The classification sheet for race two of a historic motor-racing meeting, stopped by a red flag with four laps to run.
The third brand specimen. The reference is a car maker's own site; this is what its language looks like on a results sheet somebody reads standing up in a paddock, looking for one car.
Where the style comes from
Transcribed from awesome-design-md/design-md/ferrari/DESIGN.md, read 2026-09-16, and emitted by:
python3 tools/brand_tokens.py ferrari-brand > systems/ferrari-brand/tokens.css
python3 tools/brand_tokens.py ferrari-brand --check
python3 tools/brand_tokens.py --self-test
| Decision | Transcribed | |
|---|---|---|
| Ramp | #ffffff #f7f7f7 #ebebeb #d2d2d2 #969696 #8f8f8f #666666 #303030 #181818 | |
| Chromatic | one | #da291c, with published hover #9d2211 and active #b01e0a |
| Semantics | info #4c98b9, success #03904a, warning #f13a2c | |
| Type | 80 · 56 · 36 · 26 · 18 · 16 · 14 · 13 · 12 · 11 px | |
| Tracking | negative on every display step, positive on the 11px label | |
| Weights | 400 and 500 | |
| Radii | 0 · 2 · 4 · 6 · 8 · 12 · 9999 | |
| Spacing | 4 → 128px | |
| Nav | 64px, flat | |
| Depth | a ground change, canvas to elevated. No shadow ladder | |
| Modes | dark first, with a full published light palette beside it | |
| Motion | not published |
It is the only DARK-FIRST brand in this family, so dark is declared first and lands on bare :root, the reference's own default rather than this portfolio's convention.
The finding that shaped the page. The brand red reads 4.87:1 under white, so the reference's own primary button is legal as published. As TEXT it is a different story: it reaches 4.55:1 on the light soft surface and only 2.71:1 on the elevated grey in the dark setting, so accent-text is solved a little lighter there. The page carries both, the fill on the button, the solved red on a link, and the design-notes panel says why they differ rather than hoping nobody looks.
The direction
- Archetype. Editorial spread.
- Data stance. Table-first.
- Density. Generous, a row takes twice the height a console would give it.
- The one bold decision. The table carries two orders at once. Every results sheet publishes the classification and lets the crowd work out for themselves why the car they watched lead is third. This one gives what each car was doing when the flag came out its own column, in words, "was second", with an arrow, so the sheet answers the question instead of starting it. It is a column the timekeeper already has and never publishes.
The second state
The red flag. A results table's whole promise is that the order is the order, and this is the case where it is not: the race stopped on lap 22 of 26, the classification reverts to the end of lap 21, and three of twenty-four cars are classified behind cars they had already passed. A table that showed only the final column would be correct and would start a protest.
Nothing about the moved places rests on colour, the words and an arrow glyph carry it, so the sheet survives a photocopier, a grey printout and a screen reader. That constraint came from the brand rather than from taste: this language allows one chromatic colour, and it is spent on the action.
Deviations from CONTRACT.md
- §2, the typeface. FerrariSans is proprietary. Archivo is substituted : the closest open-licensed grotesque that holds at weight 500 across an 80px-to-11px ramp with the negative tracking this language sets on display.
- §5, motion. The reference publishes no duration, easing or transition anywhere. The two rungs in
tokens.cssare the portfolio's, and the emitted file states that on its face. - The reference is photography-led and this page has no photographs. CONTRACT.md §2 admits inline SVG and
data:URIs only. What is carried across is the other half of the identity: the display step at full size with its tracking closed, one red rule under it, and everything else left as space.
The uplift round (2026-10-29)
What changed and why. The archetype and stance are unchanged; the first fold and the second state were rebuilt.
- Masthead. The headline now sits beside a drawn, unlabelled copy of the lap chart (eight traces running into a red flag and a chequered strip), so the story is told by shape before any word is read. Under it, a spec strip in the maker's own idiom: four numbers set at the display-mega step.
- A lap chart drawn by hand in SVG. Position by lap for the top eight, laps 0 to 22, with lap 22 hatched as the lap that does not count and laps 23 to 26 left empty. Labelled at the line ends, no legend; the four cars that moved are the heavy lines. A second, narrower drawing (laps 12 to 22) replaces it below 56rem so the type stays legible on a phone. The lap-by-lap orders were drawn by a throwaway script and are baked into the SVG; they agree with the classification: the order at the end of lap 21 is the table, and the road order at the flag is the "was third" column.
