A searchable catalog of aesthetic visual directions for web interfaces, each documented with a live specimen, a copyable agent prompt, and a copyable Markdown specification.
MinimalistOverviewIndexAbout
minimalism
Minimalist
Reduces an interface to only what is functionally necessary, using negative space and a restrained palette so content and hierarchy carry the design instead of decoration.
Primary actionSecondary
Reduction over addition
Minimalist is a web-interface design style defined by extreme reduction of visual elements to only what is functionally necessary. Hierarchy, spacing, and typography — not color, ornament, or decoration — carry the interface. Negative space is treated as an active structural element rather than unused margin.
RestActive
Catalog entry 01
Minimalist
Tags: minimalism · whitespace · restraint
BrutalistOverviewIndexAbout
brutalism
Brutalist
Foregrounds raw, unornamented structure with thick borders, square corners, and stark color blocking, rejecting decorative smoothing to communicate directness.
Primary actionSecondary
Structure over polishHonest materialsConfrontational hierarchyDeliberate rawness
Catalog entry 02
Brutalist
Tags: brutalism · raw · high-contrast
EditorialOverviewIndexAbout
editorial
Editorial
Borrows print-publication conventions — a serif and sans typographic pairing, a structured column grid, and imagery treated as content — to produce a considered, authored reading experience.
Primary actionSecondary
Typographic hierarchy as structure
A serif (or otherwise distinctive) display typeface for headlines, paired with a clearly contrasting sans-serif for body and interface text.
A defined column grid with controlled line length for body copy.
Captions or credit lines set visually apart from body copy, adjacent to their images.
Full-bleed or art-directed imagery integrated into the layout rather than treated as background filler.
Type contrast, not color or containers, is the primary organizing device.
Typographic hierarchy as structure
Editorial
Editorial is a web-interface design style modeled on print-publication design. It builds a clear typographic hierarchy from a serif-and-sans (or comparably contrasting) type pairing, structures content on a considered reading grid, treats imagery as narrative content rather than decoration, and paces dense and open sections deliberately, producing an interface that reads as authored rather than assembled.
Catalog entry 03
Editorial
Tags: editorial · typography · magazine-layout
SwissOverviewIndexAbout
grid
Swiss
Organizes an interface on a strict modular grid with a single neutral sans-serif, expressing hierarchy through scale and position alone so the underlying structure stays visible.
Primary actionSecondary
The grid precedes the compositionOne typographic voiceHierarchy without color
Swiss
Swiss is a web-interface design style in which a declared modular grid governs every placement decision and remains legible in the finished result. A single neutral sans-serif carries the entire typographic hierarchy through size, weight, and position; color plays no structural role. The style is defined by the presence of a system, not by the quantity of content, so a dense page and a sparse page can both be Swiss.
Catalog entry 04
Swiss
Tags: grid · typography · sans-serif
maximalism
Abundance is designed, not accumulatedDepth through overlapExtreme typographic rangeThe surface is filledDensity never costs legibility
Maximalist
Abundance is designed, not accumulated
Maximalist is a web-interface design style that treats abundance as the organizing principle. It composes with a large saturated palette, genuinely overlapping layers, extreme typographic range, and a filled rather than empty background, producing a dense and energetic surface. It is distinguished from visual noise by two requirements it never relaxes: a single dominant focal point, and legible text everywhere.
Catalog entry 05
Maximalist
Tags: maximalism · layering · saturated-color
GlassmorphismOverviewIndexAbout
glassmorphism
Glassmorphism
Builds surfaces from translucent, blurred panels that let a colored background show through, using light edges and depth ordering to separate layers.
Primary actionSecondary
The background is the material
Glassmorphism is a web-interface design style in which surfaces behave like frosted glass: partially transparent, blurring the content behind them, and edged with a thin band of light. Layer order is communicated by seeing through one surface to another. The style depends on a background with real chromatic variation, and it is only shippable when text contrast is verified against the worst-case background pixel rather than against the panel's nominal fill.
RestActive
Catalog entry 06
Glassmorphism
Tags: glassmorphism · translucency · blur
NeubrutalismOverviewIndexAbout
neubrutalism
Neubrutalism
Takes brutalism's raw geometry and makes it playful — thick dark outlines, hard unblurred shadows, and bright saturated blocks set with deliberate, cheerful imbalance.
Primary actionSecondary
Outline defines the objectShadow is structure, not lightFlat saturated color
01Thick dark outlines on every distinct object, including images and inputs.02Hard drop shadows with no blur, offset consistently in one direction.03Bright, flat, saturated block fills with no gradient anywhere.
Arranges content as a tight mosaic of rounded tiles at varying spans, each holding exactly one idea, so a dense overview stays scannable.
Primary actionSecondary
One idea per compartmentSize communicates importanceThe mosaic is fullExplicit placementTiles stand alone
Catalog entry 08
Bento
Tags: bento · grid · modular
NeumorphismOverviewIndexAbout
neumorphism
Neumorphism
Extrudes controls from the background itself using paired light and dark shadows, so elements look pressed into or raised out of a single continuous surface.
Primary actionSecondary
One continuous material
Neumorphism is a web-interface design style in which elements appear extruded from, or pressed into, a single continuous surface. The element and its container share one fill color, and separation is produced entirely by a pair of opposing shadows — one lighter than the surface, one darker — with no borders present. Because depth is the sole visual separator, the style is only shippable when contrast requirements are satisfied independently of the shadows.
RestActive
Catalog entry 09
Neumorphism
Tags: neumorphism · soft-ui · depth
FlatOverviewIndexAbout
flat-design
Flat
Removes every simulation of depth and lets solid color, geometric shape, and clear type carry the interface, using color itself as the primary structural device.
Primary actionSecondary
No simulated depthColor is structureGeometry over rendering
01Solid color regions with no gradient or texture anywhere.02A palette of several saturated colors distinguishing regions and states.03Two-tone geometric icons without shading.
Catalog entry 10
Flat
Tags: flat-design · solid-color · geometric
Y2KOverviewIndexAbout
y2k
Y2K
Revives the glossy techno-optimism of the turn of the millennium: chrome surfaces, inflated type, specular highlights, and a palette that looks lit from inside.
Primary actionSecondary
Surfaces are manufactured objects
Y2K is a web-interface design style that revives the glossy consumer-technology optimism of the turn of the millennium. Surfaces read as polished and reflective: multi-stop chrome fills, capsule shapes carrying specular highlights, high-luminance saturated color, heavy display typography, and sparkle ornament. Its visual reference is early-2000s product photography rather than any screen convention.
Catalog entry 11
Y2K
Tags: y2k · chrome · retro-futurism
01One-pixel hairline guides at low contrast, connecting or framing elements.02Uppercase monospaced labels at small sizes with positive letter spacing.03A tick scale, numbered index, or grid reference along an edge.
blueprint
Technical Blueprint
Presents an interface as an engineering drawing — hairline guides, monospaced micro-labels, coordinate marks, and line diagrams instead of photography.
Primary actionSecondary
The drawing is the metaphorOrnament is quiet, information is loudMeasurement is visibleInk, paper, and one revision colorDecoration never impersonates data
Technical Blueprint
Technical Blueprint is a web-interface design style that presents an interface as an engineering document. Hairline guides, uppercase monospaced micro-labels, a visible coordinate or tick system, line-drawn illustration, and a palette of two neutrals plus one accent combine to produce a register of precision and measurement. Because the style is built from labels and figures, it carries a specific obligation: ornament must never present itself as data.
Catalog entry 12
Technical Blueprint
Tags: blueprint · monospace · diagrammatic
HandcraftedOverviewIndexAbout
handmade
Handcrafted
Signals that a person made this — variable hand-drawn strokes, material texture, slight rotations, and unpolished imagery set against otherwise disciplined structure.
Primary actionSecondary
Imperfection is evidence, not error
Hand-drawn underlines, arrows, circles, or brackets with varying stroke thickness.
Paper, ink, or fabric texture visible behind the content at normal viewing size.
Irregularity signals that a person made specific decisions.
Imperfection is evidence, not error
Handcrafted
Handcrafted is a web-interface design style that adds visible evidence of human authorship to an otherwise disciplined interface: strokes of variable thickness, material texture, small deliberate rotations, and unpolished imagery. Its defining constraint is that imperfection is confined to the surface. Interactive mechanisms — controls, hit areas, focus indicators — remain fully regular, which is what separates the style from carelessness.
Catalog entry 13
Handcrafted
Tags: handmade · texture · imperfection
Dark ModeOverviewIndexAbout
dark-mode
Dark Mode
Builds the entire interface on a dark ground, expressing elevation through luminance steps rather than shadow and desaturating color so it survives on a low-light surface.
Primary actionSecondary
Elevation is luminance
Dark Mode is a web-interface design style built on a dark ground, in which elevation is expressed by luminance steps rather than by shadow, color is desaturated to survive on a low-light surface, and contrast is targeted at a strong rather than maximal ratio. It is not an inversion of a light interface: the mechanisms that establish depth, color, and legibility all differ.
RestActive
Catalog entry 14
Dark Mode
Tags: dark-mode · low-light · elevation
Barely-There UIOverviewIndexAbout
hyperminimal
Barely-There UI
Reduces the interface almost to its data: light type weights, boundaries near the threshold of perception, and vast air, with charts doing the visual work.
Primary actionSecondary
Weight is what gets removed
Barely-There UI is a web-interface design style that reduces the visual weight of an interface rather than the quantity of its contents. Type is set at light weights, container boundaries are drawn just above the threshold of perception, the palette holds to a single hue, and most of the surface is empty — while real data displays carry the visual interest. It differs from reductive styles in that elements are quieted rather than removed, so a data-dense screen can satisfy it.
RestActive
Catalog entry 15
Barely-There UI
Tags: hyperminimal · light-weight · data-dense
RetrowaveOverviewIndexAbout
retrowave
Retrowave
Renders an imagined 1980s future: a perspective grid running to a sunset gradient, neon glow on dark ground, and chrome display type.
Primary actionSecondary
A synthetic landscape
Retrowave is a web-interface design style that renders the future as imagined in the 1980s. A perspective grid recedes toward a horizon, a sunset gradient runs through magenta and orange, neon glow radiates against a near-black ground, and display typography is chromed, extruded, or outlined. The style depends structurally on darkness, since neon is emitted light, and it confines its glow to headlines and ornament so that reading text remains legible.
Catalog entry 16
Retrowave
Tags: retrowave · neon · retro-futurism
cut-paper
The seam is the subjectScale is dislocated deliberatelyLayering is orderedMaterials are mixedText gets clean ground
Collage
The seam is the subject
Collage is a web-interface design style that assembles a composition from visibly heterogeneous fragments — photographic cutouts, illustration, treated type, texture — layered at clashing scales and angles with their seams left exposed. Nothing is blended between materials. Because fragment surfaces are visually unpredictable, running text is placed on its own clean ground rather than over the assembly.
Catalog entry 17
Collage
Tags: collage · cut-paper · layering
constructivism
The diagonal is the structureGeometry carries massSeverity of palette
Constructivist
The diagonal is the structure
Constructivist is a web-interface design style organized around a strong diagonal axis, in which primitive geometric shapes act as structural mass, the palette is restricted to black, white, and one saturated primary, and visual weight is deliberately loaded to one side. Elements bleed past the frame rather than being contained by it. The resulting register is urgent and assertive rather than neutral.
Constructivist
Constructivist is a web-interface design style organized around a strong diagonal axis, in which primitive geometric shapes act as structural mass, the palette is restricted to black, white, and one saturated primary, and visual weight is deliberately loaded to one side. Elements bleed past the frame rather than being contained by it. The resulting register is urgent and assertive rather than neutral.
Catalog entry 18
Constructivist
Tags: constructivism · diagonal · geometric
VaporwaveOverviewIndexAbout
vaporwave
Vaporwave
Recombines discarded consumer-digital material — pastel gradients, obsolete interface chrome, classical statuary, and image artifacts — into a nostalgic, faintly ironic surface.
Primary actionSecondary
Secondhand material
Vaporwave is a web-interface design style assembled from discarded consumer-digital material: pale magenta and cyan gradients, obsolete interface chrome, classical statuary, and visible image artifacts, arranged in still, symmetrical compositions. Its register is nostalgic and faintly ironic, and its material is always secondhand. It is distinguished from neon retro-futurism by being pale, static, and backward-looking rather than dark, kinetic, and speculative.
Catalog entry 19
Vaporwave
Tags: vaporwave · pastel · glitch
Art DecoOverviewIndexAbout
art-deco
Art Deco
Composes with strict axial symmetry, stepped and radiating motifs, and restrained metallic accent over deep ground, reading as machine-age luxury.
Primary actionSecondary
Symmetry is the composition
Bilateral symmetry about the vertical center.
Stepped ziggurat forms, fans, and radiating ray motifs.
The layout mirrors about a vertical axis.
Art Deco
Art Deco is a web-interface design style built on strict axial symmetry and exact geometry, using stepped and radiating motifs, a restrained metallic accent over a deep ground, condensed letterspaced capitals, and rules drawn in groups. It is decorative but never freehand: every arc is a circle section and every angle a clean fraction, which distinguishes it from the organic ornament of the style that preceded it.
Catalog entry 20
Art Deco
Tags: art-deco · symmetry · metallic
Anti-Design
Anti-Design is a web-interface design style that deliberately subverts the conventions of polished interface work: colors clash, typography is uncomfortable, hierarchy is flattened or inverted, and no shared grid governs placement. It is a stance rather than an absence of decisions — each transgression names the convention it breaks. Its defining constraint is that aesthetic convention may be refused while access conditions may not.
Every transgression is namedDiscomfort is the effectHierarchy is a target
Anti-Design
Every transgression is named
Anti-Design is a web-interface design style that deliberately subverts the conventions of polished interface work: colors clash, typography is uncomfortable, hierarchy is flattened or inverted, and no shared grid governs placement. It is a stance rather than an absence of decisions — each transgression names the convention it breaks. Its defining constraint is that aesthetic convention may be refused while access conditions may not.
Anti-DesignOverviewIndexAbout
Catalog entry 21
Anti-Design
Tags: anti-design · subversion · clashing
MemphisOverviewIndexAbout
memphis
Memphis
Scatters flat primitive shapes, squiggles, and confetti pattern in colors that refuse to harmonize, producing cheerful, weightless visual noise.
Primary actionSecondary
Weightless scatteringDeliberate disharmonyFlatness throughout
Memphis
Memphis is a web-interface design style that scatters flat primitive shapes — circles, triangles, squiggles, confetti — across a composition in colors chosen not to harmonize, each outlined heavily in near-black, over a speckled or patterned ground. Nothing has depth and nothing is aligned. The scattered material surrounds the content and never crosses it.
Catalog entry 22
Memphis
Tags: memphis · playful · pattern
EtherealOverviewIndexAbout
ethereal
Ethereal
Holds the interface in high, even light — soft low-saturation gradients, diffuse boundaries, thin large type, and mostly empty space — to read as calm and weightless.
Primary actionSecondary
High-key throughout
Ethereal is a web-interface design style that holds the surface in high, even light: pale low-saturation gradients transition without visible edges, regions separate through tonal shift rather than through lines, type is thin and large, most of the page is empty, and motion is slow. Its register is calm and weightless. Because every characteristic property reduces contrast, the style requires that text and control colors be chosen deliberately dark against the pale ground.
Catalog entry 23
Ethereal
Tags: ethereal · airy · soft-gradient
No styles match “”
Search covers style names, visible tags, and authored aliases. It does not search inside the prompt or the specification text.
Applying a Minimalist visual direction
You are being asked to apply a Minimalist visual direction to a web interface. This document gives you framing and recommendations for that direction. It does not tell you what page, product, or feature to build — apply the guidance below to whatever interface you are already working on, at whatever scope that work requires.
What “Minimalist” means here
Minimalist is not “plain” or “unfinished.” It is a disciplined reduction: every element that remains is there because removing it would cost the user something. The interface communicates through hierarchy, spacing, and typography rather than through color, ornament, or decoration. Negative space is treated as an active structural material, not as leftover room.
Recommendations
Start by listing every element you intend to add, then remove anything that does not serve comprehension, navigation, or the primary action. If you are unsure whether an element earns its place, leave it out and add it back only if a real gap appears.
Limit yourself to one neutral base (near-white or near-black, plus a small range of grays) and at most one accent color, reserved for the single most important action or state on a screen.
Let spacing do the work that borders and boxes would otherwise do. Prefer a generous gap between two sections to a divider line between them; prefer a divider line to a filled container.
Choose one primary typeface family and use weight and size — not color or decoration — to express hierarchy. If a second family is needed for a genuinely different purpose (such as code or tabular data), keep it visually quiet.
Keep interactive elements recognizable through simple, consistent cues (underline, weight, or a single accent color on hover/focus) rather than shape changes, shadows, or animation.
Favor a single clear call to action per screen or section. If a screen seems to need several equally prominent actions, that is usually a sign the content should be split or re-prioritized, not a reason to add visual noise to distinguish them.
Resist the instinct to fill empty space. An area with nothing in it is doing its job if the surrounding content reads more clearly because of it.
Avoid
Multiple competing accent colors, gradients, or decorative background textures.
Drop shadows, glows, or blur effects used for visual interest rather than a specific, necessary purpose (such as a modal that must visually separate from a busy background).
Icons or illustrations added for decoration where text or spacing already communicates the meaning.
More than one font pairing, or more than a small number of distinct type sizes on a single screen.
Dense information layouts that rely on visual crowding rather than clear grouping and whitespace.
How to check your own work
Before considering the result finished, look at it and ask:
Could I remove any single element without losing information or function? If yes, remove it.
Is there more than one color competing for attention besides the neutral base? If yes, reduce it.
Does every remaining element sit in a clear, deliberate position, or did some of it just end up wherever there was room?
If I removed all color, would the hierarchy still be legible through spacing and type alone? If not, the hierarchy is relying on decoration instead of structure, and that is the part to fix.
Minimalist — Specification
Definition
Minimalist is a web-interface design style defined by extreme reduction of visual elements to only what is functionally necessary. Hierarchy, spacing, and typography — not color, ornament, or decoration — carry the interface. Negative space is treated as an active structural element rather than unused margin.
Core principles
Reduction over addition. An element is included only if removing it would cost the user comprehension or function.
Negative space as structure. Spacing separates and groups content in place of borders, boxes, and dividers.
Restraint as hierarchy. Emphasis is expressed through scale, weight, and position, not through color or decoration.
Typographic clarity as the primary decoration. Type is the main expressive surface; everything else stays quiet.
Visual characteristics
A near-monochrome palette: one neutral base plus at most one accent color.
Abundant, deliberate whitespace between and within sections.
Thin or absent borders; dividers used sparingly and only where spacing alone is insufficient.
Minimal typographic variation: few font sizes, a small weight range, one or two type families.
Grid-based alignment with consistent margins.
Absence of shadows, gradients, textures, and non-functional icons or illustration.
Rules
Palette restraint. The interface uses no more than one neutral base (a near-white or near-black plus its supporting grays) and at most one accent color across the entire surface.
Acceptance criterion: Inspecting the rendered page’s computed background, text, and border colors yields at most one hue beyond pure black, white, and gray.
Minimum section spacing. Negative space between primary content sections is at least one line-height of the body text.
Acceptance criterion: The measured margin or padding between adjacent content sections is greater than or equal to the body text’s line-height (e.g., at least 24px for 16px/1.5 body text).
No decorative depth effects. The interface contains no decorative box-shadow, gradient background, or background texture.
Acceptance criterion: Inspecting computed CSS across the page finds no gradient background-image and no box-shadow other than a functionally necessary one (such as a modal that must separate from background content), and any such exception is documented.
Typographic economy. No more than two font families and no more than three distinct font sizes are used on a single screen.
Acceptance criterion: Counting distinct font-family and font-size computed values on a rendered screen yields at most 2 families and at most 3 sizes.
Quiet interaction states. Interactive elements signal hover and focus through color, weight, or underline changes only, not through shape, size, or shadow changes.
Acceptance criterion: Comparing computed styles between the resting and hover/focus states of an interactive element shows no change to border-radius, box-shadow, or element dimensions greater than 2px.
Hairline borders only. Any border present in the interface is no more than 1px wide.
Acceptance criterion: Inspecting computed border-width on every visible bordered element returns a value of 1px or less.
Single primary action per section. Each screen section presents exactly one visually dominant call to action.
Acceptance criterion: Counting elements within a section that use the accent color and largest interactive size together yields exactly one.
Avoid
Multiple competing accent colors or decorative background textures.
Drop shadows, glows, or blur used for visual interest rather than a documented functional need.
Decorative icons or illustration where text or spacing already communicates meaning.
More than one font pairing, or more than a small number of distinct type sizes on one screen.
Dense layouts that rely on visual crowding rather than grouping and whitespace.
Acceptance criteria
Rule 1: At most one accent hue is present beyond black, white, and gray.
Rule 2: Inter-section spacing is at least one body line-height.
Rule 3: No decorative shadow, gradient, or texture is present without documented functional necessity.
Rule 4: At most 2 font families and 3 font sizes appear on a single screen.
Rule 5: Hover/focus states change only color, weight, or underline — not shape or size.
Rule 6: All borders are 1px or thinner.
Rule 7: Exactly one dominant call to action exists per section.
Blocking failures
More than one accent color is used for non-semantic decoration anywhere on the page.
A decorative shadow, gradient, or texture is present with no documented functional purpose.
A single section presents two or more equally prominent, competing calls to action, so a visitor cannot identify the intended next step.
These conditions make the work unacceptable to ship; they are not stylistic suggestions and are not subject to designer discretion.
Responsive adaptation guidance
Negative space may compress proportionally on narrower viewports but must never fall below a defined minimum (for example, 16px between sections).
Layouts collapse to a single column before spacing or type would become illegible; content is never truncated to preserve whitespace.
The one-hairline-border rule and the single-dominant-action-per-section rule hold at every breakpoint.
Typographic economy (Rule 4) is evaluated per breakpoint independently, since responsive type scales may legitimately introduce a different, still-limited, set of sizes.
MinimalistOverviewIndexAbout
minimalism
Minimalist
Reduces an interface to only what is functionally necessary, using negative space and a restrained palette so content and hierarchy carry the design instead of decoration.
Primary actionSecondary
Reduction over addition
Minimalist is a web-interface design style defined by extreme reduction of visual elements to only what is functionally necessary. Hierarchy, spacing, and typography — not color, ornament, or decoration — carry the interface. Negative space is treated as an active structural element rather than unused margin.
RestActive
Applying a Brutalist visual direction
You are being asked to apply a Brutalist visual direction to a web interface. This document gives you framing and recommendations for that direction. It does not tell you what page, product, or feature to build — apply the guidance below to whatever interface you are already working on, at whatever scope that work requires.
What “Brutalist” means here
Brutalist web design borrows its name and attitude from brutalist architecture: structure is exposed rather than hidden, materials are shown honestly rather than smoothed over, and the result reads as direct and confrontational rather than polished. In an interface this means visible grid seams, unrounded geometric shapes, hard-edged typography, and a rejection of the soft, softly-shadowed conventions of most contemporary UI. It is a deliberate, controlled raw look, not an unfinished or default one — every rough edge should be an intentional decision, not neglect.
Recommendations
Use square corners everywhere. Do not round containers, buttons, inputs, or images.
Build visible structure: thick, high-contrast borders around distinct content blocks, and visible boundaries at grid or column divisions instead of hiding them behind whitespace alone.
Push typographic scale hard. Headlines should be dramatically larger than body text, and the jump between the largest and smallest text on a screen should be obvious at a glance, not subtle.
Restrict color to true black, true white, and a small number of flat, fully saturated colors used as blocks rather than tints or gradients. Let color feel like a structural material (a colored block, a colored border) rather than a decorative wash.
