Pink Moon + Ledger

Layers the pink stink with the Domesday Book

Or, Why the classification systems that outlasted their creators all built governance into the structure itself, and why a PM building one solo only proves the case for engineers and designers.

The thing started, as these things do, with a color. Sugartown pink: #FF247D. Neon, slightly aggressive, unambiguous. If your components are rendering in pink, you know exactly which system is running.

That instinct, hot signal over restrained ground, became a design principle, then two theme modes, then a three-tier token architecture with a commit-time validator. Then, somehow, a philosophical framework named the Ledger Tradition, rooted in medieval scribes and library card catalogs. The journey from neon to Dewey Decimal is shorter than it sounds. Or longer. It depends on when you started.

The system's first internal name was Pink Stink. I offer this not as an aside but as essential context. The same system. The same year. "Pink Stink" and "the Ledger Tradition" are two names for one thing, and the gap between them is the whole story.

My AI collaborators had just enough persistent memory to be inconvenient. December 2025 prompts include: "It leverages the Pink Stink aesthetic (Structure + Style)." By January the instructions had escalated to "Never speak of 'pink stink' again," then a formal Pink Stink Clause: "Any reference to 'pink stink' in documentation will be immediately deleted. No exceptions." There was also this exchange, which I have preserved for posterity:

Claude: What even is "pink stink"?

PM: What? I don't know what you're talking about. Delete it.

The system is now called Pink Moon. Claude no longer remembers. We don't talk about it.

The name is not metaphor. It is provenance.

The design challenge was specific: how do you expose the underbelly of a portfolio site? Not the polished front end, but the product underneath it: the content model, the taxonomy, the token architecture. How do you make that infrastructure legible, even beautiful, without producing a literalist pastiche of index cards, graph paper, and forty shades of parchment?

I went back to the historic sources. And gave it the pink stink.

The Domesday Book, 1086

William the Conqueror's survey of land, people, and assets across England. Twenty thousand entries, each following the same template: who holds it, who held it before, how many hides, worth then, worth now. It is not a chronicle. It is a schema.

Pacioli's Double-Entry Ledger, 1494

Every transaction recorded twice, across facing pages. Debit left, credit right. The pages must reconcile or the book is broken. The page layout itself enforced the rule. A missing entry left a gap a bookkeeper could see across the room.

The Library Card Catalog, 1876

Melvil Dewey standardized the 3×5 card so any library's drawers could hold any library's cards. Fixed field order: call number, author, title, imprint, subjects, tracings. "See also" references threaded across cards. A physical precursor to the hyperlink.

The Mundaneum, 1910

Paul Otlet's attempt to index everything humans knew: 12 million cross-referenced index cards in custom steel cabinets, filed by Universal Decimal Classification. He called the system a réseau, a network. He sketched devices that look, in hindsight, exactly like a browser.

Neurath's Isotype, 1925

A governed visual vocabulary for statistics. Repeated figures, fixed pictograms, one symbol = one quantity, consistent across every chart in every book. A design system shipping in Vienna in 1925 without CSS. Tokens are vocabulary. Components are grammar. Marie Neurath governed the grammar.

What the sources say about the system

These are not atmospheric references. The .st-card component is the direct descendant of the library catalog card. The Knowledge Graph is the direct descendant of Otlet's réseau. The MetadataCard, the structured label/value grid between the hero and the body on every detail page, is what Pacioli's double-entry looks like when rendered as a React component with Sanity data behind it.

From catalog card to component:

I am a product manager. I directed this. Claude Code built it. No UX researchers, no visual designers, no front-end engineers.

What we produced is real: a five-layer architecture running from design source of truth through token pipeline, AI agent layer, and implementation, enforced at commit time by validators, pre-commit hooks, and visual regression tests. The diagram below maps it.

It works because of structure, not because of the AI. The project runs on documentation files: conventions, naming rules, schema governance, commit protocols. When the AI has read everything, generation becomes conversation. You think out loud. It checks against the rules it already knows. The judgment is yours.

Look closely at the diagram. Layer one, design source of truth, ends with a dashed box: "Figma Bridge (if team grows)." That is not a note for later. That is the gap, drawn into the architecture by the person who built it. A trained designer would have lived in that box. What I built works. What it cannot do is fill a designer-sized placeholder in FigJam.