PROTOCOL v0.1 · OPEN
Design, assayed.
An open protocol for evolving design systems the way labs screen molecules. Generate widely. Measure everything. Keep what survives.
Illustration of a 96-well assay plate. Seven wells remain highlighted after a scan. These are illustrative values, not real assay results.
THE PRINCIPLE
Variation is cheap.
Agents produce thousands of credible variants in the time a team makes three.
Judgment becomes a function.
What a senior designer checks gets measured, scored and logged.
Selection is the craft.
The designer defines what fit means and steers what survives.
Receipts or it did not happen.
Every assay publishes variants, hits, hit rate and gain over baseline.
RECEIPTS
Assay 001, in progress.
| Metric | Value |
|---|---|
| Variants generated | pending |
| Generations | pending |
| Hits | pending |
| Hit rate | pending |
| Baseline composite | pending |
| Best composite | pending |
| Gain over baseline | pending |
| Wall-clock time | pending |
| Compute cost (EUR) | pending |
| Human review time | pending |
First results publish here with a full run record.
Research basis.
Variation and selective retention as a model of creativity: Campbell (1960), Simonton. Design Assay turns the theory into an engineering practice.
Read the research basisBuilt for agents.
- /llms.txt
A plain-text index for agents.
- /protocol.json
The protocol as structured data.
- .md mirrors
The same docs, in Markdown.
Bring a target.
Have a screen, flow or design system worth assaying? Assays 002 and 003 are open.