Let interactive elements look like what they are: a button should read as a hard-edged, bordered box, not a soft rounded pill. Default or near-default form control styling is acceptable and often correct here.
Use monospace or plain system-style typefaces where they reinforce the “unstyled, structural” feeling, especially for labels, metadata, or anything that benefits from reading as raw information.
Let asymmetry and density coexist deliberately: a very large headline next to a dense block of small text is a valid and expected composition in this direction.
Avoid
Rounded corners of any kind, on any element.
Soft box shadows, blur, glow, or any effect whose purpose is to simulate depth or softness.
Gradients, pastel tints, or more than two or three flat accent colors.
Smooth, springy, or decorative motion. Transitions, where used at all, should be abrupt rather than eased.
Hiding the grid: do not let column gutters, section boundaries, or structural divisions disappear into undifferentiated whitespace.
How to check your own work
Before considering the result finished, look at it and ask:
Does any element have a rounded corner? If yes, square it off.
Does any element have a soft shadow, blur, or gradient used for depth or polish? If yes, replace it with a hard border or a flat color block, or remove it.
If I squint at the layout, can I still see where the structural divisions are — the grid, the sections, the boundaries between blocks? If the structure disappears, add a visible border or contrasting block to make it explicit again.
Is the type scale dramatic enough that the largest and smallest text on the screen are obviously, not subtly, different in size?
Brutalist — Specification
Definition
Brutalist is a web-interface design style that foregrounds raw, unornamented structure. It exposes grid seams, uses unrounded geometric shapes, sets hard-edged typography at dramatic scale contrast, and rejects the soft, shadowed conventions of contemporary interface polish in favor of a direct, controlled, deliberately raw appearance.
Core principles
Structure over polish. Grid lines, boundaries, and containers are shown, not hidden.
Honest materials. Default or near-default browser affordances are used deliberately rather than smoothed away.
Confrontational hierarchy. Contrast — in scale, weight, and color — is dramatic rather than subtle.
Deliberate rawness. Every rough edge is a considered decision, not an omission.
Visual characteristics
Thick, high-contrast borders around distinct content blocks.
Square, unrounded corners on every container, button, and image.
A restricted palette of true black, true white, and a small number of flat, fully saturated colors used as blocks, not tints or gradients.
Oversized, dramatic typographic scale contrast between headline and body text.
Visible grid seams and section boundaries rather than boundaries implied by whitespace alone.
Default or near-default form control styling in place of custom-styled soft controls.
Rules
No rounded corners. No container, card, button, input, or image has a border-radius greater than 0.
Acceptance criterion: Inspecting computed border-radius across every container and interactive element on the page returns 0px.
Visible block borders. Every distinct content block has a border of at least 2px, at high contrast against its adjacent surface.
Acceptance criterion: Measuring border-width on each block-level container returns at least 2px, with a measured contrast ratio of at least 7:1 against the adjacent background.
Dramatic type scale. The ratio between the largest and smallest rendered text on a given screen is at least 4:1.
Acceptance criterion: Measuring the largest and smallest computed font-size values present on a rendered screen yields a ratio of 4:1 or greater.
Flat, restricted palette. All color on the page is true black, true white, or one of a documented set of at most two additional flat, fully saturated colors; no gradients or tints are used for depth.
Acceptance criterion: Inspecting rendered color swatches finds every color present matches black, white, or a declared saturated color, and no linear-gradient, radial-gradient, or box-shadow with nonzero blur is present.
Structural interactive controls. Buttons and inputs are visually box-like, with square corners and a border of at least 2px, distinguishable from body content at a glance.
Acceptance criterion: A rendered screenshot review confirms every button and input has square corners and a border-width of at least 2px.
Exposed grid seams. Every major grid or section division has a visible boundary — a rule line or a contrasting color block — rather than being implied by whitespace alone.
Acceptance criterion: Inspecting the rendered layout confirms a visible border or background-color change exists at each documented major grid division.
No smoothing effects. No blur, soft shadow, or eased opacity transition is used to simulate depth or softness.
Acceptance criterion: Inspecting computed CSS finds no filter: blur, no box-shadow with a blur radius greater than 0, and no opacity transition longer than an abrupt (near-instant) duration used for polish.
Avoid
Rounded corners on any element.
Soft box shadows, glow, or blur used to simulate elevation or softness.
Gradients, pastel tints, or more than two flat accent colors beyond black and white.
Smooth, springy, or decorative motion; eased transitions used for polish rather than function.
Grid gutters or section boundaries that disappear into undifferentiated whitespace.
Acceptance criteria
Rule 1: border-radius is 0px on every container and interactive element.
Rule 2: Every content block has a border of at least 2px at 7:1+ contrast.
Rule 3: The largest-to-smallest font-size ratio on a screen is at least 4:1.
Rule 4: All colors are black, white, or one of at most two declared saturated colors, with no gradients.
Rule 5: Buttons and inputs are square-cornered with borders of at least 2px.
Rule 6: Every major grid division has a visible border or color-block boundary.
Rule 7: No blur, soft shadow, or eased polish transition is present.
Blocking failures
Any container or interactive element has a border-radius greater than 0.
A soft box-shadow, blur, or gradient is used for depth or polish anywhere on the page.
The color palette includes a gradient fill or more than two non-black/white saturated colors.
These conditions make the work unacceptable to ship; they are not stylistic suggestions and are not subject to designer discretion.
Responsive adaptation guidance
Grid seams remain visible at every viewport width; borders never collapse to a hairline that reads as invisible.
The minimum 4:1 type-scale ratio (Rule 3) may compress to no less than 2.5:1 on narrow viewports, documented as an explicit exception rather than a silent drift.
Border widths may reduce on narrow viewports but never below 2px (Rule 2 and Rule 5 continue to apply).
Square corners (Rule 1) and the restricted flat palette (Rule 4) hold unchanged at every breakpoint.
BrutalistOverviewIndexAbout
brutalism
Brutalist
Foregrounds raw, unornamented structure with thick borders, square corners, and stark color blocking, rejecting decorative smoothing to communicate directness.
Primary actionSecondary
Structure over polishHonest materialsConfrontational hierarchyDeliberate rawness
Applying an Editorial visual direction
You are being asked to apply an Editorial visual direction to a web interface. This document gives you framing and recommendations for that direction. It does not tell you what page, product, or feature to build — apply the guidance below to whatever interface you are already working on, at whatever scope that work requires.
What “Editorial” means here
Editorial design borrows the conventions of print publications — magazines, newspapers, and printed journals — and applies them to a screen. The defining move is treating typography and layout as the primary structuring devices, the way a print art director would, rather than treating them as a neutral container for content. Text is set on a considered grid, imagery is integrated as part of the story rather than as generic decoration, and the whole composition should feel authored: like someone made deliberate editorial choices about what is emphasized, what is quiet, and how the reader’s eye should move.
Recommendations
Pair a distinctive display or serif typeface for headlines with a clean, functional sans-serif (or a second serif with a clearly different character) for body text and interface labels. The contrast between the two should be legible at a glance.
Control line length: body paragraphs read best between roughly 45 and 75 characters per line. Constrain the column width rather than letting text stretch across the full viewport.
Build a real column grid for long-form content. Where content is dense enough, consider multiple columns at wide viewports rather than one very long single column.
Treat images as content: give them intentional placement (full-bleed, art-directed crops, or a defined position within the grid) and pair them with a caption or credit line set in a visually distinct style from body text.
Use thin rules and generous, purposeful spacing to separate sections — a masthead from an article, one article from another — the way a printed page uses margins and rules rather than boxes.
Keep color restrained: let typography, imagery, and layout carry the visual interest, and use no more than one accent color for links, emphasis, or interactive states.
Vary pacing deliberately. A large pull quote, a wide image, or a section break can serve the same purpose a print spread uses to give the reader’s eye a place to rest.
Avoid
Sans-only or serif-only pairings that erase the contrast between headline and body voice.
Unbroken walls of text with no typographic pacing, section breaks, or pull quotes.
Treating images as generic background decoration rather than captioned, art-directed content.
More than one accent color competing with the reading experience.
Styling captions identically to body text, which erases the distinction between narrative copy and supporting material.
How to check your own work
Before considering the result finished, look at it and ask:
Do headlines and body text clearly read as two different typographic voices?
If I measure a line of body text, does it fall roughly within a comfortable reading length, rather than stretching edge to edge or being cramped into a narrow sliver?
Does every image have a caption or credit treated as distinct from body copy?
Does the page have a clear rhythm — passages of dense reading broken up by rules, spacing, images, or pull quotes — rather than one undifferentiated block of content?
Editorial — Specification
Definition
Editorial is a web-interface design style modeled on print-publication design. It builds a clear typographic hierarchy from a serif-and-sans (or comparably contrasting) type pairing, structures content on a considered reading grid, treats imagery as narrative content rather than decoration, and paces dense and open sections deliberately, producing an interface that reads as authored rather than assembled.
Core principles
Typographic hierarchy as structure. Type contrast, not color or containers, is the primary organizing device.
Grid-based composition. Content is set on a considered column and measure system borrowed from print layout.
Imagery as content. Images are captioned, art-directed, and integrated into the narrative, not used as generic decoration.
Restrained color in service of legibility. Color supports reading; it does not compete with it.
Deliberate pacing. Dense reading passages are broken up intentionally, the way a print spread uses rules, images, and pull quotes.
Visual characteristics
A serif (or otherwise distinctive) display typeface for headlines, paired with a clearly contrasting sans-serif for body and interface text.
A defined column grid with controlled line length for body copy.
Captions or credit lines set visually apart from body copy, adjacent to their images.
Full-bleed or art-directed imagery integrated into the layout rather than treated as background filler.
Thin rules and generous, purposeful spacing separating sections, in place of heavy containers.
A restrained, mostly neutral palette with a single accent color for links and emphasis.
Rules
Controlled measure. Body copy line length stays between 45 and 75 characters per line at the primary reading viewport.
Acceptance criterion: Measuring the average characters-per-line of rendered body paragraphs at the primary reading viewport falls between 45 and 75.
Distinct headline voice. Headline elements (h1, h2) use a typeface family visibly different from the body-text family.
Acceptance criterion: Inspecting computed font-family for headline elements shows a different value than the computed font-family used for body paragraph text.
Captioned imagery. Every content image is accompanied by a caption or credit line set in a visually distinct style from body copy.
Acceptance criterion: For each content image or figure, an adjacent caption element exists whose computed font-size or font-weight differs from the surrounding body paragraph’s computed value.
Marked section breaks. Section boundaries are marked by either a visible thin rule or vertical spacing of at least twice the body line-height.
Acceptance criterion: Measuring the gap between two consecutive sections confirms either a visible rule element or a spacing value of at least 2x the body text’s line-height.
Single accent color. No more than one accent color is used for links, emphasis, or interactive states outside the near-black/neutral text palette.
Acceptance criterion: Inspecting rendered color swatches for links and emphasized text finds exactly one non-neutral hue in use across the page.
Multi-column long-form layout. At desktop width, body content exceeding 400 words is laid out either in two or more columns or within a single column whose measure satisfies Rule 1.
Acceptance criterion: For a body section exceeding 400 words at desktop width, inspecting the rendered layout shows either a column-count of 2 or more, or a single-column container whose measured line length satisfies the 45–75 character range from Rule 1.
Distinguished pull quotes. Any pull quote is set at a larger font size than body text and is visually separated from surrounding paragraphs by indentation, a rule, or additional whitespace.
Acceptance criterion: Inspecting a pull-quote element’s computed font-size shows a value greater than the body paragraph’s computed font-size, and a visible margin, indent, or rule separates it from adjacent paragraphs.
Avoid
Sans-only or serif-only pairings that erase the contrast between headline and body voice.
Long, unbroken text blocks with no typographic pacing, section breaks, or pull quotes.
Images used as generic background decoration without a caption or credit mechanism.
More than one accent color competing with the reading experience.
Captions styled identically to body text.
Acceptance criteria
Rule 1: Body line length measures 45–75 characters at the primary reading viewport.
Rule 2: Headline and body text use visibly different font families.
Rule 3: Every content image has an adjacent, visually distinct caption.
Rule 4: Every section break has a visible rule or spacing of at least 2x body line-height.
Rule 5: Exactly one accent color is used for links and emphasis.
Rule 6: Long-form body content (>400 words) uses multi-column layout or a measure-compliant single column at desktop width.
Rule 7: Pull quotes are visibly larger than body text and separated from surrounding paragraphs.
Blocking failures
Body text renders at a measure outside 40–90 characters per line at the primary reading viewport, making it materially unreadable.
No typographic distinction exists anywhere between headline and body type — a single font family, weight, and size are used across the entire page.
A content image is presented with no caption or credit mechanism in a context where the entry’s provenance metadata requires attribution.
These conditions make the work unacceptable to ship; they are not stylistic suggestions and are not subject to designer discretion.
Responsive adaptation guidance
Multi-column layouts (Rule 6) collapse to a single column before the reflow would push line length outside the 45–75 character range required by Rule 1.
Margins and section spacing reduce proportionally on narrower viewports but never below a defined minimum that would violate Rule 4.
Captions remain immediately adjacent to their associated image after any reflow, preserving Rule 3.
The single-accent-color rule (Rule 5) and the headline/body typographic distinction (Rule 2) hold unchanged at every breakpoint.
EditorialOverviewIndexAbout
editorial
Editorial
Borrows print-publication conventions — a serif and sans typographic pairing, a structured column grid, and imagery treated as content — to produce a considered, authored reading experience.
Primary actionSecondary
Typographic hierarchy as structure
A serif (or otherwise distinctive) display typeface for headlines, paired with a clearly contrasting sans-serif for body and interface text.
A defined column grid with controlled line length for body copy.
Captions or credit lines set visually apart from body copy, adjacent to their images.
Full-bleed or art-directed imagery integrated into the layout rather than treated as background filler.
Type contrast, not color or containers, is the primary organizing device.
Typographic hierarchy as structure
Editorial
Editorial is a web-interface design style modeled on print-publication design. It builds a clear typographic hierarchy from a serif-and-sans (or comparably contrasting) type pairing, structures content on a considered reading grid, treats imagery as narrative content rather than decoration, and paces dense and open sections deliberately, producing an interface that reads as authored rather than assembled.
Applying a Swiss visual direction
You are being asked to apply a Swiss visual direction to a web interface. This document gives you framing and recommendations for that direction. It does not tell you what page, product, or feature to build — apply the guidance below to whatever interface you are already working on, at whatever scope that work requires.
What “Swiss” means here
Swiss design — also called the international typographic style — treats layout as a system rather than a composition. Before anything is placed, a grid is defined; everything afterward is positioned by that grid, and the grid remains legible in the finished result. Typography does the expressive work through size, weight, and position, using a single neutral sans-serif rather than a pairing. Color is not a hierarchy device: if the interface were rendered in grayscale, the reading order would be unchanged.
The common misreading is that Swiss means sparse. It does not. A Swiss page can be extremely dense — full of text, tables, and images — and still be Swiss, because what defines it is the system, not the amount of material. Restraint applies to the number of decisions, not to the amount of content.
Recommendations
Define the grid first and write it down: how many columns, what gutter, what baseline unit. Every element you place afterward should begin and end on a column edge, and every vertical space should be a multiple of the baseline unit.
Use one sans-serif family for everything. Neo-grotesque faces suit this direction because they are neutral enough to disappear behind the structure. Build hierarchy from three or four sizes and two weights, not from a second family.
Set text flush left with a ragged right edge. Avoid justified text, which creates uneven word spacing, and avoid centered blocks, which have no grid position.
Compose asymmetrically. A block of text in the left two columns with the remaining four empty is more Swiss than the same block centered — the grid is what makes the emptiness feel placed rather than left over.
Treat photography as a grid element: crop images to a whole number of columns, and leave them untreated. No filters, no color overlays, no gradient scrims.
Let hierarchy survive without color. Assign emphasis through size, weight, and position, then check the result in grayscale. If something becomes ambiguous, fix the typographic hierarchy rather than reintroducing color.
Keep rules and lines structural. A horizontal rule should mark a real division in the grid, not decorate a section header.
Avoid
A second typeface family introduced for contrast or personality.
Centered or justified body text.
Color used to distinguish levels of a hierarchy that size and position should be carrying.
Elements positioned by eye, sitting between column edges.
Photographic treatments: duotone, gradient overlay, heavy vignette, decorative masking.
Arbitrary vertical spacing that does not resolve to the baseline unit.
How to check your own work
Before considering the result finished, look at it and ask:
Can I overlay the grid on a screenshot and see every element land on a column edge? Anything that does not is either misplaced or an undeclared exception.
If I render the page in grayscale, does the reading order stay the same? If it does not, color is carrying hierarchy that typography should carry.
How many typeface families are actually rendering? If the answer is more than one, decide which one is doing real work and remove the other.
Are the left and right margins of the main content block different? If they are identical, the composition is centered rather than composed, and the grid is doing nothing.
Swiss — Specification
Definition
Swiss is a web-interface design style in which a declared modular grid governs every placement decision and remains legible in the finished result. A single neutral sans-serif carries the entire typographic hierarchy through size, weight, and position; color plays no structural role. The style is defined by the presence of a system, not by the quantity of content, so a dense page and a sparse page can both be Swiss.
Core principles
The grid precedes the composition. Columns, gutters, and a baseline unit are defined before any element is placed, and everything resolves to them.
One typographic voice. A single sans-serif family expresses every level of hierarchy; contrast comes from scale and weight, not from a second family.
Hierarchy without color. Reading order survives in grayscale because size and position, not hue, establish emphasis.
Asymmetry as structure. Balance is achieved through deliberate unequal distribution across the grid rather than through centering.
Untreated material. Photography and rules appear as themselves, cropped and positioned by the grid rather than styled.
Visual characteristics
A visible underlying column structure, with content blocks starting and ending on column edges.
One neutral sans-serif family at a small number of sizes and two or three weights.
Text set flush left with a ragged right edge.
Asymmetric composition with unequal left and right margins.
Color photography cropped to whole column widths, without filters or overlays.
Horizontal rules marking real structural divisions rather than decorating headings.
Vertical spacing that resolves to a single baseline unit throughout.
Rules
Declared column grid. Every content block begins and ends on a column edge of a documented grid.
Acceptance criterion: Measuring the left and right edge offsets of every content block resolves them to a finite set of column positions from the declared grid, with no block landing between edges.
Single typeface family. The entire interface uses one sans-serif family.
Acceptance criterion: Collecting the computed font-family values across every rendered text node yields exactly one family.
Hierarchy survives grayscale. Emphasis is carried by size, weight, and position rather than by color.
Acceptance criterion: Rendering the page in grayscale leaves the reading order unchanged, and the number of distinct computed text colors is 2 or fewer.
Flush-left text. Body copy and headings are set flush left with a ragged right edge.
Acceptance criterion: The computed text-align of every paragraph and heading is left or start; no text block computes to justify or center.
Asymmetric composition. The optical axis of the layout does not coincide with the center of the viewport.
Acceptance criterion: The measured left and right margins of the main content block differ by at least one full column width.
Grid-derived spacing. Vertical rhythm follows the declared baseline unit.
Acceptance criterion: Every vertical margin and padding value in the layout is an integer multiple of the documented baseline unit.
Untreated photography. Images are cropped by the grid and presented without decorative treatment.
Acceptance criterion: No content image has a CSS filter or a color overlay, and each image’s rendered width equals a whole number of columns plus their intervening gutters.
Avoid
A second typeface family introduced for contrast or personality.
Justified or centered body text.
Color used to separate hierarchy levels that size and position should establish.
Elements positioned optically, landing between column edges.
Duotone, gradient scrim, vignette, or decorative masking on photography.
Vertical spacing values that do not resolve to the baseline unit.
Acceptance criteria
Rule 1: Every content block starts and ends on a declared column edge.
Rule 2: Exactly one typeface family renders across the interface.
Rule 3: Reading order is unchanged in grayscale and at most 2 text colors are used.
Rule 4: All paragraphs and headings compute to flush-left alignment.
Rule 5: Left and right margins of the main content block differ by at least one column.
Rule 6: All vertical spacing is an integer multiple of the baseline unit.
Rule 7: No content image carries a filter or overlay, and each spans whole columns.
Blocking failures
No grid is declared, so column positions cannot be verified and every placement is an isolated judgment.
Two or more typeface families render across the interface, dissolving the single typographic voice the style depends on.
Color is the only mechanism distinguishing a hierarchy level, so the structure collapses in grayscale or for a color-blind reader.
These conditions make the work unacceptable to ship; they are not stylistic suggestions and are not subject to designer discretion.
Responsive adaptation guidance
The column count reduces at narrower viewports — for example six columns to four to two — but the grid is never abandoned; every block still resolves to a column edge in the reduced grid (Rule 1).
The baseline unit may shrink proportionally, but vertical spacing continues to resolve to whole multiples of whichever unit is active (Rule 6).
Asymmetry (Rule 5) may weaken as the column count falls and must be evaluated per breakpoint; at a two-column grid, a one-column offset satisfies the rule.
Images re-crop to a whole number of columns in the reduced grid rather than scaling to an arbitrary fraction of the viewport (Rule 7).
The single-family (Rule 2) and flush-left (Rule 4) rules hold unchanged at every breakpoint.
SwissOverviewIndexAbout
grid
Swiss
Organizes an interface on a strict modular grid with a single neutral sans-serif, expressing hierarchy through scale and position alone so the underlying structure stays visible.
Primary actionSecondary
The grid precedes the compositionOne typographic voiceHierarchy without color
Swiss
Swiss is a web-interface design style in which a declared modular grid governs every placement decision and remains legible in the finished result. A single neutral sans-serif carries the entire typographic hierarchy through size, weight, and position; color plays no structural role. The style is defined by the presence of a system, not by the quantity of content, so a dense page and a sparse page can both be Swiss.
Applying a Maximalist visual direction
You are being asked to apply a Maximalist visual direction to a web interface. This document gives you framing and recommendations for that direction. It does not tell you what page, product, or feature to build — apply the guidance below to whatever interface you are already working on, at whatever scope that work requires.
What “Maximalist” means here
Maximalism treats abundance as the organizing principle. Where a reductive direction asks what can be removed, this one asks what else the composition can carry — more color, more pattern, more layers, more typographic range. The surface is meant to feel full, energetic, and personally authored rather than systematically calm.
The distinction that matters is between abundance and noise. Noise is what happens when nothing is deliberate: colors that arrived by accident, layers that overlap because nothing was positioned, text that is hard to read because no one checked. Maximalism is the opposite of accidental. Every one of those colors is chosen, every overlap is placed, and the result still has a single thing your eye lands on first. If a viewer cannot tell you what the page is about after two seconds, the composition has failed on its own terms, not just on conventional ones.
Recommendations
Build a real palette of five or more saturated colors and use all of them. Assign them before you compose, so the abundance is designed rather than accumulated.
Layer elements so they genuinely overlap. Blocks that merely sit next to each other read as a busy grid, not as maximalism; the depth comes from things crossing over one another.
Push the typographic range hard. Set the largest element several times the size of the smallest, and use three or more families with genuinely different character rather than three weights of the same one.
Fill the background. A pattern, a texture, or a gradient field gives the layers something to sit on; a flat white background under a dense composition reads as an unfinished layout rather than a chosen one.
Establish one clear focal point despite everything else. Make one element unmistakably the largest, and let the density surround it rather than compete with it.
Check contrast on every piece of text, against the actual pixels underneath it rather than against an assumed background. This is where maximalist work fails most often, and it is entirely avoidable.
Vary the angles and scales of secondary material — small photographs, badges, stickers, rules — so the composition has internal rhythm rather than uniform density.
