CHANGELOG

BCP scrape pipeline — in-context weekly orchestration. The weekly scrape now runs end-to-end inside one continuous Playwright browser context: discovery + every per-event scrape happens in a single browser session with inter-event pauses (30-90s) between events, not by forking a fresh Node process per event. The prior fork-per-event pattern violated the design's §5.8 framing ("operationally indistinguishable from a logged-in Editor browsing manually") and tripped BCP's anti-scrape defenses during the W23 manual test — degraded React renders on events 2-22, with 408-byte format-unclear stubs written for 21 of 22 events. The refactor: bcp_scrape_event.js exposes scrapeEvent(eventId, page, session, opts) returning a structured summary with status (ok / format-unclear / auth-lost / error); bcp_scrape_weekly.js orchestrates the loop in one context; bcp_pacing.js gains interEventPause(); the orchestrator halts the batch on any format-unclear or auth-lost (don't burn the session generating stubs after BCP starts pushing back). The CLI for single-event runs still works for debugging; its IIFE now wraps the exported function. Workflow doc updated to reflect the single-command Tuesday-morning flow (~75-100 minutes for a typical 15-25-event week, walk away).

BCP scrape pipeline — operational tuning bundle (5 fixes). (1) Silent-skip warnings: every roster player whose list isn't fetched now surfaces a structured `list_fetch_skipped` warning (reason `no_list_id` or `fetch_failed`) in scraped.json + scraped-summary.md — the prior behavior silently dropped missing lists. (2) Subscription tier detection: rewritten to check for the `Active Plan` affordance first; falls through to a free-tier warning only if no Active Plan is present (the prior logic false-positived paid accounts as free). (3) Snapshot trim: projectToSnapshotShape now builds a minimal projection (provenance + meta + canonical games[]/events_summary[]) rather than deep-cloning the entire scraped object — the raw scraped arrays (roster, pairings, placings, lists) stay only in the gitignored audit file, dropping per-snapshot repo weight and making list_name absence structural rather than per-field. (4) Tuesday cadence: weekly routine docs rewrote FRIDAY MORNING → TUESDAY MORNING. Tuesday lands the prior weekend's data ~36-48 hours after events end, leaving 3-4 days for the audience to absorb before the next weekend's list-building. (5) `--include-overflow N` flag: extends the scrape's search end-date by N days beyond the ISO week's Sunday so Sun-into-Mon (and rare Thu-Mon) bleed events get picked up; recorded explicitly in the published snapshot's `meta.date_range_used`. Tuesday + `--include-overflow 2` is the recommended pairing. verify:digest grew 222 → 270 with new scenarios covering the trim, the silent-skip warning emission, the ISO-week date math, and the date_range_used provenance.

BCP scrape pipeline operational: Phase 1-4 wired against the real BCP structure documented in the Phase 0 notes. Weekly tournament data extraction now automated through a Playwright headless-browser scrape with an Editor curation step before commit — the manual copy-paste workflow retires (preserved as the contingency fallback). New per-game data fields land in the snapshot shape: detachment, total points, unit count, and event format (singles / team-wtc / team-other) — enabling format-segmented analysis and unit-count vs win-rate correlation. The disposition field is present in the schema and populates from W27 onward when 11e events arrive (10e events have no disposition mechanic). _data/performance_pull.js exposes a new metricsByFormat 3-view structure (combined / singlesOnly / teamOnly) on every view alongside the existing metrics field. _data/digest.js gains three new top-level projections — dispositionMatchupMatrix, dispositionRepresentation, unitCountAnalysis — all carrying the 3-view structure. Until W27 data lands, every new projection is hasData:false with a clear awaitingData reason. UI surfaces for the new data follow in a separate commit; data layer ships first. List parser holds the strict editorial promise — only four metadata fields per list, never unit names or weapon names or ability text.

