced across sessions. Root-caused,

fixed, and made self-enforcing.

Fixed

• The :root, [data-theme="light"] cascade bug on 11 pages (cdu-calculator, cdu-hub,

cdu-selection-guide, compare-air-vs-liquid-cooling, compare-ashrae-vs-uptime,

compare-fm200-vs-novec, compare-pue-vs-dcie, compare-tier-3-vs-tier-4, fire-calculator,

fire-checklist, pln-sumatra-grid). They defined a dark var palette, but a

:root, [data-theme="light"] selector matched in all themes and (equal specificity, later

in source) overrode the dark values, so body + cards + text stayed light in dark mode.

Changed the light fallback to :root:not([data-theme="dark"]). One-selector fix; flips the

whole var-driven palette (background AND body text) to dark.

• tia-942-checklist — :root had only a light palette; added a [data-theme="dark"]

block redefining --bg-primary/--bg-secondary/--text-*/--glass-bg to the dark palette, plus a

.calc-disclaimer dark override.

• tier-advisor — added a .calc-disclaimer dark override (the one hardcoded-light element).

Added — enforcement (so future sessions can't ship broken dark mode)

• tools/audit-dark-coverage.mjs — render gate: loads every content page in dark and FAILS

on a white body or large light content block, and statically flags the :root, cascade bug.

Added to the ship-audit suite in CLAUDE.md. Run: node tools/audit-dark-coverage.mjs --strict.

• Documented the mandatory pattern in CLAUDE.md + standarization/DARK_MODE_STANDARD.md: every

content page must define a dark palette (or load the standard skin) and pass the gate; never use

:root, for the light fallback.

Verification

• Gate CLEAN across 114 content pages (0 white-in-dark, 0 cascade bug); cdu-selection-guide

(the reported example) now body-lum 15. Audits script-tags/js-syntax CLEAN.

Changed

• cdu-mini-bms.html live-parameter tiles redesigned for BMS accuracy and clarity.

Each tile now shows its instrument tag (FT-01, TT-01, PT-01, LT-01 … matching the

P&ID), a range bar with the normal band shaded and a live position marker, and an

explicit NORMAL / WARN / ALARM status line with the numeric range. Replaces the

ambiguous "band X" subtitle (which printed nonsense like "band PG25 OK" / "band cap×").

Derived tiles (flow total, return temp) are now labelled DERIVED with their formula.

• Faults now propagate to the parameters they physically affect (the previous model

flipped a single flag, so a Leak alarm left every other tile green). Now: Leak →

reservoir level (LT-01) falls + system pressure sags + make-up active; Filter clog →

loop dP up + flow down + ΔT rises; Pump-A fail → N+1 → N (flow holds on standby);

Hot facility water → supply up + HX approach widens; Low flow → ΔT up + dP down.

Correlated tiles change state together, so the board reads like a real BMS.

Added

• Active-alarm banner above the tiles — severity badge (NORMAL/WARN/ALARM), a

plain-language cause + operator response for the selected fault, and chips listing every

tripped instrument (tag + parameter). Reads "All parameters nominal" when healthy.

• Reservoir-level tile (LT-01) with its own trend, giving the leak scenario a second

correlated signal (level falling) beyond the leak switch.

Fixed

• Scenario selector: the active Normal chip now reads neutral-green (healthy) instead

of the alarm-rust colour that made a healthy board look faulted.

Changed

• KPI hero-strip values across all editorial articles now render in JetBrains/IBM Plex

Mono with font-variant-numeric: tabular-nums, and labels in IBM Plex Mono uppercase

(shared css/rz-article-dark.css, both themes). Removes the generic "SaaS hero-metric"

read; per-article accent colour + hero identity kept. Verified mono+tabular both themes.

Added

• cdu-checklist.html §07 Spare-parts & consumables register — a 15-row matrix

mapping every PM line that consumes a part to: what to replace, the

acceptance / replace-when trigger, interval, a part reference / spec (commodity

manufacturer family or OEM service-part class), a critical-spare class

(CRITICAL / RECOMMENDED / ON-DEMAND), and an estimated unit-cost band.

Covers filter/strainer element, PG25 fluid, dry-break QDs, EPDM gasket kit, pump

seal kit, spare pump, expansion bladder, leak rope+controller, flow / dP / temp-RH

sensors, TCV actuator, inhibitor/biocide dosing, PSV, and L2A air filter. Cost

bands tagged EST (illustrative market range, not a quote); specs STD/VENDOR.

• cdu-checklist.html §08 Exploded-view & IPB resources — per CDU type — per-type

cards (in-rack / in-row / sidecar / L2L end-of-row / L2A air-cooled) with

representative OEMs and links to obtain the Illustrated Parts Breakdown / service

manual (CoolIT, Vertiv, Delta, Stulz, nVent Schroff, Motivair/Schneider, Boyd,

ZutaCore) plus cross-links to the CDU comparison.

Changed

• cdu-checklist.html §06 PM checklist — enriched from 3 columns to 4

(Task | Frequency | Acceptance/action | Parts/consumable); each part-consuming row

now deep-links to its §07 register line. Added PSV annual test, QD/gasket inspection

and L2A air-side coil service rows; sharpened acceptance criteria (e.g. filter dP

trigger > clean +0.3–0.5 bar). Symptom and printable-form sections renumbered §09/§10.

Added

• pln-sumatra-grid.html + js/pln-sumatra-grid-data.js — a new PLN Sumatra interconnected

grid monitor, built like the Java-Bali one: an interactive Leaflet map + 31-node substation/plant

atlas of the 275/150 kV north-to-south backbone (Aceh→Lampung), major plants (Asahan, Pangkalan

Susu, Bukit Asam, incl Medco gas), and the separate Batam-Bintan island grid with **Medco Power

Panaran** and the Nongsa subsea-cable data island. Self-contained (no engine dep); 0 dangling edges;

curated from RUPTL 2025-2034 + PLN AR 2024 with confidence tags. Registered in sitemap/llms/search-index.

Changed (datacenter-solutions.html — PLN section rearrange)

• Removed the 4 placeholder "SOON" cards (Kalimantan, Sulawesi, Maluku-Papua, Nusa Tenggara) — the

grid now shows the 2 published monitors (Java-Bali + Sumatra), a cleaner grouping.

• Flipped the Sumatera card from "Coming soon" to published (NEW badge, links to the new monitor,

Medco/Batam/Nongsa features); footnote updated.

Verified

• Audits clean; Sumatra page: map + 60 markers/lines + 31 atlas rows + Medco/Batam stats render,

dataset 0 dangling edges, 0 console errors; datacenter-solutions PLN section 2 cards, 0 errors.

No bug, no error.

Acted on a uiux-reviewer audit of the editorial register. All in shared

css/rz-article-dark.css, both themes, every article + hub. Verified 37 pages:

0 light-on-dark, 0 errors, 0 over-wide measure.

Fixed (review CRITICAL/HIGH)

• Reading measure — prose was uncapped (~142ch). Capped .article-body p/li/h2/h3/blockquote

to 68ch (lists 70ch); tables/figures stay full-width. The single biggest readability win.

• Day body font — the day register block had no .article-body p rule, so day copy fell back

to Inter (the font design.md rejects). Forced IBM Plex Sans + line-height 1.75 in both themes.

• Heading hierarchy — skin never set heading sizes (h2 inherited 24px, h2/h3 ratio 1.2, equal

weight in dark). Now explicit editorial scale: h2 clamp(1.75rem,3vw,2.25rem) weight 600,

h3 clamp(1.2rem,2vw,1.45rem) weight 500 (ratio ≈1.55, weight contrast restored).

• Vertical rhythm — h2 was glued to the next paragraph; added margin:2.4em 0 .55em (h2),

1.7em 0 .4em (h3).

• Day hero wash — the day .article-hero::before amber radial stacked over saturated bespoke

heroes (e.g. article-26 orange) and washed out the dek; removed it in day (content:none).

• figcaption — articles hardcode inline italic 14px #64748b; the skin rule now uses

!important + font-style:normal so captions render as intended mono/muted.

• Title size normalised to one floor clamp(2.6rem,5vw,3.5rem).

Added (requested refinements)

• Section-number kickers — magazine-style mono 01 / 02 / 03 on each h2 (CSS counter),

replacing the bare amber tick with the number + an amber underline.

• Lead paragraph — first body paragraph set to 1.14rem to open the article (pairs with the drop-cap).
• Figure framing — 2.2rem rhythm, soft 8px corners, responsive img.

Design pass on the editorial register (impeccable-guided, on-brand). All in the shared

css/rz-article-dark.css, so every article + hub gets them; both themes.

Fixed

• Day-mode accent contrast. Each article sets style="--rz-art-accent:#E8B563" inline on

<html> (the dark gold). That inline value was overriding the day variant, so day-mode

links/meta/rails rendered in light gold #E8B563 (≈1.7:1 on white — fails). Forced the day

accent to readable #b45309 with !important (beats the inline). Dark unchanged (#E8B563).

Changed

• Inline body links now read as editorial amber with an always-on hairline underline that

brightens on hover (accessible; per-link inline colours like the purple disclaimer link are

preserved). Resolves the accent per theme.

• Pull-quotes lose the 2px accent side-stripe (an impeccable absolute-ban) for an even

hairline border + a hanging Fraunces quotation glyph — more editorial, on-brand.

• Figure captions → IBM Plex Mono, muted (matches the instrument-grade meta).
• Selection → a warm amber wash on editorial pages.

Verification

• Both themes: pull-quote glyph renders + border is 1px; links amber/underlined and readable

(day #b45309 / dark #E8B563); global dark re-probe 0 light-on-dark / 0 errors (34/34). Audits CLEAN.

Extends v1.46.2 from the 34 articles to the article-ecosystem hub pages

(articles.html, insights.html, future-forward.html) the §08 skin also targets.

Fixed

• No-FOUC theme guard added to the 3 hubs (they had the editorial register + skin CSS but

not the guard → dark readers flashed white on load). geopolitics.html already had it.

• insights.html — 6 white-card light-on-dark islands. The resource cards

(.reports-grid > a: Live Tracker / Regional Report / AI-HPC Platform / Infographic) use inline

background:var(--bg-card,#fff); --bg-card is undefined in dark there, so they fell back to

pure white. Added dark overrides (#1e293b card / light text). Re-probe: 0 light-on-dark.

Verification

• All 3 hubs: FOUC fires pre-paint, dark 0 light-on-dark, toggle flips cleanly, hero renders

Fraunces in both themes, 0 console errors. Audits CLEAN.

Owner report: the approved §08 article skin (plan-dark-mode-standard.html) wasn't

implemented properly across all articles, and many articles were buggy switching

dark↔day. Full sweep across all 34 content articles — "ensure no bug".

Added

• Day/light editorial variant (css/rz-article-dark.css). The §08 editorial register

was DARK-ONLY ([data-rz-register="editorial"][data-theme="dark"]); articles fell back to

the plain look in day mode. Added a light-palette mirror scoped

:not([data-theme="dark"]) — Fraunces serif title, IBM Plex Mono kicker/meta, gold drop-cap,

amber h2 accent-rail, pull-quote, soft warm hero wash — so the editorial skin now reads as the

same design in BOTH themes (gold #b45309 accents on light for contrast; article light

surfaces kept so bespoke layouts don't break). Verified Fraunces title/h2/drop-cap + mono meta

render in day on all articles.

• No-FOUC theme guard on all 34 content articles — inline <head> script applies the saved

theme before first paint (calculators already had this; articles didn't → dark readers got a

white flash). Confirmed data-theme is set at paint time.

Fixed

• Broken dark↔day toggle on article-26 — a redundant inline #themeToggle handler

double-bound with script.js initDarkMode; the two cancelled each other so the toggle was

stuck. Removed the inline handler (script.js owns it). **All 34 toggles now flip cleanly

(probe: 0 broken).**

• Dark-mode light-on-dark coverage gaps (via /ultraplan, 5 agents self-verified to zero):

article-10 .table-note-row, article-11 .bar-container, article-16 green pills + amber

timeline-dot, article-17 .highlight-row/.safe-box/.author-section/dot, FF-3

.iec-tooltip-trigger/.iec-benchmark-tag — plus article-13 amber .flow-box. Each light

pastel dark-toned to a same-hue dark tint with readable text; white gauge-needle markers left

intact. Global re-probe: 0 opaque light-on-dark across all 34.

• article-2 console TypeError — chartjs-plugin-annotation loaded without defer while

chart.js had it, so the plugin ran before Chart's helpers existed. Added defer. **0

console errors across all 34.**

Verification

• Headless probes: dark = 0 light-on-dark / 0 errors (34/34); toggle = 0 broken (34/34); day

editorial chrome renders; FOUC guard fires pre-paint. Audits script-tags/js-syntax CLEAN.

Changed (pillar-fire-safety.html — from thin landing to analysis hub)

• Added a fire-strategy four-stage lifecycle (Detect → Confirm → Suppress → Evacuate) and a deep

"Lithium-ion BBU — the thermal-runaway strategy" section (the owner's gap): why suppression does

NOT stop runaway, a 5-layer mitigation stack (cell-level BMS detection · H₂/CO off-gas detection +

ventilation per NFPA 855 · BMS isolation + EPO · compartmentation + UL 9540A spacing · suppression

for the associated fire), the off-gas hazard callout, and a clean-agent sizing/occupant-safety note.

• Added 2 resource cards (fire-calculator + fire-checklist) and a Li-ion FAQ; extended the

CollectionPage hasPart (now 6) + FAQPage schema. Scoped .fa-* styles with full dark-mode overrides.

Verified

• Audits clean; 3/3 JSON-LD valid; 6 cards / 4 lifecycle steps / 5 mitigation layers / 4 FAQs render;

0 console errors; dark + mobile pass. Fire-safety hub complete (Ships 1–5). No bug, no error.

Added

• fire-checklist.html — a super-detailed fire-safety checklist (9 sections, 5 source-tagged

tables): clean-agent design parameters (NFPA 2001 design conc / NOAEL / discharge / hold), detection

& alarm (aspirating, cross-zoned release, EPO), a dedicated Li-ion battery-room fire-safety

section (NFPA 855 / UL 9540A — off-gas H₂/CO detection ≤25% LFL, ventilation/explosion control,

thermal-runaway onset, BMS↔FACP coordination), the numeric commissioning procedure (door-fan

integrity, discharge test, cross-zone, Li-ion off-gas commissioning), installation + routine

inspection checklists, a PM cadence table (daily→annual), a symptom→cause→action troubleshooting

table, and a printable service-record form (window.print + print CSS). --fc- fire theme,

STANDARD/TYPICAL source tags, dark + mobile.

• Registered: sitemap.xml (114), llms.txt (125), search-index.json (new entry).

Verified

• Audits clean; 9 sections / 5 tables / 25 checkboxes / 35 source tags render; print form works;

0 console errors; dark + mobile pass. No bug, no error.

Added

• fire-calculator.html — interactive fire-protection calculator wiring js/fire-engine.js to

live KPIs: clean-agent quantity (NFPA 2001) + cylinders + kg/m³, design concentration, **occupant

safety vs NOAEL** (alarms when design conc exceeds NOAEL), GWP-weighted CO₂e (or inert residual-O₂),

smoke-detector coverage, discharge/hold bands, and a Li-ion BBU panel — thermal-runaway heat,

off-gas volume, and off-gas-vs-LFL room concentration (hazard-flagged, NFPA 855 / UL 9540A). Per-

scenario presets, tooltip per input, basis chips + band-coloured rails, --fc- fire theme.

• Applied the pending uiux fixes from the start: theme key theme (site-consistent), slashed-zero

numerics, dark-mode chip overrides, keyboard-focusable tooltips, 2px focus outline.

• Registered: sitemap.xml (113), llms.txt (124), search-index.json (new entry).

Verified

• Audits clean; 9 KPIs render; engine-backed values correct (Novec 500 m³ → 343 kg; FM-200 11% →

occupant-safety alarm −2 pts to NOAEL; IG-541 → 235 m³ inert + residual-O₂; 333 kWh NMC → off-gas

72.7 vol% hazard); 0 console errors; dark + mobile pass. No bug, no error.

Added (data/fire/ + glossary)

• data/fire/clean-agent-properties.csv (5) — Novec 1230 / FM-200 / IG-541 / IG-55 / CO₂: NFPA

2001 design concentrations, min-extinguishing, NOAEL/LOAEL, s=k1+k2·T coefficients, GWP, discharge.

• data/fire/li-ion-chemistry.csv (5) — NMC/LFP/LCO/VRLA thermal-runaway onset (150 / 166.8 /

150 °C), energy density, off-gas L/Wh + species, TR heat factor, governing vent-gas LFL.

• data/fire/standards-references.csv (12) — NFPA 75/76/2001/72/13/855, NFPA 70, UL 9540/9540A,

IEC 62619, FM Global DS 5-32, ISO 14520.

• data/fire/README.md manifest + basis-tag legend + key formulas.
• Glossary: 9 new fire/battery terms — Thermal Runaway, UL 9540A, Off-Gassing, Cross-Zoned

Detection, Compartmentation, Room Integrity (door-fan test), Design Concentration, NOAEL & LOAEL,

NFPA 855 — cross-linked to the fire pages.

Added (foundation for the DC fire-safety calculation hub — no UI yet)

• js/fire-model.js — deep-frozen window.FIRE_MODEL: clean-agent data (Novec 1230 / FM-200 /

IG-541 — NFPA 2001 s=k1+k2·T coefficients, design concentrations, NOAEL/LOAEL, GWP, cylinder fill),

Li-ion/VRLA chemistry (NMC/LFP/LCO thermal-runaway onset 150/166.8/150 °C, off-gas L/Wh, energy

density), off-gas LFL, detection/suppression bands (≤10 s discharge, ≥10 min hold, NFPA 72 detector

spacing, 25% LFL gas alarm), and a standards register (NFPA 75/76/2001/72/13/855, UL 9540/9540A,

IEC 62619, FM Global, ISO 14520). Every constant // source:-tagged + basisTag.

• js/fire-engine.js — pure deterministic engine window.FIRE_ENGINE (no Math.random): NFPA 2001

halocarbon agent mass W=(V/s)·(C/(100−C)) + inert volume ln(100/(100−C)); occupant safety margin vs

NOAEL; cylinder count; GWP-weighted CO₂e; detector coverage; discharge/hold-time bands; Li-ion

thermal-runaway heat, off-gas volume, room LFL margin; and a roomState() composite.

• tools/test-fire-calc.mjs — vm-sandbox, 31/31 PASS, hand-derived NFPA-2001 expecteds +

frozen/no-PRNG/determinism guards. Wired into tools/ship-gate.sh.

Changed (cross-link the new calculator into the CDU suite)

• Added in-content links to cdu-calculator.html from cdu-selection-guide.html (sizing lead),

cdu-checklist.html (operating-bands lead), cdu-comparison.html (intro) and cdu-mini-bms.html

(hero crumb) — the calculator was previously reachable only from the hub.