Avoid
A palette that grew by accident rather than being chosen, especially unsaturated colors mixed in with saturated ones.
Elements packed tightly but never overlapping, which reads as crowding rather than layering.
Three weights of one typeface standing in for three families.
Text placed over a busy image or pattern without checking its contrast at the worst point.
A composition with no dominant element, where several things are equally large and the eye has nowhere to land.
Motion added on top of the density; a maximalist static composition is already working hard, and animating it usually makes it unreadable.
How to check your own work
Before considering the result finished, look at it and ask:
If I look at the screen for two seconds and then look away, what do I remember? If the answer is “a lot of stuff,” there is no focal point, and one element needs to get much larger.
Is every color in the palette one I chose, or did some arrive because a component brought its own default with it?
Do the layers actually cross each other, or are they just close together?
Take the single worst piece of text on the page — smallest, over the busiest background — and measure its contrast. If it fails, the density is being paid for by the reader, which is the one cost this style cannot justify.
Maximalist — Specification
Definition
Maximalist is a web-interface design style that treats abundance as the organizing principle. It composes with a large saturated palette, genuinely overlapping layers, extreme typographic range, and a filled rather than empty background, producing a dense and energetic surface. It is distinguished from visual noise by two requirements it never relaxes: a single dominant focal point, and legible text everywhere.
Core principles
Abundance is designed, not accumulated. The palette, the layers, and the typographic range are decided in advance rather than arrived at by addition.
Depth through overlap. Elements cross one another; proximity alone produces crowding, not layering.
Extreme typographic range. Contrast between the largest and smallest text is dramatic, and multiple families with distinct character coexist.
The surface is filled. Background is treated as material — pattern, texture, or gradient — rather than as empty ground.
Density never costs legibility. Every piece of text remains readable against whatever is actually behind it.
Visual characteristics
Five or more saturated colors present simultaneously.
Overlapping blocks, images, and type rather than adjacent ones.
One element dramatically larger than everything around it.
Three or more typeface families with genuinely different character.
A patterned, textured, or gradient background rather than a flat fill.
Secondary material — small photographs, badges, rules — set at varied scales and angles.
Very little uncovered area within the initial viewport.
Rules
Abundant saturated palette. At least five distinct saturated colors are present at once.
Acceptance criterion: Sampling the rendered colors within the initial viewport yields at least five distinct hues with a saturation of 60% or greater.
Genuine overlap. Elements cross one another rather than merely sitting adjacent.
Acceptance criterion: At least three pairs of rendered elements have intersecting bounding boxes.
Extreme typographic range. The scale jump between largest and smallest text is dramatic, across multiple families.
Acceptance criterion: The ratio between the largest and the smallest rendered font-size is 6:1 or greater, and at least three distinct computed font-family values are present.
Deliberate density. The composition fills its surface.
Acceptance criterion: Rendered content covers at least 60% of the initial viewport area.
Filled background. The background is material rather than a flat single color.
Acceptance criterion: The primary background region has a background-image, a repeating pattern, or a gradient; it does not compute to a single solid color.
One dominant focal point. A single element leads the composition despite the density.
Acceptance criterion: The largest rendered element occupies at least twice the area of the second largest.
Legibility is not traded away. All text remains readable over whatever sits behind it.
Acceptance criterion: Every text element measures a contrast ratio of at least 4.5:1 against the least favorable background pixel beneath it, not against an average.
Avoid
A palette that accumulated from component defaults rather than being chosen.
Tight packing without overlap, which reads as crowding rather than layering.
Three weights of a single family substituting for three families.
Text placed over pattern or imagery without checking its worst-case contrast.
A composition where several elements are equally large and no focal point exists.
Animation layered on top of an already dense composition.
Acceptance criteria
Rule 1: At least five saturated hues are present in the initial viewport.
Rule 2: At least three pairs of elements genuinely overlap.
Rule 3: The largest-to-smallest type ratio is at least 6:1 across at least three families.
Rule 4: Content covers at least 60% of the initial viewport.
Rule 5: The background is patterned, textured, or gradient rather than a flat color.
Rule 6: The largest element is at least twice the area of the second largest.
Rule 7: Every text element meets 4.5:1 against its worst-case background pixel.
Blocking failures
Any text falls below 4.5:1 against the actual pixels behind it, making the density a cost paid by the reader.
No dominant element exists, so the composition has no entry point and a visitor cannot determine what the page is about.
Color, layering, or typographic variation is used to convey meaning that exists nowhere else, so the information is unavailable to anyone who cannot resolve the visual distinction.
These conditions make the work unacceptable to ship; they are not stylistic suggestions and are not subject to designer discretion.
Responsive adaptation guidance
Density (Rule 4) is evaluated per breakpoint. At narrow viewports the 60% coverage target may fall to 45% as layers are removed rather than compressed; removing a layer is correct, shrinking every layer is not.
The typographic ratio (Rule 3) may compress to no less than 4:1 at narrow viewports, documented as an explicit exception rather than allowed to drift.
Overlap (Rule 2) is preserved by reducing the number of overlapping pairs, never by separating all layers into a stacked single column.
The dominant focal point (Rule 6) and the contrast floor (Rule 7) hold unchanged at every breakpoint, and Rule 7 is re-measured after any reflow, since text frequently lands over different background material at narrower widths.
maximalism
Abundance is designed, not accumulatedDepth through overlapExtreme typographic rangeThe surface is filledDensity never costs legibility
Maximalist
Abundance is designed, not accumulated
Maximalist is a web-interface design style that treats abundance as the organizing principle. It composes with a large saturated palette, genuinely overlapping layers, extreme typographic range, and a filled rather than empty background, producing a dense and energetic surface. It is distinguished from visual noise by two requirements it never relaxes: a single dominant focal point, and legible text everywhere.
Applying a Glassmorphism visual direction
You are being asked to apply a Glassmorphism visual direction to a web interface. This document gives you framing and recommendations for that direction. It does not tell you what page, product, or feature to build — apply the guidance below to whatever interface you are already working on, at whatever scope that work requires.
What “Glassmorphism” means here
Glassmorphism builds surfaces that behave like frosted glass: partially transparent, blurring whatever passes behind them, and edged with a thin band of light where the material catches. The effect communicates layering — you can see that one thing is in front of another because you can see through it.
Two things make or break this direction, and both are frequently missed. The first is that glass needs something behind it. Over a flat gray background there is nothing to blur, and the panels read as slightly translucent boxes rather than as glass; the background is not decoration here, it is the material the effect is made of. The second is that translucency puts text on an unpredictable surface. A label that reads comfortably over the pale part of a gradient can become unreadable where the gradient darkens, and the only way to know is to check the worst point rather than the average. Most published glassmorphism fails exactly there.
Recommendations
Build the background first: soft, out-of-focus color shapes, a broad gradient, or a blurred photograph. Give it real variation — at least two hues — so there is something worth seeing through the glass.
Blur meaningfully. A backdrop blur of eight pixels or more actually frosts the background; anything less looks like a rendering artifact.
Keep the panel fill barely there. A background alpha somewhere between ten and forty percent lets the material show through while still separating the panel from what is behind it.
Give every panel a single thin light edge, brighter than its fill and only a pixel wide. That edge is what reads as the cut of the glass, and without it the panel dissolves into the background.
Limit the stack. Two layers of glass over one another still read as depth; three or more turn into visual mud and stop communicating order.
Choose text colors against the worst case. Find the darkest and lightest point of the background that falls under a text block and verify contrast at both, not against the panel’s nominal color.
Decide what happens when backdrop blur is unavailable. A solid, opaque panel in a color that preserves the same text contrast is a good outcome; a transparent panel with no blur is a broken one.
Avoid
Glass panels over a flat, single-color background, where the effect has nothing to work with.
A panel fill so opaque that no background detail is visible through it, which is just a tinted card.
Three or more translucent layers stacked at any point on the screen.
Text contrast evaluated against the panel’s tint rather than against the background actually showing through it.
Heavy drop shadows added under the glass to force separation; the light edge and the blur should be doing that work.
Placing critical small text — labels, values, legal copy — directly over the busiest part of the background.
How to check your own work
Before considering the result finished, look at it and ask:
Can I see what is behind each panel, blurred? If not, either the blur or the transparency is not actually applied.
If I replaced the background with a flat gray, would anything remain of the effect? If the answer is no, that is correct — and it means the background deserves as much attention as the panels.
Find the single worst pixel behind each block of text and check the contrast there. If it fails, either move the text, strengthen the panel fill, or change the background under it.
Turn backdrop blur off entirely and look at the result. If text becomes unreadable, the fallback does not exist yet and needs to.
Glassmorphism — Specification
Definition
Glassmorphism is a web-interface design style in which surfaces behave like frosted glass: partially transparent, blurring the content behind them, and edged with a thin band of light. Layer order is communicated by seeing through one surface to another. The style depends on a background with real chromatic variation, and it is only shippable when text contrast is verified against the worst-case background pixel rather than against the panel’s nominal fill.
Core principles
The background is the material. The effect is produced by what shows through the glass, so the background carries as much design weight as the panels.
Translucency communicates order. Depth is legible because a front surface reveals a blurred version of what sits behind it.
Light defines the edge. A single thin bright border reads as the cut of the glass and keeps the panel from dissolving into its surroundings.
Depth is shallow by design. Two layers read as order; more read as mud.
Contrast is measured at the worst point. Text sits on a variable surface, so legibility is evaluated against the least favorable pixel under it.
Visual characteristics
Translucent panels with a visibly blurred background showing through.
A background of soft gradient fields, out-of-focus color shapes, or a blurred photograph, in at least two hues.
A single hairline border on each panel, lighter than the panel fill.
Low-opacity panel fills that tint rather than obscure.
At most two overlapping glass layers at any point on screen.
Depth expressed by transparency and edge light rather than by heavy shadow.
Rules
Real backdrop blur. Each glass surface blurs what is behind it.
Acceptance criterion: Every glass panel has a computed backdrop-filter containing a blur radius of 8px or greater.
Chromatic background behind the glass. The region under each panel has something worth seeing through it.
Acceptance criterion: Sampling the background region beneath each panel yields at least two distinct hues or a gradient; it does not resolve to a single flat color.
Bounded transparency. The panel fill tints without obscuring.
Acceptance criterion: The alpha channel of each panel’s computed background-color falls between 0.1 and 0.4 inclusive.
Hairline light edge. Each panel is cut from its surroundings by a single bright border.
Acceptance criterion: Every panel has a 1px border whose color has an alpha between 0.1 and 0.5 and is lighter than the panel’s own fill.
Shallow stacking. Glass does not accumulate into unreadable depth.
Acceptance criterion: At no point on the rendered screen do more than two translucent panels overlap.
Worst-case contrast. Text legibility is verified against the actual background, not an average.
Acceptance criterion: Every text element over glass measures at least 4.5:1 against the least favorable pixel of the blurred background beneath it.
Defined degradation. The interface remains legible where backdrop blur is unsupported.
Acceptance criterion: With backdrop-filter disabled, each panel renders an opaque background that preserves the same text contrast ratios required by Rule 6.
Avoid
Glass panels over a flat, single-color background, where no material shows through.
Panel fills opaque enough that no background detail is visible, producing a tinted card rather than glass.
Three or more translucent layers overlapping at any screen position.
Contrast evaluated against the panel’s nominal tint rather than the background showing through it.
Heavy drop shadows added beneath panels to force separation that the edge and blur should provide.
Critical small text placed over the busiest region of the background.
Acceptance criteria
Rule 1: Every glass panel has a backdrop blur of at least 8px.
Rule 2: The background beneath each panel contains at least two hues or a gradient.
Rule 3: Every panel fill has an alpha between 0.1 and 0.4.
Rule 4: Every panel has a 1px border, lighter than its fill, with alpha between 0.1 and 0.5.
Rule 5: No more than two glass layers overlap at any point.
Rule 6: All text over glass meets 4.5:1 against the worst-case background pixel.
Rule 7: With backdrop blur disabled, panels render opaque and contrast is preserved.
Blocking failures
Any text over glass falls below 4.5:1 against the worst-case pixel behind it, which is the most common failure of this style and makes the interface unusable for part of its audience.
No fallback exists for absent backdrop-filter support, so the interface renders as transparent panels over an unblurred background and content becomes unreadable.
A control’s only boundary is its translucent fill against a background of similar luminance, leaving the control undiscoverable.
These conditions make the work unacceptable to ship; they are not stylistic suggestions and are not subject to designer discretion.
Responsive adaptation guidance
Blur radius (Rule 1) may reduce at narrow viewports for performance but never below 8px, since a smaller radius reads as a rendering artifact rather than as glass.
The background (Rule 2) is re-composed rather than scaled at narrow widths, so each panel still has chromatic variation beneath it after reflow.
Stacking (Rule 5) tends to resolve naturally as layouts collapse to one column; where it does not, layers are removed rather than made more transparent.
Rule 6 is re-measured after every reflow without exception, because a panel that moves to a different part of the background sits over different pixels and its contrast result does not carry over.
The fallback required by Rule 7 applies identically at every breakpoint.
GlassmorphismOverviewIndexAbout
glassmorphism
Glassmorphism
Builds surfaces from translucent, blurred panels that let a colored background show through, using light edges and depth ordering to separate layers.
Primary actionSecondary
The background is the material
Glassmorphism is a web-interface design style in which surfaces behave like frosted glass: partially transparent, blurring the content behind them, and edged with a thin band of light. Layer order is communicated by seeing through one surface to another. The style depends on a background with real chromatic variation, and it is only shippable when text contrast is verified against the worst-case background pixel rather than against the panel's nominal fill.
RestActive
Applying a Neubrutalism visual direction
You are being asked to apply a Neubrutalism visual direction to a web interface. This document gives you framing and recommendations for that direction. It does not tell you what page, product, or feature to build — apply the guidance below to whatever interface you are already working on, at whatever scope that work requires.
What “Neubrutalism” means here
Neubrutalism keeps the exposed structure and hard edges of brutalist work but drops its severity. Blocks are outlined in thick dark borders and cast hard, unblurred shadows offset to one side, as though cut from colored paper and laid slightly above the page. Colors are bright and flat. Elements sit a few pixels out of alignment with each other on purpose, so the composition feels hand-placed rather than snapped to a grid.
The difference from brutalism proper is worth stating, because the two get conflated. Brutalism is austere: black, white, a restricted palette, no shadow, nothing rounded. Neubrutalism is cheerful: it welcomes bright color, it depends on drop shadows, and it is perfectly happy with rounded corners. What the two share is the refusal to smooth anything over — no blur, no gradient, no soft elevation.
The shadow is doing structural work here, not decorative work. It has no blur and a real offset, so it reads as a solid object displaced behind the element rather than as light falling on it. That is why pressing a button can move the element into its own shadow: the interaction is physical and the feedback is obvious.
Recommendations
Outline everything that is a distinct object — cards, buttons, inputs, images — with a thick dark border, two pixels or more. The outline is what makes the block read as a cut shape.
Give each block a hard shadow: zero blur, offset by at least four pixels on both axes, in a solid dark color. Keep the offset direction identical across the whole interface so the composition feels coherent.
Fill blocks with flat saturated color. No gradients, no tints of the same hue used to fake depth. If you want variety, use more colors, not softer ones.
Pick one corner radius and use it everywhere, or use none at all. Mixed radii are the fastest way to make this direction look accidental.
Nudge things out of alignment deliberately — a few pixels, not dozens. The composition should look placed by hand, not broken.
Make buttons feel physical: on press, remove the shadow and move the element by exactly the shadow’s offset, quickly. The element appears to be pushed down onto the page.
Check text contrast against the actual block color. Bright saturated fills are exactly where this fails, particularly yellow and lime.
Avoid
Soft or blurred shadows, which immediately return the interface to conventional elevation.
Gradients or tinted fills used to suggest depth; depth here comes only from outline and offset.
Several different corner radii across the interface.
Misalignment large enough to read as a layout bug rather than as a decision.
Thin or light-colored borders, which cannot hold a bright fill.
Dark text on a saturated fill without measuring the contrast first.
How to check your own work
Before considering the result finished, look at it and ask:
Does any shadow have blur? If yes, set it to zero — that single property is most of the style.
Do all the shadows fall in the same direction? Inconsistent offsets make the composition feel unresolved rather than playful.
Are the corner radii consistent across every element, or did some component bring its own?
Press a button. Does anything physical happen? If the only feedback is a color change, the interaction is not carrying its share of the direction.
Take the brightest fill on the page and measure the contrast of the text on it. This is where the style most often fails.
Neubrutalism — Specification
Definition
Neubrutalism is a web-interface design style that retains brutalism’s exposed structure and hard edges while replacing its severity with bright, playful color. Objects are outlined in thick dark borders, filled with flat saturated color, and displaced by hard unblurred shadows, producing surfaces that read as cut paper laid slightly above the page. Unlike brutalism, it permits rounded corners and depends on drop shadow as a structural device.
Core principles
Outline defines the object. A thick dark border, not a background change, is what makes a block read as a distinct shape.
Shadow is structure, not light. Offsets are hard and unblurred, describing displacement rather than illumination.
Flat saturated color. Variety comes from adding colors, never from softening them into gradients or tints.
Deliberate imbalance. Small, intentional misalignments make the composition read as hand-placed.
Physical interaction. Controls respond by moving into their own shadow, so feedback is spatial rather than merely chromatic.
Visual characteristics
Thick dark outlines on every distinct object, including images and inputs.
Hard drop shadows with no blur, offset consistently in one direction.
Bright, flat, saturated block fills with no gradient anywhere.
A single corner radius applied uniformly, or none at all.
Sibling blocks offset from one another by small, visible amounts.
Press states that remove the shadow and displace the element by the same offset.
Rules
Thick dark outline. Every container and control is bounded by a heavy dark border.
Acceptance criterion: Computed border-width is 2px or greater on every container and interactive element, with a border color whose relative luminance is below 0.15.
Hard offset shadow. Blocks are displaced by shadow rather than lit by it.
Acceptance criterion: Each block has a box-shadow with a blur radius of exactly 0 and an offset of at least 4px on both axes.
Flat saturated fills. Color is applied as solid blocks.
Acceptance criterion: Each primary block has a solid computed background-color with saturation of 50% or greater, and no linear-gradient or radial-gradient appears anywhere on the page.
Uniform corner radius. Rounding, where present, is consistent.
Acceptance criterion: The set of non-zero computed border-radius values across the page reduces to exactly one value.
Deliberate misalignment. The composition reads as placed rather than snapped.
Acceptance criterion: At least two sibling blocks have positional offsets differing by between 4px and 24px, documented as intentional rather than incidental.
Physical press feedback. Controls move into their own shadow when activated.
Acceptance criterion: In the :active state, the control’s shadow resolves to 0 and the element translates by exactly its resting shadow offset, within 100ms or less.
Contrast over saturated fills. Text remains readable on bright blocks.
Acceptance criterion: Every text element measures at least 4.5:1 against the solid fill color of the block containing it.
Avoid
Soft or blurred shadows, which restore conventional elevation and dissolve the style.
Gradients or same-hue tints used to imply depth.
Multiple corner radii across the interface.
Misalignment large enough to read as a layout defect rather than a decision.
Thin or light-colored borders that cannot hold a saturated fill.
Text placed on a bright fill without a measured contrast result.
Acceptance criteria
Rule 1: Every container and control has a dark border of at least 2px.
Rule 2: Every block has a zero-blur shadow offset at least 4px on both axes.
Rule 3: All fills are solid and saturated, with no gradients present.
Rule 4: Exactly one non-zero corner radius is used across the page.
Rule 5: At least two sibling blocks are intentionally offset by 4–24px.
Rule 6: Active controls lose their shadow and translate by its offset within 100ms.
Rule 7: All text meets 4.5:1 against its containing block’s fill.
Blocking failures
Shadows carry a blur radius greater than zero, which removes the single property the style is built on and returns the interface to conventional soft elevation.
Text on a saturated fill falls below 4.5:1, which occurs most often on yellow and lime blocks and is not correctable by adjusting the border.
A control’s press state produces no spatial change, so the physical interaction model the style promises does not exist and pressed state is conveyed by color alone.
These conditions make the work unacceptable to ship; they are not stylistic suggestions and are not subject to designer discretion.
Responsive adaptation guidance
Border width (Rule 1) may reduce at narrow viewports but never below 2px, since a thinner outline cannot hold a saturated fill.
Shadow offset (Rule 2) may reduce proportionally to no less than 4px; the blur radius stays at 0 at every breakpoint without exception.
Deliberate misalignment (Rule 5) is reduced or removed at the narrowest widths, where a single-column layout leaves no siblings to offset against; this is an acceptable relaxation, unlike the others.
The uniform radius (Rule 4), the press feedback (Rule 6), and the contrast floor (Rule 7) hold unchanged at every breakpoint.
NeubrutalismOverviewIndexAbout
neubrutalism
Neubrutalism
Takes brutalism's raw geometry and makes it playful — thick dark outlines, hard unblurred shadows, and bright saturated blocks set with deliberate, cheerful imbalance.
Primary actionSecondary
Outline defines the objectShadow is structure, not lightFlat saturated color
01Thick dark outlines on every distinct object, including images and inputs.02Hard drop shadows with no blur, offset consistently in one direction.03Bright, flat, saturated block fills with no gradient anywhere.
Applying a Bento visual direction
You are being asked to apply a Bento visual direction to a web interface. This document gives you framing and recommendations for that direction. It does not tell you what page, product, or feature to build — apply the guidance below to whatever interface you are already working on, at whatever scope that work requires.
What “Bento” means here
Bento organizes content into a mosaic of rounded tiles at different sizes, packed tightly into a filled rectangle with an identical gap on every side — the arrangement of a compartmented lunch box, which is where the name comes from. Its purpose is to make a lot of separate things scannable at once: instead of a long scroll of sections, a viewer sees the whole set in a single glance and picks what to read.
The discipline is the one-idea rule. Each tile holds exactly one thing: one claim, one metric, one image, one feature. The moment a tile contains two headings of equal weight, it stops being a compartment and becomes a small page, and the format’s entire advantage disappears. Tile size is how you express importance — a bigger compartment for the thing that matters more — not a container to be filled with more material.
The second discipline is that the mosaic must actually be full. A bento arrangement with holes in it reads as a broken grid rather than a designed one, and fixing that with automatic dense packing is not an option: it makes what a viewer sees diverge from what a screen reader announces.
Recommendations
Define the grid explicitly — column count, row count, and the span of every tile — rather than letting tiles flow and land where they may.
Vary spans meaningfully. If every tile is the same size it is a card grid, not a bento; give the most important item two columns or two rows and let the rest fit around it.
Put one idea in each tile. If a tile needs two headings, it is two tiles.
Use one corner radius for every tile, generous enough to read as a compartment — eight pixels or more.
Keep the gap identical horizontally and vertically. An uneven gap is what makes a mosaic look accidental.
Fill the rectangle. If a hole appears, resize a neighboring tile or promote something into it; do not leave the gap and do not turn on dense auto-placement.
Make each tile self-contained: its own heading, its own meaning, understandable without reading its neighbors. Viewers do not read a bento in order.
Order the tiles in the document the way you want them read, and let the visual order follow. Explicit placement keeps those two the same.
Avoid
Uniform tiles at a single size, which is a card grid wearing the name.
Two headings of the same level inside one tile.
Different corner radii or different gap values across the mosaic.
Empty cells left inside the mosaic’s bounding rectangle.
Dense automatic packing, which makes visual order diverge from document order.
Tiles whose content only makes sense in sequence with the tile beside them.
How to check your own work
Before considering the result finished, look at it and ask:
Are all the tiles the same size? If so, nothing is being said about relative importance and the format is doing no work.
Does any tile contain two things I would describe separately? Split it.
