Status: law — the desktop chapter is normative in docs/chapters/desktop.md (ratified v1).

principle. small surfaces first. the glanceable surface is the product; the big window is where things get built, not where they get used.

rule (law).

• the full window is a fallback and a construction space: building and editing scenes, authoring automations, settings that do not fit the popover’s glance.
• the window is a destination the user chooses, never the front door. first run and double-click favour the small surface.
• one organizer per surface. a surface has exactly one primary noun, and that noun gets the primary area. everything else is subordinate, state, or absent.
• contextual relevance: a control renders only if the current state calls for it. new controls must argue their relevance to this state, not to the app in general.
• interaction weighting: primary actions sit nearest the user’s point of entry (the anchor-adjacent edge on the popover, the thumb zone on touch, none on ambient surfaces — state only).

candidate — window choreography. a window never just appears. the frame arrives first, content lands a beat behind it, exits run the entrance in reverse rather than cutting, and the frame itself carries identity: any cropped screenshot of any window is recognisably glaze. frame grammar is shared across windows so hands learn how a window will behave before it finishes opening; choreography must stay legible at high content density, and the pretty never taxes the reading. (candidate for this chapter, on the record in the audit fold-in patterns; not law until it graduates.)

named gap. hierarchy under extreme density — hundreds of controls and content regions that stay legible, with consistent iconography and a budgeted motion register — is this chapter’s hardest open research problem, flagged in the same candidates record.

---

Canonical chapter: docs/chapters/desktop.md; map and spine in §13 of the design language document.