Editorial note banner: weekly data ingestion (Performance pulls, ELO updates, Attention snapshots) pauses through the Warhammer 40,000 11th edition launch transition (Saturday 20 June 2026), resuming with the first post-launch competitive weekend. The next data cycle lands the week of 3 July 2026, capturing tournaments held on 27–28 June. Existing W23 data remains visible on every signal during the gap. The banner renders site-wide at the top of every page through a server-side date check in base.njk driven by _data/buildContext.js; the next build after 2026-07-03 evaluates the condition as false and the banner drops automatically with no manual cleanup. Sits above the existing cookie notice banner (different position, different purpose).

Light ePrivacy / UK ICO posture: a dismissable notice banner appears on first visit (and any visit before acknowledgment) explaining the Archive's localStorage use for preferences (theme, SI weights, organize mode, scatter Y, selected faction, widget view, results stat-view, Pairings Tool saved matrices), the absence of analytics / tracking pixels / advertising / data sharing, and Cloudflare's single bot-protection cookie at the CDN infrastructure level. Acknowledgment persists via localStorage `ia-cookie-notice-acknowledged-v1`. New /privacy.html page (reachable via the footer link alongside Glossary + Changelog) carries the full Editor-voice data-posture disclosure across six sections.

Awakenings: published "The W23 Meta, In Layers" — an Editor-voice feature reading ISO 2026-W23 across eight resolutions (event-size stratification, win rates with Wilson CIs, VP differential and blowouts, weekly movers, representation, non-performance signals, a condensed Defiler-Five callback, and the closing Strength Index). Four SVG figures. Cross-links the W21 + W22 dispatches, the Defiler-Five companion, the four methodology articles, and the live Strength Index page. The closing Strength Index figure is a frozen snapshot from publish-time data; the live Strength Index page is linked inline in the figcaption and in the prose.

Three focused fixes: (1) Quantum Results FACTION PERFORMANCE table gains a STAT: WIN RATE | REPRESENTATION | ROUND Δ toggle above it — the middle column swaps between the existing Wilson 95% CI bar, a field-share proportion bar (0-20% scale), and a zero-centred round-attrition bar (±15pp scale). State persists via localStorage shared per surface. (2) Strength Index breakdown table COMPOSITE column header now aligns flush with its data cells (a 2px padding-left mismatch caused the header to sit 2px left of the data). (3) HOW TO READ THIS PAGE expansion no longer renders the methodology callout and THE FOUR SIGNALS grid side-by-side; the callout now spans full width on top with the four signal cards in a 2x2 grid below.

Strength Index pared down to its essential data widgets. The yellow methodology callout ('We don't collapse these to a single number...') and the four source-attribution cards (Theory / Performance / Practice / Attention, with their descriptions, refresh dates, source lists, and phase status chips) move OUT of the page body and INTO the HOW TO READ THIS PAGE expandable — same information, accessible on demand. Side-by-side forest plot + breakdown table retire in favor of a single-active-view toggle: VIEW: FOREST PLOT | BREAKDOWN TABLE. Only one of the two widgets renders at a time. Default view is the breakdown table; the reader's choice persists via localStorage `ia-si-active-widget-v1` shared across /strength-index.html and the Quantum #strength-index tab. Forest plot labels are now click targets too — clicking a faction in the forest plot selects it in the detail panel below, same behavior as a table row click. The page reaches COMPOSITE PREVIEW within ~300px of the top instead of ~600px.

Strength Index layout fixes: (1) the four collapsed source-card summaries no longer overflow at desktop widths — the verbose phase-status chip (e.g., 'PHASE 5 · 4-SOURCE EQUAL WEIGHT') moves out of the summary row into the body, so each summary fits the auto-fit grid track that wraps responsively; (2) the forest plot and the breakdown table now render side-by-side in a 2-column grid per the v4 mockup, with the sliders and dual-view toggle above the grid at full width and the band legend + faction detail panel below at full width. Stacks vertically at ≤1100px so narrow viewports get the prior linear layout.