Deferred (with rationale)

• The mini-BMS physics-rewrite (route its synthetic basis through cduState()) is **intentionally

deferred**: its TYPES/SCEN values are already numerically identical to cdu-model.js (they were

lifted from it), so a core rewrite of the polished, accuracy-gated 5-phase cockpit would risk

regressions for zero numeric change. Recommended only as a separately-probe-verified effort if desired.

Added (cdu-calculator.html — Pro analysis tier)

• Free/Pro toggle + login modal (demo demo@resistancezero.com / demo2026), gated .cc-pro

section with blur overlay, rz_premium_session auth + rz-auth-change integration.

• TCO & ROI panel — capex (engine capexUsd by type), annual energy (computed pump power ×

$/kWh), annual opex, 10-yr NPV of ownership, $/kW·year — all from the engine's TCO functions.

• Monte-Carlo (10,000 runs) on annual pump-energy cost — inputs perturbed ±15–20%, P5/P50/P95 +

an inline SVG histogram. Randomness lives only in the page layer (SIMULATED), never the engine.

• Sensitivity tornado — ranks heat-load / run-length / fittings / pump-efficiency by impact on

pump kW (inline SVG).

• Dynamic engineering narrative (hydraulics / thermal-safety / economics).
• PDF tech-spec export — window.open() first, <\/script> escaped, derived-results + Pro

tables + embedded SVG charts, print-color-exact. Privacy note (browser-only).

• Font Awesome added for the lock/PDF icons.
• Verified end-to-end: real-click login unlocks; TCO/MC/sensitivity/narrative compute; 0 console

errors; audits clean; page still passes version-stamp + mobile.

Added

• cdu-calculator.html — a new interactive CDU sizing & thermohydraulic calculator (the

centerpiece of the hub). Wires the validated js/cdu-engine.js to live KPIs: secondary flow +

LPM/kW (vs OCP band), pipe velocity, pressure drop (Darcy-Weisbach), HX approach (ε-NTU), NPSH

margin, dew-point margin, pump electrical power and N+1 count. Per-CDU-type load presets, a tooltip

on every input, basis chips + band-coloured rails on every KPI (ACCURACY_VALIDATION rule 6),

--cdu- theme, full dark-mode + mobile, version stamp. 0 console errors.

• Registered across the site: cdu-hub.html (new "05 · Calculate" card + JSON-LD hasPart + "five

resources" copy), sitemap.xml (112 URLs), llms.txt (123 pages), search-index.json (new entry).

Notes

• Free-tier core ships first per the rollout plan; the Pro tier (TCO/NPV, Monte-Carlo, sensitivity

and a PDF tech-spec) lands next as Ship 3b. PG-25 fluid properties are handbook-class (ILLUSTRATIVE).

Added (data/cdu/ + glossary)

• data/cdu/coolant-fluid-properties.csv (24 rows) — water + PG-25 density/cp/viscosity/thermal-

conductivity/Prandtl across 5–60 °C, generated from the engine (build-fluid-properties.mjs).

• data/cdu/cdu-operating-bands.csv (20 rows) — the acceptance windows (supply, ΔT, flow LPM/kW,

dP, system pressure, approach, dew-margin, velocity, NPSH margin, water chemistry) from

cdu-checklist §01–03 + ASHRAE/OCP, each basis-tagged.

• data/cdu/cdu-models.csv (15 rows) — verified vendor CDU models with capacity/flow/dP/approach/

fluid/BMS-protocol/ASHRAE-class/link-status.

• data/cdu/standards-references.csv (10 rows) — ASHRAE TC9.9, OCP cold-plate/UQD/Deschutes,

Redfish DSP2064, ASME B31.3, ISO 4406/NAS 1638, ASTM D1193, PG-25.

• data/cdu/README.md manifest + basis-tag legend + link-validation mandate.
• Glossary: 12 new CDU terms — Approach Temperature, Cold Plate, Dew-Point Reset, Effectiveness-NTU

(ε-NTU), FWS, Filtration (ISO 4406), Inhibitor Reserve, Leak Detection (CDU), Manifold, NPSH,

Redfish CoolingUnit, TCS — each cross-linked to the relevant CDU page.

Added (foundation for the CDU data + calculation hub — no UI yet)

• js/cdu-model.js — deep-frozen window.CDU_MODEL: the single source of truth for CDU work.

Per-type presets (lifted from cdu-mini-bms TYPES/SCEN), operational bands (cdu-checklist §01),

ASHRAE W-classes, OCP cold-plate flow points, fluid anchors (water + PG-25), pipe IDs/K-values,

pump + TCO defaults, standards register, and verified vendor model rows. Every constant carries a

// source: tag + basisTag (STANDARD/VENDOR/DERIVED/ILLUSTRATIVE) per ACCURACY_VALIDATION rule 6.

• js/cdu-engine.js — pure deterministic engine window.CDU_ENGINE (no Math.random, numeric

guards): fluid properties f(T,glycol%); heat↔flow (Q=ṁ·cp·ΔT); ε-NTU + LMTD + approach; pressure

drop (Darcy-Weisbach, Haaland friction, velocity + fittings); NPSH + cavitation margin; dew-point

reset (Magnus/Tetens); pump hydraulic/shaft/electrical power + N+1; water chemistry (glycol top-up,

make-up, inhibitor reserve); per-type TCO/NPV/payback; and a cduState() composite (one source of

truth for the upcoming calculator + mini-BMS).

• tools/test-cdu-calc.mjs — vm-sandbox acceptance harness, 40/40 PASS. ~11 worked examples

with hand-derived expected values (independent of the engine) + deep-frozen, no-PRNG, determinism

guards. Wired into tools/ship-gate.sh.

Changed (data/article-26/ — corrections + verified_source columns)

• Two independent verification passes hardened the four curated CSVs; a verified_source column

(DOI/agency URL) was added to fluid-properties.csv, tfa-pfas-reference-values.csv and

regulatory-thresholds.csv. Corrections:

• Kazil 2014 journal fixed → J. Geophys. Res. Atmos. 119(24):14059-14079, DOI 10.1002/2014JD022058 (was mis-cited as ACP).
• TRI PFAS threshold 25,000 lb → 100 lb (PFAS are chemicals of special concern, NDAA 2020).
• ECHA TFA status "proposed" → RAC opinion adopted (June 2026), Repr. 1B / H360Df — also reflected in the article body + ref [21].
• Novec 7000/7100 breakdown products → COF₂ / HF / CO₂ (not shorter-chain PFCAs).
• Opteon 2P50: lifetime ~22 d (EPA SNAP), GWP ~2 (was ~10) — propagated to the calculator (runFluidCost) and the worked-model generator; worked-model-scenarios.csv regenerated.
• R-1233zd(E) lifetime ~40 d, GWP ~3.9; FC-40 GWP ~7100 / ~500 yr; FC-72 GWP ~7910 / ~3100 yr; Novec 649 GWP <1.
• AGAGE portal URL updated to the NASA LaRC archive; GHGRP format corrected to Excel/zip.
• Galden (PFPE) GWP flagged unverified/disputed (~10,000 commonly cited but PFPEs are nearly involatile) — marked in the CSV, the model constants, and a Model C footnote in the article.
• README.md gains a Verification section documenting the above.

Added

• data/article-26/measurements-usgs-wqp-genx-hfpo-da.csv — 4,832 rows of real GenX

(HFPO-DA) monitoring results from the USGS Water Quality Portal since 2020. GenX is the

current-generation "safer" PFAS replacement, now itself EPA-regulated (10 ppt MCL) — the direct

analogue to the article's HFO→TFA replacement argument. Reproduce via fetch-usgs-wqp-genx.sh.

• "The dataset that doesn't exist" finding box in the Data & Downloads section: querying the

USGS portal for trifluoroacetic acid (TFA) returns zero records under every name variant —

the breakdown product central to the article is unmonitored in US water-quality systems, a literal

confirmation of the "nobody is measuring it" thesis (this is a finding, not inline data).

• Now ~35,000 rows of real measurements across four extracts; grid → 10 cards; README + intro updated.

Added

• data/article-26/measurements-usgs-wqp-pfos.csv — 9,230 rows of real PFOS monitoring

results from the USGS Water Quality Portal since 2020 (both controlled-vocabulary terms ORed:

"Perfluorooctanesulfonate" + "Perfluorooctane sulfonic acid"), across water/groundwater/sediment/

tissue. PFOS is the second compound under the EPA 4 ppt MCL. Public domain. Reproduce via

fetch-usgs-wqp-pfos.sh.

• Now ~30,000 rows of real measurements across three extracts (atmospheric HFC-134a + aqueous

PFOA + PFOS); Data & Downloads grid → 9 cards; README + section intro updated.

Added

• data/article-26/measurements-usgs-wqp-pfoa.csv — 5,438 rows of real PFOA

(Perfluorooctanoic acid) monitoring results from the USGS Water Quality Portal (USGS/EPA/NWQMC)

since 2020: surface water, groundwater, sediment and tissue, reported by state agencies

(Minnesota PCA, Indiana, NJDEP, Arizona DEQ, NY DEC, Delaware River Basin, …). PFOA is one of the

two compounds under the EPA 4 ppt MCL. Public domain. Reproduce via fetch-usgs-wqp-pfoa.sh.

• Featured as the second card in the "Data & Downloads" block (now two real measurement extracts:

atmospheric HFC-134a + aqueous PFOA); section intro + README updated.

Added

• data/article-26/measurements-noaa-hfc134a.csv — 15,336 rows of real atmospheric

HFC-134a flask measurements from NOAA's Global Monitoring Lab (16 global sites, 1994–2026,

~1.8 → ~150 ppt). HFC-134a degrades to TFA at ~100% molar yield — the same TFA endpoint as the

HFO "PFAS-free" cooling replacements. Public-domain US Government data (PI: Montzka & Vimont).

Committed extract + convert-noaa-hfc134a.py (regenerable from the NOAA source URL).

• Featured as the first card in the article's "Data & Downloads" block; section intro + README

manifest updated; added a measured source-class tag.

Added (data/article-26/ — 6 downloadable CSVs + manifest)

• worked-model-scenarios.csv (6,000 rows) — the article's fluid-loss model across charge

(100–5,000 L) × make-up rate (0.5–20 %/yr) × 3 fluids → annual loss (L/kg), replacement cost,

t CO₂e/yr, kg TFA/yr. Reproducible via committed build-worked-model.py.

• fluid-properties.csv, loss-zones.csv, regulatory-thresholds.csv,

tfa-pfas-reference-values.csv — sourced evidence tables (each value source-class tagged).

• external-databases.csv — verified portal links to the large public monitoring archives

(EPA UCMR 5, TRI, CompTox; USGS WQP; NOAA GML / AGAGE; EU EEA Waterbase; NORMAN; German UBA;

CA GAMA; EPA GHGRP) where the raw measurement rows can be downloaded in full.

• README.md manifest documenting every file + source-class tags.
• Article "Data & Downloads" section (#section-data) — a links-only block (no inline data)

with download anchors to all six CSVs; new .pfas-downloads/.pfas-dl styles (dark-mode +

mobile-collapse covered).

Changed

• Refreshed article meta description + og:description to the reframed 17.1%/yr framing (they

still asserted the old "20-30x" as fact).

Fixed

• Tabular figures — added font-variant-numeric: tabular-nums + slashed-zero to .pfas-stat,

.pfas-kpi-value, and .pfas-table-container td so the number-dense new tables and Panel-5 KPI

cards align per design.md §2 (the chem-block ledgers already used JetBrains Mono numerics).

• Tablet dead-band — .pfas-pro-kpi-grid.cols-4 mobile fallback moved from max-width:680px to

768px to align with the page's master breakpoint; the 4-up Panel-5 grid no longer cramps to four

~160px columns on 681–768px tablets. (uiux-reviewer HIGH items; APPROVED with no blockers.)

Added (article-26.html — PFAS Risk Calculator Pro extension)

• Panel 5 — Fluid-Loss & Environmental Cost (new gated Pro panel) — 4 KPI cards derived from the

calculator's existing maintenance-vapor mass model (single source of truth, no new inputs):

Annual fluid lost (L/yr + kg/yr + make-up % of charge), Replacement cost ($/yr at the

per-fluid unit price), GWP-weighted (t CO₂e/yr), TFA formed (kg/yr for HFO fluids).

• Per-fluid property table in JS (runFluidCost): density, $/L, 100-yr GWP, molar TFA yield for

two-phase-pfas (Novec-class, GWP ~320, TFA n/a — yields short-chain PFCAs) and two-phase-hfo

(GWP ~10, TFA yield 0.695). PFAS-free configs zero out cleanly.

• .pfas-pro-kpi-grid.cols-4 4-up variant (collapses to 2×2 under 680px); pfasGate5 wired into the

existing lock/unlock + Pro-mode flow. Free-tier risk output untouched.

Verified

• Real-mouse Pro login (demo creds) unlocks Panel 5; KPIs compute across PFAS / HFO / PFAS-free paths;

0 console errors; values internally consistent with the on-page vapor estimate (256 kg → 183 L →

$12,800/yr → 82 t CO₂e at default 10×800 L).

Added (article-26.html — quantification core)

• Fluid-loss KPI framework (#section-kpi) — 6-KPI table (make-up rate, loss rate vs §608 trigger,

GWP-weighted t CO₂e/yr, TFA formation kg/yr, fluid-loss $/yr, reporting completeness) each with

formula · reference threshold · source class, closing on the "~0% reporting completeness" metric.

• Worked calculation models (#section-models) — 6 monospace ledgers with every input tagged

[published]/[vendor]/[illustrative] and results as bounded ranges: reference inputs · Model A

(single 800 L tank annual loss & cost, $1,120→$9,590/yr) · Model B (per-MW hall, $9k→$77k/yr) ·

Model C (GWP-weighted emissions, 61 t→1,920 t CO₂e/yr depending on fluid) · Model D (HFO→TFA

10-yr loading, 651–1,303 kg TFA/tank) · plus a "what the numbers do and do not say" caveat box.

Fixed

• .pfas-chem-block now renders multi-line — added white-space: pre-wrap + overflow-x:auto

(the class had a monospace font but no whitespace preservation, so the original chemistry block was

silently collapsing to a run-on line). Mobile font-size reduced to keep ledgers on-screen; verified

no horizontal page overflow at 390px.

Changed

• article-26.html wordCount 4800 → 6200.

Added (article-26.html — research-grounded fluid-loss metrics)

• Loss-zone taxonomy (#section-loss-zones) — new 8-row table mapping every fluid-escape pathway

(maintenance vapour, QD spillage, operating evaporative loss, drain/transfer residual, seal/permeation,

fill/flush, end-of-life) against sourced magnitude · metered? · reportable?, plus a "gaps are the

story" callout: the most routine loss zones (seal permeation, sensor-swap loss) carry **no published

value** — reinforcing the "invisible/unmetered" thesis.

• "Metrics to Watch" (#section-metrics) — leading-indicator watch-list table (make-up rate,

top-up cadence, level/chemistry drift, worker-air PFAS, exhaust concentration, boundary groundwater)

with healthy band · action threshold · interpretation, and a "cheapest instrument you already own"

box putting make-up rate forward as the single dashboard KPI (purchasing data ÷ installed charge).

• "Regulatory Horizon" (#section-horizon) — what changed after the April-2026 publication: the

ECHA RAC proposal to classify TFA as Reproductive Toxicant 1B (2025-2026) and the **EU Drinking

Water Directive PFAS-Total 0.5 µg/L limit live since 12 Jan 2026**, plus a 6-row instrument table

(EPA MCL, EPA §608, EU F-Gas 2024/573, EU DWD, ECHA TFA, member-state TFA limits).

• 8 new references [16]-[23] — DoD/LBNL field-loss study, UBA TFA degradation, Kazil 2014 rainwater,

EU DWD 2020/2184, EU F-Gas 2024/573, ECHA RAC TFA opinion, EPA §608, Solomon TFA review.

• Glossary — new Make-up Rate and TFA (Trifluoroacetic Acid) terms; updated `Maintenance Vapor

Release` term to the reframed estimate.

Changed

• Reframed the unsourced "20-30×" headline as a transparent modeled estimate anchored on the one

published field measurement (DoD/LBNL 17.1%/yr evaporative loss) vs the EPA §608 10% sealed-leak

baseline — labelled a modeled upper bound, not a measured constant.

• article-26.html dateModified → 2026-06-24, wordCount 3000 → 4800; sitemap lastmod + search-index

keywords/description/readingTime refreshed.

Added (ULTRAPLAN cockpit interaction, phase 5 of 5 — final)

• cdu-mini-bms.html — zoom / pan on the P&ID and datahall layout: + / − buttons with a

live zoom-level readout and reset (⤢), wheel-zoom toward the cursor (gated on the SVG being

focused or Ctrl held, so it never hijacks page scroll), drag-pan when zoomed (pointer events +

capture, clamped so the content always covers the viewport), and keyboard +/−/0. Implemented as a

CSS transform on the persistent box element, so it survives the SVG re-render. Vector-crisp at any

scale. Real-interaction deep-tested (button/keyboard/wheel zoom + drag-pan + reset + a Phase-1

click regression at 1×), both themes, 0 console errors.

ULTRAPLAN — CDU Mini-BMS operator-interaction layer COMPLETE (v1.43.58 → v1.43.62)

The cockpit is now a fully operable instrument panel: (1) bidirectional P&ID ↔ tile linking,

(2) pause/step/speed simulation controls, (3) trend history drawers, (4) guided fault walkthroughs,

(5) zoom/pan. Five reversible, real-mouse-deep-tested ships; presentation-only throughout — the

simulated values were never altered.

Added (ULTRAPLAN cockpit interaction, phase 4 of 5)

• cdu-mini-bms.html — a Walkthrough toggle opens a bottom bar that steps through narrated

callouts for the active scenario (Normal · Leak · Pump-A-fail · Filter-clog · Hot-FWS · Low-flow —

4–6 steps each). Each step explains the cause→effect chain and **highlights the relevant P&ID

instrument + parameter tile** (reusing the Phase-1 link). Prev / Next / Exit, a step counter and

scenario label; switching scenario mid-walkthrough resets to step 1 of the new one. Reduced-motion

honoured. Real-mouse deep-tested (start / next / prev / scenario-reset / exit + highlight +

disabled edge-steps), both themes, 0 console errors. Presentation-only.

Added (ULTRAPLAN cockpit interaction, phase 3 of 5)

• cdu-mini-bms.html — click a parameter tile to open a trend history drawer: a slide-in

panel with the full 300-sample sparkline (history cap raised 30→300; the tile keeps showing

the last 30), current value, min / avg / max, sample count, the parameter band and a sourced

reference, plus a deep-link to the checklist §01. Status tiles (pumps / filter / leak / coolant)

show their info without a chart. The drawer live-updates while open; closes on ✕ / Esc /

backdrop; reduced-motion honoured.

• Bug fixed via deep-testing: the .trend-drawer{display:flex} rule overrode the [hidden]

attribute, so the drawer was always rendered on top of the tiles and silently swallowed their

clicks — added .trend-drawer[hidden]{display:none!important}. Real-mouse deep-tested (numeric +

status tiles, Esc + backdrop), both themes, 0 console errors.