- Two orders, one switch. A sticky segmented control (native radios) flips the page between the official result and the order as the grandstand saw it. Table rows are reordered in the document and flown to their new place (FLIP, one
--fr-duration-basex 1.5); the chart's labels glide with them; the last column and the caption change their wording. Reduced motion keeps a short opacity pulse instead of the move. - Pit-wall table. Race-number roundels (a pre-66 touch) that are buttons: select one and its trace is the only one left in the chart. Interval and lap columns added; every figure is tabular and checks against its neighbour (each interval is the difference of two gaps).
- Ruling band. The red flag statement is the maker's livery band, full bleed.
- Publish as provisional now does something: it stamps the post time and starts the protest window (the clock on the sheet is fixed at 14:52).
- Buttons are sharp (0px), as the maker's CTA is. They were 4px.
- Corrected a number. The sheet said three cars were affected, but four places change (two passes). It now says two passes erased, four cars moved. The protest window moved with the posting time, so 15:07 no longer appears.
- Design notes set as a spread with two contrast plates; the white-on-red pair and the red word on the page ground, measured with
contrast.py. - Nav now points at four real sections; the old name-drift exception in
copy-lint.ignoreis no longer needed but is left in place. - boson on this page. Run against
index.htmlalone, boson copies the file withouttokens.css, so everyvar(--fr-*)is undefined and the contrast and keyboard probes report false findings (no focus ring, black text). Run against a copy with the tokens inlined, the page is clean except the two warnings below.
Findings kept, with reasons
slop_scan·default-ai-palette, near-black plus one bright accent. The rule allows it where the brief asked for it, and here the brief is the brand.craft_lint·disabled-unstyled, the rule looks for utility-class variants; the disabled control is styled by.btn[disabled]with a changed background, a changed label colour andcursor: not-allowed, and it carriesaria-describedbynaming what would make it live.audit_page·apple-touch-icon-missing,manifest-missing: portfolio-wide.copy_lint·name-drift, seecopy-lint.ignore.- boson
text-too-small(warning): the maker's published label step is 11px uppercase with positive tracking; every.labeland nav item uses it. It is the brand's own size and clears 4.5:1, so it is kept rather than raised. - boson
share-image-missing,no-404(launch_check): portfolio-wide.
Round four
- The masthead's lap chart is now a Rosso field: the maker's red as a ground, the chart reversed out in white. It is drawn in script to the width of its own panel, so type stays at reading size at 390, 1024 and 1440 and nothing crops. Four cars carry direct labels (21 Vance, 3 Larssen, 2 Whitmore, 18 Brill, placed at where they ran when the flag fell), positions 1 to 8 stand on the left, laps 10, 15, 21, 22 and 26 run along the foot, and the red flag is named at the pole. The data is the same eight cars the chart below it draws.
- The red now arrives in the first screen at every width. The livery band lower down keeps its role as the ruling itself; the field above it is the same colour doing a different job.
- At 390 the header is one row (wordmark, theme, Portfolio). The four section links moved to a bar at the foot of the screen, where a thumb reaches, with the two long names shortened to Laps and Results there only.
Round eight
- Headline slid under the sticky Show bar (1440, grandstand). Real. The bar was the first child of the stage, above the chart section, so the chart's headline and lede had to scroll through it. The chart section is now two parts with the bar between them: the headline and lede above it, the drawn chart below it. The bar still sticks across the chart and the classification, which is why it sticks, but nothing above it can pass under it. The skip link and the nav anchor still land on the headline.
- Design notes a plain list (1440 light). Real. "The one bold decision" is now a red slab across the entries, white on the red fill (the same 4.87:1 the notes measure), cut from the same livery the ruling above it is. It is one block, not a grid of cards, and it is the sheet's own red doing its one job, pole position. Dark theme checked: the slab is the same red.
- Declined: nothing.
Round nine
Scored before and after on the ten heuristics (0 to 4): 31 before, 34 after. Findings, severity-sorted.
- Sev 2, tracing a car answered somewhere you could not see. Selecting a roundel faded the other traces in a chart a screen or more above the table. The selected row now opens its own lap-by-lap line directly under it (grid to lap 21, the dashed stroke to the flag position), with a sentence built from the same drawing: where it started, laps led, where it was classified, where it was running. It survives the Show switch, because the row is re-seated after the reorder, and it is announced.