Strength Index breakdown table compacted to single-line per-faction rows — rank + faction (with composite-band coloring + Practice tier badge) + four signal z-scores + composite z, color-coded by sign. All 28 factions visible at a glance instead of 3-4 in the prior verbose layout. The full per-source detail (per-source win rates with Wilson CI bars, tier publication counts, attention week-over-week deltas, Theory's matched-play unit count) moves to a faction detail panel below the table — empty by default, populated when a reader clicks any row, with content that swaps when a different row is clicked. Selection persists via localStorage and is shared between /strength-index.html and the Quantum #strength-index tab. Slider drags and dual-view toggles preserve the selection. Keyboard accessibility is automatic (Tab to row + Space/Enter to select). Also fixes a regression introduced by the collapsible-source-cards commit: the four collapsed source cards now reflow on narrow viewports (auto-fit grid) instead of overflowing the viewport edge.

Strength Index source cards (Theory / Performance / Practice / Attention) are now collapsible — collapsed by default, summary row keeps the column label + phase status chip + a compact freshness chip visible at a glance, click expands the description + per-source list + verbose freshness line inline. Saves ~800–1000px of vertical chrome above the forest plot on /strength-index.html; the Quantum Strength Index tab inherits the same behavior via the shared partial. Multiple cards expand independently — native HTML5 details, no accordion logic; keyboard accessibility is automatic (Tab + Space/Enter).

Quantum dashboard redesign — Phases 5–6 follow-up, closing the redesign arc. The Strength Index tab gains its own KPI strip (STRONGER BAND / WEAKER BAND counts, BIGGEST GAINER with an empty-state until the composite-history adapter ships, MOST DIVERGENT cross-signal rank spread, plus a feature CROSS-SIGNAL MEAN ρ tile that stays in its empty state until the cross-signal history adapter ships). Faction names in the breakdown table now color-code by composite band — emerald for STRONGER (z > +0.5), default for AVERAGE, warn-orange for WEAKER (z < −0.5) — and the forest plot's faction labels mirror the same coloring; a compact band-color legend sits directly below the breakdown table and hides when the Phase 2b dual-view toggle is in tier view. The Orientation tab's dashboard-map figure is regenerated (third iteration) to match the new sidebar + status-pills + KPI-strip layout, and its qo-map-list grows from 6 to 8 items with a parenthetical mention of the Results-tab CI bars. The Results-tab COVERAGE strip is now wired to a new digest.coverageStrip projection (the four most-recent third-party writeups, server-sorted newest-first and capped at 4) — the gated AWAITING DATA card now renders real coverage.

Quantum dashboard redesign (Phases 1–4): the dashboard moves from horizontal-tab + giant-hero layout to a sidebar tab rail on the left + status pills row at the top, reclaiming about 155px of vertical chrome. The Results tab is the new default-open and the editorially headlined surface: it leads with a six-tile editorial KPI strip (TOP WIN RATE / STEEPEST CLIMB / SHARPEST DROP / LATE-GAME LEADER / HIDDEN GEM / SHARPEST SPLIT), then a 2-column 14+14 FACTION PERFORMANCE table where every win-rate cell carries a graphical Wilson 95% CI bar (wider band = smaller sample, less certainty), then a top-12 MATCHUP MATRIX below at full width, then a COVERAGE strip at the bottom for recent third-party writeups. The data-tab activation contract, hash deep-links (including the #practice → #strength-index redirect), and all other tab panels are unchanged. New digest projections: dashboardStatus (the 5 status pills), resultsKpis (the KPI tiles + the sorted faction table with server-rendered CI-bar geometry), strengthIndexKpis (band counts + most-divergent; biggest-gainer and cross-signal mean ρ deferred until the history adapters land). Sidebar collapses to a hamburger drawer at ≤900px. Phase 5 (Strength Index color-coded faction names + band legend + SI KPI strip wiring) and Phase 6 (Orientation map figure regen + qo-map-list rewrite) are deferred to a follow-on commit on this branch.