Is there a hole anywhere inside the mosaic’s rectangle? Fill it by resizing, not by packing.
Read the tiles in document order with the visuals hidden. Is that the order they appear on screen? If not, the layout is reordering content and that has to be fixed at the placement level.
Pick any single tile and read it alone. Does it still mean something? If it needs its neighbor, the content is not compartmented yet.
Bento — Specification
Definition
Bento is a web-interface layout style that arranges content as a tightly packed mosaic of rounded tiles at varying spans, filling a rectangle with a uniform gap on every side. Each tile carries exactly one idea, and tile size expresses relative importance. Its purpose is to make many separate items scannable simultaneously rather than sequentially.
Core principles
One idea per compartment. A tile holds a single claim, metric, image, or feature; two ideas mean two tiles.
Size communicates importance. Span is the mechanism for emphasis, not a container to be filled with more material.
The mosaic is full. Holes inside the bounding rectangle read as a broken grid rather than a composed one.
Explicit placement. Every tile’s span and position is declared, which is what keeps visual order identical to document order.
Tiles stand alone. Each is understandable without its neighbors, because a bento is not read in sequence.
Visual characteristics
Rectangular tiles of varying width and height packed into a filled rectangle.
A single, generous corner radius shared by every tile.
An identical gap horizontally and vertically between all tiles.
One heading plus its supporting content inside each tile.
No empty cells within the mosaic’s bounds.
Larger tiles for higher-priority items, surrounded by smaller ones.
Rules
Explicit grid with varying spans. The mosaic is a declared grid in which tiles occupy different spans.
Acceptance criterion: The container declares explicit columns and rows, and at least two tiles occupy different span values.
One idea per tile. Each compartment carries a single subject.
Acceptance criterion: No tile contains more than one heading at the same heading level.
Uniform corner radius. Tiles share a single rounding value.
Acceptance criterion: The computed border-radius is identical across every tile and measures 8px or greater.
Constant gap. Separation is even in both axes.
Acceptance criterion: Computed row-gap and column-gap are equal to each other and identical throughout the mosaic.
No holes. The mosaic fills its own rectangle.
Acceptance criterion: No empty grid cell exists within the bounding rectangle of the rendered mosaic.
Visual order equals document order. Placement never reorders content.
Acceptance criterion:grid-auto-flow does not include dense, and the visual sequence of tiles matches their order in the DOM.
Self-contained tiles. Each tile is meaningful in isolation.
Acceptance criterion: Every tile has its own accessible name, and its content is comprehensible without reading any adjacent tile.
Avoid
Uniform tiles at a single size, which is a card grid rather than a bento.
Two same-level headings inside one tile.
Varying corner radii or gap values across the mosaic.
Empty cells left inside the mosaic’s bounding rectangle.
Dense automatic packing, which separates visual order from document order.
Tile content that only makes sense in sequence with its neighbor.
Acceptance criteria
Rule 1: The grid is explicitly declared and at least two tiles have different spans.
Rule 2: No tile contains two headings of the same level.
Rule 3: All tiles share one corner radius of at least 8px.
Rule 4: Row gap and column gap are equal and constant throughout.
Rule 5: No empty cell exists inside the mosaic’s bounding rectangle.
Rule 6: Dense packing is not used and visual order matches DOM order.
Rule 7: Every tile has its own accessible name and stands alone.
Blocking failures
Dense automatic packing is enabled, so the order a sighted visitor sees differs from the order a screen reader or keyboard user encounters.
A tile’s content is unintelligible without an adjacent tile, which breaks the format’s premise that items can be read in any order.
Every tile has an identical span, so the layout communicates nothing about relative importance and the format has been applied as decoration.
These conditions make the work unacceptable to ship; they are not stylistic suggestions and are not subject to designer discretion.
Responsive adaptation guidance
The column count reduces at narrower viewports and tile spans are re-declared for the reduced grid; spans are never left to reflow automatically, since that is what introduces holes and reordering.
Span variation (Rule 1) must survive the reduction: at a two-column grid, at least one tile spans both columns while others span one.
The no-holes requirement (Rule 5) is re-verified at every breakpoint, because a span that fits a four-column grid frequently leaves a gap in a three-column one.
At the narrowest widths the mosaic legitimately collapses to a single column of stacked tiles; Rules 2, 3, 4, 6, and 7 continue to apply, while Rules 1 and 5 are satisfied trivially.
BentoOverviewIndexAbout
bento
Bento
Arranges content as a tight mosaic of rounded tiles at varying spans, each holding exactly one idea, so a dense overview stays scannable.
Primary actionSecondary
One idea per compartmentSize communicates importanceThe mosaic is fullExplicit placementTiles stand alone
Applying a Neumorphism visual direction
You are being asked to apply a Neumorphism visual direction to a web interface. This document gives you framing and recommendations for that direction. It does not tell you what page, product, or feature to build — apply the guidance below to whatever interface you are already working on, at whatever scope that work requires.
What “Neumorphism” means here
Neumorphism makes every element look pushed out of, or pressed into, one continuous surface. A button is not a colored shape placed on a background; it is the same material as the background, extruded. That illusion is produced by a pair of shadows — a pale one on the side the light comes from, a dark one opposite — with the element and its surroundings sharing exactly the same fill color. Nothing has a border. Depth is the only thing distinguishing one region from another.
That is also the danger, and you should treat it as the central design problem rather than an afterthought. If depth is the only signal, then anyone who cannot resolve a subtle two-tone shadow — on a dim screen, in sunlight, with low vision — cannot find the controls at all. Neumorphic work published without addressing this is genuinely unusable for some of its readers, and it is by far the most common outcome.
The resolution is not to abandon the style. It is to keep the extruded surfaces while making sure that every control, and all text, still meets ordinary contrast thresholds through something other than the shadow: the label, the icon, or a fill shift. Get that right and the direction is pleasant and distinctive. Skip it and the result is decorative and broken.
Recommendations
Choose a background in the middle of the luminance range. Neumorphism does not exist on white or black, because there is no room for a shadow lighter than the surface or darker than it.
Give the element and its container the same fill. If the button is a different color from what it sits on, you are building a card, not an extrusion.
Use two shadows per surface, offset in opposite directions: one lighter than the background, one darker. Keep the light direction identical everywhere on the page, or the surface stops reading as continuous.
Drop borders entirely. A border immediately reads as a separate object laid on top and destroys the effect.
Build a genuine pressed state using an inset shadow, and make sure the resting state is not inset. Raised and recessed are the style’s only two states, so both need to exist and differ clearly.
Verify text and control affordances against the background directly, ignoring the shadows. Shadows are too soft and too local to count as contrast.
Where a control cannot reach the required contrast through shadow alone — which is most of them — give it a text label, a high-contrast icon, or a slight fill shift. That addition is part of the style done properly, not a compromise of it.
Avoid
White or black backgrounds, where one of the two shadows becomes invisible.
Elements filled with a different color from their container.
Borders, outlines, or hard edges of any kind.
Multiple light directions across the same page.
Relying on the shadow alone to indicate that something is interactive.
Fine or small text on the surface, which the low ambient contrast makes disproportionately hard to read.
Nesting extruded surfaces inside extruded surfaces more than one level deep; the illusion stops resolving.
How to check your own work
Before considering the result finished, look at it and ask:
Are the element and its background the same color? If not, this is a card with soft shadows, not neumorphism.
Do all the light shadows fall on the same side across the whole page?
Turn the shadows off entirely, in your head or in the inspector. Can you still tell what is a control and what is not? If the answer is no, add a label, an icon, or a fill shift until the answer is yes.
Look at the screen at reduced brightness or from an angle. Do the controls disappear? That is what a real portion of your readers will experience permanently.
Is there a visible difference between a resting element and a pressed one? If both look raised, the interaction model is missing.
Neumorphism — Specification
Definition
Neumorphism is a web-interface design style in which elements appear extruded from, or pressed into, a single continuous surface. The element and its container share one fill color, and separation is produced entirely by a pair of opposing shadows — one lighter than the surface, one darker — with no borders present. Because depth is the sole visual separator, the style is only shippable when contrast requirements are satisfied independently of the shadows.
Core principles
One continuous material. Elements are not placed on the surface; they are formed from it.
Paired opposing shadows. A light and a dark shadow, offset in opposite directions, produce the extrusion.
A single light source. One consistent light direction across the page keeps the surface reading as continuous.
Two states only. Raised and recessed are the style’s vocabulary, and both must be visibly present.
Depth is not contrast. Legibility and control discoverability are established by something other than the shadow.
Visual characteristics
A mid-luminance solid background, never white or black.
Controls and panels filled with exactly the background color.
Two shadows per surface: one pale, one dark, offset oppositely.
No borders, outlines, or hard edges anywhere.
Inset shadows expressing pressed or selected states.
Generous rounding, which strengthens the impression of moulded material.
Rules
Shared fill. Elements and their container use the same color.
Acceptance criterion: The computed background-color of each control differs from its containing surface by 3% or less in relative luminance.
Paired opposing shadows. Extrusion is produced by exactly two shadows.
Acceptance criterion: Each neumorphic surface has a box-shadow containing exactly two shadows whose offsets are opposite in sign, one lighter and one darker than the background color.
No borders. Separation never uses an edge.
Acceptance criterion: Computed border-width is 0 on every neumorphic surface.
Mid-luminance ground. The background leaves room for both shadows.
Acceptance criterion: The page background is a solid color with a relative luminance between 0.15 and 0.85.
Single light direction. The surface reads as continuous.
Acceptance criterion: Every light shadow on the page shares the same sign of offset on both axes.
Raised and recessed both exist. The interaction model is expressed spatially.
Acceptance criterion: The pressed or selected state uses an inset shadow while the resting state does not, and both states are present in the interface.
Contrast independent of relief. Legibility does not depend on the shadow.
Acceptance criterion: All text measures at least 4.5:1 and every essential control boundary or indicator measures at least 3:1 against the background, with shadows excluded from the measurement.
Avoid
White or black backgrounds, which eliminate one of the two shadows.
Elements filled differently from their container.
Borders, outlines, or hard edges used to reinforce separation.
More than one light direction on a page.
Shadow alone signalling that an element is interactive.
Small or fine text, which the low ambient contrast makes disproportionately hard to read.
Extruded surfaces nested more than one level deep.
Acceptance criteria
Rule 1: Controls and their containers differ by 3% or less in luminance.
Rule 2: Every surface has exactly two opposing shadows, one lighter and one darker.
Rule 3: No neumorphic surface has a border.
Rule 4: The page background luminance falls between 0.15 and 0.85.
Rule 5: All light shadows share one offset direction.
Rule 6: Pressed states use inset shadows and resting states do not.
Rule 7: Text meets 4.5:1 and control indicators meet 3:1 excluding shadows.
Blocking failures
A control’s only indication that it is interactive is its shadow, measured below 3:1, leaving it undiscoverable for readers who cannot resolve soft two-tone depth. This is the defining failure of the style and the reason Rule 7 exists.
Text falls below 4.5:1 against the surface, which the low ambient contrast of the style makes easy to miss and impossible to compensate for with relief.
No pressed state exists, so activation is unconfirmed and the style’s only interaction affordance is absent.
These conditions make the work unacceptable to ship; they are not stylistic suggestions and are not subject to designer discretion.
Responsive adaptation guidance
Shadow offset and blur may scale down proportionally at narrower viewports, but both shadows remain present; dropping to a single shadow flattens the surface and ends the style.
Rule 7 is re-verified at every breakpoint and at increased text scaling, since smaller rendered text on a low-contrast surface is where the style fails first.
Rounding may reduce at narrow widths without affecting any rule.
The shared-fill (Rule 1), border-free (Rule 3), and single-light-direction (Rule 5) rules hold unchanged at every breakpoint.
Where a narrow viewport forces controls close together, spacing is increased rather than adding separators, since a separator line would violate Rule 3.
NeumorphismOverviewIndexAbout
neumorphism
Neumorphism
Extrudes controls from the background itself using paired light and dark shadows, so elements look pressed into or raised out of a single continuous surface.
Primary actionSecondary
One continuous material
Neumorphism is a web-interface design style in which elements appear extruded from, or pressed into, a single continuous surface. The element and its container share one fill color, and separation is produced entirely by a pair of opposing shadows — one lighter than the surface, one darker — with no borders present. Because depth is the sole visual separator, the style is only shippable when contrast requirements are satisfied independently of the shadows.
RestActive
Applying a Flat visual direction
You are being asked to apply a Flat visual direction to a web interface. This document gives you framing and recommendations for that direction. It does not tell you what page, product, or feature to build — apply the guidance below to whatever interface you are already working on, at whatever scope that work requires.
What “Flat” means here
Flat design removes every attempt to imitate physical material. Nothing has a shadow, a bevel, a gradient, or a texture suggesting it exists in three dimensions. What remains is solid color, simple geometry, and clear type — and color takes over the structural job that depth used to do. A region is distinguishable because it is a different color, not because it appears raised.
Flat is not the same as minimal, and conflating them produces a worse version of both. A minimal interface reduces how much is on screen and restricts color to a single accent. Flat can be extremely colorful — several saturated colors distinguishing regions, states, and categories is normal and desirable. What it removes is the illusion of depth, not the quantity of material.
The style has one well-documented failure, and the correction is not optional. When everything is flat, buttons can stop looking like buttons: a piece of colored text and a clickable control become visually identical. The fix is straightforward — every interactive element gets a fill or an outline with real contrast against what surrounds it — and applying it is what separates flat design that works from the version that frustrated people when the style first spread.
Recommendations
Remove all shadows, gradients, bevels, and textures, including the subtle ones a component library may add by default. Check the computed styles rather than trusting the design file.
Let color do the structural work. Give each major region, category, or state its own solid color, and build a palette of at least three saturated colors so the system has range.
Draw icons as flat geometry in one or two tones. No gradient fills, no drop shadows, no faux depth in the icon.
Separate regions with a color change or a thin line, never with a shadow. If two adjacent regions are the same color, they are one region.
Give every control a fill or an outline with a contrast ratio of at least 3:1 against its immediate background. This is the rule that keeps controls findable without depth.
Express state through color and opacity rather than elevation. Hover changes the fill; it does not lift the element.
Use a clean sans-serif with no effects on the text itself — no shadows, no outlines, no gradient fills.
Avoid
Any shadow, including a barely visible one inherited from a UI library default.
Gradients used to add “just a little” depth, which is the most common way this style erodes.
Icons with gradients, shadows, or dimensional shading.
Regions separated by elevation rather than by color.
Text-only controls with no fill or outline, indistinguishable from surrounding copy.
Hover or active states that raise an element rather than changing its color.
How to check your own work
Before considering the result finished, look at it and ask:
Search the computed styles for any box-shadow, text-shadow, or gradient. Every hit is a violation, including the ones you did not write.
Ask someone to point at everything on the screen they believe is clickable. If they miss something or point at something that is not, the controls need stronger fills or outlines.
How many saturated colors are actually doing structural work? If the answer is one, this is closer to a minimal interface than a flat one, and either the palette should grow or the direction should change.
Do the icons still look flat when you enlarge them? Dimensional shading tends to hide at small sizes.
Hover over a control. Does anything change other than color or opacity? If the element moves or lifts, that is depth returning.
Flat — Specification
Definition
Flat is a web-interface design style that eliminates every simulation of physical depth — shadow, gradient, bevel, and texture — and transfers the structural work of separation onto solid color, simple geometry, and clear typography. It is distinguished from reductive styles by its palette: flat interfaces are frequently colorful, because color is the mechanism that replaces depth.
Core principles
No simulated depth. Nothing imitates elevation, material, or light.
Color is structure. Regions, categories, and states are distinguished chromatically rather than spatially.
Geometry over rendering. Icons and shapes are drawn as flat forms, not shaded objects.
Controls announce themselves without relief. Interactivity is carried by fill and outline contrast.
State changes are chromatic. Hover and active alter color or opacity, never height.
Visual characteristics
Solid color regions with no gradient or texture anywhere.
A palette of several saturated colors distinguishing regions and states.
Two-tone geometric icons without shading.
Boundaries formed by color change or a thin line.
Filled or outlined controls that read as controls at a glance.
Clean sans-serif type with no applied effects.
Rules
No simulated depth. The interface contains no device that imitates elevation.
Acceptance criterion: Inspecting computed CSS across the page finds no box-shadow, no text-shadow, and no gradient background-image on any element.
Color forms the boundary. Separation between regions is chromatic or linear.
Acceptance criterion: Each boundary between adjacent regions is produced by a background-color change or a border of 1–2px; no boundary is produced by a shadow.
Flat iconography. Icons are drawn without dimensional rendering.
Acceptance criterion: Every icon uses one or two flat tones and contains no gradient, shadow, or shading.
Broad saturated palette. Color carries structural information.
Acceptance criterion: At least three distinct colors with a saturation of 40% or greater are used to identify regions, states, or categories.
Unaffected typography. Type is rendered plainly.
Acceptance criterion: All text uses a sans-serif family and carries no text-shadow, outline, or textured fill.
Controls announce themselves without relief. Interactivity is visible at rest.
Acceptance criterion: Every interactive control has a fill or border measuring at least 3:1 against its immediate background.
Chromatic state changes. States never reintroduce depth.
Acceptance criterion: Between resting and hover or active, the computed background-color or opacity changes, and no shadow appears in any state.
Avoid
Any shadow, including near-invisible defaults inherited from a component library.
Gradients used to reintroduce a small amount of depth.
Icons with gradients, shadows, or dimensional shading.
Regions separated by elevation instead of color.
Text-only controls with no fill or outline, visually identical to surrounding copy.
Hover or active states that lift or translate an element.
Acceptance criteria
Rule 1: No shadow, text-shadow, or gradient exists anywhere on the page.
Rule 2: Every region boundary is a color change or a 1–2px line.
Rule 3: All icons use flat tones with no shading.
Rule 4: At least three saturated colors carry structural meaning.
Rule 5: All text is sans-serif with no applied effects.
Rule 6: Every control has a fill or border at 3:1 or better against its background.
Rule 7: State changes alter color or opacity only, with no shadow in any state.
Blocking failures
An interactive control is visually indistinguishable from static content, so a visitor cannot tell what can be acted on. This is the historical failure of the style and Rule 6 exists to prevent it.
Color is the only carrier of a distinction that has no textual or shape equivalent, making that information unavailable to a color-blind reader — a risk this style raises specifically because color does so much structural work.
Shadows or gradients are present, which contradicts the definition of the style rather than merely weakening it.
These conditions make the work unacceptable to ship; they are not stylistic suggestions and are not subject to designer discretion.
Responsive adaptation guidance
Region boundaries (Rule 2) may switch from color blocks to thin lines as regions stack into a single column, since adjacent same-color regions would otherwise merge.
The palette (Rule 4) is evaluated per breakpoint: colors dropped along with the regions they identified do not count against the requirement, provided at least three remain structural.
Control contrast (Rule 6) is re-verified after reflow, because a control moving onto a differently colored region changes its measured ratio.
The prohibition on depth (Rules 1 and 7) holds unchanged at every breakpoint; no shadow may be introduced at narrow widths to compensate for lost separation.
FlatOverviewIndexAbout
flat-design
Flat
Removes every simulation of depth and lets solid color, geometric shape, and clear type carry the interface, using color itself as the primary structural device.
Primary actionSecondary
No simulated depthColor is structureGeometry over rendering
01Solid color regions with no gradient or texture anywhere.02A palette of several saturated colors distinguishing regions and states.03Two-tone geometric icons without shading.
Applying a Y2K visual direction
You are being asked to apply a Y2K visual direction to a web interface. This document gives you framing and recommendations for that direction. It does not tell you what page, product, or feature to build — apply the guidance below to whatever interface you are already working on, at whatever scope that work requires.
What “Y2K” means here
Y2K revives the visual optimism of the turn of the millennium, when consumer technology was being marketed as bright, friendly, and imminent. Surfaces look polished and reflective: chrome lettering with a highlight running across it, buttons shaped like liquid drops with a specular gleam on top, silver and electric blue and hot pink at high brightness, small sparkle motifs catching the light. Everything appears to be lit from inside or freshly manufactured.
The reference points are objects, not screens — the translucent plastic and brushed metal of early-2000s consumer electronics, and the way they were photographed for advertising. That is the register to aim for: product gloss, not the muted flat surfaces of contemporary interfaces.
The practical difficulty is that chrome is a gradient, and text on a gradient has no single contrast value. A label that sits comfortably on the light band of a chrome fill can vanish on the dark band a few pixels away. Chrome belongs on display type and ornament, where size carries the legibility; body text belongs on solid ground.
Recommendations
Build chrome as a multi-stop gradient with at least three stops in a light–dark–light sequence. That reversal is what the eye reads as a reflection; a simple two-stop gradient reads as a tint.
Reserve chrome and metallic fills for display type, logos, and ornament. Set body text on a solid color.
Give buttons and highlighted containers a bubble shape: a corner radius of at least half their height, so they read as capsules or drops rather than rectangles.
Add a specular highlight — a distinct paler band across the upper portion of a filled element, separate from the base fill. This is what makes a surface look glossy rather than merely colored.
Push brightness up. Several colors at high luminance and saturation, with at least one silver or chrome value present, is the palette; muted tones are a different decade.
Set headlines heavy: bold weights plus a display family with unusual width, or an outline or dimensional treatment. Restrained type does not carry this direction.
Include a sparkle or starburst motif somewhere, and mark it as decorative so assistive technology skips it.
Measure text over metallic fills at both extremes of the gradient — the lightest stop and the darkest — and treat the worse result as the real one.
Avoid
Body copy set over chrome or any multi-stop gradient.
Two-stop gradients standing in for chrome, which read as a flat tint.
Muted, dusty, or desaturated palettes, which belong to a different era.
Rectangular buttons with small radii, which flatten the bubble language.
A single flat fill on an element that is supposed to look glossy; without the highlight band it is just a colored shape.
Sparkle and starburst motifs left exposed to screen readers as meaningless content.
How to check your own work
Before considering the result finished, look at it and ask:
Does the chrome actually reverse direction — light, then dark, then light? If it only goes light to dark, it is a tint and will not read as metal.
Is there any body text sitting on a gradient? Move it onto a solid fill.
Do the filled elements have a separate, visibly paler band across the top? Without it the surface is not glossy.
Is anything on the page actually bright? A palette of medium-luminance colors will not carry this direction no matter what shapes you use.
Take the darkest stop of every metallic fill that has text on it and measure the contrast there, then repeat with the lightest. Both must pass.
Y2K — Specification
Definition
Y2K is a web-interface design style that revives the glossy consumer-technology optimism of the turn of the millennium. Surfaces read as polished and reflective: multi-stop chrome fills, capsule shapes carrying specular highlights, high-luminance saturated color, heavy display typography, and sparkle ornament. Its visual reference is early-2000s product photography rather than any screen convention.
Core principles
Surfaces are manufactured objects. Elements look moulded, polished, and lit rather than drawn.
Reflection over tint. Metal is expressed by a reversing light–dark–light gradient, not by a single directional fade.
Brightness is the palette. Colors sit high in luminance and saturation, with metal present alongside them.
Type is inflated. Display typography is heavy, wide, or dimensional; restraint belongs to another direction.
Gloss never costs legibility. Reflective fills carry display type and ornament, not reading text.
Visual characteristics
Chrome and metallic gradients on logotypes and headlines.
Capsule and bubble shapes with radii at half their height or more.
A distinct paler highlight band across the upper portion of filled elements.
High-brightness palettes including silver, electric blue, and hot pink.
Heavy display type, often outlined, extruded, or unusually wide.
Four-point sparkles and starburst motifs as ornament.
Rules
Chrome reads as reflection. Metallic surfaces reverse their light direction.