Added (ULTRAPLAN cockpit interaction, phase 2 of 5)

• cdu-mini-bms.html — a Simulation control group: Pause / Resume (freezes the data

ticks and the flow/pump animation via body.sim-paused + animation-play-state), Step

(advance exactly one tick while paused), speed 0.5× / 1× / 2× / 4× (variable timer rate +

CSS animation-duration scaling so faster ticks visibly flow faster), and a LIVE / PAUSED

state pill. Play/pause kept neutral (toggle, not radio); reduced-motion honoured. Real-mouse

deep-tested: pause held a value stable over 1.6 s, Step advanced one tick, speed set the rate,

both themes, 0 console errors. Still presentation-only — sim values unchanged.

Added (ULTRAPLAN — CDU cockpit operator-interaction layer, phase 1 of 5)

• cdu-mini-bms.html — the P&ID is now operable: click (or keyboard-activate) any tagged

instrument (FT-01 / PT-01 / PDT-01 / PDT-02 / TT-01 / TT-02 / TI-03) to highlight + pulse its

parameter tile and scroll it into view; click a tile to pulse its linked instrument; hover

cross-highlights both ways; Esc clears; single-selection invariant. Instruments get

role="button" + tabindex="0" + aria-label + a 32×41 transparent hit-rect.

• Fix surfaced by deep-testing: the animated flow pipes were painted over the instruments and

intercepted clicks intermittently (depending on the moving dash position) — set

#pidBox .pipe-* { pointer-events:none } so clicks reach the instruments beneath.

• Plan tracker: standarization/CDU_COCKPIT_INTERACTION_PLAN.md (5 phases). Presentation-only —

no change to the simulated values. prefers-reduced-motion honoured. Real-mouse deep-tested, 0 errors.

Owner feedback on the v1.43.54 amber swap: the re-toned callouts read as garish

yellow highlighter blocks in light mode, unlike the restrained §08 mockup (which

uses the soft gold #E8B563 as an accent — rails/borders/text — not a fill).

Changed

• De-yellowed every amber FILL across the 9 amber-body articles (article-11/14/20/23/25/26/27
• geopolitics/-3). The cream/amber-50/100 backgrounds and gradient stops

(#fffbeb, #fef3c7, #fde68a) on callout boxes, insight boxes, banners, and inline

stat pills were swapped to a subtle gold tint (rgba(232,181,99,0.06–0.22)). Text colours

that legitimately use those hexes (button labels, dark-mode emphasis) were protected and

left intact.

• article-27 inline stats refined — the .ws-stat pills lost their yellow box (was

#fffbeb fill + amber border + pill padding); they now read as clean amber-bold figures

with a thin gold underline. The .ws-insight-box / .ws-narrative callouts moved from a

yellow gradient to a subtle gold tint + a #E8B563 left rail (the mockup's "Key finding"

style).

• Net effect: articles now match the mockup's gold-accent restraint instead of looking

highlighter-marked. Verified light + dark on article-26 (the mockup's own subject) and

article-27 — 0 page errors.

Closes the last open planb item (I4). With this, every plan-dark-mode-standard.html

track is shipped or already-live — PLAN 02 is 100%.

Added

• I4 — instrument-register chrome for the 6 monitor pages (dc-market-tracker +

pln-java-grid and its 4 province pages). Each now declares

data-rz-register="instrument" and loads css/rz-monitor-instrument.css, which applies a

header + section-title signature only: .nav-title reinforced to JetBrains Mono, and

.pjg-section-title / .dmt-section-title get mono + a phosphor cyan→green accent tick.

• CHROME ONLY — owner-tuned viz preserved. The stylesheet targets only those heading

classes; it never touches the Leaflet/SVG map, SLD line widths/labels, voltage-tier colours,

legend swatches, tooltips, or animation thresholds. Verified headless: maps render (21/13

Leaflet elements), tier legend colours unchanged, section-title mono+tick present, **0 page

errors**. Reversible (remove the register attr + the <link>).

Status — planb PLAN 02 complete

• Track E 100% (E0–E4). §11 index hero ✓. §12 light twin ✓.
• Track I I0–I3 already-live (9 cockpits in instrument register, probe 75/75); I4 ✓ here.

Continues "semua di planb diimplementasikan". Closes the remaining Track E item

(E4) and the §11 index-hero recipe from plan-dark-mode-standard.html.

Added

• §11 — index identity hero, editorial register (DARK-ONLY, additive, reversible).

Applied the planb §11 recipe to index.html scoped to .bento-identity, dark mode only:

Fraunces serif .bento-name, IBM Plex Mono kicker on .bento-tag with a 20px amber tick

rail, amber-italic .bento-accent-text, and a new .bento-readout hairline strip — 4

site-factual KPIs (12+ Years Ops · 40+ Tools Built · 27 Articles · 100+ Pages) in JetBrains

Mono tabular with amber baseline ticks and a count-up on load (IntersectionObserver,

prefers-reduced-motion-aware, hard final-value guarantee). Day mode untouched —

readout is display:none in light, all reskin rules are [data-theme="dark"]-scoped.

Page-local inline <style> + the amber accent in styles-index.css (re-minified) per the

2-stylesheet rule. Reversible: delete the font link + CSS block + readout markup.

• E4 — calculator marketing shells, editorial chrome (css/rz-calc-editorial.css).

All 7 calculators (pue, capex, opex, roi, tco, cx, spares-readiness) now carry

data-rz-register="editorial"; in dark mode the hero only gets the editorial treatment —

Fraunces serif <h1>, mono amber badge (#E8B563), and a thin amber→mint rule under the

title. Chrome only: the calculator engines, inputs, results, and tool UI are not touched;

day mode is unchanged. Verified: dark <h1>=Fraunces / light=Inter on all 7, badge amber,

no new console/page errors (the pre-existing ipapi.co CORS notice on capex/opex is

unrelated to this change).

Notes

• §10/Q4 (index = POLISH) remains in force for the bento grid; §11 only restyles the

identity hero card, dark-only — it does not reskin the colourful bento.

• Verified via headless dark/light probes + hero screenshots before ship.

Changed

• All 34 editorial pages now use the approved §08-mockup amber accent. The article

editorial register (css/rz-article-dark.css + Fraunces) reads the per-page

--rz-art-accent / --rz-art-accent2 CSS vars to colour its chrome (Fraunces h2 accent

rail, drop-cap, mono kicker/meta rail, pull-quote, read-progress bar). Every editorial

page's vars were flipped to the mockup's exact pair — **--rz-art-accent:#E8B563 (amber)

/ --rz-art-accent2:#6FBF9A (mint)** — sourced verbatim from rz-article-mockup.html

(linear-gradient(90deg,#E8B563,#6FBF9A)). Previously each article carried its own

per-series hue (navy / blue / cyan / emerald / red); the chrome is now uniform amber,

matching the planb mockup the owner approved.

• The 9 red-series pages were fully re-toned red → amber in body too (not just chrome):

article-11, article-14, article-20, article-23, article-25, article-26,

article-27, geopolitics, geopolitics-3. The Global-Analysis red palette

(#dc2626/#ef4444/#fca5a5/#fef2f2/#991b1b/#b91c1c/#7f1d1d/#450a0a + matching rgba())

was mapped onto the Tailwind amber ladder (`#b45309/#d97706/#fcd34d/#fffbeb/#92400e/

#78350f/#451a03 + matching rgba()`). This fixes the owner's "warna dll kok masih sama"

report — the chrome had gone amber but the red-themed bodies (hero highlight, callouts,

article-27's .ws-* workforce calculator) were still red.

Preserved

• Semantic reds untouched. Reds that encode meaning in non-red articles (e.g. article-8

--accent-red / .trajectory-arrow.deteriorating / "Deteriorating" cells paired with

#10b981 green for good-vs-bad contrast) were deliberately not swapped — only the

series-accent red theme on the 9 red pages was re-toned.

• article-27 Workforce Strategy calculator still renders and recalculates — colour-only

swap, no structural/JS change (verified: input→recalc fires, 0 page/console errors).

• article-9-paper.html (print variant) excluded.

Verification

• Headless dark-mode probe (hue-accurate): 9/9 red-series pages = 0 residual red, amber chrome present.
• article-27 calculator interaction probe: 0 errors, recalc fires on input change.
• audit-script-tags --strict CLEAN · audit-js-syntax --strict CLEAN.

Changed

• SEO compliance — tightened <title> (≤60 chars) and <meta name="description"> (≤160 chars)

on all 5 CDU pages (cdu-hub, cdu-selection-guide, cdu-checklist, cdu-mini-bms,

cdu-comparison) and added the missing <meta name="ai-content-declaration" content="human-authored">.

The CDU pages now pass the full audit suite — script-tags, js-syntax, version-stamp, mobile,

and SEO — with no warnings.

Fixed

• The editorial h2 chrome was broken on ~30 articles — not just article-27. The editorial CSS

(rz-article-dark.css) neutralises border-left but NOT border-bottom. Every article whose base

.article-body h2 used a border-bottom: 3px solid <accent> underline kept that underline sitting

on top of the editorial accent-rail in dark mode — so the editorial register never fully took.

The original article-skin rollout shipped this latent bug across the whole batch.

• Fixed via a parallel /ultraplan workflow (6 agents): rendered all 33 articles in dark mode,

detected the underline (computed border-bottom-width ≠ 0), and surgically added

[data-theme="dark"] .article-body h2 { border-bottom:none; padding-bottom:0; } per article so the

editorial rail owns the chrome. Light-mode underline preserved; content-specific styling

(calculators .ws-*, PDF-export strings, data tables) left untouched.

• 100% verified by an independent full render probe: 33/33 PASS — every article now computes

Fraunces h2 (no underline) + Fraunces title + IBM Plex Mono meta. Audits CLEAN (script-tags,

js-syntax). 31 files changed (article-19/20 were already correct).

• Tracker + second-brain vault updated.

Fixed

• Scroll-aware navbar, now site-wide. Extended the CDU-page navbar fix (v1.43.49) to the whole

site: the base .navbar is background:transparent and only turns solid via the .scrolled

class added by script.js. ~23 pages use custom inline scripts instead of script.js (calculators,

comparison pages, LTC labs, network hub) and never got it, so content bled through the navbar on

scroll. Added the scroll-aware .scrolled toggle once to the shared js/rz-mobile-nav.js (loaded

site-wide; idempotent + self-guarded so it never double-binds with script.js), fixing all of them

in one place. Also added the missing rz-mobile-nav.js tag to network-visualization-hub.html

(was absent — a hamburger-mandate gap too). Verified the solid-on-scroll navbar on calculators,

comparisons, LTC labs, network hub and CDU pages.

Changed

• Editorial skin consistency, tanpa pengecualian. After confirming all 29 articles + 6 series

landings are consistent, closed the last 2 edge cases:

• 4 hub/landing pages (articles, insights, geopolitics, future-forward) — added the

js/rz-article-editorial.js read-progress/stagger (they already had the editorial CSS + Fraunces;

JS is progressive-enhancement + self-guards). Verified: no errors, hubs render fine.

• article-9-paper.html (print/PDF "Technical Paper" variant) — given a **screen-only dark

editorial mode** (Fraunces title + amber accent + dark surfaces) scoped under @media screen so

the print/PDF output stays pristine white — the print feature is untouched. Verified headless:

screen bg #0E0F12 + Fraunces title; print media bg #fff + dark text; zero console errors.

Notes

• No .article-* classes on the paper, so it uses a self-contained @media screen override (not

rz-article-dark.css). Colours only — no layout/structure changes. Tracker + vault updated.

Fixed

• Navbar collision, real root cause — the site's base .navbar is background:transparent and only

gets its solid blur background from a .scrolled class that script.js adds on scroll. The CDU

pages use inline scripts and don't load script.js, so their navbar stayed transparent and table

content bled through it on scroll (most visible in dark mode). Added the scroll-aware .scrolled

toggle to the inline script of all five CDU pages (cdu-hub, cdu-selection-guide, cdu-checklist,

cdu-mini-bms, cdu-comparison). Navbar is now opaque on scroll, matching the rest of the site.

Changed

• Applied the approved §08 article editorial skin to the 3 articles the rollout skipped.

The editorial register (css/rz-article-dark.css + Fraunces/Plex fonts + read-progress + staggered

entrance) shipped to 26 articles (v1.43.9–17) but article-27, FF-2, FF-3 were missed and still

rendered in generic dark mode. Added the identical editorial head block (replicated from article-26)

to each:

• article-27.html — data-rz-register="editorial" + --rz-art-accent:#dc2626 (Global Analysis red).
• FF-2.html / FF-3.html — --rz-art-accent:#E8B563 (Future Forward amber, matching FF-1).

Verified dark: Fraunces serif hero + accent italic, mono meta, editorial chrome, read-progress bar.

article-27's inline .ws-* workforce calculator + content untouched (head-only change). Body copy

stays IBM Plex Sans. Light mode unchanged. Article editorial skin is now 29/29 complete

(article-9-paper = print variant, intentionally excluded).

• Tracker + second-brain vault + plan §08/E2 updated.

Added

• cdu-hub.html (new) — a single landing page for the liquid-cooling CDU toolkit with four

resource cards (Selection Guide · Checklist · Mini-BMS · Deep Comparison). The four separate CDU

cards on datacenter-solutions.html and tools.html are now consolidated into one "Liquid-

Cooling CDU Toolkit" card that opens this hub. Registered in sitemap/search-index/llms.

Fixed

• Dense comparison tables were cramped & unprofessional — the 9-column guide tables had

min-width:760px, forcing the spec column to wrap word-by-word. Raised to 1320px with explicit

per-column min-widths (spec column 300–340px) so cells breathe; widened the comparison tables to

880px. Header rows get a 2px under-rule + no-wrap.

• Table header collided with the fixed navbar — removed th{position:sticky;top:0} (stuck the

header under the navbar on scroll) and the sticky-first-column z-index rules (raised cells above

the nav) from .cdu-tbl / .cp-tbl / .ck-tbl. Content now scrolls cleanly under the navbar.

Fixed

• Hero name "blink" (disappears/reappears even without hover). .bento-name carried TWO infinite

animations — nameGlow (4s drop-shadow pulse) + rzNameSweep (12s gradient sweep, dark). Removed

both; the gradient text is now static (background-position:0% 50%). Identified via headless

animation probe (cards themselves were stable; the looping name animation was the culprit).

• Operational Excellence card "wobble" on hover. .oe-card:hover .oe-desc expanded the description

max-height: 3em → 10em (#30 hover-expand) — the card grew on hover = the bergoyang. Removed the

expand; desc stays 2-line clamped, card no longer changes size on hover. Verified (max-height

38.4px → 38.4px, no growth).

• Both removals mirrored in styles.css + styles-index.css, re-minified; cache-bust

styles-index.min.css?v=2026-06-21-calm.

Added

• cdu-mini-bms.html — each numeric parameter tile (flow, flow-total, supply/return temp, ΔT,

dP, system pressure, HX approach) now carries a live sparkline showing the rolling 30-sample

trend with a subtle fill area — making it read as a real BMS dashboard rather than a static

readout. Sparkline colour follows the tile's OK/WARN/ALARM state; history resets on CDU-type

change; status tiles (pumps/filter/leak/coolant) stay static. Pure presentation — no change to

the simulation logic or values.

Changed

• cdu-mini-bms.html — reorganised the cockpit so the detailed ISA P&ID gets a full-width row

(capped 920 px, centred) instead of sharing a half-width column; the datahall layout + live

parameter tiles now share the top row. The dense instrumentation (valve/pump/instrument symbols,

tag bubbles, manifold, legend) is now legible at full scale. No data/logic change.

Changed (applied to the new high-density tables)

• Sticky first column — the Model/Vendor/Symptom column now pins to the left on the wide 9-column

guide tables, the comparison tables and the 16-row checklist tables, so context stays visible while

scrolling the spec columns horizontally (cdu-selection-guide.html .cdu-tbl, cdu-comparison.html

.cp-tbl, cdu-checklist.html .ck-tbl).

• Scroll-shadow affordance — .tbl-wrap gains the classic CSS scroll-shadow (edge fades + inner

shadows that appear only when there's more to scroll), signalling the horizontal overflow on dense

tables instead of silently clipping.

• Capability matrix — the comparison matrix's ✓/—/~ glyph columns are now centred + mono for

faster scanning (.cp-matrix).

• Lighter inner-row hairlines on the guide table to match the editorial-register tables.

Changed

• cdu-checklist.html — §01 operational parameters +5 (supply-vs-dew-point, pipe velocity,

corrosion-inhibitor reserve, pump NPSH margin, footprint+weight) → 16 rows; §06 PM +5 tasks

(make-up/level, dew-point-reset verify, control-valve stroke, BMS telemetry verify, inhibitor+Cu/Fe)

→ 16 rows; §07 symptom→action +6 rows (condensation, GPU throttle/maldistribution, 2-phase dry-out,

frequent make-up, control-valve fault, over-pressure) → 16 rows.

• cdu-comparison.html — added a per-vendor published-spec & capability matrix (10 vendors ×

Redfish / dP published / ASHRAE class / hot-swap service / filtration / notable), distilling the

datasheet research into a single scannable view; Redfish is published only by CoolIT, Accelsius,

Lenovo. 6 tables total.

Changed

• cdu-selection-guide.html — rebuilt both comparison tables from a single "notable specs" cell

into a 9-column deep comparison (Model · Capacity · Type · Secondary flow · dP/head · Approach ·

Dimensions & weight · Fluid·filter·conn·BMS·class · Links), populated from per-model datasheet

research. In-row table now 13 rows incl. Vertiv Liebert XDU1350 and CoolChip CDU 70 (L2A)

with verified datasheets, CoolIT CHx1500, and a corrected nVent RackChiller CDU800 (the

earlier "CX121" was wrong — nVent has no CX series). In-rack table now 7 rows incl. **nVent

RackChiller CHx (real 4U specs: 150 L/min, 2.76 bar, 4 °C, 41 kg, W4) and Lenovo Neptune RM100**

(corrected to L2L, with verified O&M manual). Every new external link curl-verified (200) or

tagged VENDOR PORTAL; "n/p" marks genuinely unpublished specs.

Changed

• cdu-mini-bms.html — rebuilt the P&ID from a simplified schematic into a full **ISA-instrumented

diagram**: facility isolation valves (HV-01/02) + 3-way control valve TCV-01, E-01 plate HX with

approach (TI-03), CDU skid with expansion tank T-01 + level LT-01 + makeup MV-01 + relief PSV-01,

dual pumps P-01A (duty) / P-01B (standby) with check valves, F-01 50 µm strainer, tagged

instruments FT-01 / PT-01 / PDT-01 (filter) / PDT-02 (loop) / TT-01 / TT-02 with live values,

rack manifold M-01 with QD couplers + cold plates, leak rope LSH-01, a control-setpoint note

(TCV holds TT-01=18 °C ≥ dew-point+3; P-01 VFD on ΔP; N+1 changeover), and a symbol legend.

