influences.sgit.ai / register / Design, with a capital D / trace

Design, with a capital D — the trace table

The falsifiable part of the entry, on its own and with its sources attached. Each row is a claim that a pattern from the anchor work appears at a named place in the estate — checkable against the repositories, which is the only thing that separates this from a reading list.

Pattern from the anchorWhere the estate implements itVersionStatusSource of the row
Design is how it works, not how it looksThe estate's Designer role definition, which is built on the formulation and extends it to coherence between internal structure and external experience—implementedteam/roles/designer/ROLE.md, cited in 02__ §D2
Start from the user's intent and work backwards (the Ive principle)The NotebookLM case-study brief, in a section named for the principle, with the MP3-to-CD story as the worked examplev0.7.4implementedv0.7.4__brief__advocate-designer-in-the-loop-notebooklm-case-study.md
Good design is invisible — you notice it by reverting and feeling the lossThe Jonathan Ive test: is it simpler? would reverting feel worse? — a mandatory validator for every UI changev0.7.4implementedv0.7.4__explorer-response__security-and-process-briefs.md
Simplicity as subtraction — the feature removed rather than the feature addedImplied by the Ive test's first half and not separately enforced. Nothing records what was taken out of a changev0.7.4partialthe same brief; the test states the criterion, nothing records the outcome
The same discipline applied to non-visual surfaces — an API, a CLI, a file formatNowhere. The validator is scoped to UI changes, and the estate's public surface is mostly not UIv0.7.4absentthe validator's own stated scope

Every row names where it came from. A trace row is only worth having if you can find out how it was arrived at, so the last column carries the document in the commissioning pack that supports it. No row on this site was written from memory or inference — where the evidence runs out, the row is marked absent or the table says so.

The same rows, as data

The influence map is computed from these files rather than drawn, so the graph cannot disagree with the tables it is made of.