Acceptance criterion: At least one primary element uses a gradient of three or more stops in a light–dark–light sequence.
Inflated display type. Headlines carry weight and unusual form.
Acceptance criterion: Headline elements compute a font-weight of 700 or greater and use a display family of non-standard width, or carry an outline or dimensional treatment.
High-brightness palette. Color sits at the bright end of the range.
Acceptance criterion: At least three colors present measure a relative luminance of 0.6 or greater with saturation of 50% or greater, and at least one silver or chrome value is present.
Specular highlight. Filled surfaces look glossy rather than colored.
Acceptance criterion: At least one filled element has a distinct paler band across its upper portion, rendered separately from its base fill.
Bubble geometry. Containers and buttons read as capsules.
Acceptance criterion: Highlighted containers and buttons have a computed border-radius of at least 50% of their rendered height.
Sparkle ornament, marked decorative. The period motif is present without polluting the accessibility tree.
Acceptance criterion: At least one sparkle or starburst motif is rendered and carries aria-hidden or an equivalent exclusion from assistive technology.
Legibility over metal. Text on reflective fills passes at both extremes.
Acceptance criterion: Text over a metallic surface measures at least 4.5:1 against both the lightest and the darkest stop of that gradient.
Avoid
Body copy set over chrome or any multi-stop gradient.
Two-stop gradients used as chrome, which read as flat tints.
Muted, dusty, or desaturated palettes.
Rectangular buttons with small radii.
Flat single fills on elements intended to look glossy.
Sparkle motifs left exposed to assistive technology as meaningless content.
Acceptance criteria
Rule 1: A three-or-more-stop light–dark–light gradient is present.
Rule 2: Headlines are weight 700+ in a non-standard-width or treated display family.
Rule 3: Three colors meet the brightness threshold and a metallic value is present.
Rule 4: At least one filled element carries a separate specular highlight band.
Rule 5: Buttons and highlighted containers are radiused to half their height or more.
Rule 6: A sparkle motif is present and hidden from assistive technology.
Rule 7: Text over metal passes 4.5:1 at both the lightest and darkest gradient stop.
Blocking failures
Reading text is set over a metallic or multi-stop gradient and fails contrast at one extreme, which cannot be corrected by adjusting the text color because a single color cannot satisfy both ends of the gradient.
Decorative sparkle and starburst motifs are announced by assistive technology, injecting meaningless content into the reading experience.
No metallic or specular treatment is present at all, in which case the work is a bright colorful interface rather than this style.
These conditions make the work unacceptable to ship; they are not stylistic suggestions and are not subject to designer discretion.
Responsive adaptation guidance
Chrome treatments (Rule 1) are preserved on the primary logotype and dropped from secondary elements at narrow viewports rather than being scaled down, since a compressed gradient loses its reversal and stops reading as metal.
Display type (Rule 2) may reduce in size but not in weight or character; substituting a plain family at narrow widths ends the style.
Bubble geometry (Rule 5) is recalculated after reflow, because a button that becomes taller when its label wraps needs a larger radius to remain a capsule.
Rule 7 is re-measured after every reflow, since text moving across a gradient encounters different stops.
Ornament (Rule 6) may be reduced in quantity at narrow widths; whatever remains stays excluded from assistive technology.
Y2KOverviewIndexAbout
y2k
Y2K
Revives the glossy techno-optimism of the turn of the millennium: chrome surfaces, inflated type, specular highlights, and a palette that looks lit from inside.
Primary actionSecondary
Surfaces are manufactured objects
Y2K is a web-interface design style that revives the glossy consumer-technology optimism of the turn of the millennium. Surfaces read as polished and reflective: multi-stop chrome fills, capsule shapes carrying specular highlights, high-luminance saturated color, heavy display typography, and sparkle ornament. Its visual reference is early-2000s product photography rather than any screen convention.
Applying a Technical Blueprint visual direction
You are being asked to apply a Technical Blueprint visual direction to a web interface. This document gives you framing and recommendations for that direction. It does not tell you what page, product, or feature to build — apply the guidance below to whatever interface you are already working on, at whatever scope that work requires.
What “Technical Blueprint” means here
This direction dresses an interface as an engineering document. Faint hairline guides run between elements the way leader lines run across a drawing. Small monospaced labels sit beside things, set in capitals with wide letter spacing. A tick scale or coordinate index runs along an edge. Where a photograph would normally go, there is a line drawing — an exploded view, a schematic, a cross-section. The palette is two neutrals and a single accent, the way a printed drawing is ink, paper, and one revision color.
The register it produces is precision and seriousness: the impression that the thing being described has been measured. That suits work that genuinely has a technical dimension, and it looks borrowed on work that does not.
There is one thing to be careful about, and it is the reason this direction can go wrong in a way most cannot. The aesthetic is built from labels and figures, and it is tempting to scatter labels that look meaningful but say nothing — reference numbers, tolerances, coordinates invented because they complete the picture. On a poster that is harmless. On a product, a number that looks like a measurement will be read as one. Keep ornament clearly ornamental: lines, ticks, and marks that carry no digits. Anything that presents as data must actually be data.
Recommendations
Draw hairline guides at one pixel with deliberately low contrast, so they sit below the text in the visual hierarchy. They connect and frame elements; they are not container borders.
Set micro-labels in a monospaced family at twelve pixels or smaller, uppercase, with positive letter spacing. That combination is most of what makes the register read.
Put a coordinate system somewhere visible: a tick scale along an edge, a numbered index down a column, a grid reference. It is the single element that most signals “measured.”
Hold the palette to two neutrals and one accent. Blueprint blue and drafting white are the obvious reference, but any two-neutral scheme works as long as the accent stays alone.
Replace photography with line work. An exploded diagram, a wireframe view, or a cross-section is the correct illustration here; a photograph collapses the whole effect.
Keep decorative marks non-numeric. Ticks, arrows, brackets, and crop marks are safe ornament. Invented part numbers, dimensions, and tolerances are not.
Keep real information at full contrast while ornament stays faint. The hierarchy should be unambiguous: what you can read matters, what you can barely see is structure.
Avoid
Invented figures — dimensions, tolerances, reference codes, coordinates — presented as though they described the actual subject.
Hairlines drawn at the same contrast as body text, which turns structure into noise.
Photographs used as the primary illustration.
A second accent color, which breaks the ink-paper-revision logic.
Monospaced type used for body copy, where its wide spacing makes reading slower without adding meaning.
Guides so faint they disappear entirely; below roughly 1.5:1 they read as rendering artifacts.
How to check your own work
Before considering the result finished, look at it and ask:
Point at every number on the screen. Can you say where each one comes from? Any number you cannot source is decoration pretending to be data, and it has to go.
Are the guides clearly quieter than the text? If a hairline competes with a label, the hierarchy has inverted.
Is there a coordinate scale, tick index, or grid reference anywhere? Without one the drawing register does not establish.
How many colors are there beyond the two neutrals? If more than one, the palette has drifted.
Is the main illustration a line drawing? If it is a photograph, the rest of the treatment is decoration on top of an ordinary layout.
Technical Blueprint — Specification
Definition
Technical Blueprint is a web-interface design style that presents an interface as an engineering document. Hairline guides, uppercase monospaced micro-labels, a visible coordinate or tick system, line-drawn illustration, and a palette of two neutrals plus one accent combine to produce a register of precision and measurement. Because the style is built from labels and figures, it carries a specific obligation: ornament must never present itself as data.
Core principles
The drawing is the metaphor. Guides, ticks, and labels behave like annotation on a technical drawing rather than like interface chrome.
Ornament is quiet, information is loud. Structural marks sit below text in contrast; anything readable is meant to be read.
Measurement is visible. A coordinate system, index, or scale establishes the register more than any other single element.
Ink, paper, and one revision color. The palette stays at two neutrals plus a single accent.
Decoration never impersonates data. Marks and lines may be meaningless; numbers may not.
Visual characteristics
One-pixel hairline guides at low contrast, connecting or framing elements.
Uppercase monospaced labels at small sizes with positive letter spacing.
A tick scale, numbered index, or grid reference along an edge.
Line-drawn illustration — exploded views, wireframes, cross-sections — in place of photography.
Two neutral colors plus a single accent.
Brackets, crop marks, arrows, and other non-numeric annotation.
Rules
Hairline guides. Structural marks are thin and quiet.
Acceptance criterion: One-pixel lines are present with a measured contrast ratio between 1.5:1 and 3:1 against their background, and are distinct from container borders.
Monospaced micro-labels. Annotation uses drafting typography.
Acceptance criterion: Label elements compute a monospaced font-family at 12px or smaller with a positive letter-spacing.
Visible coordinate system. The interface declares that it has been measured.
Acceptance criterion: At least one tick scale, numbered index, or grid reference is rendered on the page.
Two neutrals and one accent. The palette follows drawing convention.
Acceptance criterion: The rendered palette resolves to at most two neutral colors plus exactly one accent color.
Line work replaces photography. Illustration is drawn, not captured.
Acceptance criterion: The primary illustration is a line drawing, schematic, or diagram; no photograph occupies the primary illustration position.
Ornament carries no figures. Decoration never impersonates measurement.
Acceptance criterion: No decorative label, mark, or annotation displays a numeric value that could be read as a real measurement, specification, or identifier of the subject.
Information outranks ornament. The contrast hierarchy is unambiguous.
Acceptance criterion: All informational text measures at least 4.5:1 while all decorative lines and marks measure below 3:1.
Avoid
Invented dimensions, tolerances, reference codes, or coordinates presented as descriptions of the subject.
Hairlines drawn at the same contrast as body text.
Photography used as the primary illustration.
A second accent color.
Monospaced type used for body copy.
Guides below roughly 1.5:1, which read as rendering artifacts rather than structure.
Acceptance criteria
Rule 1: Hairline guides measure between 1.5:1 and 3:1 and are distinct from borders.
Rule 2: Labels are monospaced, 12px or smaller, with positive letter spacing.
Rule 3: A tick scale, index, or grid reference is present.
Rule 4: The palette is two neutrals plus exactly one accent.
Rule 5: The primary illustration is line work, not a photograph.
Rule 6: No decorative element displays a numeric value readable as real data.
Rule 7: Informational text is at or above 4.5:1 and decorative marks below 3:1.
Blocking failures
A decorative label presents a number that a visitor could reasonably read as a measurement, specification, or identifier of the actual subject. This is the defining risk of the style: on a poster it is harmless, in a product it is false information.
Informational text is rendered at the same low contrast as the structural guides, so what must be read is indistinguishable from what is merely structure.
No coordinate system, tick scale, or index exists anywhere, in which case the work is a monospaced layout rather than this style.
These conditions make the work unacceptable to ship; they are not stylistic suggestions and are not subject to designer discretion.
Responsive adaptation guidance
Guide density reduces at narrower viewports — fewer leader lines, shorter runs — but the remaining guides keep the contrast band required by Rule 1 rather than being made fainter to reduce clutter.
Micro-labels (Rule 2) may drop to a smaller subset at narrow widths, but never grow past 12px, since a larger monospaced label stops reading as annotation and starts competing with body text.
The coordinate system (Rule 3) may shorten or move to a single edge but must remain present at every breakpoint.
Line illustration (Rule 5) is re-cropped or simplified rather than replaced by a photograph at any width.
Rules 6 and 7 hold unchanged at every breakpoint; reducing ornament must never promote a decorative mark into a position where it reads as data.
01One-pixel hairline guides at low contrast, connecting or framing elements.02Uppercase monospaced labels at small sizes with positive letter spacing.03A tick scale, numbered index, or grid reference along an edge.
blueprint
Technical Blueprint
Presents an interface as an engineering drawing — hairline guides, monospaced micro-labels, coordinate marks, and line diagrams instead of photography.
Primary actionSecondary
The drawing is the metaphorOrnament is quiet, information is loudMeasurement is visibleInk, paper, and one revision colorDecoration never impersonates data
Technical Blueprint
Technical Blueprint is a web-interface design style that presents an interface as an engineering document. Hairline guides, uppercase monospaced micro-labels, a visible coordinate or tick system, line-drawn illustration, and a palette of two neutrals plus one accent combine to produce a register of precision and measurement. Because the style is built from labels and figures, it carries a specific obligation: ornament must never present itself as data.
Applying a Handcrafted visual direction
You are being asked to apply a Handcrafted visual direction to a web interface. This document gives you framing and recommendations for that direction. It does not tell you what page, product, or feature to build — apply the guidance below to whatever interface you are already working on, at whatever scope that work requires.
What “Handcrafted” means here
This direction adds evidence of a human hand to an otherwise ordinary interface. An underline that wobbles and changes thickness. An arrow drawn rather than drawn from an icon set. Paper or ink texture behind the content. A photograph that looks taken rather than art-directed. Something rotated by a degree or two, as though placed by hand rather than positioned by a layout engine.
The purpose is warmth and specificity — the sense that a particular person made this, in a landscape of interfaces that all look machine-assembled. The underlying philosophy is the one that treats imperfection as evidence of authorship rather than as a defect.
The distinction the whole style depends on is between human and careless, and they are easy to confuse. The rule that keeps them apart is simple: imperfection belongs to the surface, never to the mechanism. Textures, strokes, crops, and slight rotations are the surface. Buttons, form fields, focus indicators, and hit areas are the mechanism, and they stay exactly as regular and predictable as they would be in any other interface. A wobbly underline is charming; a button you cannot reliably hit is a defect wearing a charming coat.
Recommendations
Introduce at least one genuinely hand-drawn stroke — an underline, a circle around a word, an arrow, a bracket. It must have variable thickness along its length. A one-pixel CSS border curved with a radius does not read as handmade.
Put a material texture behind the content: paper grain, ink bleed, fabric weave. It should be visible at normal viewing size, not a texture that only exists at 400% zoom.
Rotate one or two elements by a small amount — half a degree to three degrees — or offset them from the grid by a few pixels. Beyond that it stops reading as hand-placed and starts reading as broken.
Let photography be unpolished: real grain, an imperfect crop, an off-center subject. Studio-corrected imagery contradicts everything else on the page.
Add exactly one handwritten or irregular typeface, and use it only for accents — a callout, a signature, a label. Never for body copy, which it makes slower to read for everyone and much harder for some.
Keep every interactive element rigorously regular: rectangular hit areas of at least forty-four by forty-four pixels, with a clearly visible focus indicator. This is not a compromise with the style; it is what makes the style trustworthy.
Check text contrast against the texture, at its darkest point and its lightest. Texture is exactly the kind of variable background that quietly breaks legibility.
Avoid
Rotations or offsets large enough to read as a layout bug.
Handwritten typefaces used for body copy or for anything a reader must parse quickly.
Irregular, rotated, or oddly shaped interactive controls and hit areas.
Texture applied so heavily that text sits on unpredictable ground.
Generic stock illustration standing in for hand-drawn marks; the wobble has to be real.
Removing or weakening focus indicators because they look too mechanical against the rest.
How to check your own work
Before considering the result finished, look at it and ask:
Does any hand-drawn element have genuinely variable stroke width? If every line is uniform, it is CSS pretending to be a hand.
Are the interactive elements perfectly regular? Every rotation, irregular edge, or unpredictable hit area on a control is the style leaking into the mechanism, and it needs to be pulled back.
Tab through the page. Is the focus indicator obvious at every stop, against the texture?
Take the smallest text over the busiest part of the texture and measure its contrast at the worst pixel.
Would a stranger describe this as “made by someone” or as “slightly broken”? If there is any doubt, reduce the rotations and keep the strokes.
Handcrafted — Specification
Definition
Handcrafted is a web-interface design style that adds visible evidence of human authorship to an otherwise disciplined interface: strokes of variable thickness, material texture, small deliberate rotations, and unpolished imagery. Its defining constraint is that imperfection is confined to the surface. Interactive mechanisms — controls, hit areas, focus indicators — remain fully regular, which is what separates the style from carelessness.
Core principles
Imperfection is evidence, not error. Irregularity signals that a person made specific decisions.
The hand must be real. Variable stroke width, genuine texture, and true photographic grain cannot be simulated by uniform CSS shapes.
Restraint in the deviation. Small rotations and offsets read as placement; large ones read as breakage.
The mechanism stays regular. Controls, targets, and focus indicators are unaffected by the style.
Warmth never costs legibility. Text over texture is verified at the worst point, not the average.
Visual characteristics
Hand-drawn underlines, arrows, circles, or brackets with varying stroke thickness.
Paper, ink, or fabric texture visible behind the content at normal viewing size.
One or two elements rotated slightly or offset from the grid.
Photography with real grain and imperfect, off-center cropping.
A single handwritten or irregular typeface used only for accents.
Regular, rectangular interactive controls with clear focus states.
Rules
Genuine hand-drawn stroke. At least one mark shows a real hand.
Acceptance criterion: At least one underline, arrow, or frame has measurably variable stroke width along its length; a uniform CSS border does not satisfy this.
Material texture. The ground is a material rather than a fill.
Acceptance criterion: The primary background region carries a texture that is discernible at 100% zoom.
Restrained deviation. Placement reads as manual, not broken.
Acceptance criterion: At least two elements have a rotation between 0.5° and 3°, or an offset from the grid between 2px and 8px.
Unpolished imagery. Photography is captured rather than corrected.
Acceptance criterion: Content photographs retain visible grain or an irregular crop and are not fitted to a perfect geometric frame.
One handwritten counterpoint. Irregular type is an accent only.
Acceptance criterion: Exactly one handwritten or irregular family is present, and it appears only in accent positions, never in body copy.
Imperfection never reaches the mechanism. Controls remain fully regular.
Acceptance criterion: Every interactive target is a regular rectangle of at least 44 × 44px with a visible focus indicator, carrying no rotation or irregular edge.
Legibility over texture. Text is verified against the real background.
Acceptance criterion: Every text element measures at least 4.5:1 against both the lightest and the darkest point of the texture beneath it.
Avoid
Rotations or offsets large enough to read as a layout defect.
Handwritten typefaces used for body copy or for anything requiring fast parsing.
Rotated, irregular, or unpredictably shaped interactive controls.
Texture heavy enough that text sits on unpredictable ground.
Generic stock illustration substituting for genuine hand-drawn marks.
Focus indicators removed or weakened because they look mechanical.
Acceptance criteria
Rule 1: At least one mark has genuinely variable stroke width.
Rule 2: A texture is discernible behind the content at 100% zoom.
Rule 3: Two or more elements carry a 0.5–3° rotation or a 2–8px offset.
Rule 4: Content photography retains grain or an irregular crop.
Rule 5: Exactly one handwritten family is present, used only for accents.
Rule 6: All interactive targets are regular rectangles of 44 × 44px or larger with visible focus.
Rule 7: All text meets 4.5:1 against the texture’s lightest and darkest points.
Blocking failures
An interactive control is rotated, irregularly shaped, or has an unpredictable hit area, so the style’s decorative logic has been applied to a mechanism that people must be able to operate reliably.
A focus indicator has been removed, weakened, or rendered invisible against the texture, making the interface unusable by keyboard.
Body copy is set in a handwritten or irregular face, which slows reading for everyone and prevents it for some readers.
These conditions make the work unacceptable to ship; they are not stylistic suggestions and are not subject to designer discretion.
Responsive adaptation guidance
Texture (Rule 2) may reduce in scale or intensity at narrow viewports, but Rule 7 is re-measured afterward, since a compressed texture changes the pixels beneath any given text block.
Rotations and offsets (Rule 3) reduce toward the lower bound at narrow widths, where a 3° rotation is far more likely to cause overlap or clipping than it is at desktop width.
Hand-drawn marks are re-drawn or cropped rather than scaled non-uniformly, since stretching a stroke destroys the variable-width quality Rule 1 requires.
Rule 6 strengthens rather than relaxes at narrow viewports: the 44 × 44px minimum is a floor for touch input, not a desktop convenience, and it applies at every breakpoint without exception.
HandcraftedOverviewIndexAbout
handmade
Handcrafted
Signals that a person made this — variable hand-drawn strokes, material texture, slight rotations, and unpolished imagery set against otherwise disciplined structure.
Primary actionSecondary
Imperfection is evidence, not error
Hand-drawn underlines, arrows, circles, or brackets with varying stroke thickness.
Paper, ink, or fabric texture visible behind the content at normal viewing size.
Irregularity signals that a person made specific decisions.
Imperfection is evidence, not error
Handcrafted
Handcrafted is a web-interface design style that adds visible evidence of human authorship to an otherwise disciplined interface: strokes of variable thickness, material texture, small deliberate rotations, and unpolished imagery. Its defining constraint is that imperfection is confined to the surface. Interactive mechanisms — controls, hit areas, focus indicators — remain fully regular, which is what separates the style from carelessness.
Applying a Dark Mode visual direction
You are being asked to apply a Dark Mode visual direction to a web interface. This document gives you framing and recommendations for that direction. It does not tell you what page, product, or feature to build — apply the guidance below to whatever interface you are already working on, at whatever scope that work requires.
What “Dark Mode” means here
Dark mode is not a light interface with the colors inverted. Almost every rule that governs a light interface changes when the ground goes dark, and applying the light-mode logic to a dark surface is what produces the harsh, glaring results people complain about.
Three things change in particular. Shadow stops working — a dark shadow on a dark surface is invisible — so elevation has to be expressed by making raised surfaces lighter rather than by casting anything. Saturated color stops working — a color that reads correctly on white will vibrate uncomfortably on near-black — so accents have to be desaturated and lightened. And maximum contrast stops being the goal: pure white text on pure black is the most contrast available and one of the least comfortable things to read, because the bright glyphs bleed into the dark ground. The target is a strong contrast ratio, not the highest one obtainable.
Recommendations
Use a dark grey as your base surface, not pure black. Something in the region of a very dark neutral gives you room to build lighter surfaces above it, which pure black does not.
Express elevation by luminance. Each level closer to the viewer is a measurable step lighter than the level behind it. A modal is lighter than a panel, which is lighter than the page.
Soften the text. An off-white, or white at reduced opacity, reads more comfortably than pure white and reduces the halo effect around glyphs.
Rebuild the accent colors rather than reusing them. Take each accent from the light theme, reduce its saturation and raise its lightness until it reaches 4.5:1 against the dark surface without glowing.
Aim for a contrast ratio somewhere in the strong-but-not-maximal range. Extremely high contrast is measurably compliant and genuinely tiring over long reading sessions.
Leave content images alone. Never invert them. If a bright image feels aggressive against the dark surface, reduce it slightly with an overlay, and only if it does not distort the content.
Carry semantic states across deliberately. Success, warning, and error need their own dark-theme values that still pass contrast, and they must keep their icons and labels so that meaning does not rest on hue alone.
Avoid
Pure black as the base surface, which leaves no room to express elevation.
Drop shadows used as the primary separator between layers.
Accent colors carried over unchanged from a light theme.
Pure white body text.
Inverting images, illustrations, or charts.
Semantic colors that pass on a light background but fail on the dark one.
How to check your own work
Before considering the result finished, look at it and ask:
Is the base surface actually pure black? If yes, lighten it — everything else gets easier.
Take two stacked surfaces and measure them. Is the nearer one lighter? If they are the same and separated only by a shadow, the elevation model is still the light-mode one.
Is the body text pure white? Soften it and read a paragraph. The difference is immediately noticeable.
Look at each accent color. Does it appear to glow or vibrate against the background? That is oversaturation, not brightness.
Check every semantic state at its dark value, not just the neutral text. These are the ones most often carried over unmodified.
Dark Mode — Specification
Definition
Dark Mode is a web-interface design style built on a dark ground, in which elevation is expressed by luminance steps rather than by shadow, color is desaturated to survive on a low-light surface, and contrast is targeted at a strong rather than maximal ratio. It is not an inversion of a light interface: the mechanisms that establish depth, color, and legibility all differ.