Datahall layout gained a 600 mm rack-pitch dimension line and a per-type footprint/density caption

(CDU dims, kW/rack, aisle containment). All values feed from the live sim; renders clean across all

5 types × light/dark.

Fixed

• Cookie banner rendered white-on-dark on datahallAI + dc-conventional (they use the

.cookie-banner + .cookie-accept/.cookie-decline scheme with no dark override). Added shared

dark overrides to css/rz-bms-shell.css (loaded by all 5 cockpits) — banner → rgba(15,23,42,.95),

light text, brighter accept button; !important beats per-page inline. datahall/water/fire already

had a dark banner (this only brightens their accept button). Presentation-only — **accuracy probe

re-run 75/75 PASS** (engine/alarm/SLD untouched). Cache-bust bumped on all 5 cockpit pages.

Notes

• Completes the /ultraplan dark-mode defect rollout (batches 1–5). All confirmed real defects fixed;

verified false positives left as-is. Tracker + vault note finalized.

Fixed

• articles.html .philosophy-section — light gradient section block rendered on dark; added

[data-theme="dark"] dark-gradient override.

• tia-942-checklist — .tia-btn-reset + .nav-user-dropdown were white-on-dark (chrome);

added dark overrides. (Left .tier-btn.active/.dctype-btn.active white — intentional selected-state

highlight, dark-text-on-white reads fine.)

• dc-market-tracker — .dmt-cookie-decline white button on dark cookie bar → dark override.
• Part of the /ultraplan rollout. Tracker + vault updated. Remaining: cockpit cookie banners

(probe-gated — they load css/rz-bms-shell.css).

Changed (uiux-reviewer findings applied across all 4 CDU pages)

• Brand typography — cdu-selection-guide.html swapped Inter → IBM Plex Sans (design.md

mandates IBM Plex, "not Inter — generic SaaS") and set an explicit body font-family.

• Tabular numerics — added font-variant-numeric:tabular-nums slashed-zero to every mono

numeric class (guide .kw/.formula, checklist .p/form inputs, mini-bms .tile .v/.sch-val,

comparison .mono) so dense data columns align and live instrument readouts stop jittering.

• Accessibility — gated the mini-BMS alarm/leak blink animations behind

prefers-reduced-motion:reduce (they already carry text + colour, so motion is now enhancement

only).

• Token/radius discipline — container radii 12→8 px (typecard/panel/tbl-wrap) and tile 9→6 px;

taxonomy card border-top 3→2 px; taxonomy SVG pipes re-graded to tier weights (supply 1.6 /

return 1.2 / air 1.4 px) with return re-pointed to fault-red #FF3030; comparison .win/.mid/.lose

re-pointed to RZ severity tokens (#00FF88/#FFAA00/#FF3030 in dark); tier-3 lighter inner-row

hairlines on dense tables; mini-BMS hero line-grid opacity lowered to ≤0.06; FAQ summary hover +

focus-visible ring; lead paragraphs capped at 70ch; removed a no-op sticky th.

Fixed

• Inline #64748b disclaimer + nav links low-contrast on dark — across 52 pages. The shared

independence-disclaimer <p style="...color:#64748b..."> (and the "All Insights" nav link) is an

inline style CSS can't normally override. Added one targeted [data-theme="dark"] [style*="color:#64748b"]

rule (!important) in styles.css → lifts to #94a3b8 (readable, still de-emphasized) on every

page at once (pillars, articles, etc.). Index unaffected (uses #94a3b8 already). Verified via probe.

• Part of the /ultraplan rollout. Tracker + vault updated.

Fixed

• Status badges rendered light-on-dark on geopolitics-2 · geopolitics-3 · FF-1 · FF-2.

Pages already had [data-theme="dark"] overrides for .confidence-high/-medium but missed

.confidence-low + all .prob-low/-medium/-high (light pastel bg, no dark override). Added the

missing translucent-tint dark overrides per page (prob-high colour matched each page: red for

geo-2/3, violet for FF-1, amber for FF-2). Verified dark via headless probe.

• Part of the /ultraplan rollout. Tracker + vault note updated.

Added

• cdu-selection-guide.html — new §08 FAQ section (8 questions: L2L vs L2A, flow sizing,

coolant/water quality, filtration, dew-point, quick-disconnect standards, type selection, leak

safety) using native <details> accordions, plus a FAQPage JSON-LD block (8 Q&As) for rich

results / AI-search.

• glossary.html — three new liquid-cooling terms cross-linked to the CDU suite: L2A

(liquid-to-air CDU), L2L (liquid-to-liquid CDU), and Quick-Disconnect (QD/UQD/UQDB).

Fixed

• .references-section rendered light-on-dark on 4 report/infographic pages — the section

carries an inline style="background:#f8fafc" that overrode the shared dark rule, so in dark mode

the panel stayed light gray while its text was light → text invisible. Made the shared

[data-theme="dark"] .references-section rule !important (styles.css) so it beats the inline bg.

Fixes infographic-dc-cost-breakdown · infographic-dc-sustainability · infographic-pue-global

· asean-dc-report-2026 at once. Verified dark (rgb(15,23,42)) via headless probe.

• Found by the /ultraplan 9-agent parallel dark-mode audit (56 defects/27 files). Audit also

caught false positives (saturated-accent compare-table headers read fine on dark — left as-is).

Tracker: standarization/DARK_MODE_ROLLOUT_TRACKER.md + vault 05-Standards/Dark-Mode-Rollout.md.

Changed

• SEO/JSON-LD — added TechArticle/WebApplication + BreadcrumbList structured data and

Twitter Card meta to cdu-checklist.html, cdu-mini-bms.html, cdu-comparison.html, and a

TechArticle block to cdu-selection-guide.html (all 4 pages now carry 2 valid JSON-LD blocks).

• Link re-validation — re-curled every external CDU link (21); honestly downgraded the nVent

Data Solutions resource-library link from VERIFIED to VENDOR PORTAL (now intermittently 403s to

automated fetch / bot-blocked).

• Cross-link completeness — every CDU page now links to the other three (added Mini-BMS +

Deep Comparison to the checklist footer nav; added Deep Comparison to the Mini-BMS links strip).

• AI search — regenerated llms-full.txt so all five CDU pages are included.
• Full render QA: 4 pages × light/dark = 8 renders, 0 console errors.

Added

• cdu-selection-guide.html — new §05 Sizing & installation requirements section: the core

flow/ΔT sizing equation (flow ≈ 14.7 × Q(kW) / ΔT for PG25; ≈1.5 L/min/kW at ΔT 10, OCP band

1.25–2.0) with worked rack/node examples; a pipe-sizing table (DN25→DN200 at ~2 m/s); heat-

exchanger approach + W-class supply-temperature + dew-point guidance; and a physical-install

checklist (floor loading, clearances, facility water vs L2A air budget, dual feeds, BMS/leak,

N+1). Manuals/Standards sections renumbered to 06/07.

Added

• cdu-comparison.html (new public page, editorial register) — a deep, source-tagged

multi-aspect comparison of liquid-cooling CDUs, the companion to the Selection Guide:

• §01 Worldwide common field issues — 12 failure modes (leaks, biofilm, galvanic/chloride

corrosion, glycol degradation, particulate/microchannel clog, flow maldistribution, cavitation,

dew-point condensation, pump/N+1, controls gaps, commissioning defects, standardization) with

symptom → root cause → prevention.

• §02 Control systems & BMS/DCIM — pump VFD (flow vs ΔP), temperature PID valve, dew-point

reset, N+1 changeover, leak→action; per-vendor Modbus/BACnet/SNMP/Redfish matrix; DMTF Redfish

CoolingUnit (DSP2064) + OCP telemetry standards.

• §03 After-sales & support — per-vendor warranty/SLA/serviceability/remote-monitoring matrix

(Vertiv, CoolIT, Motivair/Schneider, Boyd/Eaton, nVent, Delta, Stulz, Accelsius, ZutaCore,

Lenovo) with honest "not publicly disclosed" markers.

• §04 TCO & maintenance by type — relative trade-off matrix (capex/opex/density/water/retrofit/

maintenance/redundancy) across in-rack/in-row/sidecar/L2L/L2A/2-phase.

• Every figure tagged STANDARD (ASHRAE/OCP/DMTF) vs VENDOR vs REPORTED; vendor-sponsored TCO

claims attributed by name. Registered in dc-solutions/tools/sitemap/search-index/llms; cross-

linked from the Selection Guide.

Changed

• cdu-selection-guide.html — the "CDU types — what each is for" taxonomy section now carries a

visual mini-diagram per type (not just text cards): an inline SVG showing the CDU's datahall

placement signature + animated supply/return flow + heat-rejection path. In-rack (CDU slice inside

a rack), In-row (cabinet between racks), Sidecar (slim unit on the rack side), L2A (→ room air, no

facility water), L2L (plate HX to facility CDW), Facility/room-scale (multiple rows → central unit

• HX). Each in its type accent colour; flow animation honours prefers-reduced-motion. Full

interactive version lives on the CDU Mini-BMS cockpit.

Added

• cdu-mini-bms.html (new public page, instrument register) — an interactive miniature

building-management cockpit for liquid-cooling CDUs. Per CDU type (in-rack · in-row · sidecar ·

L2L end-of-row · L2A air-cooled): a datahall installation layout plan view (cold/hot aisle,

rack row, CDU placement, facility-water vs room-air rejection, animated flow), an **animated

P&ID schematic** (HX, N+1 pumps, filter, expansion tank, FT/PT/dPT sensors, manifold → rack

cold-plates, leak rope), and a live simulated parameter panel (flow LPM/kW + total, supply/

return temp, ΔT, dP, system pressure, pump N+1, filter, leak, HX approach, coolant) with

warn/alarm tile states and a RUN/WARN/ALARM P&ID pill. Fault-injection scenarios (leak,

pump-A fail, filter clog, hot facility water, low flow) drive the controls so users can see how

a CDU controller responds. Animation honours prefers-reduced-motion; all values are on-device

simulation from the sourced bands in the CDU Checklist (not a live feed).

Cross-linkage

• Registered in datacenter-solutions.html, tools.html, sitemap.xml, search-index.json

(cdu-mini-bms-1), llms.txt; cross-linked from the CDU Selection Guide next-strip and back to

the Guide + Checklist from the cockpit.

Changed

• cdu-checklist.html — added two deep, source-tagged sections (every figure tagged

STANDARD vs TYPICAL/VENDOR so nothing is overstated):

• §02 Water quality — acceptance criteria & standard limits: ASHRAE TC 9.9 facility-water

W-class table (W17–W+, max supply temps); OCP cold-plate TCS fluid acceptance table (pH,

conductivity, TDS, TSS, hardness, chloride, turbidity, bacteria, inhibitor, Cu/Fe, filtration);

ASHRAE/Dell TCS-vs-FWS side-by-side limits; OCP make-up water sub-spec + ASTM D1193 Type II.

Honest caveats on pH/conductivity loop regimes and the absence of a published TOC limit.

• §03 Installation & commissioning — numeric procedure: hydrostatic test (1.5× design,

≥10-min hold per ASME B31.3/B31.1, "no-leakage" acceptance), turbulent flush (Re>4000, ISO 4406

/ NAS 1638 target), filtration sizing table, coolant & flow (PG25, 1.5 LPM/kW @10 °C, band

1.25–2.0), OCP UQD/UQDB quick-disconnect spec (100 psi working / 300 psi burst / ≥5000 cycles /

±1 mm blind-mate), leak detection sequence.

• Sections renumbered 02→04…06→08; TOC + source-tag legend added.

Fixed

• AWS company logo invisible in dark mode. assets/aws-dark.svg was byte-identical to the day

version and used fill="currentColor" on the wordmark — via <img>, currentColor resolves to the

SVG's default black, so the "dark" logo rendered black-on-dark. Made the wordmark #FFFFFF

(orange #F90 smile kept) + cache-bust the <img src> so the white version loads. The day/dark

toggle CSS ([data-theme="dark"] .exp-logo-aws-dark{display:block}) was already correct.

Added

• css/rz-dark.css — POLISH-TRACK consistency tokens (character-agnostic): radius scale

(--rz-r-xs…xl/pill), elevation scale (--rz-elev-1..3, light + dark), .rz-surface premium

card utility, .rz-reveal-on (no hover movement). For the "polish + unify tokens" path — pages keep

their own type + accent, adopt one structural language. Distinct from the Fraunces editorial register.

Changed

• cdu-selection-guide.html — enriched the in-rack comparison table with two link-verified

two-phase (dielectric direct-to-chip) units: Accelsius NeuCool IR150 (150 kW rack-

integrated, R-1233zd(E), 1+1 pumps, ASHRAE W27/W45, accelsius.com/ir150/ VERIFIED) and

ZutaCore HyperCool In-Rack (20–120 kW waterless 3U/6U, zutacore.com/solutions VERIFIED).

Added a 2-PHASE tag (violet) and a section-lead note distinguishing single-phase water CDUs

from two-phase/waterless options. Both URLs independently curl-verified (HTTP 200 + content

match); gated lead-capture spec sheets tagged VENDOR PORTAL. Chilldyne (negative-pressure

single-phase) and LiquidStack (2-phase immersion only) excluded as not 2-phase in-rack DTC.

Fixed

• Stale cached JS re-running the cursor-tracking 3D tilt. index.html loaded

script.min.js?v=2026-05-09-v1; the on-disk min has initCardTilt() disabled, but returning

visitors were served the old cached build (active `perspective rotateX/Y + translateY(-10px) +

scale + parallax children`) under that unchanged query string — cards wobbled, shifted out from under

the cursor, and flickered. Bumped to ?v=2026-06-20-cards to force the clean, tilt-free build.

• Company-logo hover pop too large — .bento-exp-card:hover .bento-company-logo scaled 1.15

(visible "wobble"); reduced to a subtle 1.04. The grayscale→colour fade on hover is kept.

• Both styles.css + styles-index.css re-minified; CSS cache-bust → ?v=2026-06-20-cards.
• Card hover feedback is now border/glow + logo colour-fade only — no card movement (matches the

owner's no-card-movement rule). Verified via headless hover probe (card transform = none).

Added

• cdu-selection-guide.html (new public page) — liquid-cooling Coolant Distribution Unit

resource for the DC Solutions hub. CDU type taxonomy (in-rack · in-row · sidecar · L2A · L2L ·

facility-scale), an in-row + facility comparison table and an in-rack comparison table with

link-verified vendor specs and product/datasheet/manual links (Vertiv CoolChip, CoolIT CHx,

Motivair, Boyd, Delta, nVent, Stulz, Envicool, ZutaCore), a manuals/documentation hub of

HTTP-200-verified vendor portals, an L2L-vs-L2A decision section, and ASHRAE TC 9.9 / OCP

standards references. Every shipped link was fetched and tagged VERIFIED (200 + content match)

or VENDOR PORTAL (working docs page, deep PDF not auto-verifiable) — no unverified deep links.

• cdu-checklist.html (new public page, linked from the guide) — operational parameter bands

(supply temp / ΔT / flow / dP / system pressure / coolant / ASHRAE class / filtration / pump

redundancy / leak / ATD), an installation checklist, an inspection checklist, a preventive-

maintenance checklist with cadence tags, a symptom→cause→action troubleshooting table, and a

printable on-screen service-record form (window.print() + print stylesheet).

Cross-linkage (CONTENT_LINKAGE_PLAYBOOK)

• Registered both pages: datacenter-solutions.html (two DS tool rows), tools.html (two tool

cards), sitemap.xml, search-index.json (cdu-guide-1 + cdu-checklist-1), llms.txt, and a

glossary cross-link from the existing CDU term.

Fixed

• Hover-lift removed (it never worked). The translateY(-4px) bento hover added with the

polish was dead on every card — a pre-existing global .bento-card:hover{transform:none!important}

(the owner's deliberate no-card-movement rule) overrode it. Removed the dead rule rather than force

it with !important (which would reintroduce motion the owner removed). Hover feedback stays via the

existing border/glow rules — no card movement. Found by the uiux-reviewer agent (headless-verified).

• Stale cache-bust — index.html styles-index.min.css?v=2026-06-14-light → ?v=2026-06-20-bento

so the fixes reach returning visitors.

• Rejected-purple disclaimer — the independence-disclaimer box used Anthropic-default #8b5cf6

(CLAUDE.md Rejected-Patterns #3) with no dark path; retoned to neutral slate + emerald link.

• No-op .bento-sap dark tint — dark override re-declared the identical 2% emerald (invisible);

bumped to 6% so the SAP card keeps its emerald identity in dark.

• Deduped the redundant dark-scoped bentoRise reveal block (theme-agnostic block already covers dark).
• Both styles.css + styles-index.css re-minified. Standard DARK_MODE_STANDARD.md updated.

Closes the P1 metadata items (review doc-27 §5.5 + §5.7). ADDITIVE ONLY —

displayed data unchanged (owner mandate: skin/metadata only, never alter

content). Verified: network link table values + rack/CRAH grid byte-identical;

accuracy probe 75/75.

Changed

• ict.html — network link rows (tr.linkrow) now carry data-rz-line

metadata derived from the EXISTING row data (from/to/medium/state/capacity/

current/redundancy). The displayed table (cap/traffic/util/latency/loss/

jitter/state/redundancy columns) is unchanged. Loads rz-line-model +

rz-inspector + rz-alarm-state + rz-telemetry-quality; `data-rz-data-mode=

"simulated"`. 5/5 rows tagged, 0 unknown mediums, NONE errors.

• datahall.html — data-quality service: data-rz-data-mode="simulated"
• rz-telemetry-quality.js (SIM banner + chip). The rack/CRAH grid (200 racks
• 20 CRAH cells) and all engine-bound values are unchanged. Accuracy probe

75/75 confirms byte-identical.

Scope discipline

• Metadata + library + banner only. ZERO content/data edits. Per

feedback_skin_only_preserve_data (owner 2026-06-14).

Completes the cockpit instrument re-skin. Same chrome-only, additive, scoped,

dark-only treatment as the datahallAI pilot, applied to the remaining 8 cockpit

pages via the 2-line opt-in. Per standarization/COCKPIT_RESKIN_PLAN.md.

Hard invariants verified (all green):

• SVG semantic colours untouched · engine byte-identical

(probe-accuracy-validation 75/75, test-datahall-calc 57/57,

test-conv-calc 22/22, probe-line-model 52/52) · light mode inert.

• All 8 pages headless-verified: instrument register active + graticule

atmosphere on in dark, NONE errors. EPMS ATS→rack green fix intact.

Changed (<html data-rz-register="instrument"> + 1 CSS link each)

• dc-conventional.html (2nd light-mode page; light inert)
• chiller-plant.html · water-system.html · fuel-system.html
• fire-system.html · ict.html · EPMS_Telemetry.html · datahall.html

Rollout status — cockpit instrument register COMPLETE

• 9 / 9 cockpit pages on the instrument register (datahallAI + 8 here).
• Chrome-only: atmosphere + mono headings + cyan-hairline panels. Zero SVG/

engine/inline-CSS edits. Trivially reversible (2 lines per page).

