# Anti-Design — style guide for coding agents

> Generated from the Aesthetary record for **Anti-Design**.
> Research: **Partly sourced** · Build guidance: recorded in 5 of 8 research dimensions.
> 10 sources attached.

## How to use this

Paste this file into your agent as context before asking it to design or build something in this visual language. It states what is established, what is our interpretation, and what nobody has established. Treat all three differently.

**The most important section is "Not researched yet".** Its purpose is to stop you and your agent inventing rules that sound right and are not recorded anywhere.

## What this language is

**Defining:** No shared edge, crowded type that won't align, harsh colour, pages that disorient on purpose

**Recognise it by:** Alignment broken so that nothing settles onto a shared edge, type set to be difficult rather than merely unusual, clashing colour with no arbitration, and interactions that behave unexpectedly — controls that move, resist, or do something other than what they advertise. The diagnostic condition is that the friction is functional: something is genuinely harder than it needs to be, and that cost was accepted deliberately. Web Brutalism exposes structure and leaves it working. New Wave Typography violates a grid that stays inferable; here there is no grid to infer.

**Family:** Raw & Anti-Design
**Era:** postmodern, contemporary
**Native medium:** screen, print

**Scope:** Anti-design was an Italian critique movement (1960s–80s) working in furniture and objects. This record covers the later web and graphic trend sharing the name: non-aligning type, loud colour; the rest is Aesthetary's reading.

## What it is arguing

Usability conventions are a commercial value, not a natural law. Breaking alignment, legibility and interaction on purpose refuses the assumption that a page exists to convert a visitor, and treats the visitor as someone who came to look rather than to be processed.

**It refuses:** Refuses smooth interaction, predictable alignment, and any optimisation aimed at conversion.

## What it is not

Commonly confused with these. Getting the difference right matters more than getting the surface right:

- Web Brutalism
- New Wave Typography

## Established rules

Grouped by what kind of claim each is. A requirement of the language and our own interpretation are not interchangeable.

**How to read each rule.** Every rule states two separate things, and neither stands in for the other:

- **Kind of claim** (the heading): From our catalogue entry · From our catalogue entry · Source attached · Our reading · disputed · Our reading · Documented · Source attached.
- **Evidence** (how it is established): Graded by our research brief · No source attached (`DOCUMENTED`) · From our catalogue entry (`SUPPORTED`) · Our reading (`CURATORIAL`) · Not researched yet (`UNKNOWN`) · No grade recorded (no grade in the record). Where a claim has its own canonical label, the Evidence line gives that label instead of the grade.
- **Product area**: which of the twelve areas the rule belongs to (Layout, Typography, Color, Imagery, Surface, Interface, Motion, Motif, Technique, Form, Hierarchy, Density), mapped from its raw area label in bold. "Context" means the rule is about the Aesthetic (its name, origin, lineage) rather than a part of the design. "Not mapped" means the raw area has no single area yet.
- The code in brackets is the `grade` field at `implementationConstraints.tokenRules.value.rules[].provenance.grade` in https://aesthetary.com/data/anti-design.json. "From our catalogue entry" means: Stated in Aesthetary’s existing catalogue entry. An outside source is named only where one is linked to the claim.

### From our catalogue entry

_Carried over from Aesthetary's original catalogue entry. The reasoning behind it was never recorded._

- **layout** — Composition: broken grid.
  - Kind of claim: From our catalogue entry · Evidence: From our catalogue entry · Product area: Layout

### From our catalogue entry · Source attached

_Carried over from Aesthetary's original catalogue entry. The reasoning behind it was never recorded._

- **color** — Colour: clashing.
  - Kind of claim: From our catalogue entry · Source attached · Evidence: From our catalogue entry · Source attached · Product area: Color
- **motion** — Rare, and it is the record's sharpest asset. "Unexpected interaction" is a claimed behaviour, not an inference, which makes this one of the few records where motion is evidenced rather than invented.
  - Kind of claim: From our catalogue entry · Source attached · Evidence: From our catalogue entry · Source attached · Product area: Motion

### Our reading · disputed

- **hierarchy** — Deliberately contested — competing signals that never resolve. Stating that there is no stable ranking is the instruction.
  - Kind of claim: Our reading · disputed · Evidence: Our reading · disputed · Product area: Hierarchy
- **typography** — Typography: type that refuses to align, set to be hard work rather than unreadable.
  - Kind of claim: Our reading · disputed · Evidence: Our reading · Product area: Typography

### Our reading

- **composition** — Breakage, not exposure: alignment and legibility are deliberately broken. If the structure stays legible the artefact is Web Brutalism, and if the system stays inferable it is New Wave Typography.
  - Kind of claim: Our reading · Evidence: Our reading · Product area: Layout

## Found in research

What research established from sources Aesthetary opened and checked. Each line names its area, how it is known, and where it is scoped. "Seen in the work" means the source documents particular works or products rather than the Aesthetic as a whole.