Core principles
Elevation is luminance. Surfaces nearer the viewer are lighter, because shadow is invisible on a dark ground.
Color must be rebuilt, not reused. Accents from a light theme oversaturate against darkness and require new values.
Maximum contrast is not the target. Very high contrast produces glare and halation; a strong bounded ratio reads better.
The base is grey, not black. Pure black leaves no room to express levels above it.
Content is never inverted. Images and illustrations keep their own values.
Visual characteristics
A dark neutral base surface rather than pure black.
Stacked surfaces stepping progressively lighter toward the viewer.
Off-white or reduced-opacity body text rather than pure white.
Desaturated, lightened accent colors.
Semantic states re-derived for the dark surface, each with its icon and label.
Content imagery left at its natural values, at most gently attenuated.
Rules
Dark grey base. The ground leaves room for elevation.
Acceptance criterion: The page background is a solid color with a relative luminance between 0.02 and 0.08; #000000 is not used as a base surface.
Elevation by luminance. Depth is expressed by lightening, not by shadow.
Acceptance criterion: Each elevated surface measures a higher relative luminance than the surface beneath it, and shadow is not the primary mechanism of separation.
No pure white text. Body copy is softened.
Acceptance criterion: The primary text color has a lower relative luminance than #FFFFFF, or is applied at an alpha of 0.87 or less.
Desaturated accents. Color is rebuilt for the dark surface.
Acceptance criterion: Each accent color has a lower saturation than its light-theme equivalent and measures at least 4.5:1 against the dark background.
Bounded contrast. Legibility is strong rather than maximal.
Acceptance criterion: The measured contrast ratio between primary text and its background falls between 7:1 and 15:1.
Images are not inverted. Content keeps its own values.
Acceptance criterion: No content image, illustration, or chart carries filter: invert; attenuation, where present, is applied as an overlay only.
Semantic parity. Meaning survives the theme change.
Acceptance criterion: Each semantic state retains its icon and text label, and its color measures at least 4.5:1 against the dark background.
Avoid
Pure black as the base surface.
Drop shadows used as the primary separator between layers.
Accent colors carried over unchanged from a light theme.
Pure white body text.
Inverted images, illustrations, or charts.
Semantic colors validated only against a light background.
Acceptance criteria
Rule 1: The base surface luminance falls between 0.02 and 0.08 and is not pure black.
Rule 2: Each elevated surface is measurably lighter than the one beneath it.
Rule 3: Primary text is softer than pure white.
Rule 4: Every accent is desaturated relative to its light equivalent and meets 4.5:1.
Rule 5: Primary text contrast falls between 7:1 and 15:1.
Rule 6: No content image is inverted.
Rule 7: Every semantic state keeps its icon and label and meets 4.5:1.
Blocking failures
Elevation is expressed only by shadow, so layer order is invisible on the dark ground and stacked surfaces cannot be distinguished.
A semantic state fails contrast on the dark background or has lost the icon and label it carried in the light theme, leaving its meaning dependent on a hue that no longer reads.
Text or an essential control indicator falls below its required ratio because a light-theme color was reused without re-derivation.
These conditions make the work unacceptable to ship; they are not stylistic suggestions and are not subject to designer discretion.
Responsive adaptation guidance
The elevation scale (Rule 2) compresses at narrow viewports as layers collapse, but every surface pair that remains stacked keeps a measurable luminance difference; separators are not substituted for the step.
Rule 5 is re-verified at increased text scaling and at 200% zoom, where larger glyphs make halation more pronounced and may require softening the text further within the permitted band.
Accent and semantic values (Rules 4 and 7) are independent of viewport and hold unchanged at every breakpoint.
Where a narrow layout places content over an attenuated image, Rule 6’s overlay is re-measured so it neither distorts the image nor drops text below its ratio.
Dark ModeOverviewIndexAbout
dark-mode
Dark Mode
Builds the entire interface on a dark ground, expressing elevation through luminance steps rather than shadow and desaturating color so it survives on a low-light surface.
Primary actionSecondary
Elevation is luminance
Dark Mode is a web-interface design style built on a dark ground, in which elevation is expressed by luminance steps rather than by shadow, color is desaturated to survive on a low-light surface, and contrast is targeted at a strong rather than maximal ratio. It is not an inversion of a light interface: the mechanisms that establish depth, color, and legibility all differ.
RestActive
Applying a Barely-There UI visual direction
You are being asked to apply a Barely-There UI visual direction to a web interface. This document gives you framing and recommendations for that direction. It does not tell you what page, product, or feature to build — apply the guidance below to whatever interface you are already working on, at whatever scope that work requires.
What “Barely-There UI” means here
This direction reduces the weight of an interface rather than its contents. Type is set light — thin weights, generous size, almost no bold. Container edges are drawn just above the threshold of perception, so you can find them when you look but never notice them when you do not. Color is held to one hue with variations in lightness. The page is mostly air. And in the middle of all that quiet, there is usually real data: a chart, a series, a set of live numbers doing the visual work that decoration would do elsewhere.
The distinction from a minimal interface is worth being precise about, because they look superficially similar and are built on opposite moves. A minimal interface removes elements: fewer things on the page, one accent color, nothing that does not earn its place. This one keeps the elements — including dense data displays that a minimal treatment would consider excessive — and instead removes their visual force. A screen full of numbers can be barely-there. It cannot be minimal.
The risk is obvious and specific. If boundaries sit below the threshold of attention by design, controls can fall below the threshold of discovery, which is a different and much worse thing. The resolution is to exempt interactive elements: containers stay whisper-quiet, but anything you can act on keeps a real, measurable edge and an unmistakable focus state.
Recommendations
Set the interface type light. A weight of 300 or below for body and interface text, and no heavier than 400 for headings; hierarchy comes from size and space, not from weight.
Draw container boundaries deliberately faint — findable on inspection, invisible at a glance. A very low contrast against the surrounding surface is the intent, not an accident.
Hold the palette to one hue with lightness variations, plus at most one accent. Everything neutral, one thing that is not.
Give the page real emptiness. Content covering less than half the initial viewport is the target, and more air is usually better.
Put actual data on the screen. A chart, a sparkline, a live metric. Without it this direction has nothing to look at and reads as an unfinished page rather than a considered one.
Remove anything that is neither a control, a datum, nor content. There is no room for decorative elements here; the emptiness is the decoration.
Exempt every interactive element from the quietness. Controls keep a fill or border with real contrast, and focus indicators are unambiguous. Nothing about the direction requires controls to be hard to find.
Avoid
Bold weights used to reintroduce hierarchy that size and space should establish.
Container edges so faint they cannot be found even when looked for.
Interactive controls styled with the same near-invisible edges as static containers.
A second accent color, which immediately breaks the single-hue discipline.
Decorative elements added to fill the emptiness.
A page with no data on it, which leaves the direction with nothing to justify its restraint.
How to check your own work
Before considering the result finished, look at it and ask:
Is there any bold text on the page? If so, ask whether size or spacing could do that job instead.
Ask someone to point at everything clickable. If they hesitate on anything, that control needs a real edge — and giving it one costs the direction nothing.
Tab through the interface. Is every focus state obvious? Faint borders make weak focus rings much worse.
How much of the first screen is covered? If it is more than half, the composition is dense rather than quiet.
Is there real data anywhere? If the page is only text and whitespace, this is a minimal interface, and the direction should be named accordingly.
Barely-There UI — Specification
Definition
Barely-There UI is a web-interface design style that reduces the visual weight of an interface rather than the quantity of its contents. Type is set at light weights, container boundaries are drawn just above the threshold of perception, the palette holds to a single hue, and most of the surface is empty — while real data displays carry the visual interest. It differs from reductive styles in that elements are quieted rather than removed, so a data-dense screen can satisfy it.
Core principles
Weight is what gets removed. Elements remain; their visual force does not.
Boundaries sit at the threshold. Edges are findable on inspection and unnoticed at a glance.
One hue, varied by lightness. Color carries no structural load beyond a single accent.
Emptiness is the composition. Air, not ornament, is what the surface is made of.
Quietness never applies to controls. Interactive elements are exempt, without exception.
Visual characteristics
Light typographic weights throughout, with hierarchy from size and space.
Container edges at very low contrast against their surroundings.
A single-hue palette varying only in lightness, plus at most one accent.
Large uncovered areas across the initial viewport.
Data visualizations — charts, series, live metrics — as the primary visual material.
Clearly bounded interactive controls that stand apart from the quiet containers.
Rules
Light weight as the default voice. Type carries no unnecessary force.
Acceptance criterion: Interface and body text compute a font-weight of 300 or less, and headings compute no more than 400.
Boundaries below the threshold of attention. Container edges are deliberately faint.
Acceptance criterion: Container borders measure a contrast ratio between 1.1:1 and 1.8:1 against their adjacent surface.
Real data present. The direction has something to show.
Acceptance criterion: At least one genuine data visualization is rendered in the primary view.
Single-hue palette. Color does no structural work.
Acceptance criterion: All colors on the page share one hue and vary only in lightness, with at most one additional accent color.
Air dominates. The surface is mostly empty.
Acceptance criterion: Rendered content covers 40% or less of the initial viewport area.
Nothing non-functional. Emptiness replaces ornament.
Acceptance criterion: Every non-text element on the page is a control, a datum, or a content image.
Controls are exempt from the quietness. Interactivity remains discoverable.
Acceptance criterion: Despite Rule 2, every interactive control has a fill or border measuring at least 3:1 against its background, and a focus indicator measuring at least 3:1.
Avoid
Bold weights reintroducing hierarchy that size and space should carry.
Container edges so faint they cannot be found on deliberate inspection.
Controls styled with the same near-invisible edges as static containers.
A second accent color.
Decorative elements added to occupy the emptiness.
A page with no data on it, which leaves the restraint unjustified.
Acceptance criteria
Rule 1: Interface text is weight 300 or below; headings 400 or below.
Rule 2: Container borders measure between 1.1:1 and 1.8:1.
Rule 3: At least one real data visualization is present in the primary view.
Rule 4: All colors share one hue, with at most one accent.
Rule 5: Content covers 40% or less of the initial viewport.
Rule 6: Every non-text element is a control, a datum, or a content image.
Rule 7: Every control has a 3:1 boundary and a 3:1 focus indicator.
Blocking failures
An interactive control is drawn at the same sub-threshold contrast as a static container, making it undiscoverable. Rule 2 applies to containers only, and applying it to controls is the failure this style is most likely to produce.
A focus indicator measures below 3:1, which the faint surrounding boundaries make disproportionately damaging, since there is no other visual anchor for a keyboard user.
Body text is set at a light weight and a small size such that it falls below 4.5:1, trading legibility for lightness.
These conditions make the work unacceptable to ship; they are not stylistic suggestions and are not subject to designer discretion.
Responsive adaptation guidance
Coverage (Rule 5) is evaluated per breakpoint. At narrow widths the 40% target may rise to 60% as content stacks, since vertical stacking necessarily fills more of a short viewport; this is a legitimate relaxation.
Type weight (Rule 1) may increase to 400 for body text at small sizes or high zoom levels, where thin strokes degrade; this is documented as an explicit exception rather than allowed to drift.
Boundary contrast (Rule 2) holds unchanged, since it is a perceptual threshold rather than a spatial one.
Rule 7 holds unchanged at every breakpoint and is re-verified at 200% zoom, where thin focus rings are most likely to be lost.
Data visualizations (Rule 3) are simplified rather than removed at narrow widths; removing the last one leaves the style without its subject.
Barely-There UIOverviewIndexAbout
hyperminimal
Barely-There UI
Reduces the interface almost to its data: light type weights, boundaries near the threshold of perception, and vast air, with charts doing the visual work.
Primary actionSecondary
Weight is what gets removed
Barely-There UI is a web-interface design style that reduces the visual weight of an interface rather than the quantity of its contents. Type is set at light weights, container boundaries are drawn just above the threshold of perception, the palette holds to a single hue, and most of the surface is empty — while real data displays carry the visual interest. It differs from reductive styles in that elements are quieted rather than removed, so a data-dense screen can satisfy it.
RestActive
Applying a Retrowave visual direction
You are being asked to apply a Retrowave visual direction to a web interface. This document gives you framing and recommendations for that direction. It does not tell you what page, product, or feature to build — apply the guidance below to whatever interface you are already working on, at whatever scope that work requires.
What “Retrowave” means here
Retrowave renders a future that never happened: specifically, the future as imagined in the 1980s. A wireframe grid recedes toward a horizon. A banded sun sits in a gradient running from magenta through orange. Neon glows against a nearly black sky. Display type is chromed, extruded, or outlined. Palm trees, wireframe mountains, and speed lines appear as landscape.
It is not nostalgia for the actual 1980s — that would be a different palette and a different set of objects entirely. It is nostalgia for that decade’s science fiction, which is why everything looks synthetic, and why the sun is always setting.
Two constraints matter technically. The direction requires a dark ground: neon is light emitted in darkness, and on a pale background the glow simply disappears. And the glow itself must stay away from anything you actually have to read — a soft colored halo around a headline is the effect working, while the same halo around a paragraph makes every letter blur into the next.
Recommendations
Build a perspective grid or a set of receding lines somewhere in the composition. It is the single most recognizable element of the direction, and without it the rest reads as generic neon.
Run a sunset gradient through magenta or violet into orange or cyan. Broad, smooth, and unmistakable — this gradient is the light source for the whole scene.
Produce glow with a colored shadow at zero offset and a generous blur, so the light radiates evenly rather than falling to one side. An offset shadow reads as a drop shadow, which is the wrong effect entirely.
Keep the background near black. A dark navy or deep purple works; anything approaching mid-grey kills the neon.
Set headlines in a period display face — chrome-filled, extruded, or outlined — sized large enough for the treatment to be legible.
Include at least one landscape motif: a banded sun, palms, wireframe mountains, or speed lines. One is enough; four is a poster rather than an interface.
Keep glow off body copy. Reading text sits plain on the dark background at full contrast, and the light effects stay on headlines and ornament.
Avoid
A light background, on which the entire direction stops functioning.
Offset shadows standing in for glow.
Glow applied to paragraphs, labels, or form fields.
A palette of neon colors with no gradient anywhere, which loses the sunset that motivates the light.
Every element treated with a glow, which leaves nothing dark to glow against.
Grid, sun, palms, mountains, and speed lines all at once.
How to check your own work
Before considering the result finished, look at it and ask:
Is the background dark enough? If you can imagine reading a document on it comfortably, it is too light for neon.
Does the glow have any offset? Set it to zero — glow radiates in all directions, shadows do not.
Is there anything receding into the distance? Without a grid or perspective lines this is a neon interface rather than this direction.
Is any body text glowing? Remove it and check that the plain version passes contrast on the dark background.
How many landscape motifs are on screen? More than two and the interface is competing with its own background.
Retrowave — Specification
Definition
Retrowave is a web-interface design style that renders the future as imagined in the 1980s. A perspective grid recedes toward a horizon, a sunset gradient runs through magenta and orange, neon glow radiates against a near-black ground, and display typography is chromed, extruded, or outlined. The style depends structurally on darkness, since neon is emitted light, and it confines its glow to headlines and ornament so that reading text remains legible.
Core principles
A synthetic landscape. The composition depicts a place — grid, horizon, sun — rather than arranging abstract shapes.
Emitted light, not cast shadow. Glow radiates evenly from the element; nothing falls to one side.
Darkness is structural. The direction ceases to function on a pale background.
The sunset is the light source. A broad magenta-to-orange gradient motivates every other bright value.
Effects stay off reading text. Glow belongs to display type and ornament only.
Visual characteristics
A wireframe perspective grid or receding line set.
A broad gradient from magenta or violet into orange or cyan.
Colored glow at zero offset with generous blur.
A near-black background, sometimes tinted navy or deep purple.
Chrome-filled, extruded, or outlined display typography.
Period landscape motifs: banded sun, palms, wireframe mountains, speed lines.
Rules
Perspective grid present. The scene recedes.
Acceptance criterion: A perspective grid or a set of converging vanishing lines is rendered as a compositional element.
Sunset gradient. The light source is depicted.
Acceptance criterion: At least one primary gradient runs from magenta or violet into orange or cyan.
Glow, not shadow. Light radiates rather than falls.
Acceptance criterion: Highlighted elements use a text-shadow or box-shadow in a saturated color with a blur radius of 8px or greater and an offset of 0 on both axes.
Period display type. Headlines carry the era.
Acceptance criterion: Headline elements use a display family with a chrome fill, outline, or extrusion treatment.
Dark ground required. Neon has something to glow against.
Acceptance criterion: The page background has a relative luminance of 0.1 or less.
Landscape motif present. The scene is identifiable.
Acceptance criterion: At least one of a banded sun, palm forms, wireframe mountains, or speed lines is rendered.
Reading text stays plain. Effects do not reach body copy.
Acceptance criterion: Body text carries no colored shadow and measures at least 4.5:1 against the dark background.
Avoid
A light background, on which the direction stops functioning entirely.
Offset shadows substituted for glow.
Glow applied to paragraphs, labels, or form fields.
Neon colors used with no gradient present, losing the sunset that motivates them.
Glow applied to every element, leaving nothing dark for it to register against.
All available landscape motifs used simultaneously.
Acceptance criteria
Rule 1: A perspective grid or converging line set is present.
Rule 2: A magenta-or-violet to orange-or-cyan gradient is present.
Rule 3: Glow uses zero offset with a blur of 8px or more.
Rule 4: Headlines use a treated period display family.
Rule 5: The background luminance is 0.1 or less.
Rule 6: At least one period landscape motif is rendered.
Rule 7: Body text is unglowed and meets 4.5:1 on the dark ground.
Blocking failures
Body copy carries a colored glow, which blurs adjacent glyphs into one another and makes sustained reading impossible regardless of the measured contrast ratio.
The background is light, so no glow registers and the direction has been applied as a set of colors rather than as a lighting model.
A form field or control is rendered with glow as its only boundary, making its edges indeterminate and its focus state indistinguishable from its resting state.
These conditions make the work unacceptable to ship; they are not stylistic suggestions and are not subject to designer discretion.
Responsive adaptation guidance
The perspective grid (Rule 1) is re-drawn for the narrower aspect rather than scaled, since a horizontally compressed grid loses its convergence and stops reading as depth.
Landscape motifs (Rule 6) reduce to one at narrow viewports; at least one remains at every breakpoint.
Glow blur (Rule 3) may reduce for performance but never below 8px, and the offset stays at 0 at every breakpoint.
The dark ground (Rule 5) and the plain-body-text requirement (Rule 7) hold unchanged, and Rule 7 is re-verified whenever reflow moves body text over the gradient rather than the flat background.
Display type (Rule 4) may reduce in size but keeps its treatment; substituting a plain face at narrow widths ends the style.
RetrowaveOverviewIndexAbout
retrowave
Retrowave
Renders an imagined 1980s future: a perspective grid running to a sunset gradient, neon glow on dark ground, and chrome display type.
Primary actionSecondary
A synthetic landscape
Retrowave is a web-interface design style that renders the future as imagined in the 1980s. A perspective grid recedes toward a horizon, a sunset gradient runs through magenta and orange, neon glow radiates against a near-black ground, and display typography is chromed, extruded, or outlined. The style depends structurally on darkness, since neon is emitted light, and it confines its glow to headlines and ornament so that reading text remains legible.
Applying a Collage visual direction
You are being asked to apply a Collage visual direction to a web interface. This document gives you framing and recommendations for that direction. It does not tell you what page, product, or feature to build — apply the guidance below to whatever interface you are already working on, at whatever scope that work requires.
What “Collage” means here
Collage assembles a composition from fragments that visibly came from somewhere else. A photographic cutout with a torn edge overlaps a hand-drawn shape, which overlaps a piece of printed type; a person is larger than the building behind them; several pieces sit at angles. Nothing is blended, and that is the point — the seams between materials are the subject, not a flaw to be smoothed away.
The register this produces is handmade and slightly surreal, closer to a physical scrapbook or a cut-paper poster than to anything rendered. It carries editorial and cultural work well, and it fights any interface that needs to feel systematic.
The one place it needs discipline is text. Running text over a busy cutout is unreadable no matter what color you pick, because the background changes every few pixels. In a physical collage the solution is what a designer would actually do: paste a clean piece of paper over the fragments and write on that. The same move works here.
Recommendations
Assemble at least three fragments with genuinely irregular edges — torn, cut, or hand-traced. A rectangular crop is not a cutout, and neither is a circle.
Break scale on purpose. Take two things that would be similar in size in reality and make one several times larger. That mismatch is what makes a collage read as a collage instead of as a busy layout.
Overlap in a stated order. Fragments should cross each other with a deliberate front-to-back sequence, not merely sit near one another.
Mix material kinds. Photography with illustration, illustration with treated type, type with texture. Two kinds is enough; the point is that they are visibly not the same substance.
Rotate pieces off-axis by more than a few degrees. Unlike directions where a slight tilt reads as a hand, here the angles want to be obvious.
Give text its own clean ground. A solid panel, a band of flat color, or an area of the composition kept deliberately empty — then set the text there at full contrast.
Consider letting a fragment run past the edge of the frame. Collage tends to look composed rather than contained.
Avoid
Rectangular or circular crops used as fragments, which read as a photo grid rather than a collage.
Fragments arranged neatly with no overlap.
All fragments at similar scale, which loses the dislocation that defines the style.
Running text placed directly over a busy cutout.
Blending, feathering, or blurring the seams between fragments.
A single material — a page of photographs alone is a gallery, not a collage.
How to check your own work
Before considering the result finished, look at it and ask:
Are any of the edges perfectly straight or perfectly round? Those pieces are not reading as cut, and they will pull the whole composition toward a conventional layout.
Is anything unexpectedly large or small relative to the things around it? If everything is proportionate, the surreal quality is missing.
Do the fragments actually cross one another, or are they merely adjacent?
Find every piece of running text. Is it on a clean ground? If any sits on a fragment, move it or give it a panel.
How many different material kinds are present? If the answer is one, the seams have nothing to be between.
Collage — Specification
Definition
Collage is a web-interface design style that assembles a composition from visibly heterogeneous fragments — photographic cutouts, illustration, treated type, texture — layered at clashing scales and angles with their seams left exposed. Nothing is blended between materials. Because fragment surfaces are visually unpredictable, running text is placed on its own clean ground rather than over the assembly.
Core principles
The seam is the subject. Boundaries between materials are shown, never smoothed.
Scale is dislocated deliberately. Mismatched proportions produce the surreal quality that defines the style.
Layering is ordered. Fragments cross one another in a stated front-to-back sequence.
Materials are mixed. A single material produces a gallery, not a collage.
Text gets clean ground. Reading copy sits on a solid surface rather than on the assembly.
Visual characteristics
Three or more fragments with torn, cut, or hand-traced edges.
Objects rendered at proportions unrelated to their real relative sizes.
Overlapping pieces with explicit stacking order.
Photography, illustration, treated type, and texture coexisting.
Fragments rotated visibly off-axis.
Text confined to solid panels or deliberately empty regions.
Rules
Irregular cut edges. Fragments read as cut rather than cropped.
Acceptance criterion: At least three images have irregular, non-geometric edges; rectangular and circular crops do not satisfy this rule.
Dislocated scale. Proportion is broken on purpose.
Acceptance criterion: At least two elements representing objects of comparable real-world size differ in rendered scale by a factor of 3 or more.
Ordered overlap. Layering is explicit.
Acceptance criterion: At least three elements overlap with a declared z-index.
Mixed materials. The assembly is heterogeneous.
Acceptance criterion: At least two distinct material treatments coexist among photography, illustration, type-as-image, and texture.
Off-axis rotation. Pieces are visibly angled.
Acceptance criterion: At least two elements carry a rotation greater than 3°.