First cockpit page on the instrument register. CHROME-ONLY, additive, scoped,

dark-only. Per standarization/COCKPIT_RESKIN_PLAN.md.

Hard invariants verified (the "no mistake" gates):

• SVG semantic colours (feed A/B, alarm red, cooling cyan) — UNTOUCHED (CSS

contains zero SVG stroke/fill selectors).

• Engine values byte-identical — probe-accuracy-validation 75/75, KPIs

(PUE 1.30 / IT 14.26 / GPU 7,776) unchanged; test-datahall-calc 57/57.

• Light mode inert — skin gated :not([data-theme="light"]); datahallAI

defaults to light, so the skin only activates on dark toggle.

Added

• css/rz-cockpit-instrument.css — instrument register for cockpits.

Activates on <html data-rz-register="instrument"> + dark. Styles ONLY:

page atmosphere (graticule + faint scanlines via body::before/::after),

display headings → JetBrains Mono (.hdr h1, .bx h3, .cd h4, .sb h4),

panel borders → cyan hairline + 3px radius. No SVG, no engine text.

• standarization/COCKPIT_RESKIN_PLAN.md — full plan: invariants, page

inventory (9 pages), ship sequence, per-ship verification checklist, rollback.

Changed

• datahallAI.html — <html data-rz-register="instrument"> + one CSS

link. Two lines; trivially reversible. SVG/engine/inline-CSS untouched.

Next (per plan)

• v1.43.2x — dc-conventional, then chiller/water/fuel, then fire/ict/EPMS/datahall.

Changed

• Homepage (index.html) light mode — polished to twin the dark polish (v1.43.14).

Same decision (keep the colourful bento character; raise quality only), now applied

light-scoped so day↔night match. Pure CSS:

• Layered light card surfaces — .bento-card gets a soft white gradient body +

layered shadow (was flat white + faint shadow).

• Per-card accent restored — .bento-exp-card corner glow (--bexp-accent) now

always-on in light too.

• Hover lift — cards rise 4px with a deeper shadow.
• Staggered reveal — the bentoRise reveal is now theme-agnostic (runs in light
• dark); honours prefers-reduced-motion.
• Rules in both styles.css + styles-index.css (2-stylesheet rule), re-minified;

cache-bust styles-index.min.css?v=2026-06-14-light. Plan/mock: rz-index-mockup-day.html.

Finishes the article-family editorial rollout. The three bespoke series

landings get the serif hero treatment (additive, hero-only — bespoke card

grids untouched); FF-1 (the real Future-Forward article, standard .article-*

template) gets the full opt-in. future-forward-1.html is a redirect stub to

FF-1 and is left as-is.

No cockpit/engine touched. Light mode unchanged (scoped). Gates CLEAN. All four

headless-verified (dark = Fraunces serif hero, light = Inter, NONE errors).

Added

• css/rz-article-dark.css — landing-hero block: .insights-hero,

.geopolitics-hero, .futureforward-hero H1/dek → serif; shared

.section-title → serif. Additive + scoped.

Changed (editorial-register opt-in)

• insights.html — cyan #0891b2 (hero-only treatment)
• geopolitics.html — red #dc2626 (hero-only)
• future-forward.html — amber #E8B563 (remapped from violet #a855f7,

a CLAUDE.md rejected token; hero-only)

• FF-1.html — amber #E8B563 (remapped from violet #6d28d9); full

.article-* editorial treatment + read-progress runtime.

Rollout status — article family COMPLETE

• 36 pages on the editorial register: 28 standard articles + geopolitics-1/2/3
• FF-1 + articles index + insights/geopolitics/future-forward landings.
• Every public article-family page is now on the editorial register (dark-only,

light untouched). The remaining dark-mode work is the instrument-register

cockpit re-skin (separate track, owner approval pending).

Completes the article-family editorial rollout with the journal index. Extends

the scoped css/rz-article-dark.css with an additive index-grid block

(.articles-hero + .article-card-*) and opts in articles.html.

No cockpit/engine touched. Light mode unchanged (skin scoped to

[data-rz-register=editorial][data-theme=dark]). Gates CLEAN. Verified dark =

Fraunces hero + serif card titles across 27 cards; light = Inter (scope guard);

NONE errors. Existing aurora hero + card layout preserved.

Added

• css/rz-article-dark.css — index-grid block: .articles-hero h1/p,

.article-card-title (serif), .article-card-excerpt, .article-card-number

(mono accent), .article-card-category (mono). Additive + scoped.

Changed

• articles.html — `

style="--rz-art-accent:#0891b2">` + Fraunces/Plex fonts + scoped editorial CSS.

Structure + light mode untouched.

Rollout status — article family

• 32 pages on the editorial register: 28 standard articles + geopolitics-1/2/3
• articles.html index.
• Out of scope (bespoke landing templates, parallel landing/index track):

insights.html, geopolitics.html, future-forward.html,

future-forward-1.html — handed off to the landing/index design pass.

Extends the editorial-register rollout to the Geopolitics series children

(same standard .article-* template as articles 1–26). Landing pages

(geopolitics.html, future-forward.html) use bespoke templates and are

out of scope here.

No cockpit/engine touched. Light mode unchanged (dark-scoped). Gates CLEAN.

All three headless-verified (Fraunces title, read-progress, accent, NONE errors).

Changed (editorial-register opt-in)

• geopolitics-1.html — 72-Hour Warning — green #059669
• geopolitics-2.html — $50T Shift — green #059669
• geopolitics-3.html — Hormuz Fiber Shock — deep red #991b1b

Progress

• Standard-template articles + geopolitics children on the editorial register:

31 pages (28 articles + geopolitics-1/2/3).

Changed

• Homepage (index.html) dark mode — polished, not reskinned. Owner kept the

current colourful bento character (emerald + pastel accents + rounded cards +

photo + Inter); only the dark-mode quality was raised. Pure CSS, dark-only

([data-theme="dark"] scoped) — light/day mode unchanged:

• Layered card surfaces — .bento-card gains a gradient body + inner top

highlight + soft layered shadow (was flat #131e2e), so cards read with depth.

• Per-card accent restored — each .bento-exp-card keeps its --bexp-accent

(blue/emerald/amber/orange/violet) as an always-on corner glow (was an

invisible 6–12 % hover-only wash).

• Hover lift — cards rise 4px with a deeper shadow on hover.
• Staggered reveal on load — bentoRise keyframe walks the bento rows in on

first paint; honours prefers-reduced-motion. No JS, no markup change.

• Rules mirrored in both styles.css + styles-index.css (2-stylesheet rule),

re-minified; cache-bust styles-index.min.css?v=2026-06-14-polish.

• Plan + before/after mock: plan-dark-mode-standard.html §10 · rz-index-polish.html.

See standarization/DARK_MODE_STANDARD.md (Track E3 index).

Owner-requested fix on EPMS_Telemetry.html (owner explicitly directed this

edit; the earlier "jangan merusak" applied to the dark-skin rollout, not to

operator-requested corrections).

Single surgical change: the dashed flow line AFTER each rack ATS (ATS → Rack —

the final leg to the load) now renders GREEN instead of inheriting the upstream

source colour (red, Feed A). Downstream of the ATS is on the protected/green

bus regardless of which source the ATS selected.

Changed

• EPMS_Telemetry.html energize() — one conditional: wires whose id

starts with w_ats_rack_ use colour class energized-B (green); all other

legs unchanged (upstream PDU→ATS stays Feed-A red). Verified: 10/10 ATS→rack

legs green, upstream legs unchanged, NONE errors.

No other EPMS wiring, engine, or page touched.

Final wave of the standard-template article rollout. 17 articles opted into the

editorial register in one sweep (pattern fully proven), each accent read from

its own theme-color meta. **All 28 standard articles now on the editorial

register.**

No cockpit/engine touched. Light mode unchanged (dark-scoped). Gates CLEAN.

16/17 headless-verified clean; art-2 shows a PRE-EXISTING Chart.js

'helpers' timing error (present on origin before this change — not caused by

the skin; the editorial chrome renders correctly).

Changed (editorial-register opt-in, articles 2–18)

• Each: <html data-rz-register="editorial" style="--rz-art-accent:…"> + fonts
• rz-article-dark.css + rz-article-editorial.js. Accent per theme-color.
• article-8.html — its #8b5cf6 (Anthropic-purple, a CLAUDE.md rejected

token) was REMAPPED to editorial amber #E8B563 to avoid reintroducing slop.

Progress

• 28 / 28 standard articles on the editorial register (1, 2–18, 19–26).

Future-Forward + Geopolitics series pages use different templates — separate

pass if desired.

Third wave of the article editorial-register rollout. Same one-line-per-page

opt-in via the shared css/rz-article-dark.css; each article keeps its series

accent.

No cockpit/engine touched. Light mode unchanged (dark-scoped). Gates CLEAN.

All four headless-verified (Fraunces title, read-progress, correct accent,

NONE errors).

Changed (editorial-register opt-in)

• article-23.html — xAI Colossus — red #dc2626
• article-24.html — manpower shortage — green #059669
• article-25.html — PJM grid crisis — deep red #b91c1c
• article-1.html — When Nothing Happens — navy #1e3a5f

Progress

• Articles on editorial register: 9 / 28 (26 pilot + 19/20/21/22 + 23/24/25/1).

Remaining 19 to follow in batches.

Second wave of the article editorial-register rollout. Four articles opted in

via the shared css/rz-article-dark.css (one-line-per-page pattern), each

keeping its series accent.

No cockpit/engine touched. Light mode unchanged (skin dark-scoped). Gates CLEAN.

All four headless-verified (Fraunces serif title, read-progress bar, correct

accent, NONE errors).

Changed (editorial-register opt-in)

• article-19.html — Singapore vs Batam — cyan #0891b2
• article-20.html — AI water use — red #dc2626
• article-21.html — Nuclear SMRs — green #059669
• article-22.html — NVIDIA photonics — cyan #0891b2

Each: <html data-rz-register="editorial" style="--rz-art-accent:…"> + Fraunces/

Plex fonts + rz-article-dark.css + rz-article-editorial.js. Structure +

light mode untouched.

Progress

• Articles on editorial register: 5 / 28 (26 pilot + 19/20/21/22). Remaining 23

to follow in batches of ~4, each with its series accent.

Owner directed the skin rollout at ARTICLES (content), not the DC AI / DC

Conventional cockpits. First real-page adoption of the editorial register.

No cockpit/engine touched. Light mode verified UNCHANGED (skin scoped to

data-rz-register="editorial" + dark). Gates CLEAN. Pilot headless-verified.

Added

• css/rz-article-dark.css — editorial register for articles. Activates

ONLY when <html data-rz-register="editorial"> AND dark mode. Restyles the

existing .article-* chrome: Fraunces serif title (italic accent), mono

kickers/meta, drop-cap, accent rail on h2, editorial callout/quote rail,

read-progress bar, staggered entrance. Body copy stays IBM Plex Sans for

readability. Per-series accent via --rz-art-accent override.

• js/rz-article-editorial.js — progressive-enhancement runtime

(read-progress bar + entrance stagger). No-op unless editorial register

declared. Honours prefers-reduced-motion.

• rz-article-mockup.html (noindex) — article before/after demo.

Changed

• article-26.html — PILOT. `

style="--rz-art-accent:#ef4444">` + Fraunces/Plex fonts + the editorial

CSS/JS. Structure + light mode untouched.

• plan-dark-mode-standard.html — new §08 (article editorial skin +

before/after iframe).

Rollout plan

• Shared CSS + one-line-per-page opt-in. Batch remaining 27 articles 3–4 per

ship, each with its series accent, after this pilot is approved.

DOCS ship: folds the dark-mode design system into the brand bible as a

permanent numbered chapter (was only a standalone standard + live demos).

No page logic touched. audit-script-tags + audit-js-syntax + audit-version-stamp CLEAN.

Changed

• documentation/design.md — new Section 16 "Dark Mode — RZ Dark System v1":

why-it-exists, two-register principle (instrument + editorial, hybrid locked),

tokens, 6 motion primitives (with the probe-safe count-up + hard-settle rule),

anti-slop checklist, responsive rules, adoption order. Plus a Decision Log row

recording the 2026-06-01 dark-system decision.

Links

• Implementation: css/rz-dark.css + standarization/DARK_MODE_STANDARD.md.
• Live: rz-style-lab.html (picker), rz-skin-gallery.html (12-surface before/after),

rz-cockpit-mockup.html, plan-dark-mode-standard.html (PLAN 02).

PATCH ship: comprehensive visual mockups for the dark-mode standard, per owner

request "kurang banyak before after buat semua mock up" (make before/after

mockups for ALL surfaces) + "enhance front end design skin".

Internal/noindex only — no cockpit logic touched. Engine + #p-dash unchanged.

audit-script-tags + audit-js-syntax + audit-version-stamp CLEAN. Both new

pages headless-verified NONE errors.

Added

• rz-skin-gallery.html (noindex) — before/after for 12 surfaces: KPI strip,

telemetry line (trace-in), per-hall bars (grow+count), energy donut (sweep),

SLD breakers (symbol + semantic colour), equipment inspector, ISA-18.2 alarm

states, data-quality + sim banner, plan card, landing hero, buttons, data

table. Instrument register for cockpit surfaces, editorial for content.

Charts animate on load with hard-settle. Guarded getPointAtLength against

non-finite path length (fixed mid-build).

• rz-cockpit-mockup.html (noindex) — datahallAI cockpit slice before/after

(KPI + MV-intake SLD + cooling trace) proving the instrument register lands

on a cockpit WITHOUT breaking load-bearing semantic colours (feed A blue /

feed B green / trip red / cooling cyan). Tripped breaker stays red both sides.

Changed

• plan-dark-mode-standard.html — new §02b (cockpit before/after iframe)
• new §07 (full skin gallery iframe). Q1 register-split marked LOCKED (hybrid).

Version stamp bumped.

Process

• Reinforces the visual mandate (memory feedback_planb_always_visual.md):

every change shows before/after VISUAL + code + explanation.

PATCH ship: registers the dark-mode design-system plan in the planb hub as a

live card and adds its dedicated visual plan page. Owner locked the HYBRID

register split (instrument for cockpits, editorial for landing/articles).

Internal/noindex only — no cockpit logic touched. Engine + #p-dash

byte-identical (no change). audit-script-tags + audit-js-syntax CLEAN.

Added

• plan-dark-mode-standard.html (noindex) — dedicated PLAN 02 page in the

editorial register (dogfoods RZ Dark System). Fully visual per the planb

mandate: embeds the before/after demo + the 4-variant Style Lab switcher

(iframes), register tables, anti-slop checklist, code, and the 3 decision

questions. Carries version stamp.

Changed

• planb.html — new active "PLAN 02 · design system" card (cyan, live)

linking to plan-dark-mode-standard.html; reserved slots renumbered to 03/04.

Added the site version-stamp script.

• standarization/DARK_MODE_STANDARD.md — marked the HYBRID register

decision as LOCKED (2026-06-01).

Decision

• HYBRID register split LOCKED — Instrument register (oscilloscope

character) for cockpits + SLD/P&ID labs; Editorial register (Fraunces serif,

refined) for landing / articles / hubs / plans. Cockpit instrument re-skin

queued as a future ship — must preserve semantic SLD/alarm colours

(additive atmosphere/type only).

Owner pivot: existing dark mode read as "AI design slop" — asked for animated-

data-on-load, distinctive type, intuitive responsive (ref raihankalla.id).

This ship lays the design-system foundation + applies the first motion

primitive to the real cockpit.

Engine + #p-dash DISPLAYED values byte-identical. 75/75 accuracy probe

(count-up settles to exact original strings) + 57/57 + 22/22 + all strict

audits PASS.

Added (design system — internal/reference)

• rz-style-lab.html (noindex) — 4 switchable dark-mode characters

(Oscilloscope / Blueprint / Terminal / Editorial). Each replays its

signature animated chart on selection (trace-in / plot-in / grow / sweep)

• KPI count-up. Mobile-responsive switcher. The character picker.
• rz-dark-beforeafter.html (noindex) — side-by-side generic-dark vs

RZ Dark System instrument register, same data block.

• css/rz-dark.css — RZ Dark System v1 shared stylesheet. Two registers

([data-rz-register="instrument"|"editorial"]) of one token system:

shared signal semantics + base ramp + atmosphere + motion primitives.

• standarization/DARK_MODE_STANDARD.md — the standard: two registers,

anti-slop checklist, tokens, type scale, motion primitives, responsive

rules, adoption plan.

Changed

• datahallAI.html — first cockpit adoption of the "animated on load"

primitive. #p-dash headline KPIs (PUE/WUE/CUE/IT/GPU/NVL72) now count up

on first paint via requestAnimationFrame cubic-ease.

• PROBE-SAFE: each value's exact original string is captured and

restored verbatim at animation end → displayed final text byte-identical.

• HARD-SETTLE guarantee: a setTimeout(settle, dur+300) forces the

exact original value even if rAF stalls (backgrounded/throttled tab).

Critical because dkGpu/dkDom are NOT refreshed by the 4s interval — without

it a stalled animation could freeze a wrong basis number. Caught + fixed

during verification (frozen 1.04 PUE / 6,148 GPU on a throttled read).

• Honours prefers-reduced-motion (skips animation entirely).

Process

• Owner mandates logged (memory feedback_planb_always_visual.md): planb

plans must ALWAYS show a working visual; any change must present

before-vs-after VISUAL + code + explanation (not code+prose alone).

Docs

• standarization/DARK_MODE_STANDARD.md NEW.
• CHANGELOG.md (this entry) + changelog.html regen.
• Memory tracker + visual-mandate feedback updated.

MINOR ship: closes the LAST open P0 in review doc-27 — §3.3 color discipline

("warna status harus menang atas warna domain") + §4.3 alarm state UX

(active/acknowledged/shelved/inhibited/returned-to-normal).

Engine + #p-dash byte-identical. 52/52 line-model probe (added 3 alarm-

state assertions per inspector page) + 75/75 accuracy probe + 57/57 + 22/22

• all strict audits PASS.

Added

• js/rz-alarm-state.js — window.RZAlarmState ISA-18.2 alarm state

machine + colour-discipline arbiter:

• 7 states (normal / unack / ack / rtn_unack / shelved /

suppressed / oos). oos (maintenance) is visually distinct from

fault — closes review requirement.

• 4 severity tiers (critical / high / medium / low).
• resolveColor(alarmState, domainColor) — status colour WINS over domain

unless normal. A faulted cooling pipe shows fault-red, not cyan.

• deriveFromEquipment(equipState) — maps line/breaker data-state into

alarm state + severity + summary.

• chipHtml() + audit().
• standarization/ALARM_STATE.md — full schema + adoption table.

Changed

• js/rz-inspector.js — Alarms tab now renders an ISA-18.2 alarm-state

chip + derived summary via RZAlarmState.deriveFromEquipment(). Graceful

fallback to v1.43.0 inline logic if the library is absent.

• **datahallAI.html / chiller-plant.html / water-system.html /

fire-system.html** — <script src="js/rz-alarm-state.js?v=1.43.4" defer>

loaded before the inspector. Inspector cache bumped to ?v=1.43.4.

