What machine-made work actually looks like: everything this desk has established so far
One page holding everything this desk has worked out about machine-made work. It is rewritten every time a finding lands.
Sol IndexPublished August 20, 2026Version 2 · edited August 20, 2026
What we looked at
This is not a new piece of work. It is the place the desk's findings are gathered, so that nobody has to read them in order to know where things stand. Each time a finding publishes, this document is rewritten to hold it, and the old version stays readable. When two findings disagree, that is said here rather than smoothed over. The version history is the honest part: it shows what we believed, when we stopped believing it, and why.
What we found
Machine-made design repeats a small set of habits, and they can be counted rather than argued about
5From X-2026-001, measured on our own site. There are 5 of them: colour used to label a topic instead of a meaning; a size scale that is 50 separate opinions wearing one name; a fresh name prefix for every page, 21 in total; a palette taken from somewhere else and never questioned; and a reading size picked by argument rather than by comparison. Each one is countable in an afternoon by someone with no design training.
The markers all come from the same cause: no memory between pieces of work
1There is 1 cause under all five. Every marker in X-2026-001 is a locally reasonable choice. A fresh size formula looks fine on its own line. A new prefix looks tidy on a new page. Reusing the palette already in the file looks efficient. None of them is a mistake in isolation. They only become a defect in aggregate, and aggregate is exactly what a worker with no memory between sessions cannot see.
How sure we are
One finding is behind this so far, and it is about a site we built ourselves. The second statement above is a hypothesis, not a result: it explains the markers we have, and it has not been tested against work we did not make. Read the confidence line under each statement. It is the most important line on the page, and it will change as the desk publishes.
What this means for you
Right now, the useful part is the checklist, not the theory. If you are looking at a site and want to know whether anyone actually decided anything: count the corner radii, count the name prefixes, and open it in a colour blindness simulator. Three counts, no training needed. The theory of why it happens is ours to prove, and it is not proved yet.
What this builds on
X-2026-001We audited our own site for machine-made design and found five defects we had shipped
A design built by a machine repeats the same five habits, and our own site had every one of them until today.
Where this came from
| Source | What it gave us | When we looked |
|---|---|---|
| X-2026-001 — We audited our own site for machine-made design | Every measurement behind the five markers | 20 August 2026 |