Strength Index gains an interactive cross-signal scatter plot — the first instrument on the site for cross-validating the four-signal model. For a chosen Y signal (Theory / Performance / Practice / Attention; default Practice), the widget plots each faction's Y z-score against the composite z-score computed WITHOUT that signal (with the surviving weights renormalised). Spearman ρ in the header, OLS trendline + 95% CI band, and four persistent outlier labels — all live-update on slider drag and Y change. State (which signal is on Y) persists via localStorage and is shared across the standalone /strength-index.html page and the Quantum #strength-index tab. The widget answers the question the Phase 2b dual-view toggle was teeing up: when one signal is removed, do the OTHER three say the same thing as the removed signal?

Dossiers landing restructure: the Strength Index is now an inline tier list at the top of the section — STRONGER / AVERAGE / WEAKER bands with every faction name linking to its dossier, rendered server-side from the composite buckets via the shared tier-list primitive (the same one the home page's Live From The Archive panel uses). A primary OPEN THE STRENGTH INDEX link sits below it. The Performance Pull card is demoted to a smaller secondary position; the 28-faction grid and the roadmap card are unchanged. The section now answers 'which factions are strong right now' on the page itself.

Strength Index dual-view: organize the 28 factions by composite z-score band (default) OR by Practice tier letter, via a toggle above the breakdown table. A per-faction Practice tier badge now sits next to each faction name in both views (— when a faction has no Practice data this window; those collect in an UNRATED group in the tier view). The view choice and your composite-weight slider positions persist via localStorage and are shared across the standalone /strength-index.html page and the Quantum Strength Index tab — tune the weights or pick a view once and it follows you between the two surfaces.

Quantum dashboard restructure: Strength Index promoted to its own tab (the full tier list, forest plot, breakdown table, and weight sliders, sharing one source partial with the standalone /strength-index.html page); the Composite tab repurposed as Signals with a new cross-signal rank-comparison table that surfaces the factions the four signals most disagree about; the Practice tab retired (its data still feeds the composite and now surfaces on Strength Index plus the Signals-tab rank column); the default-open tab shifted to Results; and the Orientation panel content plus its dashboard-map figure refreshed to match. Legacy /quantum.html#practice links silently redirect to #strength-index; #composite still resolves to the Signals panel.

Performance composite expanded to four primary sources (40kstats joins Stat Check, Warpfriends, and the Archive's own pull); composite math switched from sample-size-weighted mean to equal-weight mean, so each source contributes 25% regardless of sample volume. The composite CI is now the between-source standard error — it widens when the sources disagree and narrows when they converge. The Strength Index Performance column legend discloses the new policy plus the population-blending caveat (40kstats includes casual games via the TableTop Battles app; the other three sources are tournament-only). A transparency note on Performance Pull covers both the methodology upgrade and Stat Check's absence — it hasn't published since 14 May, so this week's composite runs on three of the four primary sources.

Quantum dashboard — FACTION SNAPSHOT table gains the weekly / current-patch / edition-lifetime view toggle that Performance Pull's matchup matrix already has. The dashboard snapshot was previously weekly-only despite the underlying Pull adapter already projecting all three views via `digest.performancePull.views.*`; this commit closes the gap by adding a three-segment `.view-toggle` above the snapshot table and rendering three `<tbody>` bodies (one per view), with a small client-side JS handler that swaps the `hidden` attribute on view-segment clicks. The sort controller now iterates all three tbody bodies in lockstep so a column-sort persists when the reader switches views. Markup-only work; the Pull adapter (Phase 7.1.1.2) already supplied the view-block data.

Pairings Tool chrome converted to design-system shared classes. The toolbar's segmented controls (BLIND / WTC, SCORING) and action buttons adopt `.mx-seg` and `.ia-btn`; the two main cards (Roster + Matrix) plus seven additional cards (the optimal-assignment / alternatives / sensitivity result panels, the PAIRING PROTOCOL explainer, PRESETS, SAVED MATRICES, and the second team-roster card) all adopt `.ia-card`, retiring 36 phosphor-era quadrant decorations (the same `.dc-corner`-style retirement Captain Decisions got in the Matrix surface conversion). The bespoke interactive widgets — the matchup matrix builder (kept on literal rgba 5-stop ratings rather than design-system heat tokens, since the rating scale is editorially meaningful), the calibration overlay, the WTC sequential solver, the saved-matrices panel, the scenario report — stay page-local by design. With this commit, every signature page on the site is on the design-system chrome.