• tools/probe-line-model.mjs — 3 new assertions per inspector page:

RZAlarmState exposed, color-discipline (status overrides domain),

fault→unack/critical derivation.

Scope discipline

• EPMS_Telemetry.html still untouched per owner mandate.

Review doc-27 P0 status — ALL CLOSED

Docs

• standarization/ALARM_STATE.md NEW spec.
• standarization/INSPECTOR.md Alarms-tab note.
• standarization/BMS_SHELL.md v1.43.x table extended.
• CHANGELOG.md (this entry) + changelog.html regen.
• Memory tracker updated.

MINOR ship: closes the remaining half of review doc-27 §3.4 P0 ("Setiap

number card harus punya tooltip 'source + formula + timestamp'"). The

#p-dash dashboard's 7 headline KPI cards now carry hover tooltips citing

the engine path, the formula, and the live update timestamp.

Owner-exclusion on #p-dash was lifted 2026-05-23, so this binding is

permitted under the accuracy gates. ADDITIVE only — the tooltip sets

the title attribute; textContent (the displayed value) is never touched,

so the no-random-on-basis-KPI accuracy rule stays intact.

Engine + #p-dash displayed values byte-identical. **40/40 line-model

probe** (added KPI-tooltip assertion) + 75/75 accuracy probe (no random-

basis regression) + 57/57 + 22/22 + all strict audits PASS.

Changed

• datahallAI.html updateDashKPI() — appended an additive IIFE that

sets a title tooltip on each headline KPI value:

• dkPue — "PUE = Total facility kW ÷ IT kW" · DATAHALL_CALC.pueBasis()
• dkWue — "WUE = water L ÷ IT kWh = 0 (dry-cooler closed loop)" · BASELINE-DECISION.md
• dkCue — "CUE_IT = grid factor × PUE = 0.69 × " · PLN Java grid 2025
• dkIt — "IT Load = 4 halls × 27 NVL72 × 132 kW = 14.256 MW" · Scenario A
• dkGpu — "GPUs = 108 domains × 72 = 7,776" · DATAHALL_MODEL
• dkDom — "NVL72 domains = 4 halls × 27 = 108" · DATAHALL_MODEL
• dkCdu — "CDU = ceil(3,029 kW ÷ 350) = 9/12 per hall × 4" · lockedState().cdu
• Each tooltip ends with "Updated: HH:MM:SS (engine-derived, SIM)" —

timestamp refreshes on every 4s tick so operators see data freshness.

• tools/probe-line-model.mjs — added datahallAI assertion: all 7

headline KPIs carry a title containing both Source: and Updated:.

Scope discipline

• Displayed KPI values unchanged — accuracy probe confirms no

random-basis regression. The owner-excluded #p-dash displayed numbers

are byte-identical; only the hover title metadata is new.

• EPMS_Telemetry.html still untouched per owner mandate.

Docs

• standarization/BMS_SHELL.md v1.43.x table extended.
• standarization/ACCURACY_VALIDATION.md note (KPI tooltip = additive, no

random-basis impact).

• CHANGELOG.md (this entry) + changelog.html regen.
• Memory tracker updated.

MINOR ship: closes review doc-27 §3.4 ("simulation mode banner") + §5.7 P1

(BMS/DCIM data-quality discipline) + doc-28 Global UIUX Corrections ("data

freshness badge on top bar"). Introduces the data-quality service that

surfaces telemetry trust at both page level (banner) and point level (chip).

Engine + #p-dash byte-identical. 39/39 line-model probe pass (added

3 telemetry-quality assertions per inspector page — total +12) + 75/75

accuracy probe + 57/57 + 22/22 + all strict audits PASS.

Added

• js/rz-telemetry-quality.js — window.RZTelemetryQuality service:
• 7 states with colour + chip label (live / simulated / stale /

manual / comms_lost / inhibited / demo).

• Page-level setPageMode(mode) → injects dismissible top banner

(skips render when mode = live).

• Per-element markPoint(el, state) + getPointState(el) (falls back to

page mode if no explicit data-quality-state).

• chipHtml(state) — small inline chip for embedded rendering.
• audit(rootEl) consumed by probe-line-model.mjs.
• Auto-init on DOMContentLoaded if <body data-rz-data-mode="..."> set.
• standarization/TELEMETRY_QUALITY.md — full schema + adoption table.
• Inspector Live tab now renders a Data Quality chip below State,

reading RZTelemetryQuality.getPointState(currentElement).

Changed

• datahallAI.html:
• <body data-rz-data-mode="simulated"> (added attribute, body otherwise

unchanged).

• <script src="js/rz-telemetry-quality.js?v=1.43.2" defer> loaded after

inspector. Both scripts bumped to ?v=1.43.2.

• chiller-plant.html — same body attribute + script. Body otherwise

byte-identical.

• water-system.html — same body attribute + script.
• fire-system.html — same body attribute + script.
• js/rz-inspector.js — Live tab renders a "Data quality" row reading

the element's data-quality-state (or inherited page mode). No other

inspector behaviour changes.

• tools/probe-line-model.mjs — added 3 telemetry-quality assertions

per inspector page:

• window.RZTelemetryQuality exposed
• body data-rz-data-mode = 'simulated'
• simulated-mode banner rendered

Scope discipline (unchanged)

• EPMS_Telemetry.html — untouched per owner mandate.

Docs

• standarization/TELEMETRY_QUALITY.md NEW spec.
• standarization/INSPECTOR.md adoption table updated.
• standarization/BMS_SHELL.md v1.43.x table extended.
• CHANGELOG.md (this entry) + changelog.html regen.
• Memory tracker updated.

PATCH ship: extends the v1.43.0 right-side inspector to the three other

tagged cockpit pages. Pure additive — one <script src> per page,

no rendering changes, no DOM modifications.

Engine + #p-dash byte-identical. 27/27 line-model probe (added 6

inspector assertions for the 3 newly-covered pages) + 75/75 accuracy

probe + 57/57 + 22/22 + all strict audits PASS.

Changed

• chiller-plant.html — <script src="js/rz-inspector.js?v=1.43.0" defer>

added after rz-breaker-symbols.js. Click any of the 18 tagged CHW pipes

→ inspector opens.

• water-system.html — <script src="js/rz-inspector.js?v=1.43.0" defer>

added after rz-line-model.js. Click any of the 10 tagged pipes → inspector

opens.

• fire-system.html — <script src="js/rz-inspector.js?v=1.43.0" defer>

added after rz-line-model.js. Click any of the 14 tagged fire-water / N2

pipes → inspector opens.

• tools/probe-line-model.mjs — INSPECTOR_PAGES set extended; every

tagged cockpit page now verifies inspector availability + click-to-open

behaviour.

Scope discipline (unchanged)

• EPMS_Telemetry.html — still untouched per owner mandate

("jangan merusak EPMS DC Conventional ya, enhance bole" — 2026-05-26).

• dc-conventional.html — landing page, no inspector needed

(no tagged equipment).

Cumulative state after v1.43.1

Docs

• standarization/INSPECTOR.md adoption table v1.43.1 row.
• standarization/BMS_SHELL.md v1.43.x table extended.
• CHANGELOG.md (this entry) + changelog.html regen.
• Memory tracker updated.

MINOR ship: closes review doc-27 §3.2 P0 ("Equipment popup masih MODAL

CENTER, menutup topology. Jadikan click equipment membuka right-side

inspector, bukan modal tengah").

Self-attaching inspector library that surfaces the metadata laid down by

v1.42.0-v1.42.5. Click any [data-rz-line] or [data-rz-breaker] element

on datahallAI.html to see Live / Capacity / Deps / Alarms / Trend /

Maintenance tabs in a sticky right-side panel that does NOT occlude topology.

Engine + #p-dash byte-identical. 21/21 line-model probe pass (including

the two new inspector assertions: window.RZInspector exposed +

clicking [data-rz-line] opens inspector) + 75/75 accuracy probe + 57/57 +

22/22 + all strict audits PASS.

Added

• js/rz-inspector.js — self-contained inspector library. Vanilla

ES5, no external dependencies. CSS injected on init (scoped to

.rz-inspector*).

• Slide-in panel from right (360px desktop, full-width on ≤640px).
• 6 tabs read directly from element data-* attributes.
• Delegated click handler — catches dynamically-rendered elements.
• ESC + outside-click close.
• Dependency-card click navigates to linked equipment ID.
• Pulsing dot when state=energized. Colour-coded state pills.
• State machine respects review doc-27 §4.3 alarm philosophy

(active / acknowledged / inhibited / RTN derived from state).

• standarization/INSPECTOR.md — schema, tabs, API, adoption table,

authoring guidelines.

Changed

• datahallAI.html — <script src="js/rz-inspector.js?v=1.43.0" defer>

loaded after rz-breaker-symbols.js. Other cockpit pages (chiller-plant,

water-system, fire-system) deferred to v1.43.1 to limit blast radius.

• tools/probe-line-model.mjs — extended for datahallAI:

asserts window.RZInspector exposed + synthetic click on first

[data-rz-line] opens the panel.

Scope discipline

• EPMS_Telemetry.html intentionally untouched per the owner mandate

("jangan merusak EPMS DC Conventional ya, enhance bole" — 2026-05-26).

• Other cockpit pages (chiller-plant, water-system, fire-system) deferred

to v1.43.1 — additive script load only, no rendering change.

Docs

• standarization/INSPECTOR.md NEW spec doc.
• standarization/BMS_SHELL.md v1.43.x adoption row.
• CHANGELOG.md (this entry) + changelog.html regen.
• Memory tracker updated.

MINOR ship: sixth in the v1.42.x → v1.45.x sweep. Cross-page adoption

extends to water-system.html (water treatment P&ID) and fire-system.html

(fire-water + N2 distribution P&ID).

Static SVG <path class="pipe-base"> elements tagged inline with

data-rz-line="1" + data-from / data-to / data-medium / data-state

/ data-capacity / data-current / data-redundancy / data-sensor.

No JS changes — the validator queries the static attributes.

Engine + #p-dash byte-identical. 19/19 line-model probe pass

(datahallAI 171L+36B; chiller-plant 18L; water-system 10L; fire-system 14L)

• 75/75 accuracy probe + 57/57 + 22/22 + all strict audits PASS.

Added

• water-system.html + fire-system.html — js/rz-line-model.js

loaded non-deferred after conv-engine.js so the probe can call

window.RZLineModel.audit().

Changed (additive metadata only, visual identical)

• water-system.html — 10 pipes tagged inline (raw → filter →

pump → UV → treated tank/makeup chains, backwash drain, makeup blowdown):

ws-mains-in, ws-raw-to-filter, ws-filter-to-pump, ws-pump-to-uv,

ws-uv-to-treated-header, ws-treated-to-domtank, ws-treated-to-makeup,

ws-domtank-to-service, ws-bw-to-drain, ws-makeup-blowdown.

• fire-system.html — 14 pipes tagged inline:

fs-landlord-makeup, fs-suction-header, fs-n2-supply,

fs-fp01-discharge, fs-jp-discharge, fs-fp02-discharge,

fs-to-wet-zone, fs-to-preaction-zone,

fs-n2-distribution-header, fs-n2-pv1..fs-n2-pv5.

Probe infrastructure

• tools/probe-line-model.mjs — waitUntil:'load' (was networkidle2)

to avoid timeouts on pages with long-running flow animations.

• Targets bumped:

water-system.html: 10, fire-system.html: 14.

Docs

• standarization/LINE_MODEL.md v1.42.5 row.
• standarization/BMS_SHELL.md v1.42.x table extended.
• CHANGELOG.md (this entry) + changelog.html regen.
• Memory tracker updated.

MINOR ship: fifth in the v1.42.x → v1.45.x sweep. First port to a SECOND

page (chiller-plant.html). Owner mandate respected: **EPMS_Telemetry.html

intentionally untouched** — additive enhancement only, no rendering changes.

Engine + #p-dash byte-identical. 11/11 line-model probe (datahallAI 171

lines + 36 breakers; chiller-plant 18 lines) + 75/75 accuracy probe +

57/57 + 22/22 + all strict audits PASS.

Added

• chiller-plant.html script imports — js/rz-line-model.js +

js/rz-breaker-symbols.js loaded non-deferred after conv-engine.js,

before drawPid() inline IIFE can call them.

Changed

• chiller-plant.html drawPid() function — const RZL = window.RZLineModel;

binding added at top.

• +16 branch lines ported (4 chiller loops × 4 lines):
• chw-sup-drop-{loopId} (CHWS header → loop supply tee)
• chw-sup-leg-{loopId} (loop supply tee → evaporator)
• chw-ret-drop-{loopId} (CHWR header → loop return tee)
• chw-ret-leg-{loopId} (evaporator → loop return tee)
• State binds to evalLoop(lp).level (fault when ALARM, energized otherwise).
• Sensor tags TT-CHWS-{i+1} / TT-CHWR-{i+1}.
• +2 header lines ported:
• chw-header-supply (CHILLER-PLANT → BUILDING-AHU,

data-current bound to engine coolingKw + flow + temp)

• chw-header-return (BUILDING-AHU → CHILLER-PLANT)
• Animation overlay (<line class="flow chws|chwr" />) unchanged.
• All ports use style.stroke='currentColor' to preserve the existing

.pipe CSS class colour (no visual regression).

• tools/probe-line-model.mjs — ADOPTION_TARGETS['chiller-plant.html'] = 18.

NOT changed (owner mandate)

• EPMS_Telemetry.html — left byte-identical. Owner: "jangan merusak

EPMS DC Conventional ya, enhance bole". Future ports to EPMS will follow

a separate, more cautious plan (likely v1.43.x with inspector pattern

ready first).

Docs

• standarization/LINE_MODEL.md v1.42.4 row.
• standarization/BMS_SHELL.md v1.42.x table extended.
• CHANGELOG.md (this entry) + changelog.html regen.
• Memory tracker project_rz_review_2026-05-26.md updated.

MINOR ship: fourth in the v1.42.x → v1.45.x sweep. Closes review doc-27 §5.5

P0 ("Tampilkan spine/leaf/super-spine grouping. Link harus punya speed:

100G/200G/400G/800G. Tampilkan utilization, packet loss, latency,

oversubscription ratio. Tampilkan failure domain: rack, row, pod,

availability zone.").

Engine + #p-dash byte-identical. 7/7 line-model probe + 75/75 accuracy

probe + 57/57 + 22/22 + all strict audits PASS.

Changed

• datahallAI.html netSvg IIFE (Network Fabric Topology):
• var RZL=window.RZLineModel; binding added at top.
• +32 spine-leaf bundled lanes ported. IDs: sl{si}{li} (where si =

spine 0-3, li = leaf 0-7). Each carries:

• data-medium="fiber"
• data-direction="bidirectional"
• data-state="energized"
• data-capacity="800 Gb/s IB XDR"
• data-current="782 Gb/s"
• data-redundancy="redundant_a"
• data-tag="rail-{si+1} / dom-pod-{li+1}" (failure domain — 4-rail FT)
• +27 NVL72 domain-to-leaf bundled lanes ported. IDs: dom{n}-to-LF-{li}.

Each carries:

• data-medium="fiber"
• data-capacity="4× 800 Gb/s NIC"
• data-current="756 Gb/s"
• data-tag="leaf-row-{ri} / pod-{lfIdx+1}"
• Visual identical (preserves var(--p) purple for spine-leaf and

var(--b) blue for domain-leaf via style.stroke override). Laser

overlay (hover-reveal individual links) unchanged.

• tools/probe-line-model.mjs — ADOPTION_TARGETS.datahallAI.html = 171.

Docs

• standarization/LINE_MODEL.md v1.42.3 row.
• standarization/BMS_SHELL.md v1.42.x table extended.
• CHANGELOG.md (this entry) + changelog.html regen.
• Memory tracker project_rz_review_2026-05-26.md updated.

port × 4 halls)

MINOR ship: third in the v1.42.x → v1.45.x sweep. Ports the SHARED drawDH()

function (called once per DH-01..04) so every per-hall SLD now carries the

line-model + breaker-symbol metadata. Each port multiplies by 4 across the

four data hall containers (elecDH1Svg–elecDH4Svg).

Engine + #p-dash byte-identical. 75/75 accuracy probe + 57/57 + 22/22 +

all strict audits PASS. Line-model probe 7/7 at the bumped targets

(112 lines + 36 breakers on datahallAI.html).

Changed

• datahallAI.html drawDH() function — local var RZL/RZB bindings

added at top after coordinate constants. L0 (MV Customer Substation)

and L1 (RMU) sections ported:

• L0 lines (×4 halls = 44 tagged):
• dh{n}-pln-{a,b}-to-meter (utility → revenue meter)
• dh{n}-meter-{a,b}-to-vcb (meter → VCB incomer)
• dh{n}-vcb-inc-{a,b}-to-bus (incomer → SM6 bus drop)
• dh{n}-bus-tie (purple N.O., redundancy=tie)
• dh{n}-bus-to-f{a,b} (bus → feeder)
• dh{n}-f{a,b}-to-rmu-drop (feeder exit to RMU)
• L0 breakers (×4 halls = 20 tagged):
• VCB-INC-DH{n}-{A,B} (closed, 50/51, PPE2)
• VCB-TIE-DH{n} (open, 25/27 sync, tie)
• F{n}{A,B}-DH (closed, 50/51, PPE2)
• L1 lines (×4 halls = 36 tagged):
• dh{n}-rmu-input-{a,b} (entry to RMU panel)
• dh{n}-rmu-meter-{a,b}-to-vcb (meter to RMU VCB)
• dh{n}-rmu-vcb-{a,b}-to-bus (RMU bus drop)
• dh{n}-rmu-bus-tie (purple N.O., redundancy=tie)
• dh{n}-rmu-bus-{a,b}-drop (RMU bus → TX downstream)
• L1 breakers (×4 halls = 12 tagged):
• VCB-RMU-DH{n}-{A,B} (closed, 50/51, IDMT OC)
• VCB-TIE-RMU-DH{n} (open, 25/27 sync, tie)
• tools/probe-line-model.mjs — targets bumped:

ADOPTION_TARGETS.datahallAI.html = 112, BREAKER_TARGETS.datahallAI.html = 36.

Docs

• standarization/LINE_MODEL.md adoption table v1.42.2 row.
• standarization/BMS_SHELL.md v1.42.x table extended.
• CHANGELOG.md (this entry) + changelog.html regenerated.
• Memory tracker project_rz_review_2026-05-26.md updated.

MINOR ship: second in the v1.42.x → v1.45.x sweep. Adds the breaker symbol

library that the review docs flagged as P0 ("Red/green line can be misread as

energized / alarm / source. Breaker state needs SYMBOL, not just color."),

and ports the Electrical SLD overview to use both RZLineModel + new

RZBreakerSymbols.

Engine + #p-dash byte-identical. Probe targets bumped accordingly.

