Case study 02 · Concept · Product design · AI direction
A private, offline personal-finance ledger in a single HTML file — no cloud, no account, no network. I run my own money on it. It’s here because taste without shipping ability is a mood board.
The mark proofs — three rounds, in order
The executions, proofed
The system
Eleven themes, full palettes at verified parity. A finance app doesn’t have to feel like a tax form. Making it something you’d want to open is a retention decision, not decoration — and it’s the same eye, pointed at a spreadsheet’s job.
They divide into two kinds, and the difference is where the color comes from. Gulf, Riviera and Ocean were generated, proofed and selected — each one a candidate set narrowed to a shipped cut, and the palette is taken from the image. Ledger, Terminal and Matrix are references rather than photographs: a paper ledger, a monospace console, cascading green code.
The other five are my own photographs, and they behave differently. Wings, City and Route are monochrome by nature rather than by filter — the frames carry no color at all, so the accent has to be chosen rather than taken. Clouds and Landscape sit between: two hues pulled from one set and given opposite jobs, money in and money out. That constraint is the interesting half. A palette derived from a colorful photograph is a pipeline doing its work; a palette for frames that have no color is a decision, and it has to be defended against ten other themes doing the same job.

One decision, two framings
The Gulf theme takes its palette from a photograph. How that photograph is framed decides whether a viewer can tell. The interface is identical in both — the same figures, the same donut, the same bars. Only the ground moves.


When the photograph has no color
Wings, City and Landscape run on my own frames, and those frames are monochrome by nature rather than by filter. There is nothing in them to take a palette from — so the accent is a decision, and it has to hold against ten other themes doing the same job. Same dashboard, same figures, same month as the Gulf frames above; only the theme changes.




My role — concept, product design, UX, functionality, branding, AI direction.
Status — daily use on real finances, 1,595 transactions.
Built with Claude — authored, not just used. I set the product’s principles
and directed the build against them.
The product-side case study → — where the decision this app turns on lives: never write after a failed read.