No blended seams. Boundaries remain visible.
Acceptance criterion: No fragment boundary uses feathering, blur, or a blend mode that dissolves the edge into the layer beneath it.
Text on its own ground. Reading copy is never set over the assembly.
Acceptance criterion: Every text element sits on a solid panel or a uniform background region and measures at least 4.5:1 against it; no running text is placed directly over a busy fragment.
Avoid
Rectangular or circular crops used as fragments.
Fragments arranged tidily with no overlap.
Uniform scale across all fragments.
Running text placed directly over a cutout.
Feathering, blurring, or blending the seams between fragments.
A composition made from a single material.
Acceptance criteria
Rule 1: At least three fragments have irregular, non-geometric edges.
Rule 2: Two comparable objects differ in rendered scale by 3× or more.
Rule 3: At least three elements overlap with declared stacking order.
Rule 4: At least two distinct material treatments are present.
Rule 5: At least two elements are rotated more than 3°.
Rule 6: No fragment boundary is feathered, blurred, or blended away.
Rule 7: All text sits on solid ground at 4.5:1 or better.
Blocking failures
Running text is set directly over a photographic or illustrated fragment, where the background varies at the scale of individual glyphs and no single text color can remain legible across it.
Meaning depends on the spatial arrangement of overlapping fragments alone, so the content is unavailable to anyone reading the page linearly or through assistive technology.
Interactive controls are placed over fragments without their own ground, leaving their boundaries and focus states indeterminate.
These conditions make the work unacceptable to ship; they are not stylistic suggestions and are not subject to designer discretion.
Responsive adaptation guidance
Fragment count reduces at narrower viewports rather than fragments being scaled down uniformly; three irregular pieces (Rule 1) is the floor, below which the composition stops reading as an assembly.
Scale dislocation (Rule 2) is preserved by keeping the enlarged element large and shrinking its counterpart, not by normalizing both toward a middle size.
Rotations (Rule 5) may reduce toward 3° where a steeper angle would cause clipping at narrow widths, but at least two elements stay visibly off-axis.
Rule 7 is re-verified after every reflow without exception, since a text panel that clears the fragments at desktop width frequently lands on them once the layout collapses.
Text panels grow rather than shrink as the viewport narrows, since the available clean ground is what limits how much copy the composition can carry.
cut-paper
The seam is the subjectScale is dislocated deliberatelyLayering is orderedMaterials are mixedText gets clean ground
Collage
The seam is the subject
Collage is a web-interface design style that assembles a composition from visibly heterogeneous fragments — photographic cutouts, illustration, treated type, texture — layered at clashing scales and angles with their seams left exposed. Nothing is blended between materials. Because fragment surfaces are visually unpredictable, running text is placed on its own clean ground rather than over the assembly.
Applying a Constructivist visual direction
You are being asked to apply a Constructivist visual direction to a web interface. This document gives you framing and recommendations for that direction. It does not tell you what page, product, or feature to build — apply the guidance below to whatever interface you are already working on, at whatever scope that work requires.
What “Constructivist” means here
Constructivist composition is built on a diagonal. A wedge, a bar, or a line crosses the frame at a steep angle, and everything else — shapes, type, images — organizes around that axis rather than around a horizontal grid. Primitive geometry does structural work: a circle is not an ornament, it is a mass balancing the composition. The palette is severe, usually red, black, and white. Weight collects on one side rather than distributing evenly, and elements are frequently cut off by the edge of the frame rather than politely contained within it.
The effect is urgency. This is a compositional language developed for posters that had to stop someone crossing a street, and it retains that quality: assertive, kinetic, and impossible to read passively. It suits announcements and statements. It fights any interface whose job is to feel calm or neutral.
One practical note, because it is where implementations usually go wrong: the angled type is real text. Rotating a text block with CSS keeps it selectable, searchable, and available to a screen reader in the correct order. Exporting the same headline as an image throws all of that away for no visual gain.
Recommendations
Establish the diagonal axis first, somewhere between fifteen and forty-five degrees, then compose everything else in relation to it. The angle is the structure, not a decoration applied afterward.
Hold the palette to three colors: black, white, and one saturated red. If red is wrong for the context, substitute a single other primary — but it stays one color, not two.
Use primitive shapes as structural masses. A circle, a triangle, a heavy bar. They should be large enough to carry compositional weight rather than sitting in a corner as accents.
Set headlines heavy and let at least one text block share the diagonal’s angle. Type is a construction material here, not a caption on a picture.
Load one side. Concentrate the visual mass in one region and let the opposite side carry much less; symmetry is the one thing this direction cannot tolerate.
Let something bleed off the edge. A shape cut by the frame implies that the composition continues beyond it, which is where much of the energy comes from.
Rotate real text elements rather than exporting typographic images. The visual result is identical and everything else about it is better.
Avoid
A symmetrical or centered composition.
A fourth color, which immediately dissolves the severity.
Geometric shapes used as small decorative accents rather than as structural mass.
Every element contained politely inside the margins.
Headline type set as an image in order to achieve an angle.
Angles so shallow they read as a mistake — below roughly fifteen degrees, a tilt looks like a misalignment rather than a decision.
How to check your own work
Before considering the result finished, look at it and ask:
Is there one clear diagonal that organizes the page, or several unrelated angles? Multiple competing axes read as chaos rather than construction.
Draw a vertical line down the middle. Is one side clearly heavier? If the two sides are balanced, the composition is centered and the direction has not been applied.
How many colors are there? Three is the budget.
Does anything run off the edge of the frame? If everything is contained, the composition is static.
Select the angled headline with your cursor. Does it select as text? If not, it is an image, and it should not be.
Constructivist — Specification
Definition
Constructivist is a web-interface design style organized around a strong diagonal axis, in which primitive geometric shapes act as structural mass, the palette is restricted to black, white, and one saturated primary, and visual weight is deliberately loaded to one side. Elements bleed past the frame rather than being contained by it. The resulting register is urgent and assertive rather than neutral.
Core principles
The diagonal is the structure. A steep axis organizes the composition; everything else relates to it.
Geometry carries mass. Circles, triangles, and bars are structural elements, not accents.
Severity of palette. Three colors, one of which is a single saturated primary.
Weight is loaded, not balanced. Symmetry is the one arrangement the style cannot accommodate.
The frame is a crop, not a container. Elements continue past the edge.
Visual characteristics
One dominant diagonal between roughly fifteen and forty-five degrees.
Large primitive shapes acting as compositional counterweights.
Black, white, and a single saturated red or other primary.
Heavy sans-serif headlines, at least one running along the diagonal.
Visual mass concentrated in one region of the frame.
At least one element cut off by the viewport edge.
Rules
Dominant diagonal. The composition has a stated axis.
Acceptance criterion: At least one major compositional axis is rotated between 15° and 45°.
Three colors at most. The palette stays severe.
Acceptance criterion: The rendered palette resolves to black, white, and exactly one saturated primary color.
Primitives as structure. Geometry does compositional work.
Acceptance criterion: At least two primitive shapes — circle, triangle, or bar — function as compositional elements rather than as decorative accents, each occupying a substantial share of the layout area.
Heavy type, some angled. Type is construction material.
Acceptance criterion: Headlines compute a font-weight of 700 or greater, and at least one text block shares the angle of the diagonal axis.
Loaded asymmetry. Weight collects on one side.
Acceptance criterion: At least 65% of the covered area falls on one side of the vertical center axis.
Bleeding edge. The frame crops the composition.
Acceptance criterion: At least one element extends beyond the visible area rather than terminating inside the margins.
Rotated text remains text. Angled type keeps its semantics.
Acceptance criterion: Every rotated text element remains selectable text in the DOM with correct reading order; no headline is substituted by an image to achieve its angle.
Avoid
Symmetrical or centered composition.
A fourth color.
Geometric shapes used as small decorative accents.
Every element contained within the margins.
Headline type exported as an image in order to achieve rotation.
Angles below roughly fifteen degrees, which read as misalignment rather than intent.
Acceptance criteria
Rule 1: A major axis is rotated between 15° and 45°.
Rule 2: The palette is black, white, and exactly one saturated primary.
Rule 3: At least two primitive shapes act as structural mass.
Rule 4: Headlines are weight 700+ and at least one text block follows the diagonal.
Rule 5: At least 65% of covered area falls on one side of the center axis.
Rule 6: At least one element bleeds past the visible frame.
Rule 7: All rotated text is selectable text with correct reading order.
Blocking failures
Angled headline text is rendered as an image, removing it from selection, search, translation, and assistive technology for a visual result that CSS rotation achieves identically.
Rotation places text at an angle steep enough to be unreadable, or reading order in the DOM no longer matches the intended sequence of the composition.
An element bleeding past the frame carries content that exists nowhere else, so information is cropped away rather than implied to continue.
These conditions make the work unacceptable to ship; they are not stylistic suggestions and are not subject to designer discretion.
Responsive adaptation guidance
The diagonal (Rule 1) is re-angled for the narrower aspect rather than preserved numerically; an angle that reads as dynamic across a wide frame becomes nearly vertical in a narrow one.
Rotated text (Rule 4) reduces in quantity as the viewport narrows, since angled type consumes far more width; at least the primary headline retains the angle, or the rule is documented as suspended at that breakpoint.
Asymmetric loading (Rule 5) is preserved vertically when the layout stacks: mass collects toward the top or bottom rather than being distributed evenly down the column.
Bleed (Rule 6) is re-verified after reflow, since an element that crossed the edge at desktop width often fits entirely within a narrower one.
Rules 2 and 7 hold unchanged at every breakpoint.
constructivism
The diagonal is the structureGeometry carries massSeverity of palette
Constructivist
The diagonal is the structure
Constructivist is a web-interface design style organized around a strong diagonal axis, in which primitive geometric shapes act as structural mass, the palette is restricted to black, white, and one saturated primary, and visual weight is deliberately loaded to one side. Elements bleed past the frame rather than being contained by it. The resulting register is urgent and assertive rather than neutral.
Constructivist
Constructivist is a web-interface design style organized around a strong diagonal axis, in which primitive geometric shapes act as structural mass, the palette is restricted to black, white, and one saturated primary, and visual weight is deliberately loaded to one side. Elements bleed past the frame rather than being contained by it. The resulting register is urgent and assertive rather than neutral.
Applying a Vaporwave visual direction
You are being asked to apply a Vaporwave visual direction to a web interface. This document gives you framing and recommendations for that direction. It does not tell you what page, product, or feature to build — apply the guidance below to whatever interface you are already working on, at whatever scope that work requires.
What “Vaporwave” means here
Vaporwave builds its surface out of discarded digital material. Pale pink and cyan gradients. A window frame borrowed from an operating system nobody supports anymore. A marble bust dropped into the middle of a web page. A checkerboard floor receding into nothing. A heading whose color channels have slipped a few pixels apart. Compositions tend to be symmetrical and oddly still, like a screensaver.
The tone is nostalgic and slightly ironic at the same time — an affection for the commercial digital culture of the 1990s combined with an awareness that it was mostly selling you something. That doubled quality is why the material is always secondhand: obsolete interfaces, stock imagery, classical sculpture pulled out of a museum context.
It is worth distinguishing it from the neon-on-black direction it is often confused with. That one is dark, kinetic, and imagines a future. This one is pale, still, and looks backward at a past that was itself imagining a future. Almost every visual property is opposite.
Recommendations
Build the ground from pale magenta and cyan. Mid-range saturation and high lightness — washed out rather than vivid, which is what separates this from ordinary neon.
Include a real image artifact somewhere: channel separation on a heading, visible color banding in a gradient, scanlines, a compression block. The material should look degraded by having been copied too many times.
Use classical statuary — a bust, a torso, a column. Use a source you are actually entitled to use; museum open-access collections are the straightforward route and many are explicitly licensed for reuse.
Quote an obsolete interface: a window frame with a title bar, a beveled button, a scrollbar rendered the way one looked thirty years ago. One quoted component is enough to set the register.
Compose symmetrically. Center the main elements on a vertical axis. The stillness is part of the effect, and asymmetry pulls the composition toward a different style entirely.
If you use a non-Latin writing system, use real, correct text with a documented translation. The style has a long habit of treating Japanese characters as texture, which produces either nonsense or a language used as ornament. Neither is good work, and correct text costs nothing.
Keep informational text clean. Glitch and channel offset belong on display type and ornament; a paragraph with separated channels is unreadable.
Avoid
Vivid, high-saturation neon, which belongs to a darker and more kinetic direction.
A composition with no artifact anywhere, which reads as pastel decoration rather than as degraded material.
Non-Latin characters chosen for their shapes, with no meaning or verified translation.
Asymmetric, dynamic layouts.
Channel offset or glitch applied to body copy, labels, or form fields.
Classical imagery pulled from a source whose license you have not checked.
How to check your own work
Before considering the result finished, look at it and ask:
Are the colors washed out or vivid? Vivid means you have drifted toward a different style.
Is there anything on the page that looks damaged or degraded? Without it the surface is merely pastel.
Is the composition symmetrical about a vertical axis? If it is dynamic and angled, the stillness is missing.
If there is text in another writing system, can you say exactly what it means and where the translation came from? If not, remove it — that is the honest fix, not a smaller version of it.
Is any body text glitched? Clean it and keep the effect on the display type.
Vaporwave — Specification
Definition
Vaporwave is a web-interface design style assembled from discarded consumer-digital material: pale magenta and cyan gradients, obsolete interface chrome, classical statuary, and visible image artifacts, arranged in still, symmetrical compositions. Its register is nostalgic and faintly ironic, and its material is always secondhand. It is distinguished from neon retro-futurism by being pale, static, and backward-looking rather than dark, kinetic, and speculative.
Core principles
Secondhand material. Every element looks recovered from somewhere rather than made for this page.
Degradation is visible. Artifacts show that the material has been copied, compressed, or displayed on failing hardware.
Stillness over motion. Symmetrical, centered composition produces a screensaver-like calm.
Washed rather than vivid. Color sits pale and mid-saturation, not at full neon intensity.
Borrowed material is used correctly. Imagery is licensed and quoted text is real.
Visual characteristics
Pale magenta and cyan gradient grounds.
Channel separation, color banding, scanlines, or compression artifacts.
Classical busts, torsos, or columns.
Window frames, beveled buttons, and scrollbars quoted from obsolete operating systems.
Receding checkerboard planes.
Symmetrical composition about a vertical axis.
Rules
Pale magenta and cyan. The palette is washed rather than vivid.
Acceptance criterion: The dominant colors measure between 20% and 60% saturation with a relative luminance of 0.5 or greater, centered on magenta and cyan.
Visible artifact. The material shows degradation.
Acceptance criterion: At least one element carries a rendered chromatic channel offset, visible banding, scanlines, or compression blocking.
Classical reference, licensed. Statuary is present and legitimately sourced.
Acceptance criterion: At least one classical sculpture, bust, or marble motif is present, and its source and license are recorded.
Obsolete interface quotation. A discarded convention is reproduced.
Acceptance criterion: At least one component reproduces a 1990s GUI convention — title bar, bevel, or window frame.
Non-Latin text is real text. Borrowed writing systems carry meaning.
Acceptance criterion: Any non-Latin script rendered on the page is correct text with a documented translation; characters selected for their visual form alone do not satisfy this rule.
Symmetrical composition. The layout is still.
Acceptance criterion: The primary composition is symmetrical about the vertical center axis.
Artifacts stay off information. Degradation does not reach reading text.
Acceptance criterion: Informational text carries no channel offset or glitch treatment and measures at least 4.5:1 against its background.
Avoid
Vivid, high-saturation neon.
A composition with no visible artifact.
Non-Latin characters chosen for their shapes rather than their meaning.
Asymmetric or dynamic layouts.
Glitch treatment applied to body copy, labels, or form fields.
Classical imagery whose license has not been verified.
Acceptance criteria
Rule 1: Dominant colors are 20–60% saturation at 0.5+ luminance in magenta and cyan.
Rule 2: At least one visible degradation artifact is rendered.
Rule 3: A classical motif is present with recorded source and license.
Rule 4: At least one obsolete GUI convention is reproduced.
Rule 5: All non-Latin text is correct and documented.
Rule 6: The primary composition is vertically symmetrical.
Rule 7: Informational text is unglitched and meets 4.5:1.
Blocking failures
Non-Latin characters appear as decoration with no meaning or verified translation, which publishes either nonsense or another language used as ornament.
Classical or archival imagery is used without a verified license, since the material this style draws on is frequently under museum or photographer rights despite the age of the original work.
Glitch or channel offset is applied to reading text, form labels, or control text, making them illegible regardless of measured contrast.
These conditions make the work unacceptable to ship; they are not stylistic suggestions and are not subject to designer discretion.
Responsive adaptation guidance
Symmetry (Rule 6) is preserved as the layout stacks: elements remain centered on the vertical axis rather than collapsing to a left-aligned column.
Artifacts (Rule 2) reduce in quantity but at least one remains at every breakpoint; scaling an artifact down until it is invisible does not satisfy the rule.
Quoted GUI components (Rule 4) are re-drawn at a workable size rather than shrunk, since a scaled-down title bar loses the detail that makes the quotation legible.
Rules 3, 5, and 7 are independent of viewport and hold unchanged at every breakpoint.
VaporwaveOverviewIndexAbout
vaporwave
Vaporwave
Recombines discarded consumer-digital material — pastel gradients, obsolete interface chrome, classical statuary, and image artifacts — into a nostalgic, faintly ironic surface.
Primary actionSecondary
Secondhand material
Vaporwave is a web-interface design style assembled from discarded consumer-digital material: pale magenta and cyan gradients, obsolete interface chrome, classical statuary, and visible image artifacts, arranged in still, symmetrical compositions. Its register is nostalgic and faintly ironic, and its material is always secondhand. It is distinguished from neon retro-futurism by being pale, static, and backward-looking rather than dark, kinetic, and speculative.
Applying an Art Deco visual direction
You are being asked to apply an Art Deco visual direction to a web interface. This document gives you framing and recommendations for that direction. It does not tell you what page, product, or feature to build — apply the guidance below to whatever interface you are already working on, at whatever scope that work requires.
What “Art Deco” means here
Art Deco composes with symmetry and exact geometry. Everything mirrors about a vertical axis. Motifs step and radiate: ziggurat forms, fans, rays spreading from a point. A metallic accent — brass, gold, chrome — appears against a deep ground and is used sparingly enough to still read as precious. Typography is condensed, uppercase, and widely letterspaced. Thin rules appear in groups of two or three, framing regions the way inlay frames a panel.
What makes it distinctive is the combination of ornament and precision. It is decorative, but every arc is a section of a circle and every angle is a clean fraction of a right angle. Nothing is organic or freehand. That is what separates it from the flowing, plant-derived ornament of the era before it, which is a different style with almost opposite rules.
The register is machine-age luxury: confident, formal, and unmistakably designed. It suits work that wants to feel crafted and expensive, and it fights work that wants to feel casual or immediate.
Recommendations
Compose symmetrically about a vertical axis. Deco without symmetry loses most of its identity.
Build one stepped or radiating motif and make it structural — a ziggurat frame around a heading, a fan behind a title, rays spreading from a corner. It should organize a region, not sit in one.
Restrain the metallic accent. Use it for rules, motifs, and highlights, and keep it well under a fifth of the visible area so it stays valuable rather than becoming the background.
Choose a deep ground: navy, oxblood, forest, near-black. The metallic needs darkness to read as metal rather than as yellow.
Set headlines condensed and uppercase with generous letterspacing. This one typographic decision carries a large share of the style.
Draw rules in pairs or triples rather than singly. A single hairline is a divider; two parallel hairlines with a small gap is deco.
Keep the geometry exact. Snap arcs to circle sections and angles to multiples of fifteen degrees. Approximate curves make the ornament look drawn rather than constructed.
Avoid
Asymmetric or dynamically balanced composition.
Organic, floral, or freehand ornament, which belongs to the preceding style rather than this one.
Metallic accent spread across large areas, which turns precious material into wallpaper.
A pale background, on which gold reads as mustard.
Single hairline rules where paired rules are the convention.
Arbitrary angles and approximate arcs.
How to check your own work
Before considering the result finished, look at it and ask:
Fold the composition down the middle. Do the halves correspond? If not, symmetry is missing and it is the first thing to fix.
How much of the screen is metallic? If it is a large share, the accent has stopped being an accent.
Are the arcs true circle sections and the angles clean fractions? Freehand curves are visible immediately at this level of formality.
Are the headlines condensed, uppercase, and letterspaced? If they are set in an ordinary sans at default spacing, the typography is not participating.
Is there any organic or floral ornament? That is a different style, and mixing them makes both illegible.
Art Deco — Specification
Definition
Art Deco is a web-interface design style built on strict axial symmetry and exact geometry, using stepped and radiating motifs, a restrained metallic accent over a deep ground, condensed letterspaced capitals, and rules drawn in groups. It is decorative but never freehand: every arc is a circle section and every angle a clean fraction, which distinguishes it from the organic ornament of the style that preceded it.
Core principles
Symmetry is the composition. The layout mirrors about a vertical axis.
Ornament is constructed. Motifs are geometric, exact, and repeatable rather than drawn.
Metal stays precious. The metallic accent is restricted enough to retain value.
Depth of ground. A dark, saturated background is what makes metal read as metal.
Typography is formal. Condensed uppercase with wide letterspacing carries the register.
Visual characteristics
Bilateral symmetry about the vertical center.
Stepped ziggurat forms, fans, and radiating ray motifs.
Brass, gold, or chrome accents against navy, oxblood, forest, or near-black.
Condensed uppercase headlines with generous tracking.
Thin rules grouped in pairs or triples framing regions.
Arcs that are true circle sections and angles at clean intervals.
Rules
Strict axial symmetry. The composition mirrors.
Acceptance criterion: The primary composition is reflected about the vertical center axis.
Stepped or radiating motif. Ornament is structural.
Acceptance criterion: At least one ziggurat, fan, or sunburst motif functions as a structural element organizing a region of the layout.
Restrained metallic accent. Metal remains valuable.
Acceptance criterion: The metallic color is used over a dark or deeply saturated ground and occupies no more than 15% of the rendered area.
Formal display typography. Headlines carry the style.
Acceptance criterion: Headline elements are set uppercase in a condensed display family with a computed letter-spacing of 0.05em or greater.
Grouped rules. Separation follows the convention.
Acceptance criterion: Rules appear as parallel groups of two or three lines of 1–2px framing regions, rather than as single dividers.
Exact geometry. Ornament is constructed, not drawn.
Acceptance criterion: All arcs are sections of circles and all angles are multiples of 15°.
Contrast over metal. Text on metallic ground remains readable.
Acceptance criterion: Text over a metallic accent measures at least 4.5:1 against both its lightest and its darkest value.
Avoid
Asymmetric or dynamically balanced composition.
Organic, floral, or freehand ornament.
Metallic accent spread across large areas.
A pale background, on which gold reads as mustard.
Single hairline rules where paired rules are the convention.
Arbitrary angles and approximate arcs.
Acceptance criteria
Rule 1: The primary composition is symmetrical about the vertical axis.
Rule 2: A stepped or radiating motif organizes a region.
Rule 3: Metallic accent covers 15% or less of the area, over a deep ground.
Rule 4: Headlines are condensed uppercase at 0.05em tracking or more.
Rule 5: Rules appear in parallel groups of two or three.
Rule 6: All arcs are circle sections and all angles multiples of 15°.
Rule 7: Text over metallic ground meets 4.5:1 at both extremes.
Blocking failures
Text is set over a metallic gradient and fails contrast at one extreme, which no single text color can resolve.
Wide letterspacing is applied to body copy rather than to headlines alone, which degrades reading speed substantially and affects readers with dyslexia disproportionately.