- **Surface** — Anti-design adds deliberate blemishes and glitches to digital work, so that it reads as made by a person.
  - Evidence: Documented · Source attached · Claim: `kcc:w2-surface-1` · Covers Trade-press account of the current (2020s) trend..
- **Surface** — A cited example site has a handcrafted or scribbled look.
  - Evidence: Documented · Source attached · Claim: `kcc:w2-surface-2` · Covers One example..
- **Interface** — Some anti-design sites use weird cursors and unnecessary, distracting animations.
  - Evidence: Documented · Source attached · Claim: `kcc:interface-1` · Covers web, c. 2016–2019 examples.
- **Interface** — Some designers deliberately make interfaces ugly, disorienting or complex; NN/g adds that in such work interaction does not have to be smooth.
  - Evidence: Documented · Source attached · Claim: `kcc:interface-2` · Covers NN/g's second quote describes where antidesign can work (entertainment contexts), not a designer's stated rule.
- **Motif** — Some anti-design sites use disorienting patterns.
  - Evidence: Documented · Source attached · Claim: `kcc:motif-1`
- **Motif** — Built In calls anti-design 'a way of thinking more than a specific aesthetic'; the sources reviewed name no fixed motif set.
  - Evidence: Documented · Source attached · Claim: `kcc:motif-2`
- **Technique** — One interviewee in a thesis links the look to how the web looked around 1996-97, built with tables, buttons and GIF animations; then it was not irony but testing the limits.
  - Evidence: Documented · Source attached · Claim: `kcc:technique-1` · Covers one interviewee in a bachelor thesis; origin, not current method.
- **Technique** — Creative Bloq advises reaching the look by forgetting the usual rules and going to extremes (stretching images, breaking layouts, combining clashing colours), then dialling it back until it works.
  - Evidence: Documented · Source attached · Claim: `kcc:w2-technique-1`
- **Technique** — It sets itself against the cookie-cutter templates that design apps make easy.
  - Evidence: Documented · Source attached · Claim: `kcc:w2-technique-2`
- **Technique** — Creative Bloq suggests looking for anti-design inspiration in old zines, 90s grunge and early web design such as Geocities.
  - Evidence: Documented · Source attached · Claim: `kcc:w2-technique-3`
- **Form** — The name also covers furniture, architecture and fashion, and the historical movement worked through objects. The approved scope limits this record to the web and graphic trend, where no source describes three-dimensional form.
  - Evidence: Documented · Source attached · Claim: `kcc:form-1` · Covers object strand is outside approved scope.
- **Form** — The term 'anti-design' was first used for a 1960s-70s European architectural movement represented by Archigram, Archizoom and Superstudio.
  - Evidence: Documented · Source attached · Claim: `kcc:w2-form-1` · Covers The historical movement, excluded by the approved scope..

## Positions by area

The overall position recorded for each area. "No overall position recorded" means rules exist for the area but no position above them; the rules still apply. It is not the same as "Not established", which means nobody has researched the area.

- **Layout** — Specified
- **Typography** — Specified
- **Color** — Specified
- **Surface** — Specified
- **Interface** — Specified
- **Motion** — Specified
- **Motif** — Specified
- **Technique** — Specified
- **Imagery** — Outside this language. The area does not apply; this is not a gap in the research.

## Design grammar — web primitives

A starting grammar for a website in Anti-Design: nothing on a shared edge, colours left to clash, type that refuses to align but stays readable, and controls that resist, with one switch that straightens it all.

Three kinds of line: **On the record** names a rule this record holds (its wording is under Established rules) and states what it becomes on the web. **Web translation** is Aesthetary's choice for the web, derived from the rules it cites; change it if you have a reason, keep the cited rule. **Not specified by the source** means the research reached this but the sources say nothing about it; **Not researched yet** means the research has not reached it yet. Either way, follow the instruction and do not invent a rule. Structured version: `designGrammar` in data/anti-design.json.

