Capture: voice only, on purpose
No keyboard face exists. Typing on a 40mm screen isn't a real capture path, so the watch offers exactly one: tap to talk, on-device transcription, sent to the same Capture pipeline as mobile and Slack. If the mic fails (no connectivity, permission), it fails closed with a haptic + "try from phone", never a silent drop.
What's deliberately absent
Search, the You tab, Explore/graph, and the Send/Claim segmented control all stay off the watch. Claiming a Code by typing 6–10 characters on a watch face isn't a real interaction; that flow stays on phone, Slack, or chat.
Scope
Concept only, unscheduled (no epic yet (companion to RCTX-459/460/461 in the Mobile V2 PRD, not part of it), reuses the mobile shell's v3.5 tokens and Act/Aware/Capture vocabulary at watch scale.