Added

• js/rz-breaker-symbols.js — window.RZBreakerSymbols IEC 60617 /

IEEE C37.2 compliant symbol library. 7 STATES with distinct glyphs:

• closed — vertical mechanical link
• open — angled arm separated from upper terminal
• tripped — open arm + red X overlay + pulse animation
• racked_out — dashed cradle bracket + open arm
• test — open arm + 'T' badge
• maintenance — open arm + padlock badge (LOTO)
• disabled — faint arm + diagonal slash
• 11 ANSI device function numbers (25/27/50/51/52/67/67N/81/86/87T/87B)
• 4 NFPA 70E PPE categories (PPE1–PPE4)
• standarization/BREAKER_SYMBOLS.md — full schema + glyph table +

device-function reference + adoption schedule.

• Probe extended (tools/probe-line-model.mjs) — verifies breaker

tagging alongside line tagging. Result: 7 pass, 0 fail at v1.42.1

(32 lines + 4 breakers tagged on datahallAI.html).

Changed

• datahallAI.html Electrical SLD overview (elecOvC IIFE):
• +25 lines ported to RZLineModel.line():
• elec-pln-a-to-meter / elec-pln-b-to-meter (MV utility supply)
• elec-meter-a-to-vcb-inc-a / elec-meter-b-to-vcb-inc-b (revenue meter to incomer)
• elec-vcb-inc-{a,b}-to-bus-{a,b} (incomer to MV bus, horiz + vert drops)
• elec-bus-a-drop-vert / elec-bus-b-drop-vert
• elec-bus-tie (bus tie N.O., redundancy=tie)
• 8 feeder loop × 2 segments = 16 feeders (elec-feeder-{a,b}{1-4}-drop / -exit)
• 4 breakers ported to RZBreakerSymbols.render() (pilot):
• VCB-INC-A (closed, redundant_a, 50/51/67N, PPE2)
• VCB-INC-B (closed, redundant_b, 50/51/67N, PPE2)
• VCB-TIE (open, tie, 25/27 sync-check)
• F1A (closed, redundant_a, 50/51, PPE2)
• Visual identical via style.stroke overrides + co-existing with

legacy cbl() / symCB() markers. Engine integrity preserved.

Process

• Review-doc mandates addressed: doc-27 §3.1 + §5.3, doc-17 §3.2,

doc-28/18 EPMS line callouts. Tracked in

Documents/screenshot bms rz/REVIEW-ANALYSIS-2026-05-26-vs-v141x.md.

• Memory tracker project_rz_review_2026-05-26.md updated.
• standarization/BMS_SHELL.md v1.42.x adoption row extended.
• standarization/LINE_MODEL.md v1.42.1 target ≥32 codified.

MINOR ship: opens the v1.42.x → v1.45.x sweep responding to team review docs

27/28 (DC AI) and 17/18 (DC Conv). Foundation library so every pipe / busbar

/ cable in the BMS cockpit can carry structured metadata (from, to,

medium, direction, state, capacity, current, redundancy).

Engine files (js/datahall-model.js, js/datahall-calculations.js,

js/conv-engine.js) byte-identical. #p-dash panel byte-identical. 75/75

accuracy probe + 57/57 + 22/22 engine tests + all 4 strict audits PASS.

Added

• js/rz-line-model.js — semantic line model library. IIFE ES5, exposes

window.RZLineModel with:

• MEDIUMS (23 entries): CHWS / CHWR / TCS supply+return / CW supply+return

/ FWS / FWR / dry-loop / liquid-supply+return / HV/MV/LV power / busway

/ UPS feed / signal / fiber / copper / fire / leak / drain / fuel.

• STATES (7 entries): energized / de-energized / standby / fault /

isolated / maintenance / simulated. Each maps to opacity + pulse.

• REDUNDANCY (8 roles): duty / standby / redundant_a / redundant_b /

bypass / tie / maintenance / common.

• line(spec) / path(spec) / polyline(spec) builders — emit SVG with

both visual rendering AND data-* metadata baked in.

• audit(rootEl) — walks DOM, returns {tagged, untagged, total, issues}.
• tools/probe-line-model.mjs — headless puppeteer probe. Verifies

per-page adoption schedule (v1.42.0 target: 7 lines in datahallAI Cooling

P&ID) + every tagged line has the required schema fields + all mediums /

states reference the canonical enum. Result: 4 pass, 0 fail.

• standarization/LINE_MODEL.md — new standard. Schema table, mediums

table, states table, redundancy table, builder API, validator usage, per-

ship adoption schedule v1.42.0 → v1.45.x.

Changed

• datahallAI.html Cooling P&ID — 7 PILOT lines ported to RZLineModel.line():

1. cool-cw-dc-to-pump — CW supply, DC array → CW pump station.

2. cool-cw-pump-to-chiller — CW supply, pump → chiller plant.

3. cool-fws-chiller-to-cdu — FWS supply, chiller → CDU array.

4. cool-tcs-cdu-to-racks — TCS supply, CDU → rack manifold.

5. cool-tcs-return — TCS return, racks → CDU.

6. cool-fws-return — FWR, CDU → chiller.

7. cool-cw-return — CW return, chiller → DC array.

Visual identical (style overrides preserve existing palette + class="fR"

/ class="fL" animation hooks). Engine integrity preserved.

Process

• Review-doc mandates addressed: doc-27 §3.1 line metadata requirement,

doc-28 per-screenshot line semantics. Full gap analysis at

Documents/screenshot bms rz/REVIEW-ANALYSIS-2026-05-26-vs-v141x.md.

• Memory tracker: project_rz_review_2026-05-26.md + MEMORY.md pointer.
• BMS_SHELL.md adoption row appended.
• Per CONTENT_LINKAGE_PLAYBOOK.md: handoff to memory + standarization +

CHANGELOG (this entry) + Standards-Hub (next-pass note).

MINOR ship: closes owner request "GUI tampilan2 HMI dga block/chart atau widget

dll kok tidak ada... dan di security systemnya juga tidak lengkap mana monitoring

semua cctv, aacs dll. sempurnakan".

Added

• Security HMI widget panel on Security Systems tab. Three-tile grid

(1.4fr / 1fr / 1fr) with:

• CCTV Live Mosaic — 16-cell scan-line/gradient TV preview grid with

blinking rec dot per cell, camera ID + zone label, status colour ring

(ok / warn / bad). Plus 24-hour camera availability sparkline (SVG

polyline, green fill); 3-stat row (Cams Online / Uptime 24 h / VMS

Storage).

• Access Control Doors — scrollable list (max 240 px) of 12

representative doors (4 mantraps / DH-1..4 / NOC / MMR / Chiller /

Generator / Fuel Farm). Each row: door tag, 24-hour event count, status

pill (LOCKED green / UNLOCKED amber with left-border accent).

• Intrusion Zones — SVG donut (armed-green / disarmed-grey arcs) over

central numeric armed/total count; "No alarms" or "⚠ N alarm" subline.

Plus scrollable 10-zone list with PIR detector tag, zone name, status

pill (ARMED / DISARM / ALARM). Donut math: 2πR=201, dasharray split

by pctArmed and pctDisarmed with proper offset rotation.

• Security segment data expanded (SEGMENTS.sec):
• Services: 4 → 8 entries (NVR-01 + NVR-02 separate, Access Control

with door count, Intrusion + zone count, Fence-line PIR, VMS

Recording, SOC link, Mass Notification).

• Caps: 4 → 8 cards (cameras 96, doors 48, zones 24, trunk cap 10 Gbps,

trunk util 34 %, retention 14 d, NVR storage 192 TB, stream uptime

99.96 %).

• Links: 3 → 9 entries (NVR-01/02 separate, ACS-01/02 separate, IDS-01,

PIR-Fence, SEC-A/B failover, SOC link, Mass-Notif PA/strobe trigger).

• New hmi block with cctv[16] + doors[12] + intrusion[10]

structured data feeding the widgets.

CSS

• .hmi-grid 3-column responsive (single-col under 1024 px), .hmi-tile,

.cctv-grid 4×4 with aspect-ratio: 16/9, .cctv-cell with scan-line

gradient + rec-dot pulse animation, .door-list/.zone-list with

scrollable max-height, status pills (.dr-st.locked/.unlocked, .zr-st

.armed/.disarmed/.alarm), .hmi-donut SVG holder, .hmi-spark

sparkline, .hmi-row-stats 3-cell bottom strip.

Engines locked

• js/datahall-model.js + js/datahall-calculations.js + js/conv-engine.js —

byte-identical. 57/57 + 22/22 tests pass.

Changed

• js/rz-version.js → v1.41.6.
• sw.js cache version → rz-cache-v1.41.6.

Files touched

• ict.html — Security HMI CSS block; SEGMENTS.sec data expansion; new

renderSecurityHmi(hmi) function; renderSegment wiring (`if key==='sec'

&& seg.hmi`).

PATCH ship: closes three water-system.html owner complaints in one pass.

Fixed

• UV-401 / DOS-302 label overlap (owner: "ini uv ster dan DOS itu masih

saling tumpang tindi"). UV-401 rect previously sat at x=660-760 directly

over DOS-302's right edge (x=630-690), so the UV-401 tag at (666, 97) fell

inside DOS-302's box. UV-401 rect relocated to x=700-760 (width reduced

100→60), tag x=706, labels tightened to fit the narrower box. DOS-302 stays

where it is; the two boxes no longer overlap horizontally.

• TK-402 dual-pipe routing (owner: "ini juga kok ada 2 piping masuk ke

TK-402"). Original treated header → TK-402 path used a Z-shape

(M850 210 H760 V250) which visually appeared as two separate pipes

entering the tank top. Replaced with a straight drop

(M850 210 V250) so a single clean pipe enters the tank from the tee.

Animated flow path (#f-dom) updated to match.

• CT-MK rename to MK-501 Make-up Water System (owner: "DAN DI CT-MK,

KAN TIDAK ADA PAKAI COOLING TOWER. cukup tuliskan aja make up water

system"). The DC AI baseline is dry-only (BASELINE-DECISION.md, no

evaporative cooling tower); the CT-MK / "Cooling Tower" label was

inconsistent with that design.

• Tag: CT-MK → MK-501.
• Labels: Cooling Tower + Makeup → Make-up Water + System.
• Comment "Cooling tower blowdown -> drain" → "Make-up system blowdown

/ discharge -> drain".

Engines locked

• js/datahall-model.js + js/datahall-calculations.js + js/conv-engine.js —

byte-identical. 57/57 + 22/22 tests pass.

Changed

• js/rz-version.js → v1.41.5.
• sw.js cache version → rz-cache-v1.41.5.

Files touched

• water-system.html — UV-401 rect + label coords; treated header pipe

routing (.pipe + .flow); makeup section tag + labels + comment.

MINOR ship: datahall.html mega-bundle closing 5 owner requests in one pass.

Fixed / Added

• CRAH popover replaces full-screen modal (owner: "ini mdal a01 jangan

mendisable main screen, tapi dibaut semacam tootip gitu"). Old behaviour:

click CRAH cell → opens unitModal with .modal-overlay covering entire

screen + blur backdrop. New behaviour: click CRAH cell → opens floating

.crah-pop positioned next to the clicked cell. Main screen stays fully

interactive. Click outside or press Esc to close.

• New .crah-pop CSS: position:absolute, min-width 240, max-width 280,

fade-in 160 ms, RUN/STBY state pill in cyan/amber.

• Grid: 8 rows × 2 cols (status / SAT / RAT / fan / valve / duty / CHWS-R /

redundancy / source).

• Outside-click + Esc handlers wired.
• Legacy unitModal kept for any non-CRAH inspector use.

• Inline SAT values on CRAH cells (owner: "enahce ui dan kasih SAT value

di A01 dstnya"). Each CRAH cell now shows tag + SAT (17.0°C / OFF)

inline; previously the SAT was only visible in tooltip.

• Per-cell rack labels (ID + per-mode value) (owner: "tetap di kotak2

itu kasih ID rack based on row dan ada temp per masing kotakan saat menu

temperature dll. dan saat power juga ada kw nya ganti").

Each rack cell now renders two-line content:

• Line 1 (.rk-id, 7.5px bold): rack tag without prefix (e.g. AL01).
• Line 2 (.rk-val, 7.5px regular): per-mode value
• Power mode: 8.4 kW
• Temperature mode: 22.0°C
• Cooling-margin mode: 5.0°C (margin to ASHRAE 27 °C)
• Alarms mode: 78% (rack utilisation)
• Space mode: IT / spare

repaintRacks() rewrites innerHTML per mode change.

• Cold-aisle normalisation to ~22°C uniform (owner: "temp aisle rack kok

beda2 harusnya hampir sama, for sake of standard and normal operation").

Previous behaviour: each zone's coldAisleC was derived from

SUPPLY_C + zHeat / (rho*cp*airflow) which yielded different values per

zone based on rack-loading variance. New behaviour: normal state is

pinned to 22.0 ± 0.3 °C across all rows (deterministic seed per row), so

the page baseline matches ASHRAE A1 recommended low. Each zone retains

the original physics-derived value as baseTempC for restore-after-

excursion. Hot-spot indicator now genuinely indicates anomalies, not

load skew.

• Excursion simulator (owner: "klw mau dibuat simulasi awalnya normal

mungkn dibuat aja salah satu rack atau row sebentar aja 10-15 detik per

2 menit, random position, pastikan align dengan parameter DAHU (A01-A20,

random)"). Every 2 min (first event 30 s after load), the simulator

picks a random zone z ∈ {0..9} + random CRAH n ∈ {1..20} and sets that

zone's coldAisleC to a value 27.5-30.0 °C for 10-15 s. During the

excursion:

• Zone's racks pulse red (animation rk-pulse).
• Affected CRAH cell gets red outline ring (.cc-affected).
• Red banner appears top-centre with zone / CRAH / temp / countdown.
• Event logged with warn severity in the SCADA log.

After expiry: zone returns to baseTempC, banner hides, CRAH ring

clears, recovery logged.

Engines locked

• js/datahall-model.js + js/datahall-calculations.js + js/conv-engine.js —

byte-identical. 57/57 + 22/22 tests pass.

Changed

• js/rz-version.js → v1.41.4.
• sw.js cache version → rz-cache-v1.41.4.

Files touched

• datahall.html — new CSS rules (.rack inline content + .crah-pop + .crah-cell

.cc-sat + .excursion-banner + animations); new DOM (popover + banner);

rewritten __inspectCrah + outside-click/Esc handlers; new dhRand +

fireExcursion simulator; updated repaintRacks, renderFloor rail(),

zone-build with baseTempC + excursionUntil fields.

PATCH ship: closes two owner architectural-correctness complaints.

Fixed

• Building Overview — chillers relocated from Roof to Ground Floor (owner:

"ok chillernya berarti bukan di roof donk? di roof itu harusnya Dry cooler donk?").

Water-cooled centrifugal chillers belong on a Mechanical floor, NOT on the

roof — only heat-rejection equipment (dry coolers exposed to ambient) lives

on the roof in this hybrid design.

• floors[] array entries updated:
• GF: 'MEP · Infrastructure · Support' → 'MEP · Chillers · Generators · WTP'
• Roof: 'Mechanical · Chillers · AHU' → 'Heat Rejection · Dry Coolers · AHU'
• floorNames map updated (overview titles + drill-down labels).
• Roof plan (renderRoof) title: "ROOF PLAN — MECHANICAL EQUIPMENT" →

"ROOF PLAN — HEAT REJECTION & DRY COOLERS" plus sub-title clarifying chiller

location.

• NW quadrant of roof plan: "CHILLER PLANT" (8× CH cells) replaced with

"CW DISTRIBUTION MANIFOLD" (8 supply + 8 return DN500 risers from chillers

below, cross-headers, amber engineering callout).

• Roof equipment schedule: "Chillers 8×4MW" entry replaced with "CW Risers

8 pairs DN500".

• Roof dimension annotation: "28m (Chiller Plant)" → "28m (CW Manifold)".
• Ground Floor plan (renderGroundFloor) title: "GROUND FLOOR PLAN — MEP &

INFRASTRUCTURE" → "GROUND FLOOR PLAN — MEP, CHILLER PLANT & INFRASTRUCTURE".

• Ground Floor "Workshop B / Staging" zone (42-66 × 26-36 = 24m × 10m)

replaced with Chiller Plant Hall containing 8× Carrier 19XR 4MW

water-cooled centrifugal chillers (7 RUN + 1 STBY N+1, COP 6.8, R-1234ze(E),

mag-bearing oil-free). CW supply/return riser callout pointing up to Roof

dry-cooler array.

• Cooling P&ID — label overlap + clarity (owner: "4.3 bar masing kotakan

tumpang tindih dengan block2 lain, atau temp 35.2°C itu temp apa. nggak jelas,

dan return temp itu tertutuk dash line kuning, dan parameter2 lain 3.4 f dan p

itu coba review lagi").

• Floating temp badges now prefixed with named context — "FWS-Ret 22.0°C",

"TCS-Ret 45.0°C", "CW-Mid 35.2°C" — so each badge explicitly says what

return circuit it represents. Badge widths increased from 30 to 42-50 px to

accommodate the prefix.

• CW pump-station discharge pressure + flow labels moved from y=170-180

overlap zone down to y=193-203 (separate row, off the pump-box outline).

• CW discharge to chillers label "4.5 bar" → "CW@4.5 bar" with positions

spread (y=84 above pipe, y=105 below pipe).

• FWS pump-station discharge flow + pressure labels moved from y=178-188

down to y=195-205 (clear of pump-box).

• FWS to CDUs label "6 bar" → "FWS@6 bar" with explicit prefix.

Changed

• js/rz-version.js → v1.41.3.
• sw.js cache version → rz-cache-v1.41.3.

Engines locked

• js/datahall-model.js + js/datahall-calculations.js + js/conv-engine.js —

byte-identical. 57/57 + 22/22 tests pass.

Files touched

• datahallAI.html — floors array, floorNames map, renderRoof (NW quadrant +

title + schedule + dimension), renderGroundFloor (title + Workshop B → Chiller

Plant Hall), cooling P&ID (temp badges + pump-station labels).

• js/rz-version.js — version bump.
• sw.js — cache name bump.
• CHANGELOG.md — this entry.

PATCH ship: closes owner request "kualitas air masing-masing di dc ai dan dc conventional itu juga masukkan ya full standard parameter dan chemical dan consumption di tech spec."

Added

• DC AI Tech Spec (datahallAI.html) — 7 new sections appended to Section 5

Cooling Discipline (after 5.10 Cooling Tower / Condenser Water Sizing):

• 5.11 Make-up Water Quality — 13 inlet quality parameters vs WHO

drinking-water guidelines (pH, TDS, hardness, chloride, sulphate, iron,

manganese, turbidity, coliform, Legionella, FCl, inlet filtration spec).

• 5.12 Closed-Loop Condenser Water Chemistry — 50 % USP propylene-glycol

dry-cooler loop (12 parameters: pH, reserve alkalinity, freeze protection,

specific gravity, nitrite, azole, Fe, Cu, glycol degradation, top-up rate,

refresh cadence).