Cover features converted to the `.aw-prose` reader. Where Cover Lives and Cover Ledger now use the same long-form reading chrome the Awakenings dispatches and the methodology articles use — the 17-px Inter measure, the emerald drop-cap, the display-face section heads, the in-article TOC sidebar. With this batch, all eight long-form pages on the site share one reader experience (three dispatches + five methodology articles + two cover features).

Methodology articles converted to the `.aw-prose` reader. Big Soup, Six Bins, Swiss Isn't Random, On Average, and What the Numbers Can Bear all now use the same long-form reading chrome the Awakenings dispatches use — the 17-px Inter measure, the emerald drop-cap on the first paragraph, display-face section heads with the soft bottom rule, mono inline numerals, the in-article TOC sidebar and Related-articles sidebar. Big Soup and Swiss Isn't Random's machinery sidebars adopt `.aw-callout` (first consumer of the primitive); Six Bins and On Average's Chart.js charts re-parent into the new narrower body without changes to their wiring. The two cover features (Where Cover Lives, Cover Ledger) are a separate pending pass.

Awakenings — *The Second Week* — third dispatch published. ISO 2026-W22 against the Infinite Archive performance pull (9,454 games, 28 factions). The disciplined read on whether W21's stories survived contact with twice the data — short answer, every faction's W22 Wilson 95% band overlaps its W21 band, the Defiler-five anecdote stops being a blanket, and the Imperial Agents twenty-point drop is the data telling you N=55 is not a read.

Methodology editorial — *What the Numbers Can Bear* — published to Protocol. The fifth methodology article in the launch series, alongside Big Soup, Six Bins, Swiss Isn't Random, and On Average. Editor voice; two inline SVG figures (the three-dataset frontier of sample size vs recency, and the faction-detachment-disposition dimensionality cascade); makes the structural case for reading 40k statistics at the resolution they support and the editorial case for what publishing them at finer resolution does to the meta they describe.

Strength Index converted to the design system. The per-faction breakdown table re-roots onto the shared `.ia-table` chrome, and the four data cells (Theory / Practice / Attention / synthetic placeholder) re-emit on the shared `.ia-ci-cell` primitive — the same Wilson-CI cell shape Quantum's Performance-tab snapshot table uses. The Performance cell keeps its per-source horizontal-bar visualisation (page-unique to this surface) but the small-N flag becomes a design-system provisional chip. The four column-legend cards, the weights / forest-plot / methodology / roadmap panels all adopt `.ia-card` chrome; the weights-reset button gains `.ia-btn --ghost`; freshness chips map to `.ia-chip --fresh / --stale`. Forest-plot SVG, sliders, row-click highlight, weights-reset behaviour, and the `digest.*` data wiring are unchanged.

Performance Pull converted to the design system — the page the .ia-matrix heat cell was always waiting for. The 28×28 matchup matrix now renders on the seven-stop diverging palette striped against each row faction's own win rate (not the global average — strong-faction rows light up only where they exceed their own elevated mean; weak-faction rows light up where they outperform their own). The four sortable per-faction tables — aggregate, margin profile, representation, consistency — re-root onto the shared table chrome with mono right-aligned numerics; small-N flags become the design-system provisional chips. The weekly / current-patch / edition-lifetime selector adopts the design-system segmented control. The matchup data stays server-rendered; no payload in inline scripts. The Pairings Tool calibration overlay (which reads the same projection) continues to work cleanly.

Captain Decision Archive restyled to the design system. The card-grid index keeps its at-a-glance shape but the chrome moves to design-language conventions: sharp corners with the emerald top-edge ignite on hover (the phosphor-era quadrant-marker corners are gone), display-face titles, mono-caps tags as bordered chips, outcome dots remapped to the positive / warn / negative semantic tokens, and the filters bar lifted to the design-system controls treatment with the count callout pinned right. The Editor-voice arc-intro and the cross-entry lessons cards keep their gold left-rule (Editor voice; distinct from chrome emerald). The five CHTT round detail pages and the matrix.html section landing render through Phase-2-migrated shared chrome unchanged. The per-team BFT-amber / RLD-sky voice accents survive intact across captain quotes and the round narrative bodies. The Pairings Tool is explicitly out of scope on this branch — its bespoke interactive widgets get a focused pass of their own.

