The Fibonacci sequence — 1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89 — is not a design system. It is a growth pattern. The question this project asked was whether a system built from it would carry that organic quality into the work, or whether the mathematics would simply produce another set of arbitrary values dressed in natural authority.

The type scale was the first decision. Taking 13px as the base (the second occurrence of a prime in the sequence), each step multiplied by the ratio between adjacent Fibonacci numbers as that ratio converges toward φ. The result is not geometrically perfect — it is biologically approximate, and that approximation is the point.

Spacing followed the same logic, with the baseline grid set at 13px and all vertical measurements snapping to multiples thereof. The 14-column grid was a pragmatic deviation: 14 is not Fibonacci, but it divides cleanly into the common proportions the work required. Some constraints are mathematical, some are practical. The skill is knowing which is which.

The project began with a collection of approximately 4,000 documents — letters, photographs, institutional records — held by a private family archive. The material had been digitised but not organised; it existed as a flat folder of files with inconsistent naming conventions and no metadata structure.

Conventional archival interfaces impose hierarchy: series, subseries, item. This works well for retrieval — if you know what you are looking for. It works poorly for discovery — if you do not. The family wanted both.

Rather than build a single interface that attempted to serve both purposes, the project proposed two entry modes: a structured index for retrieval, and an associative view for discovery. The associative view was the interesting design problem.

The conventional model of site navigation is hierarchical — a tree, with a root and branches. This prototype asked what navigation might look like if the underlying model were topological instead: a surface with regions, adjacencies, and paths, rather than a structure with parents and children.

The practical question was whether users could orient themselves within a space that had no fixed entry point and no explicit breadcrumb. Early testing suggested they could, provided the visual language of the space was consistent enough to function as its own wayfinding system.

This remains a prototype. The interaction model is unresolved in several places, particularly around deep linking and the handling of content that belongs to multiple regions simultaneously.

Every image was graded toward what I thought of as the glass edge — that specific green-blue you see in the side of a thick piece of plate glass. Cool, precise, slightly industrial. It gave the outdoor work a quality that sat between the natural and the constructed, which felt right for clothes that existed in exactly that space.

The shoot took four days across two locations — a house in Kensington that had been in the same family for sixty years, and a suite at a hotel that had agreed to let us work in rooms between guests. Both spaces were chosen for the same reason: they held the texture of occupation without any current occupant.

Every decision in the frame was made to preserve that quality. No styling that announced itself. No objects placed for the camera that would not have been there otherwise. The work of the shoot was largely editorial — choosing what to remove rather than what to add.