Workflows / 04

Check FidelityWorks today, by hand

Judge built work against the same record that specified it: on-language, off-language, or a default.

As a manual rubric on every Build page, and in every agent guide. Automated checking of a page or screenshot is planned, not built.

What you bring

  • The built page, screens or framesRequiredWorks today, by hand
  • The Aesthetic it was meant to beRequiredWorks today, by hand
  • A URL or screenshot for automatic checkingOptionalPlanned

Before → Aesthetary → After

Before

  • Your builder or design tool
  • Specify a Direction

Aesthetary

  • Checks in five dimensions: recognition, fidelity, implementation, avoidance, accessibility
  • Which checks a person can measure and which need someone to look
  • What each failure means for this language
  • Telling an intentional trait from a median defaultRequires more research

Leaves with: Fidelity Report

After

  • Fix, then Specify again
  • Client review

What leaves Aesthetary

Fidelity Report: Planned

Fidelity Report

  • On-language findings
  • Off-language findings
  • Defaults that crept in
  • Each tied to the rule it tests

What exists today

The rubric on a Build page, filled in by a person.

Worked example

Kawaii · real data

In

Work meant to be Kawaii.

The rubric, from the record Real checks

  1. 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: Mascot characters built to consistent proportion, sparkle and heart motifs placed at controlled intervals, a pastel field carrying exactly one bright accent, and every corner and terminal rounded without exception. The diagnostic condition is precision inside softness: the construction is exact even though the vocabulary is soft, and loose execution reads as a different and lesser thing. Mascot-Led, a Module, is the layout device of building a page around a character; this is a full register with its own palette rule and craft standard.

    recognition · someone has to lookIf it fails: The work may be competent and still read as a different Aesthetic, or as none. Nothing below matters until this passes.
  2. Your work may depart from this — but knowingly, not by accident. Rank by the single bright accent and character gaze — the accent marks the one important thing, and the mascot looks toward it.

    implementation · someone has to lookIf it fails: A defensible departure, and you should be able to say why you made it.
  3. Your work may depart from this — but knowingly, not by accident. Colour is a count constraint: pastel field plus exactly one bright. A second bright breaks the register — the same shape as Seventies Revival's accent prohibition, inverted.

    implementation · you can measure thisIf it fails: A defensible departure, and you should be able to say why you made it.

3 of the checks on the Kawaii brief. Filled in by a person; the Fidelity Report is planned.

Every Aesthetic's brief carries the same checklist under “Grade your own work”. 83 Aesthetics have one.

Routes

2 · products as routes, not destinations

Frames or screenshots, checked by handWorks today, by hand

Into Aesthetary · Figma

Put the work beside the Build page rubric and answer each check. Measurable checks first, then the ones that need someone to look.

Alongside ImpeccableNeeds verification

Pairs with Aesthetary · Impeccable

Future route: run its checks, then decide which findings the chosen language actually permits. How it handles intentional exceptions has not been checked.

Failure modes

Visual examples planned

  • Scoring against design in general instead of against this language.
  • Passing a check nobody has researched. Unresearched means unscored.
  • A generic lint rule flags a trait the language requires.

What Aesthetary does not do here

  • Generic anti-slop lintingImpeccable
  • Accessibility auditing beyond the rubricDedicated accessibility tools

Try it

Grade your own work: the Kawaii brief →

Feedback