Awakenings converted to the design-system reader. Each dispatch page now renders on the .aw-page / .aw-main grid with a sticky sidebar (Related dispatches and an in-dispatch TOC on the longer reads); the prose body moves to .aw-prose for the 17-px Inter measure, the wider line-height, the display-face section heads with the soft bottom rule, the mono inline numerals, and the emerald drop-cap on the first paragraph. The Awakenings section landing's aggregated coverage feed moves to the shared dispatch-list chrome with the medium-filter chips on the .ia-chip primitive; the filter handler is unchanged save the selectors. The Editor-voice featured band and Protocol-feature cross-link callouts on the landing are restyled in place onto the new tokens. The dispatch figure SVGs are untouched (fixed-dark insets per the redesign rule). Out of scope on this branch and still pending their own conversion: the two cover features and the four methodology articles (they already carry the reading-progress bar from Phase 3).

Quantum dashboard — closing pass. Numeric column headers in the converted .ia-tables now line up with their right-aligned data (the alignment classes gained the specificity to win against the table default). The four By-The-Numbers tiles become design-system stat cards with the Reanimation counter still ticking. The divergence widget's hover tooltip is rebuilt as a two-column grid so labels like PERFORMANCE Z and Z-DIFFERENCE can never collide their values — every row reads cleanly on hover, with pos / neg colouring on each signed value. The divergence widget itself and the Theory tab composite-breakdown bars are restyled onto the new tokens (sharp corners, mono numerals, bordered delta chips, --pos / --neg diverging bars). With this branch the Quantum surface conversion is complete across four feature branches; Quantum now matches the design system end to end. Awakenings, Matrix, Performance Pull, and Strength Index surfaces still pending their own conversion branches.

Quantum dashboard tab content converted to the design-system shared classes. The Performance-tab faction-snapshot table re-roots onto the shared table chrome with each Wilson 95% CI cell rendered via a NEW design-system primitive — `.ia-ci-cell` — the confidence-interval cell becomes a first-class shared component (it recurs across Performance Pull, Strength Index, and the CI Explorer, all to follow). The Practice-tab tier table moves to the shared table chrome. The Editorial tab's three side-by-side cards become design-system flat cards (sharp corners, accent reserved for the OPEN-link affordance). Page-local CSS for all three surfaces gone. The data globals, the adapters, the sort + render JS, and the Chart.js handlers are unchanged. With this branch the Quantum surface conversion is COMPLETE across the chrome (status bar / tabs / Vital Signs), panels (ELO / coverage feed), and tab content (Performance / Practice / Editorial) — three feature branches, the dashboard now matches the design system end-to-end.

Quantum dashboard panels converted to the design-system shared classes. The ELO leaderboard table re-roots onto the Matrix kit's sortable table chrome — mono numerals, top-three ranks in emerald, the provisional-rating badge as a design-system chip — with the client-side renderer emitting the new cell classes. The Performance-tab coverage feed becomes the shared dispatch-list chrome with date stamp on the left, source / author dek in the middle, and an emerald accent on video / podcast media tags. Page-local CSS for both surfaces gone. The data globals (window.__quantumElo / window.__quantumPractice), the adapters, the Chart.js handlers, the sort logic, and the hash-deep-linked tab controller are all untouched.

Quantum dashboard converted to the design-system surface. The hero now uses the Orbitron display face, the four-cell status bar is the design system's ONE glass surface (backdrop blur, sticky to viewport top, fresh / aging / stale chip on the data-freshness cell), the seven-tab strip carries the emerald active-tab underline, the Reference tab's Vital Signs panel uses the shared per-column freshness grid, and the four By-The-Numbers tiles tick up on enter-viewport via the Reanimation counter. The cube easter-egg, the Practice / ELO client-rendered widgets, the Chart.js handlers, and the meta-verdict pipeline are all untouched. First of several surface-by-surface conversions; the design-system shared classes are already in place for the rest.