### Colour

| Role | Value | Contrast | Use | Kind |
| --- | --- | --- | --- | --- |
| ground | `#F4EFE2` | — | The page ground: a cream the other colours fight on. | Web translation |
| ink | `#17141B` | 15.9:1 | Text, borders, the dark blocks. | Web translation |
| blue | `#1F3E9E` | 8.2:1 | Links and the first colour system. | Web translation |
| lime | `#D6FF3F` | — | Blocks behind ink text (15.8:1); never text on cream. | Web translation |
| pink | `#FF3DA6` | — | Blocks behind ink text (5.6:1) and focus; never text on cream (2.8:1). | Web translation |
| lilac | `#C7D0FF` | — | The second colour system, set against the first. | Web translation |

- **On the record · From our catalogue entry · Source attached** · `token_rule:color-5` → Leave the colours clashing: two systems on one page with nothing arbitrating between them.
- **On the record · Documented** · `claim:kcc:w2-technique-1` → Reach it the documented way: combine clashing colours at the extreme, then dial back until it works.
- **Web translation** — Carry text only in ink or blue; put lime and pink behind it, never under it as text. _Why:_ The clash is in the fields; contrast is what the page cannot afford to give up. (`token_rule:color-5`, `claim:kcc:w2-technique-1`)
- **Not specified by the source** — The sources do not specify a palette. _Instead:_ These six are the room's. Choose your own clash; do not treat lime and pink as the language.

### Typography

Family: one variable grotesque pushed to both ends of its weight axis. Stack: `"WorkSansX", "Work Sans", "Helvetica Neue", Arial, sans-serif`

| Role | Size | Weight | Line height | Case | Tracking | Measure | Use |
| --- | --- | --- | --- | --- | --- | --- | --- |
| display | clamp(3rem, 9vw, 8rem) | 900 | 0.82 | not specified | -0.05em | — | Headings, with a word or two dropped to weight 100 italic. |
| thin | inherit | 100 | inherit | not specified | not specified | — | The italic that breaks into a heading. |
| text | 1.0625rem | 400 | 1.5 | not specified | not specified | 58ch | Everything read at length: straight, level, readable. |

- **On the record · Our reading** · `claim:dcc:typography-4` → Set type that refuses to align: hard work to read, never unreadable.
- **Web translation** — Use one variable face and swing it from 900 to 100 italic inside a single heading. _Why:_ Weight collision is how the room makes type difficult without making it illegible. (`claim:dcc:typography-4`)
- **Web translation** — Keep running text level and at 17px. _Why:_ The sources describe difficulty, not illegibility; the difficulty belongs to the headings and the placement. (`claim:dcc:typography-4`, `claim:kcc:interface-2`)
- **Not specified by the source** — The sources do not name a typeface. _Instead:_ Work Sans is the room's. Any face with a wide weight range will do.

### Grid and layout

No grid to infer. A twelve-column field exists only so that blocks can be placed off it: every block has its own start, its own angle, and at least one overlap.

| Viewport | Columns |
| --- | --- |
| all | 12 |

Tokens: tilt-a `-2deg` · tilt-b `3deg` · tilt-c `-1deg` · overlap `-40px`

- **On the record · From our catalogue entry** · `token_rule:layout-3` → Break the grid.
- **On the record · Our reading** · `token_rule:ad-r1` → Break alignment so nothing settles on a shared edge; if the system stays inferable it has become New Wave Typography.
- **On the record · Our reading · disputed** · `token_rule:hierarchy-13` → Leave the ranking contested: competing signals that never resolve.
- **Web translation** — Place every offset and angle by hand; nothing random. _Why:_ A placed break can be straightened by one switch; a random one cannot be designed or tested. (`token_rule:ad-r1`)

### Shape and geometry

Tokens: radius `0` · border `3px solid #17141B` · tilt-max `14deg` · text-tilt-max `3deg`

- **Web translation** — Square corners and 3px ink borders; the break is in the angle and the placement, not the shape. _Why:_ The record breaks alignment and legibility, not form; three-dimensional form is an open question. (`token_rule:ad-r1`, `claim:kcc:form-1`)
- **On the record · Documented** · `claim:kcc:w2-surface-1` → Add deliberate blemishes and glitches so the work reads as made by a person.

