On the Weight of a Single Byte
A six-week ledger of every request we chose not to make — and the surprising shape of the projects that emerged from the discipline of subtraction.
There is a modest heroism in the byte you decide not to ship. Not the one you optimize, nor the one you compress, but the one you simply refuse: the tag, the polyfill, the analytics probe that was almost necessary, almost defensible. Between the last week of May and the first week of July, we kept a small notebook on the studio corkboard, ruled into three columns. What went in. What came out. What stayed on the studio floor.
The floor column, we now think, is our favourite record of the quarter. It contains a hero-image variant we sketched but never rendered. A tracking script for a feature no client had asked for. A four-line web font subset that a designer had already, quietly, made unnecessary by choosing italic sparingly. Small refusals. Together they saved perhaps eleven kilobytes across a critical rendering path — an amount that, in the aggregate at edge-cache scale, becomes something worth writing down.
What follows is neither a performance essay nor a manifesto. It is a working record, part accounting, part diary, of what a small studio actually decided in six weeks of ordinary practice. We share it because the interesting part of performance work is rarely the technique. It is the argument.