- Sev 2, the key column scrolled away between 640 and 740px. The table was 46rem wide with a 672px well, so "On the road at the flag" sat in an inner scroll, clipped, at 200% zoom on a laptop and on a small tablet. Laps and interval now drop below 56rem; the table never scrolls at any width from 320 to 1920 (checked in 10px steps, both schemes, trace open).
- Sev 2, the phone header read as one jumbled phrase. "On the road at the flag" wrapped to three lines against "Gap / lap 21". Below 40rem it is "At the flag" (and "Official" in the other order), aligned to the cells under it.
- Sev 1, "Measured" said the word red was lighter than the fill in both themes. It is only lighter on the dark canvas. The sentence is now true in either.
- Sev 1, the gap columns looked wrong in the grandstand order. The caption now says gap and interval stay as they stood at the end of lap 21.
- Sev 1, "As the grandstand saw it" broke to an orphaned "it" at 360px. Below 25rem the option reads "Grandstand order".
- Wide screen. At 100rem and up the sheet runs to 1520px instead of leaving 320px margins either side, as the maker's own pages run to the edge of the glass. Below that nothing changed.
- Footer. A single grey line at the left edge became a colophon: the wordmark with the red rule, the four sections, and a 44px return link.
- Taste. none.
Left: the 11px label and chart axis size (brand's own step, kept as before); the narrow lap chart at 720 to 895px stays left-aligned at 30rem rather than filling the well (sev 1).
Round ten
A second, cold read against the ten heuristics, scored 28 before and 33 after (round nine had scored 34; the harder read found a severity-3 and four severity-2 defects round nine passed).
- Sev 3, a commit with no way back. "Publish as provisional" posted the sheet and disabled itself: no undo, no statement of what it does, focus dropped to the page. It now says beside the button what it posts and that it can be withdrawn; after posting, a "Withdraw posting" button takes its place and receives focus, and withdrawing returns the sheet to draft and puts focus back on Publish.
- Sev 3, Publish sat under a table that might not be the result. In the grandstand order the button looked as if it posted what was on screen. The note beside it now says it posts the official result, end of lap 21, in either order.
- Sev 2, the grandstand order did not say it was not the result on a phone. The sticky bar's note is hidden below 52rem. The table head now carries a persistent "Not the result" tag in that order.
- Sev 2, one fact two ways. The table head said "provisional" while the ruling said "Draft, not yet posted". Both now read from one state and change together on post and withdraw, along with the protest window row ("Open from 14:52 to 15:22") and the hint under Declare final.
- Sev 2, the posted line landed below the buttons, last in its panel. It now sits above them, with a red rule, and is announced.
- Sev 2, the ruling led with the cause. The sentence that decides the classification is now first; the live rows (result taken at, cars moved, status, protest window) come before "Stopped".
- Sev 2, help away from the figure. "Int" had no explanation. The header reads "Interval", and a key under the table says what gap and interval are and that positions 9 to 24 are the same in both orders. The "select a car number" instruction left the end of the caption for its own line above the table. The chart caption says the four heavy lines are the cars that changed place.
Inventions, none contradicting the brief: a posting can be withdrawn while the protest window is open and withdrawing only returns the sheet to draft and cancels the window; positions 9 to 24 are unchanged (derived from "4 of 24 moved", with all four in the top eight). Nothing else was added to the sheet's facts.
Left: Declare final never becomes available because the clock is fixed at 14:52 (sev 1, the brief's own fixed clock); the table still shows eight of twenty-four cars (sev 1, no further names invented); the 11px label size (brand's own step).
Round eleven (2026-11-10)
- Sev 2, a reader looking for one car could find 8 of 24. Reproduced at 1440 light, 390 dark, 360 light and 1920 dark: eight rows, no search or jump, and the ruling named car 14, which is nowhere on the page. This brief gives facts for the eight listed cars only (and that car 14 stopped at Colwyn), so the other sixteen were not invented. The line above the table now says plainly that the sheet lists 8 of the 24 cars and names them (7, 3, 21, 9, 18, 2, 31 and 11), and that the other sixteen are classified 9th to 24th and are not listed. The ruling now reads "Car 14, which is not one of the eight cars in the classification, stopped at Colwyn ...", and the key under the table says positions 9 to 24 are not listed.
- Invention. None beyond restating the brief and round ten: the other sixteen are classified 9th to 24th (24 cars, top eight listed) and car 14 is one of them (it is not in the eight). No search or jump was built because there is nothing to find.