### Imagery

- **On the record · Our reading** · `claim:imagery-reason` → Imagery is neither demanded nor forbidden.
- **On the record · Documented** · `claim:kcc:w2-technique-1` → Where you use images, stretch them and let them break the layout.
- **On the record · Documented** · `claim:kcc:motif-1` → Use disorienting patterns.
- **Web translation** — Build patterns in CSS, not raster files. _Why:_ A disorienting pattern must stay sharp at every width and be removable in one switch. (`claim:kcc:motif-1`)
- **Not specified by the source** — The sources describe no fixed motif set. _Instead:_ Built In calls it a way of thinking more than an aesthetic; do not turn one site's scribble into a rule.

### Friction and its exit

The friction is functional and it costs accessibility on purpose. A switch that straightens every transform keeps the cost a choice the visitor can refuse.

Tokens: resist-tries `3`

- **On the record · Our reading** · `claim:gestalt.interface` → Controls move, resist, or do something other than they advertise.
- **On the record · Documented** · `claim:kcc:interface-2` → Interaction does not have to be smooth.
- **Web translation** — Let a resisting control give in after three tries, and never resist the keyboard. _Why:_ The cost is accepted for the pointer, not for someone who cannot use one. (`claim:intent-cost`, `claim:gestalt.interface`)
- **Web translation** — Ship a switch that removes every transform. _Why:_ Accessibility is a cost the record accepts; the switch lets each visitor decide whether to pay it. (`claim:intent-cost`)

### Components

Anatomy, tokens and states below are Web translation: Aesthetary's specification for the screen. Each value line after them says whether it restates a rule this record holds.

#### Button

A lime slab knocked off true, with a pink block landing behind it on the wrong side.

- Anatomy: lime field, 3px ink border; label in Work Sans 800; pink block offset down and to the left (-8px, 8px); a placed tilt of -4deg
- Tokens: background `#D6FF3F` · color `#17141B` · border `3px solid #17141B` · offset `-8px 8px 0 #FF3DA6` · tilt `-4deg` · radius `0`
- State, default: Lime, tilted -4deg, pink block behind.
- State, hover: Inverts: ink field, lime label.
- State, focus: 4px dashed pink outline, 4px outside.
- State, active: The tilt drops to 0 for the press: the only moment it sits straight.
- State, disabled: Dashed ink border, no pink block, no tilt.
- Accessibility: Label on lime at 15.8:1.
- Accessibility: The tilt is placed, never random, and the straighten switch removes it.
- Accessibility: At least 44px tall.
- **Web translation** — Tilt the button off the page's edges and put its shadow on the side no light would cast it. _Why:_ Nothing may settle on a shared edge; a button is one more thing that refuses to. (`token_rule:ad-r1`, `token_rule:layout-3`)
- **Web translation** — Pair two colours from the set that do not agree. _Why:_ Colour is left clashing, with no arbitration between hues. (`token_rule:color-5`)

#### Card

A bordered block dropped onto the page at its own angle and its own column, overlapping its neighbour.

- Anatomy: 3px ink border; heading in Work Sans 900; a kind label on a coloured bar; body text, straight
- Tokens: border `3px solid #17141B` · tilt `one of -2deg, 3deg, -1deg, 1.5deg` · overlap `-40px` · radius `0`
- State, default: Each card at its own placed angle and column start.
- State, hover: Background turns pink.
- State, focus: 4px dashed pink outline.
- Accessibility: Body text is never rotated beyond 3deg and never set below 16px.
- Accessibility: Reading order follows the source, whatever the placement.
- **Web translation** — Give every card its own column start and angle; never let two share an edge. _Why:_ Alignment is broken so that nothing settles; if the structure stays legible it is Web Brutalism instead. (`token_rule:ad-r1`, `token_rule:layout-3`)
- **Web translation** — Rank nothing: let size, colour and position pull against each other. _Why:_ The record states that there is no stable ranking. (`token_rule:hierarchy-13`)

#### Input

A plain ruled field whose label sits off its edge in a thin italic, as if it had slid.