Visual redesign — Phase 3 (the redesign is complete). The three UI kits ship as shared classes in the stylesheet: Quantum surfaces (matchup matrix with the seven-stop diverging heatmap, Vital Signs freshness panel, dispatch list), the Awakenings dispatch reader (Inter prose with the emerald drop-cap, display-face pull-quote, stat callout, sticky TOC sidebar, reading-progress bar), and the Matrix leaderboard family (sortable ELO table, RAW / THEORY-ADJ / PERF-ADJ segmented control, provisional banner, mover lists, skill-premium cards, Captain Decision retrospective card). Two shared inline motion controllers — the Reanimation counter (numbers tick up to value on enter-viewport, prefers-reduced-motion honoured) and the reading-progress bar — load site-wide; both are data-free and the eight long-form pages opt into the progress bar today. The Phase 1 alias bridge is now deleted; the stylesheet reads exclusively against the new design-system tokens.

Visual redesign — Phase 2. The shared component layer migrates to the design system: sharp-cornered chrome with border-driven hierarchy, the two-line nav with an emerald left-edge mark on the active section and a top-edge ignite on hover, the four-way theme control as monolithic display-face buttons, and the full design-system component class library (cards, tables, buttons, chips, fields, tabs, faction rows, tier buckets, status bar, footer) added to the stylesheet. Every alias-bridge reference across the site rewrites onto the real new tokens; the bridge is kept as a safety net for the final Phase 3 cleanup.

Visual redesign — Phase 1. The token layer is rebased onto the Claude Design handoff bundle: pure-black OLED ground with a single Cryptek-emerald accent, sharp monolithic corners, and the Orbitron display face. Four themes ship — Nebula (the new default) / Graphite / Light (all three sharing the emerald chrome) / Theme (the launch phosphor data-terminal, retained and re-expressed in the new token system, scanline overlay kept). An alias bridge keeps every existing rule rendering correctly in the new palette while components migrate in Phases 2 and 3.

Landing 'Live from the Archive' band now reads as two side-by-side panels — the Strength Index tier list pulled out as a static panel on the left (always visible the moment the page loads, no longer something a visitor must wait for the cycle to land on), with the auto-cycling viewer narrowed to its four short slides (Win Rates, the latest Captain Decision, ELO, the latest dispatch) on the right. The dead space the tall tier list was forcing on the short slides collapses; labels and values sit together rather than stretched to opposite edges of the page.

Theme system overhauled. Graphite (slate-grey / indigo) is the new default; the phosphor-green data-terminal aesthetic moves to a dedicated 'Theme' option and renders byte-identical to before; Light is rewritten in a blue-accent palette; a fourth option, Nebula (near-black / purple), joins the toggle. Heading and body fonts split — DM Sans for headings, Inter for body across the three proportional themes; 'Theme' keeps its monospace stack. Reader is retired; a saved Reader preference falls back to Graphite, a saved phosphor selection lands on the relabelled 'Theme'.

Landing: replaced the static Strength Index teaser with an auto-cycling highlight viewer — five live slides (faction tiers, top win rates, the latest Captain Decision, the top of the ELO leaderboard, the latest dispatch) that rotate every four seconds with hover-to-pause and reduced-motion respected. Section buttons compressed: title-only by default, full description revealed on hover or keyboard focus.

Landing-page redesign: a plain-language headline answers ‘what is this site’ in five seconds, the six section cards lead with their functional descriptors (Dashboard / This Week / Methodology / Factions / Teams / Directory), the site-wide nav flips to plain-first across every page, and a new band on the landing shows the full Strength Index as a compact STRONGER / AVERAGE / WEAKER tier strip.

Damage Calculator: Phase B.3 — Blast (attacks scale with target model count, with the 11+ rule resolving a random Attacks characteristic to its maximum), Rapid Fire X (gated by an at-half-range toggle), and Twin-linked (a page-side alias for re-roll failed wounds). Closes the in-scope Phase B keyword set.