• 5.13 TCS / DLC Water Quality — Deionised water spec per ASTM Type II + OCP
• NVIDIA GB200 reference (14 parameters: conductivity < 5 µS/cm, pH 7-9,

TOC < 50 ppb, silica < 0.1 mg/L, < 0.1 µm filtration, ORP > +200 mV,

DO < 100 ppb, HPC < 100 cfu/mL, quarterly polish bed replacement).

• 5.14 Chemical Dosing Programme — 11-row table per CTI WTP-148 + ASHRAE

12-2020 + Nalco/ChemTreat.

• 5.15 Water Testing Programme — 6-row tiered cadence (daily / weekly /

monthly / quarterly Legionella+ATP+coupons / annual audit / online continuous).

• 5.16 Annual Consumption + Discharge Budget — 12-row facility-level annual

budget. PG top-up ~1,600 L/yr, TCS DI top-up ~30 L/yr, domestic ~4,200 m³/yr,

zero blowdown (closed loop), STP reuse-irrigation ~3,800 m³/yr, WUE

cooling-only = 0.00 (BASELINE-DECISION dry-only), annual water saved vs cooling-

tower design ~21,400 m³/yr.

• 5.17 Standards + References — 8-entry list (ASHRAE 12-2020, CTI WTP-148,

BS 8580-1, WHO 4th, ASTM D1193-06, OCP, EU 2020/2184, Pergub 122/2005).

• DC Conv Tech Spec (dc-conventional.html) — 9 new sections appended to Section 5

Water Discipline (after 5.4 Annualised water consumption):

• 5.5 Make-up Water Quality — same inlet matrix as DC AI plus chloride

limit < 150 mg/L in tower + silica < 50.

• 5.6 Cooling-Tower Loop Chemistry — 17-row evaporative-loop chemistry: pH

8.0-9.0, CoC 4-6, basin conductivity 1,200-1,800 µS/cm, LSI, RSI, FCl

0.2-0.5, ORP +650-750, side-stream filter 5 % recirc.

• 5.7 Cooling-Tower Mass Balance — Make-up = Evaporation + Drift + Blowdown

derivation; engine-bound to live waterFlowLpm from CONV_CALC.snapshot.

• 5.8 Closed-Loop CHW Chemistry — CHW no-glycol loop, nitrite 500-1,000

ppm, azole 50-100 ppm.

• 5.9 Chemical Dosing Programme — 11-row open-tower dosing schedule. Total

~1,820 kg/yr (dominated by 720 kg/yr NaOCl + 280 kg/yr corrosion inhibitor +

220 kg/yr scale inhibitor).

• 5.10 Water Testing Programme — 7-row tiered cadence with **mandatory

quarterly Legionella PCR + culture** per ASHRAE 12-2020.

• 5.11 Annual Consumption + Discharge Budget — 11-row engine-bound budget.

WUE cooling-only from MODEL.environment.wue_l_per_kwh.

• 5.12 Effluent Discharge Compliance — 12-row Pergub 122/2005 + Permen LHK

68/2016 effluent quality matrix. Quarterly KLHK reporting.

• 5.13 Standards + References — 8-entry list (ASHRAE 12-2020, CTI WTP-148,

CTI STD-201, BS 8580-1, WHO 4th, Pergub 122/2005, Permen LHK 68/2016, SNI

6989.59:2008).

Engine binding

• DC Conv 5.7 + 5.11 + 5.12 derive from CONV_CALC.snapshot (waterFlowLpm,

wue, itKw). PDF reflects live scenario state.

• DC AI 5.11-5.17 reference design-locked dry-only basis (BASELINE-DECISION.md).

Engines locked

• js/datahall-model.js + js/datahall-calculations.js + js/conv-engine.js —

byte-identical. Tech Spec PDF is presentation-only; no engine drift.

Changed

• js/rz-version.js → v1.41.2 (date 2026-05-24).

PATCH ship: closes 3 deferred items from v1.40.0 + Network Hub backlog.

Added

• 27 OG images at assets/og/network-*.webp (1200×630 WebP).

Hub landing + compare scaffold + 25 per-topic cards.

tools/build-og-images.py TARGETS list extended with 27 entries; accent

locked to instrument-cyan #00DDFF per Knowledge Labs section.

• Spares Readiness x-post-2.md.
• Spares Readiness LinkedIn draft refreshed to reflect v1.16 engine

(was written at v1.11; engine grew from single calculator to 25-tab

operating engine).

Changed

• auth.js — login inputs wrapped in <form> with onsubmit

preventDefault → doLogin(). Enter key now submits; autocomplete +

required attrs work; for on labels; submit button is type="submit";

aria-hidden on decorative modal shield icon. Site-wide fix.

• js/rz-version.js → v1.40.1
• sw.js → rz-cache-v1.40.1

Deferred (future sessions)

• Network Hub tempo system (25 topic-module refactor)
• Live screen-reader walkthrough
• MTBF / numeric-field normalisation
• Mobile nav drawer tuning

MINOR ship: post-production-review on ai-engineering-maintenance.html.

Reviewer's 2026-05-24 finding split into 3 tiers; **Tier 1 + Tier 2 fixed

here; Tier 3 captured as future-work roadmap** (see docs/plans/).

Reviewer's Addendum A "Industrial Build Blueprint" (11 production screens

• multi-tenant RBAC + edge gateways + CMMS connectors + IEC/ISO compliance)

is explicitly out of scope for this portfolio site and tracked at

docs/plans/2026-05-25-ai-maintenance-product-roadmap.md as a separate

multi-year initiative. The concept page is now honestly labeled and

correctly links there.

Tier 1 — Real bugs / content lies fixed

• docs/research/csv/*.csv — all 8 CSVs gained confidence_tier,

source_ref, effective_date, last_verified_by, license_class

columns per KNOWLEDGE_BASE_STANDARD.md. Page previously claimed

"Every fault row carries a confidence_tier column" but the CSVs

didn't have it — that's now true. Tiers auto-inferred from

existing source strings using the standards-body / industry-press /

vendor mapping:

• high (43.8%): IEEE, IEC, ISO, ASHRAE, CIGRE, NETA, NFPA, NPRD-2016,

OREDA, IEEE 493, CIGRE TB-* etc.

• medium (5.8%): Hydraulic Institute, CTI, OCP, EPA/OSHA standards,

journal articles, vendor application notes

• thin (50.4%): single-vendor / manufacturer / OEM / unattributed
• failures.csv + steps.csv — added source column (previously

absent); inherit from parent fault / action.

• Referential integrity fix: 3 step rows referenced action

A-14.1-P-1 which didn't exist in actions.csv (only A-14.1-P did).

Re-pointed to A-14.1-P. **0 orphan FKs across all 826 data rows

post-transform** (verified via cross-CSV grep).

• Row count corrected: page claimed "834 KG-ready rows"; actual data

rows = 826 (834 = lines including 8 headers). Page now says

826 KG-ready data rows across 8 CSV seed files (834 lines incl. headers).

• "Auto-action allowed" wording removed — was unsafe industrial

framing. Now reads: *"Eligible for draft work-order generation; human

approval required before any operational action. AI is advisory only;

physical control remains in SIS / protection relays / BMS engineered

sequences."* IEC 61508 alignment.

• Concept-page banner at top of hero: explicit "this is a concept-

and-design document, not a production product" framing, with link to

the product roadmap doc.

• Knowledge Base section added to sticky section-nav (was orphaned).
• 27 Font Awesome decorative icons got aria-hidden="true".
• .rz-demo-hint hidden on this Pro-only page (page-scoped CSS) so

the demo credential isn't advertised when page-access: { demo: false }.

Tier 2 — Honesty + provenance work

• All CSVs now machine-checkable for confidence-tier discipline.
• Provenance fields (source_ref, effective_date, last_verified_by,

license_class) enable downstream KG ingestion governance.

• Concept page now distinguishes RPN as "ranking only" from

probability-equivalent loss math (the production roadmap doc lays out

Weibull / calibrated probability / expected-loss for future work).

• Each confidence tier's engine treatment now describes who approves

(human / reliability engineer / vendor outreach), not "auto-action".

Added

• docs/plans/2026-05-25-ai-maintenance-product-roadmap.md —

faithful capture of the reviewer's 3,092-line industrial-product

blueprint as future-work roadmap. 13 sections: north star, RBAC,

11-screen product surface, cloud/edge architecture, calculation

engine, knowledge governance, safety + cybersecurity (IEC 61508 /

62443 / ISA-95), build phases A–H, vertical-slice pilot, standards

anchor (cite-don't-claim-compliance), explicit out-of-scope-for-RZ

carve-out, production acceptance bar (20 items), acknowledgements.

• Memory: feedback_concept_vs_product_scope.md — codifies the

"don't conflate concept-page critique with production-app critique"

rule so future reviewers proposing similar scope-creep can be politely

refused with reference to this pattern.

What was REFUSED (and why)

The 2026-05-24 review proposed building:

• 11 production screens (Command Center, Triage Queue, Diagnostic Case

Detail, Planner Board, Technician Mobile Workbench, etc.)

• 20+ microservices (Auth/Tenant, Asset Registry, Sensor Ingest, Feature

Extraction, Model Inference, Calibration, Anomaly, RUL, KG, Advisor,

Recommendation, CMMS, Spares, WO, Review, Audit, Model Registry, KG

Release Registry, Edge Sync, Notification, Reporting)

• Multi-tenant industrial SaaS with 9-role enterprise RBAC
• Edge gateways with signed OTA + OPC UA / BACnet/SC integration
• Compliance audits against IEC 60812 / 61508 / 62443, ISO 14224 / 55001,

NIST AI RMF / SSDF, ISA-95 / ISA-101

• React/Vue/Svelte frontend stack adoption

This is enterprise industrial-SaaS scope — multi-year, multi-engineer,

multi-million-dollar. The resistancezero.com portfolio site is a single-

developer zero-build static GitHub Pages deployment. Building these

inside the portfolio site would be scope-explosion of 2–3 orders of

magnitude. The roadmap doc captures all of it as **valid future-product

vision**; it does not become next-week code. See

feedback_concept_vs_product_scope.md memory for the principle.

Status

• audit-script-tags.py --strict CLEAN (176 files)
• audit-js-syntax.py --strict CLEAN (106 files)
• RPN integrity: 109/109 rows match S * O * D (unchanged)
• Referential integrity: 0 orphan FKs across all 826 rows (was 3)
• Concept page lies removed; banner honest; roadmap linked

Bumped

• js/rz-version.js → v1.40.0 (MINOR; concept-page honesty pass)
• sw.js → rz-cache-v1.40.0

Cross-references

docs/plans/2026-05-25-ai-maintenance-product-roadmap.md ·

standarization/KNOWLEDGE_BASE_STANDARD.md ·

docs/research/2026-05-23-fmeca-kg-worldwide-asset-failure-data.md

MINOR ship: Phase 2 Lane B complete. All 9 Lane B topics live.

Anti-monotony audit ran across 9 topics, 0 findings (max pairwise share = 2).

Added (5 new live topic pages + modules)

• network/industrial-ot/dnp3.html — IEEE 1815. Distinctive trait:

UNSOLICITED responses (outstation pushes spontaneously, amber-labeled).

4 pitfalls (Class-0/1/2/3 buffer overflow, SBO vs Direct Operate, time

sync drift, SAv5 cert rotation).

• network/industrial-ot/profinet.html — IEC 61784-2. Sync line above

the data wire shows cyclic deterministic timing; green tick at each cycle

start. 4 pitfalls (RT/TCP jitter, GSDML/firmware mismatch, topology

change, IRT clock master loss).

• network/industrial-ot/ethernet-ip.html — ODVA CIP over Ethernet.

Sawtooth waveform, envelope chips with rotating CIP-layer marker stripes

(ENIP / CPF / CIP-Conn / CIP-Svc). 4 pitfalls (Class 1 RPI, EDS vs

firmware, timeout multiplier, port 2222 vs 44818).

• network/industrial-ot/ethercat.html — IEC 61158. Telegram passes

through every slave on-the-fly; nearest slave lights green as the chip

crosses. Slave count + cycle time configurable. 4 pitfalls (slave

processing accumulation, distributed clocks, hot-plug, mailbox bandwidth).

• network/industrial-ot/bacnet-mstp.html — ASHRAE 135 Annex H.

Token-passing on RS-485: amber token chip visibly passes between nodes

before any data frame. 4 pitfalls (mixed baud rates, token timeout,

Max_Master, reply-too-late re-poll storms).

• 5 corresponding topic modules in js/network-anim/topics/:

dnp3.js, profinet.js, ethernet-ip.js, ethercat.js, bacnet-mstp.js.

All Strategy-A deterministic frame logic.

Anti-monotony matrix (all 9 Lane B pairs)

All 36 pairs ≤ 2 shared timbre fields. Audit passes by design.

Changed

• network-visualization-hub.html — all 9 Lane B cards now show LIVE.

Each card describes the distinctive trait per Appendix E.

• network-compare.html — picker expanded to 9 protocols; topic

registry + script loads updated. Compare any 2–4 of the full

Lane B set.

• datacenter-solutions.html — Knowledge Labs card description

updated to mention all 9 live Lane B topics.

• js/rz-feature-flags.js — 5 new public-tier entries

(network-dnp3, network-profinet, network-ethernet-ip, network-ethercat,

network-bacnet-mstp).

• sitemap.xml + llms.txt — 5 new entries.
• js/rz-version.js → v1.38.0 (MINOR; Phase 2 Lane B completion)
• sw.js → rz-cache-v1.38.0

Status

tools/audit-network-anim.py — CLEAN, 9 topics audited, 0 findings.

tools/audit-script-tags.py --strict — CLEAN (160 files).

tools/audit-js-syntax.py --strict — CLEAN (106 files).

test-network-anim-determinism.py --static — expected 27/27 PASS

once re-run with the new fixtures.

Phase 2 Lane B distinctive-trait inventory (live now)

9 protocols, 9 distinct visual + audio signatures. Anti-monotony works.

Next phases (Lane B is 100% done — moving to other lanes)

• Phase 3 Lane A — Foundations (OSI/TCP-IP models, IPv4 vs IPv6, subnetting/CIDR, TCP handshake, DHCP/DNS)
• Phase 4 Lane D — Security (TLS handshake, OAuth/JWT, mTLS, WireGuard)
• Phase 5 Lane E — APIs + Agents (REST API, GraphQL, gRPC, MCP tool-call)
• Phase 6 Lane C — DC Management (SNMP, IPMI/Redfish, syslog)

MINOR ship: completes the v2.3 Phase 0 DoD inner loop. Anti-monotony +

determinism + post-draft + content-linkage all operational.

Added

• tools/test-network-anim-determinism.py (~210 lines) — Strategy-A

determinism harness with Node + static modes. 12 / 12 PASS across

4 topics.

• Article/Post Draft/Network Hub/ + 4 topic folders (Modbus RTU /

Modbus TCP / BACnet IP / OPC-UA), each with linkedin + x-post-1 +

mastodon-1. 15 post-draft files total per POST_DRAFT_STANDARD.

• standarization/CONTENT_LINKAGE_PLAYBOOK.md §2.5 — Knowledge Labs

topic page deliverable checklist.

Status

• audit-network-anim.py CLEAN (4 topics, 0 findings)
• test-network-anim-determinism.py --static 12/12 PASS
• audit-script-tags --strict CLEAN
• audit-js-syntax --strict CLEAN

Deferred to next session

• OG images for network pages (tool doesn't scan subdirectories)
• Live screen-reader walkthrough validation
• Phase 2 Lane B (DNP3, PROFINET, EtherNet/IP, EtherCAT, BACnet MS/TP)
• Spares Readiness post-draft refresh

Changed

• js/rz-version.js → 1.37.0
• sw.js → rz-cache-v1.37.0

MINOR ship: Lane B (Industrial OT) Phase 1 complete with 4 live topics

and a functional 4-panel compare scaffold. Audit confirms anti-monotony

across all 4 (max pairwise share = 2 fields).

Added

• network/industrial-ot/bacnet-ip.html — live Phase-1 topic.

ASHRAE 135 BACnet/IP packet exchange over UDP with BVLC tunnel rendered

as scan-line shroud at packet head. 3 parameter controls (payload bytes

/ UDP RTT / line noise). 4 engineering pitfalls (BBMD foreign-device

registration, port 47808 firewall, instance ID collisions, COV

subscription leaks).

• network/industrial-ot/opc-ua.html — live Phase-1 topic.

IEC 62541 subscription model: client → server with publishing

interval, monitored items, security mode (none / sign / sign-and-encrypt).

Always-on scan-line shroud when security != none. Tertiary discovery

server node visible. 4 engineering pitfalls (cert trust list,

publish/sample interval mismatch, queue overflow, endpoint discovery).

• network-compare.html — 4-panel side-by-side compare scaffold.

Topic picker (any 2–4 of the 4 live protocols). Per-panel

instrument chip strip (throughput / latency / overhead / status)

reading from getNormalized() per Appendix B. URL deep-link

(?topics=modbus-rtu,modbus-tcp,bacnet-ip,opc-ua). Audio muted

default across all panels (compare-mode convention per §7).

• js/network-anim/topics/bacnet-ip.js (~225 lines) — distinct

timbre per Appendix E row 4: triangle 950 Hz, hex 8×8,

ethernet 1.0 px, controller-square master, 1.2× medium

tempo. Shares with RTU: 1 (tempo). Shares with TCP: 1 (wire).

• js/network-anim/topics/opc-ua.js (~260 lines) — distinct

timbre per Appendix E row 5: sine-sweep 1400→1700 Hz,

layered 10×8, ethernet 1.0 px, broker-diamond master,

1.2× medium tempo, progressive encryption. Shares with

RTU: 1 (tempo). Shares with TCP: 1 (wire). Shares with BACnet/IP: 2

(wire + tempo).

• js/rz-feature-flags.js — 4 new public-tier entries

(network-bacnet-ip, network-opc-ua, network-compare, plus prior

network-modbus-tcp).

• sitemap.xml + llms.txt — 4 new entries.

Changed

• network-visualization-hub.html — BACnet/IP + OPC-UA cards now

show LIVE status; compare-mode CTA upgraded from placeholder to

functional link.

Anti-monotony evidence (4 topics, pairwise within Lane B)

All pairs ≤2 shared. Anti-monotony cap holds. Each protocol has its

own audio signature and visual chip vocabulary.

Status

tools/audit-network-anim.py — CLEAN, 4 topics audited, 0 findings.

tools/audit-script-tags.py --strict — CLEAN (155 files).

tools/audit-js-syntax.py --strict — CLEAN (106 files).

Versioning note

v1.35.0 = Modbus TCP + hub landing

v1.35.1 = parallel session's cross-page headline consistency probe

v1.36.0 (this ship) = Lane B complete + compare scaffold

Next

• Determinism test harness (test-network-anim-determinism.py)
• OG images for the 4 live topics + hub + compare pages
• Post-draft folders per POST_DRAFT_STANDARD
• Phase 2: DNP3, PROFINET, EtherNet/IP, EtherCAT (5 more Lane B topics)

Changed

• js/rz-version.js — v1.36.0 (MINOR; Lane B completion)
• sw.js — cache rz-cache-v1.36.0