- Anatomy: 3px ink border; cream field; label above in Work Sans 100 italic, pushed 2rem right; error as a pink block beginning "Error:"
- Tokens: border `3px solid #17141B` · background `#F4EFE2` · label-offset `2rem` · radius `0` · height `2.75rem`
- State, default: 3px ink border.
- State, focus: 4px dashed pink outline.
- State, error: Ink text on a pink block (5.6:1), beginning "Error:".
- State, disabled: Dashed border.
- Accessibility: The label is real text tied to the field.
- Accessibility: The field itself is never tilted: typing into a moving target fails everyone.
- Accessibility: Error text is words, not colour alone.
- **Web translation** — Slide the label off the field's edge; keep the field straight. _Why:_ The friction has to be chosen, and the record accepts cost to conversion, not to entering text at all. (`token_rule:ad-r1`, `claim:intent-cost`)
- **Not specified by the source** — The sources do not specify an anti-design form convention. _Instead:_ Do not invent a form convention; keep the field plain and break only its alignment.

#### Navigation

Links that refuse a shared baseline, each dropped to its own height and angle, the current one on lime.

- Anatomy: links in Work Sans 800; each at its own vertical offset and tilt; current page on a lime block
- Tokens: offsets `0, 14px, -6px, 22px` · tilts `-3deg, 2deg, 0, -1deg` · current `#D6FF3F`
- State, default: Blue links, underlined 3px, off one another's baseline.
- State, hover: Pink block behind the link.
- State, focus: 4px dashed pink outline.
- State, current: Lime block, ink label.
- Accessibility: aria-current on the current page.
- Accessibility: Every link keeps a 44px target whatever its offset.
- Accessibility: Nothing moves away from the keyboard.
- **Web translation** — Drop each link to its own baseline. _Why:_ Type that does not align is the documented trait; a row of links is the easiest place to refuse it. (`claim:dcc:typography-4`, `claim:dcc:definition`)
- **Web translation** — Let one control move, resist or do something other than it advertises, for the mouse only. _Why:_ Unexpected interaction is the record's evidenced behaviour; keyboards and assistive technology never meet it. (`claim:gestalt.interface`, `token_rule:motion-14`)

### Motion

Durations, easing and the rest of this specification are Web translation: Aesthetary's choice for the screen.

- Durations: resist 150ms, flip 500ms, settle 300ms
- Easing: all `steps(n): motion lurches, it never glides`
- Transforms: translate and rotate on hover or pointer approach; the resisting control moves 80 to 220px away
- Entrance: none: the page arrives already broken
- Hover: a block flips or jumps in 2 to 4 steps
- Reduced motion: Nothing moves and nothing resists; the placed tilts stay, and the straighten switch removes them.
- **On the record · From our catalogue entry · Source attached** · `token_rule:motion-14` → Make interaction unexpected: it is a claimed behaviour, not an inference.
- **On the record · Documented** · `claim:kcc:interface-1` → Weird cursors and unnecessary, distracting animation are documented.
- **Web translation** — Move in steps, never eased. _Why:_ Easing reads as polish, and the record refuses smooth, predictable interaction. (`claim:failure-1`, `token_rule:motion-14`)

### Deliberately not added

- **A palette** (color): The record says the colours clash; it names none. The six are the room's.
- **A typeface** (typography): The record grades the behaviour of the type, not a face.
- **A motif set** (imagery): The sources name none, and one calls anti-design a way of thinking.
- **Interface conventions** (components): The research record does not establish them. The four components are contemporary translations of documented rules.
- **Three-dimensional form** (shapeGeometry): An open question for the web and graphic trend.

## Coverage — what has been looked at

One row per research dimension a research pass considered. "Not researched yet" (`UNKNOWN`) means exactly that; a dimension can carry an Aesthetary recommendation and still be unresearched, and those are different things.

| Research dimension | Evidence | Resolved by | Carries a rule |
| --- | --- | --- | --- |
| Composition | Specified · From our catalogue entry | follows the recorded claim | yes |
| Typography | Specified · From our catalogue entry | follows the recorded claim | yes |
| Hierarchy | Specified · Our reading | Aesthetary decided | yes |
| Colour | Specified · From our catalogue entry | follows the recorded claim | yes |
| Density | Not specified by the source | Aesthetary decided | no |
| Imagery | Not specified by the source | — | no |
| Form/motif/surface/technique | Open question: form · Specified: motif, surface, technique | Aesthetary decided | yes |
| Motion | Specified · From our catalogue entry | follows the recorded claim | yes |

Unresearched: Form/motif/surface/technique. Where a rule appears for one of these it is Aesthetary's implementation recommendation, not a property of the language, and its Evidence line above says so.