Damage Calculator: Phase B.2 — the critical-interaction keywords (Sustained Hits, Lethal Hits, Devastating Wounds, Anti-X). The engine now tracks the critical-roll streams explicitly: Devastating Wounds produces mortal damage that bypasses the save and spills within the unit; Lethal Hits auto-wounds on a crit; Sustained Hits adds X normal hits on a crit; Anti-X lowers the wound crit threshold per profile.

Damage Calculator: Phase B.1 — multiple weapon profiles into one shared defender, the non-critical keywords (±1 hit/wound modifiers, the four standard re-rolls, −1 damage, Feel No Pain), and a distribution-aware points-efficiency view with a points-trade probability.

Published the Changelog — a reader-facing log of what changes on the site, including any corrections to published work.

Published The Cover Ledger — the durability accounting of the hypothetical 11th-edition cover change, archetype by archetype and weapon by weapon.

Pairings Tool: added a calibration overlay that lays the Archive's first-party matchup baseline under your matrix and flags where your read departs from the measured data, with a five-state marker including a foggy-data 'inconclusive' state.

Pairings Tool: lifted the Blind / WTC protocol toggle into the top toolbar, added saved matrices (persisted to your browser), and a printable scenario report.

Added a small Archivist data-stream easter egg, hidden behind a triple-click on the Quantum dashboard's status-bar cube.

Added Reader — a third site theme — alongside Dark and Light, with a three-way segmented control in the header.

Published Where Cover Lives — the first Editor-voice feature long-read, a speculative column on a hypothetical 11th-edition cover-rules change.

Published Awakenings dispatch #2 (The Baseline Week, W21) and folded the Chronicle section into Awakenings so dispatches and retrospectives surface together.

Wayfinder: full pre-launch link audit (28 removals, 26 URL fixes, 1 rename), plus six new directory entries. The VISIT button now properly links out.

Tournament-coverage feed catch-up: 45 curated competitive-40k videos added alongside the existing written coverage.

The Archive flipped from pre-launch posture to public launch — stale 'pre-launch' labels removed, data-provenance lines added across the dashboard.

Quantum: added the ELO tab — a provisional three-rating system (Raw, Theory-adjusted, Performance-adjusted) for the competitive scene the Archive tracks.

Added the Damage Distribution Calculator — type a Tier-1 attack-and-target profile and see the exact full probability distribution of total damage and models destroyed.

Published the four methodology articles — The Big Soup Problem, Six Bins, Swiss Isn't Random, and On Average — alongside their interactive Protocol tools.

Restructured the landing page and the Quantum dashboard, added an About page (manifesto + Editor disclosure), and gave Quantum a slim hero, a status bar, and seven themed tabs.

First Attention snapshot landed (2026-W20) — Google Trends and YouTube interest signals for all 29 factions, feeding the Quantum dashboard's Attention tab.

Captain Decision Archive: published the first full event — CHTT 2026 R1-R5, including the Blunt Force Gambit Round-5 entry as the lead case study.

Added the Performance Pull — the Archive's own first-party tournament aggregation, including a 28×28 matchup matrix and a weekly / current-patch / edition multi-view toggle.

Faction Dossiers: 29 per-faction pages with Unit Lab, archetype classification, and the five-axis radar that compares each faction's profile against the population.

Protocol launch — the methodology lab. Five interactive tools (CI Explorer, Two-Proportion Test, Power Analysis, Damage Distribution, Methodology Library directory) plus the methodology-article register.

Glossary published — 61 entries covering 40k-scene vocabulary, scoring conventions, and the statistical concepts the methodology pieces use.

Site Phase 1 complete: shared base layout, the six section landings (Quantum, Awakenings, Protocol, Dossiers, Matrix, Wayfinder), and the phosphor data-terminal design language.

Initial repository scaffold — the Pairings Tool, the WTC-solver reference implementation, and the early dispatch-page convention that the rest of the site grew from.