Phase 3 — Landing pages
The section landings — Home /, /fleet, /inventory, /team — designed as 30 ft views: the shape of each domain and the doors into it, image over text. Visual bake-offs first (Team → Inventory → Fleet → Home), then composed pages in every state — 12 pages so far.
Cycle: c100 Landing Pages · ← Phase 2
Composed mocks
The landings in the qdash shell, composed from the experiments below — one world, the MOCK bar drives it; show parts names each slot's source.
The Home landing composed from the Phase 3 bake-offs, in the qdash shell — showroom + Seen as lenses; show parts to see each slot's source and swap variants
The /inventory landing composed from bake-offs #3–#4 in the qdash shell
The /team landing composed from bake-offs #1–#2 (and the tile-map / matrix / lever-strip variations) in the qdash shell
Every part promoted out of the Phase 3 bake-offs, rendered live on the shared world with its source — what the mocks are composed from
Page redesigns
Existing qdash pages re-drawn on the patterns the landings and listings settled — each its own cycle, ported to the Fable app once confirmed.
The c109 index (DataIndexSection + DataMachineMicro, 35 machines) tried as cards: each micro on white with rounded corners and a soft shadow, smaller data (13 px titles, 16 px values, 6 px bars, 12 px sparks), and the whole section on white — presets A today · B micro cards · C micro cards small data · D section card · E section card with sunken tiles, every knob live (card, radius, padding, title size / weight / tone, hairline, value, segment, spark, roll-down form, gaps, column width), plus every preset side by side
c105 — /cost re-drawn on the Organization / Inventory listing pattern: Spend · Pace · Rate band, one ledger pivoted by lens (team, user, repo, model, provider, harness, account, token kind) with the Team quick filter, costliest-session receipts, ceilings; design notes on what is knowable
Experiments — bake-offs & variations
Every team at once, through one lens — four forms on one fixture world: A tile map (area = people, fill = lens), B the c67 lever strip, C swarm rows (a dot per person), D team × lens matrix; eight lenses (AI usage · Spend · Active · Violations · Coverage · Harness incl. Claude Code · Waste · Retention), scale 20 / 200 / 1,500, attention / all-clear / first-run, period pill obeyed; click a team to focus it in every form
Bake-off #1's tile map pushed three ways — ink (greyscale, grey + status, a designer-picked hue from swatches or a free picker, stepped ramps), size (hero → card → DataMachine → table cell → 48 px glyph in a sentence) with six labelling strategies at DataMachine size (silent + note, numbered key, callouts, highlight, other bucket, paired list), and a spin (squarified / ordered strips / equal tiles / one-row bar, area by people or by the lens, solid / level / dot fills, nested people, eight-lens small multiples, morph); linked hover across every map
Bake-off #1's matrix pushed the tile-map thread's three ways — ink (grey = one picture vs per-job ink = eight charts, designer-picked hue, stepped ramps), size (table → heatmap → DataMachine texture → a one-row team fingerprint in a sentence and a listing column) with six labelling strategies at DataMachine size (silent + note, label the short axis by transposing, column key, highlight + callouts, other bucket, live readout), and a spin (normalisation per column / row / rank / deviation-from-median, Bertin ordering incl. seriation, fill / dot / bar / numeral cells, row height = people, reader groups + focus column, morph); linked crosshair across every matrix
Bake-off #1's c67 lever strip given the same treatment — ink (greyscale with a shape code for severity, designer-picked hue), size (every chip → one-row rail → top four → DataMachine rows → one pill in a table cell → the micro-form in a sentence), six selection strategies at DataMachine size (top 3 + more, worst · median · best, names only, micro only, lever verbs, ticker), and a spin (card / pill / row / tile chips, a micro-form picker, neutral / tinted / status-edge grounds, sort + pinned teams, multi-lens chips per reader, department groups, FLIP morph); lens selects (UxSelectInline) on the variations and in every DataMachine title row
Now is done well; this is the history beside it — §0 the Teams page as c103 Part A shipped it (Spend rolls down every team, Spend trend column), §1 a catalogue of 22 trend forms (number · delta · sparklines · qdash Spark · prior-window ghost · peer band · micro-bars · bar + tick · slope · heat strip · horizon · stacked area · 100 % weeks · small multiples · bump · nested harness ink · total ⇄ spark · breakdown + chart pair) plus Activity's history row (2026-10-01), §2 seven roll-downs with history, §3 table cells, §4 breakdown over time (team → user · harness type → install · provider → model), §5 toggles and title switchers, §6 a degradation lab (every form × S1–S6 + free W × H, three auto-checks per cell), §7 a keep / maybe / kill ledger; worlds = a dated seed snapshot (real Meridian per-team cost from :7882) + 20 / 200 / 1,500, window follows the pill or a fixed 90d, gaps drawn as gaps
Every person as one mark, grouped by team, toned by their own state (never the team's worst) — A team blocks (waffle mosaic), B team rows (c67 barcode), C one org field (c68 pixel grid with real geography), D coverage skyline, E the field in greyscale with the lifted "needs you" list (pawns, two-way hover); adds the identity-known, never-seen seat-holder; scale 20 / 200 / 1,500, attention / all-clear / first-run, period pill obeyed
What is our AI made of, and how concentrated is it? Four heroes on one world — 1 alluvial harness → provider → model (unseen installs as dashed stubs, Hermes as a dashed ribbon, personal accounts hatched), 2 provider bowtie (no crossings), 3 switchboard matrix (harness × provider dots, rings for personal accounts), 4 pixel body (Mark's figure, one pixel per asset; proposes feet = providers); lens tokens · spend · installs
Four supporting cards, each best-prior vs fresh: Visibility (per-harness pixel rows vs ghost pull-out), Provenance (Configured-vs-Connected vs two-witness strip), Arrivals (week strip vs swimlane beeswarm), Accounts (ledger + personal standout vs per-provider split of people with a ghost block affordance)
The factory floor drawn eight ways on one world — A floor plan (zones = teams, stations, machines; lit / dim / ghost, wall mode), B qmap isometric (team rugs; limits stated, see c101), C pixel field, D production line (teams → harnesses → providers → data out), E loom (teams × harnesses), F shift clocks, G pawn lineup, H control-room wall; lens activity · spend · violations · coverage · harness; deep-linkable ?scale=&state=&lens=&period=&harness=
Who runs what (teams × harnesses matrix + the full Claude Code lens), shifts (team × hour heat strip and punchcard), and the loading dock (floor → AI providers, personal accounts, web destinations; width = volume, sensitive data classes toned)
Home as a showroom floor of six exhibits (floor, estate, org, where are we, what good looks like, proof), each a portal, re-lit by Seen as: Security · Compliance · Engineering · Finance · IT (?as=) — lens verdict + ≤5-row Needs-you queue, the same shapes re-toned, one c67 graduate per lens (governance ledger, runway board, hygiene block, ramp panel); six-lens thumbnail strip, lens ledger
Nine sketches past today's data (X1–X9): work-type mix, seats bought vs active, data class × destination, approval lifecycle, spend forecast cone, adoption curves, before/after floor, evidence shelf, version drift — each with the data it needs, where it would live, and a keep / maybe / kill read