## Not researched yet

Nobody has established a position for this language on:

- Form

**Do not fill these in.** An absent position is absent research, not permission and not a blank for you to complete. If the work needs a decision in one of these areas, make it on other grounds and do not attribute it to this language.

## Grade your own work

17 checks: 8 mechanically testable, 8 needing a person to look, 0 needing a source Aesthetary does not hold, 1 with no recorded position. 3 further research dimensions produced no rule (Density, Imagery, Form/motif/surface/technique). This rubric is silent on them. Scores do not add up. A failure against a required rule and a departure from something we invented are not the same result, and no total should let them cancel.

### Recognition — does this read as the intended Aesthetic?

- **[visual_human · recognition signature · Evidence: No grade recorded]** Strip every logo, word and borrowed image out of your work, then show it to someone who knows the collection. Do they land on this Aesthetic? The signature to hold is: Alignment broken so that nothing settles onto a shared edge, type set to be difficult rather than merely unusual, clashing colour with no arbitration, and interactions that behave unexpectedly — controls that move, resist, or do something other than what they advertise. The diagnostic condition is that the friction is functional: something is genuinely harder than it needs to be, and that cost was accepted deliberately. Web Brutalism exposes structure and leaves it working. New Wave Typography violates a grid that stays inferable; here there is no grid to infer.
  - _If it fails:_ The work may be competent and still read as a different Aesthetic, or as none. Nothing below matters until this passes.

### Fidelity — are its defining rules present?

- **[unassessable · no recorded position · Evidence: Not researched yet (`UNKNOWN`)]** Nothing is recorded as a requirement of this Aesthetic. There is no rule here that your work must hold — which means nothing you do can fail this dimension, and nothing you do can satisfy it either. That is missing research, not freedom: our research brief does record failure modes to avoid, listed under avoidance below. They are recorded judgement, not requirements.
  - _If it fails:_ Not applicable. An empty dimension is an absence of research, never a pass and never a failure.

### Implementation — were those rules translated appropriately into this medium?

- **[visual_human · From our catalogue entry · Evidence: From our catalogue entry]** Your work may depart from this — but knowingly, not by accident. Composition: broken grid.
  - _If it fails:_ A defensible departure, and you should be able to say why you made it.
- **[visual_human · Our reading · disputed · Evidence: Our reading · disputed]** Your work may depart from this — but knowingly, not by accident. Deliberately contested — competing signals that never resolve. Stating that there is no stable ranking is the instruction.
  - _If it fails:_ A defensible departure, and you should be able to say why you made it.
- **[visual_human · From our catalogue entry · Source attached · Evidence: From our catalogue entry · Source attached]** Your work may depart from this — but knowingly, not by accident. Colour: clashing.
  - _If it fails:_ A defensible departure, and you should be able to say why you made it.
- **[visual_human · From our catalogue entry · Source attached · Evidence: From our catalogue entry · Source attached]** Your work may depart from this — but knowingly, not by accident. Rare, and it is the record's sharpest asset. "Unexpected interaction" is a claimed behaviour, not an inference, which makes this one of the few records where motion is evidenced rather than invented.
  - _If it fails:_ A defensible departure, and you should be able to say why you made it.
- **[visual_human · Our reading · Evidence: Our reading]** Your work may depart from this — but knowingly, not by accident. Breakage, not exposure: alignment and legibility are deliberately broken. If the structure stays legible the artefact is Web Brutalism, and if the system stays inferable it is New Wave Typography.
  - _If it fails:_ A defensible departure, and you should be able to say why you made it.