Ornamental motifs are placed such that they overlap or interrupt reading text, since the density of stepped and radiating forms makes any text over them unreadable.
These conditions make the work unacceptable to ship; they are not stylistic suggestions and are not subject to designer discretion.
Responsive adaptation guidance
Symmetry (Rule 1) is preserved as the layout stacks: content remains centered rather than collapsing to a left-aligned column, since a left-aligned deco layout loses its defining property.
Motifs (Rule 2) are simplified — fewer steps, fewer rays — rather than scaled down, because a reduced ziggurat quickly stops reading as one.
Letterspacing (Rule 4) may reduce toward 0.05em at narrow widths where wide tracking forces headline wrapping, but never below it.
The metallic budget (Rule 3) is re-measured after reflow, since a fixed-size ornament occupies a much larger share of a narrow viewport.
Rules 6 and 7 hold unchanged at every breakpoint.
Art DecoOverviewIndexAbout
art-deco
Art Deco
Composes with strict axial symmetry, stepped and radiating motifs, and restrained metallic accent over deep ground, reading as machine-age luxury.
Primary actionSecondary
Symmetry is the composition
Bilateral symmetry about the vertical center.
Stepped ziggurat forms, fans, and radiating ray motifs.
The layout mirrors about a vertical axis.
Art Deco
Art Deco is a web-interface design style built on strict axial symmetry and exact geometry, using stepped and radiating motifs, a restrained metallic accent over a deep ground, condensed letterspaced capitals, and rules drawn in groups. It is decorative but never freehand: every arc is a circle section and every angle a clean fraction, which distinguishes it from the organic ornament of the style that preceded it.
Applying an Anti-Design visual direction
You are being asked to apply an Anti-Design visual direction to a web interface. This document gives you framing and recommendations for that direction. It does not tell you what page, product, or feature to build — apply the guidance below to whatever interface you are already working on, at whatever scope that work requires.
What “Anti-Design” means here
Anti-design refuses the conventions of polished interface work on purpose. Colors clash instead of harmonizing. Type is set at sizes and spacings that feel wrong. Hierarchy is flattened or inverted, so the thing you expect to be prominent is not. Nothing aligns to a shared grid. The result reads as a position — a reaction against the smoothed, optimized sameness of most contemporary products — rather than as an accident.
That last word is the whole discipline. Anti-design is not the absence of design decisions; it is a set of deliberate ones that happen to run against convention. Every clash is chosen, and you should be able to name the convention each choice is breaking. Work that simply had no decisions made about it looks similar from a distance and is not the same thing at all.
There is a hard limit, and it is not negotiable. Refusing convention is a legitimate creative stance; refusing access is not. Aesthetic conventions — harmony, alignment, conventional hierarchy — are fair game. Contrast minimums, keyboard operability, visible focus, and not encoding meaning in color alone are not conventions in that sense. They are the conditions under which anyone can use the thing at all, and breaking them does not make the work more radical, only unusable by a portion of its audience.
Recommendations
Identify the specific convention you are breaking before you break it, and write it down. “The primary action is the most prominent element” is a convention; inverting it is a decision. Vagueness here produces mush.
Choose clashing colors deliberately. Near-complementary hues at high saturation placed directly against each other produce genuine visual discomfort — which is the effect — rather than mere ugliness.
Set some text uncomfortably: too large, too tight, too loose, in a face that fights its content. Apply it to secondary material, not to the copy someone actually has to read.
Flatten or invert the hierarchy on purpose. Let a caption dominate the headline. Make the choice legible as a choice.
Abandon the shared grid. Elements should not resolve to common column positions — that is what makes the page feel refused rather than merely loose.
Keep the primary task completable throughout. Someone should be able to do the main thing on the page using only the keyboard, with focus visible at every step.
Hold every accessibility floor: contrast, focus, target size, and no meaning carried by color alone. Design the discomfort within those bounds; it is entirely possible and it is what makes the work defensible.
Avoid
Randomness presented as subversion. If you cannot name what a choice is breaking, it is not a choice.
Uncomfortable typography applied to the content someone must actually read.
Text or controls that fall below contrast minimums.
Focus indicators removed because they look conventional.
Interaction patterns broken in ways that make a task impossible rather than unfamiliar.
Meaning conveyed by a color relationship with no textual equivalent.
How to check your own work
Before considering the result finished, look at it and ask:
For each transgression on the page, can you name the convention it breaks? Anything you cannot name is noise and should either become a real decision or come out.
Complete the primary task with the keyboard alone, without touching the mouse. If you cannot, the work is not shippable regardless of how good the stance is.
Tab through every control. Is focus visible at each stop, against whatever clashing background it lands on?
Measure the worst text on the page. If it fails, change the colors — there are plenty of clashing pairs that also pass.
Would a stranger describe this as “aggressive on purpose” or as “broken”? The first is the goal; the second means the decisions are not legible as decisions.
Anti-Design — Specification
Definition
Anti-Design is a web-interface design style that deliberately subverts the conventions of polished interface work: colors clash, typography is uncomfortable, hierarchy is flattened or inverted, and no shared grid governs placement. It is a stance rather than an absence of decisions — each transgression names the convention it breaks. Its defining constraint is that aesthetic convention may be refused while access conditions may not.
Core principles
Every transgression is named. A rejected convention is identified before it is rejected; unnamed disorder is not this style.
Discomfort is the effect. Clashing color and uncomfortable typography are intended to be felt, not tolerated.
Hierarchy is a target. Expected prominence is inverted or flattened as a legible decision.
The grid is refused. Elements do not resolve to shared column positions.
Access is not a convention. Contrast, keyboard operability, focus, and non-color-dependent meaning are floors, not styles.
Visual characteristics
Near-complementary saturated colors placed directly against one another.
Type set at sizes, weights, or spacings that contradict reading convention.
Secondary content rendered more prominently than primary content.
Plainly styled, fully operable controls despite the surrounding disorder.
Rules
Named subversion. Each transgression identifies its convention.
Acceptance criterion: At least one established interface pattern is subverted, and the original convention is documented alongside the deviation.
Clashing color. Discomfort is produced deliberately.
Acceptance criterion: At least two adjacent colors are near-complementary in hue with saturation of 70% or greater.
Uncomfortable typography. Type contradicts convention.
Acceptance criterion: At least one text block uses a family, tracking, or size that contradicts reading convention, and it is not the primary content of the page.
Flattened or inverted hierarchy. Prominence is misassigned on purpose.
Acceptance criterion: At least one secondary element renders more visually prominent than a primary one, documented as intentional.
No shared grid. Placement is refused.
Acceptance criterion: Element offsets do not resolve to a common column system.
The primary path completes. The task remains possible.
Acceptance criterion: The primary task can be completed using the keyboard alone, with a visible focus indicator at every step.
The access floor holds. Refusal stops at usability.
Acceptance criterion: Despite Rules 1 through 5, all text measures at least 4.5:1, all controls at least 3:1, and no meaning is conveyed by color alone.
Avoid
Randomness presented as subversion, with no convention identified.
Uncomfortable typography applied to the content a reader must parse.
Text or controls below contrast minimums.
Focus indicators removed for looking conventional.
Interaction patterns broken such that a task becomes impossible rather than unfamiliar.
Meaning carried by a color relationship with no textual equivalent.
Acceptance criteria
Rule 1: Each subverted convention is documented alongside its deviation.
Rule 2: Two adjacent near-complementary colors at 70%+ saturation are present.
Rule 3: An uncomfortable type treatment is applied to non-primary content.
Rule 4: A secondary element is deliberately more prominent than a primary one.
Rule 5: Element offsets do not resolve to a shared column system.
Rule 6: The primary task completes by keyboard with visible focus throughout.
Rule 7: All text meets 4.5:1, all controls 3:1, and no meaning is color-only.
Blocking failures
The primary task cannot be completed by keyboard, or a focus indicator is missing at any step, which converts a stylistic stance into an exclusion.
Any text or control falls below its contrast floor. Clashing color pairs that satisfy 4.5:1 are readily available, so this failure is never a cost of the style.
A subversion removes information rather than rearranging it — a control with no accessible name, an unlabeled input, a state conveyed only by hue — leaving content unavailable rather than unconventional.
These conditions make the work unacceptable to ship; they are not stylistic suggestions and are not subject to designer discretion.
Responsive adaptation guidance
Grid refusal (Rule 5) is preserved by keeping deliberate offsets in the stacking order; it does not license overlapping content that clips or becomes unreachable at narrow widths.
Uncomfortable typography (Rule 3) reduces in intensity at narrow viewports where extreme tracking or scale causes overflow, since content lost to overflow is a Rule 7 failure rather than a stylistic success.
Rule 6 strengthens at narrow and touch viewports: targets meet the platform minimum size and remain reachable in a predictable order.
Rule 7 is re-verified at every breakpoint and at 200% zoom, since reflow frequently moves text onto a differently colored block and changes its measured ratio.
Rules 1, 2, and 4 hold unchanged across breakpoints.
Anti-Design
Anti-Design is a web-interface design style that deliberately subverts the conventions of polished interface work: colors clash, typography is uncomfortable, hierarchy is flattened or inverted, and no shared grid governs placement. It is a stance rather than an absence of decisions — each transgression names the convention it breaks. Its defining constraint is that aesthetic convention may be refused while access conditions may not.
Every transgression is namedDiscomfort is the effectHierarchy is a target
Anti-Design
Every transgression is named
Anti-Design is a web-interface design style that deliberately subverts the conventions of polished interface work: colors clash, typography is uncomfortable, hierarchy is flattened or inverted, and no shared grid governs placement. It is a stance rather than an absence of decisions — each transgression names the convention it breaks. Its defining constraint is that aesthetic convention may be refused while access conditions may not.
Anti-DesignOverviewIndexAbout
Applying a Memphis visual direction
You are being asked to apply a Memphis visual direction to a web interface. This document gives you framing and recommendations for that direction. It does not tell you what page, product, or feature to build — apply the guidance below to whatever interface you are already working on, at whatever scope that work requires.
What “Memphis” means here
Memphis scatters flat geometric shapes across a composition in colors that were never meant to agree with each other. Circles, triangles, wavy lines, dots and dashes float at positions unrelated to the layout or to one another, each outlined heavily in black. A confetti or terrazzo pattern speckles the background. Nothing has depth, nothing casts a shadow, and nothing is arranged.
The feeling is deliberately weightless and cheerful — objects suspended rather than placed, colors clashing pleasantly rather than harmonizing. It is playful in a way that is genuinely hard to fake with a color palette alone; the scattering and the black outlines are doing at least as much work as the hues.
The practical constraint is simple and easy to get wrong. All that floating shape material must stay off the content. A squiggle passing behind a paragraph looks charming in a mockup and makes the paragraph unreadable in practice, because the text now sits on an unpredictable background. Shapes surround the content; they do not cross it.
Recommendations
Scatter at least five independent shapes. Fewer than that reads as a few decorative accents rather than as a scattering.
Include the characteristic pattern somewhere: squiggles, confetti dots, or terrazzo speckle. Without it the composition is just floating shapes.
Pick four or more saturated colors that do not form a harmonious scheme. Yellow next to teal next to pink is right; a carefully balanced triad is not.
Keep everything flat. No gradients, no shadows, no depth of any kind — the shapes are cut from colored paper, not rendered.
Outline the shapes heavily in near-black. This is what holds an unharmonious palette together, and without it the composition falls apart into unrelated colored blobs.
Distribute without alignment. Shapes should not line up with each other or with the content grid; regular spacing immediately reads as a pattern library rather than as scattering.
Keep every shape clear of text. Route them around the content, into margins and empty regions.
Avoid
Shapes overlapping paragraphs, labels, or any reading text.
Gradients or shadows on the shapes, which give them a weight the style does not want.
A harmonious, carefully balanced palette.
Thin or absent outlines, which let the clashing colors separate into visual noise.
Shapes aligned to a grid or evenly spaced.
So many shapes that the content has to compete for attention with its own decoration.
How to check your own work
Before considering the result finished, look at it and ask:
Does any shape touch or sit behind text? Move it. This is the only thing in this direction that can genuinely break the page.
Do the shapes have black outlines? Without them, an unharmonious palette looks accidental instead of intentional.
Is there a squiggle, confetti, or terrazzo pattern anywhere? That motif is the signature.
Are the shapes aligned with each other in any way? If they are, the scattering is too orderly.
Do the colors actually clash? A tasteful palette produces a pleasant page that is not this style.
Memphis — Specification
Definition
Memphis is a web-interface design style that scatters flat primitive shapes — circles, triangles, squiggles, confetti — across a composition in colors chosen not to harmonize, each outlined heavily in near-black, over a speckled or patterned ground. Nothing has depth and nothing is aligned. The scattered material surrounds the content and never crosses it.
Core principles
Weightless scattering. Shapes float at positions unrelated to the layout or to each other.
Deliberate disharmony. The palette clashes on purpose; a balanced scheme belongs to a different style.
Flatness throughout. No gradient, no shadow, no implied depth anywhere.
The outline binds it. Heavy black outlines hold an unharmonious palette together.
Decoration yields to content. Shapes route around text rather than over it.
Visual characteristics
Five or more independent flat shapes suspended around the layout.
Squiggle, confetti, or terrazzo pattern on the background.
Four or more saturated colors with no harmonic relationship.
Thick near-black outlines on every decorative shape.
Shapes positioned without alignment to the grid or to each other.
Content areas kept clear of decorative material.
Rules
Scattered primitives. Enough shapes to read as scattering.
Acceptance criterion: At least five independent geometric shapes are rendered as floating compositional elements.
Signature pattern present. The characteristic motif appears.
Acceptance criterion: At least one squiggle, confetti, or terrazzo pattern is rendered on the page.
Disharmonious palette. Colors clash by design.
Acceptance criterion: At least four saturated colors coexist with no documented harmonic relationship between them.
Total flatness. Nothing implies depth.
Acceptance criterion: Every shape uses a flat fill; no gradient and no box-shadow appears anywhere on the page.
Heavy outlines. Shapes are bound by black.
Acceptance criterion: Every decorative shape carries an outline of 2px or greater in a near-black color.
Unaligned distribution. Scattering is genuine.
Acceptance criterion: Decorative shapes do not align with each other or with the content grid, and are not evenly spaced.
Decoration never touches content. Text stays on predictable ground.
Acceptance criterion: No decorative shape overlaps any text element or reduces its measured contrast below 4.5:1.
Avoid
Shapes overlapping paragraphs, labels, or any reading text.
Gradients or shadows applied to shapes.
A harmonious, balanced palette.
Thin or absent outlines.
Shapes aligned to a grid or evenly distributed.
Decoration dense enough to compete with the content for attention.
Acceptance criteria
Rule 1: At least five independent floating shapes are present.
Rule 2: A squiggle, confetti, or terrazzo pattern is rendered.
Rule 3: At least four saturated colors coexist without harmonic relation.
Rule 4: All fills are flat, with no gradient or shadow anywhere.
Rule 5: Every decorative shape has a 2px+ near-black outline.
Rule 6: Shapes are neither aligned nor evenly spaced.
Rule 7: No shape overlaps text or drops its contrast below 4.5:1.
Blocking failures
A decorative shape sits behind reading text, placing it on an unpredictable background where no single text color remains legible across the block.
A decorative shape overlaps an interactive control, obscuring its boundary or focus indicator.
Decorative shapes are exposed to assistive technology as meaningful content, injecting nonsense into the reading order.
These conditions make the work unacceptable to ship; they are not stylistic suggestions and are not subject to designer discretion.
Responsive adaptation guidance
Shape count (Rule 1) reduces at narrow viewports, since the margins that hold scattered decoration disappear first; three shapes is a reasonable floor before the scattering stops reading.
Rule 7 is re-verified after every reflow without exception. A shape that sits in an empty margin at desktop width lands directly behind text once the layout collapses, which is the single most likely way this style breaks responsively.
The background pattern (Rule 2) may reduce in scale or density but remains present at every breakpoint.
Outlines (Rule 5) do not thin below 2px at any width.
Where a narrow viewport leaves no clear region for decoration, shapes are removed rather than layered behind content.
MemphisOverviewIndexAbout
memphis
Memphis
Scatters flat primitive shapes, squiggles, and confetti pattern in colors that refuse to harmonize, producing cheerful, weightless visual noise.
Primary actionSecondary
Weightless scatteringDeliberate disharmonyFlatness throughout
Memphis
Memphis is a web-interface design style that scatters flat primitive shapes — circles, triangles, squiggles, confetti — across a composition in colors chosen not to harmonize, each outlined heavily in near-black, over a speckled or patterned ground. Nothing has depth and nothing is aligned. The scattered material surrounds the content and never crosses it.
Applying an Ethereal visual direction
You are being asked to apply an Ethereal visual direction to a web interface. This document gives you framing and recommendations for that direction. It does not tell you what page, product, or feature to build — apply the guidance below to whatever interface you are already working on, at whatever scope that work requires.
What “Ethereal” means here
Ethereal holds an interface in high, even light. Pale gradients wash across the surface with no visible edge where one begins and another ends. Regions separate through gradual tonal change rather than through lines or containers. Type is thin and large, floating in space rather than sitting in a block. Most of the page is empty. Motion, where it exists, is slow enough to feel like drift.
The intended feeling is calm and weightlessness — the sense that nothing is demanding anything. It suits work about rest, care, craft, or luxury, and it is actively wrong for anything urgent or dense.
The tension to manage is that everything characteristic of this direction pushes contrast down. High-key color, thin strokes, and diffuse edges each reduce the difference between text and its background, and together they reduce it a lot. The resolution is not to compromise the look but to be deliberate: the palette can stay pale as long as the text does not, and a thin face at a large size can still be perfectly legible if its color is dark enough. Softness is a property of the surface, not of the text sitting on it.
Recommendations
Keep the overall surface bright. The page should read as high-key, with most of its area well up the luminance range.
Build gradients from low-saturation stops and let them transition over long distances — a couple of hundred pixels at least — so no band or edge is visible anywhere in the blend.
Leave most of the page empty. Around a third of the initial screen covered is the target, and less is usually better.
Set headlines thin and large. A light weight at a generous size reads as delicate; the same weight at a small size just reads as faint.
Separate regions with tonal shifts, not lines. A gradient that gradually becomes another color does the work a divider would do elsewhere, without the hard edge.
Keep motion slow. If anything animates, give it a long duration and gentle easing, and honor reduced-motion preferences.
Choose text and control colors dark enough to pass on the pale ground. This does not compromise the direction — a deep ink on a blush field is entirely consistent with it, and a pale grey on the same field is simply unreadable.
Avoid
Hard lines, borders, or containers with visible edges.
Saturated color, which breaks the high-key register immediately.
Thin type at small sizes, where delicacy becomes faintness.
Short, snappy animation, which contradicts the drift.
Pale grey text on a pale background, the failure this style is most prone to.
Dense content, which the composition cannot hold without abandoning the emptiness that defines it.
How to check your own work
Before considering the result finished, look at it and ask:
Is there a visible edge anywhere in a gradient? Lengthen the transition until it disappears.
How much of the first screen is covered? If it is more than half, the emptiness is not doing its work.
Take the lightest text on the page and measure it. If it fails, darken the text — not the background, which would cost you the whole look.
Is the thin type actually large? Thin at a small size is the most common way this direction becomes unreadable.
Turn on reduced motion. Does anything still animate? It should not.
Ethereal — Specification
Definition
Ethereal is a web-interface design style that holds the surface in high, even light: pale low-saturation gradients transition without visible edges, regions separate through tonal shift rather than through lines, type is thin and large, most of the page is empty, and motion is slow. Its register is calm and weightless. Because every characteristic property reduces contrast, the style requires that text and control colors be chosen deliberately dark against the pale ground.
Core principles
High-key throughout. The surface sits well up the luminance range.
Boundaries dissolve. Separation is tonal and gradual, never a line.
Delicacy needs size. Thin type is legible when large and faint when small.
Emptiness is the composition. Most of the surface carries nothing.
Softness belongs to the surface, not the text. The pale ground never justifies pale text.
Visual characteristics
A bright overall surface with high average luminance.
Low-saturation gradients transitioning over long distances with no visible banding.
Regions separated by gradual tonal change rather than borders.
Large headlines at light weights.
Large uncovered areas across the initial viewport.
Slow, gently eased motion where motion exists at all.
Rules
High-key surface. The page is bright overall.
Acceptance criterion: The mean relative luminance of the rendered page is 0.8 or greater.
Soft desaturated gradients. Blends have no visible edge.
Acceptance criterion: Gradient stops measure 30% saturation or less and each transition spans at least 200px.
Emptiness dominates. The composition is mostly open.
Acceptance criterion: Rendered content covers 35% or less of the initial viewport area.
Thin type at scale. Delicacy is carried by size.
Acceptance criterion: Headline elements compute a font-weight of 300 or less at a rendered size of 40px or greater.
Diffuse boundaries. Regions are not drawn.
Acceptance criterion: Separation between regions is produced by a gradient transition or soft mask; no 1px border divides primary regions.
Slow motion or none. Movement reads as drift.
Acceptance criterion: Any animation has a duration of 800ms or greater with gentle easing, and is suppressed under prefers-reduced-motion.
Delicacy costs no contrast. Legibility is unaffected by the register.
Acceptance criterion: Despite Rules 1 and 4, all text measures at least 4.5:1 and all controls at least 3:1 against their background.
Avoid
Hard lines, borders, or containers with visible edges.
Saturated color, which breaks the high-key register.
Thin type at small sizes.
Short, snappy animation.
Pale grey text on a pale background.
Dense content, which the composition cannot hold.
Acceptance criteria
Rule 1: Mean page luminance is 0.8 or greater.
Rule 2: Gradient stops are 30% saturation or less, transitioning over 200px or more.
Rule 3: Content covers 35% or less of the initial viewport.
Rule 4: Headlines are weight 300 or less at 40px or larger.
Rule 5: No 1px border separates primary regions.
Rule 6: Animations run 800ms or longer and respect reduced-motion.
Rule 7: All text meets 4.5:1 and all controls 3:1.
Blocking failures
Text is rendered in a pale color on the pale ground and falls below 4.5:1. This is the characteristic failure of the style, and it is never necessary: a deep ink on a blush field is fully consistent with the direction.
A control has no boundary or focus indicator reaching 3:1, since the absence of borders required by Rule 5 removes the fallback that would otherwise make it findable.
Animation continues under prefers-reduced-motion, which in a style built on slow drifting movement affects a large proportion of the interface rather than an isolated element.
These conditions make the work unacceptable to ship; they are not stylistic suggestions and are not subject to designer discretion.
Responsive adaptation guidance
Coverage (Rule 3) is evaluated per breakpoint; at narrow widths the target may rise to 55% as content stacks vertically, since a short viewport cannot hold the same proportion of open space.
Headline size (Rule 4) may fall below 40px at narrow viewports only if the weight increases to 400 in compensation, since thin strokes at small sizes are where legibility degrades fastest.
Gradient transition length (Rule 2) scales with the viewport rather than staying fixed, otherwise a 200px blend that reads as seamless on a wide screen becomes a visible band on a narrow one.
Rules 5, 6, and 7 hold unchanged at every breakpoint, and Rule 7 is re-verified at 200% zoom.
EtherealOverviewIndexAbout
ethereal
Ethereal
Holds the interface in high, even light — soft low-saturation gradients, diffuse boundaries, thin large type, and mostly empty space — to read as calm and weightless.
Primary actionSecondary
High-key throughout
Ethereal is a web-interface design style that holds the surface in high, even light: pale low-saturation gradients transition without visible edges, regions separate through tonal shift rather than through lines, type is thin and large, most of the page is empty, and motion is slow. Its register is calm and weightless. Because every characteristic property reduces contrast, the style requires that text and control colors be chosen deliberately dark against the pale ground.