### Avoidance — known failure modes, neighbour drift, cloning

- **[mechanical · recorded failure mode · Evidence: No grade recorded]** Check your work against this: No smooth predictable interaction.
  - _If it fails:_ A failure mode recorded in our research brief: recorded judgement about what weakens this Aesthetic, not a requirement of the language.
- **[mechanical · recorded failure mode · Evidence: No grade recorded]** Check your work against this: No settled alignment.
  - _If it fails:_ A failure mode recorded in our research brief: recorded judgement about what weakens this Aesthetic, not a requirement of the language.
- **[mechanical · recorded failure mode · Evidence: No grade recorded]** Check your work against this: No arbitration between colours.
  - _If it fails:_ A failure mode recorded in our research brief: recorded judgement about what weakens this Aesthetic, not a requirement of the language.
- **[mechanical · recorded failure mode · Evidence: No grade recorded]** Check your work against this: No conversion optimisation.
  - _If it fails:_ A failure mode recorded in our research brief: recorded judgement about what weakens this Aesthetic, not a requirement of the language.
- **[mechanical · recorded failure mode · Evidence: No grade recorded]** Check your work against this: No decorative-only awkwardness.
  - _If it fails:_ A failure mode recorded in our research brief: recorded judgement about what weakens this Aesthetic, not a requirement of the language.
- **[visual_human · recorded failure mode · Evidence: No grade recorded]** Have you made any of these changes? Each one moves the work into a neighbouring Aesthetic: Against Web Brutalism: breakage versus exposure. Against New Wave Typography: no grid versus an inferable one. Anti-signature: if everything still works smoothly, it is Web Brutalism wearing a costume.
  - _If it fails:_ You have built the neighbour. That may be fine — but name which Aesthetic you are in.
- **[mechanical · declared closure · Evidence: No grade recorded]** Check your work against this: Not open: making it secretly usable..
  - _If it fails:_ This one is closed, not open. The record says generation must not go here.
- **[visual_human · neighbour drift · Evidence: No grade recorded]** Could your work be mistaken for Web Brutalism? That language is: Plainness as a statement; Utilitarian's plainness is unexamined
  - _If it fails:_ Not necessarily wrong, but you are closer to Web Brutalism than to this Aesthetic.

### Accessibility — the Aesthetic preserved without its inaccessible practice

- **[mechanical · recorded accessibility cost · Evidence: No grade recorded]** This Aesthetic accepts "Conversion" as a cost. Your reader did not. Does your work carry that cost in the decorative layer only — display, demonstration, atmosphere — while every passage meant to be read holds a contrast ratio of at least 4.5:1?
  - _If it fails:_ You have inherited a historical cost that the language never required you to pass on. The Aesthetic survives at AA on reading matter; reproducing the cost is a choice, not fidelity.
- **[mechanical · recorded accessibility cost · Evidence: No grade recorded]** This Aesthetic accepts "Accessibility" as a cost. Your reader did not. Does your work carry that cost in the decorative layer only — display, demonstration, atmosphere — while every passage meant to be read holds a contrast ratio of at least 4.5:1?
  - _If it fails:_ You have inherited a historical cost that the language never required you to pass on. The Aesthetic survives at AA on reading matter; reproducing the cost is a choice, not fidelity.

**Not covered at all:** Density: Not specified by the source; Imagery: Not specified by the source; Form/motif/surface/technique: Open question: form · Specified: motif, surface, technique. These research dimensions produced no rule, so there is no check. Treat them as unscored, never as passed and never as failed.

## Rules for the agent

1. Do not invent rules for any area listed under **Not researched yet**.
2. Do not treat an absent position as permission to do anything you like — say that the language has no position, and decide on other grounds.
3. Keep our interpretations separable from the language's own requirements. If asked to justify a choice, cite which kind of claim it came from.
4. If you need something this record does not contain, say so rather than producing a confident answer from the aesthetic's name.
5. When you self-score against "Grade your own work", score only checks marked `mechanical`. Declare `visual_human` and `evidence_dependent` checks unjudged rather than guessing, and never score an `unassessable` one.
6. Evidence "Not researched yet" (`UNKNOWN`) is never a failure. It means nobody researched the dimension, so departing from what is written there costs nothing and must not be reported as a defect. "No grade recorded" is not a grade either: do not treat it as our decision or as an outside source.
7. In **Design grammar**, a "Web translation" line is our choice for the web and may be changed with a reason; the rules it cites may not. Never present a translation as a property of the language.

## Provenance

- Record: `anti-design` · status `draft` · research `partly_researched`
- Full structured record: https://aesthetary.com/data/anti-design.json
- Human-readable record: https://aesthetary.com/a/anti-design.html
- This file is a snapshot. The record may have changed since; check the page above.

_Aesthetary is an alpha. Where a page says nothing, nobody has done the work yet._