v3.5 neutral-first is settled — tokens, type, motion, the teal rarity rule. Nothing in these three concepts changes a value, and the standing brief already rules that the visual system is not what needs redoing. What every audit, finding register and onboarding log points at instead is the layer above: the same object wears different chrome, different vocabulary and a different grammar depending on which surface you meet it on. That incoherence — not depth, and not the palette — is what reads as overwhelming.
So 4.0 is the coherence release. One shell, one list, one card, one vocabulary, three depth tiers, across all three surfaces and both kinds of user — the person on a phone and the agent in a harness.
A live v3.5 ↔ v4.0 switch over the same data: the shipped composition with five stacked control
layers, a tag wall ending in +104 more, and three identical badges on every row —
against one control row, filters in a sheet, honest counts, and badges only on the exception.
Then the rail at 24 → 6 with a destination for every removed item, hats and lenses made
touchable, and the bounded depth door. Toggle to 390px to see the composition that produced
"I don't even know where to begin."
Open the app concept →
Dressed by the same stylesheet as the app concept, which is the whole argument: an operator learns one rail and one list and has learned both surfaces. Shows the shared table reaching all twenty screens, density as a remembered setting, preview-as-role stated as a lens that never grants, the blast-radius confirm promoted to console grammar, and dense tables becoming cards at phone width. Sections are deliberately left alone.
Open the console concept →
The same stylesheet again, and here it carries the most weight: the native app in the build today is not a density of anything, it is a separate design system. Shows the four jobs with Attend added — governance, account, hats, library, people, support — and the arithmetic that settles the nav question. A live reachability counter over the fifteen destinations the mobile PRD puts in scope: the shipped five-tab bar reaches eight and strands seven with no route at all, the drawer reaches every one, and the counter describes what the shell can reach — never what the product holds.
Open the mobile concept →
The same shell with the argument removed and the screens left in. All fifteen destinations route; approvals resolve with undo and the tallies re-derive; capture drafts and files; the hat switcher seals with the nav order held constant; both product verbs run end to end (relay a Code in — preview, then the seal spends; hand off an unsealed relay and the Code is minted at that moment, never before); the relay detail carries both Code states — hero when shared, no Code, never left this workspace when not. Error states behind a toggle, deep-linkable per screen, full-bleed at phone width. Fixture data — nothing talks to the platform.
Drive the prototype →
relay-board/docs/product/PRD.mobile-v3.md.
The design-side direction is the
mobile v4.0 creative brief.
/brand/generated/relay-tokens-v36.css with no local override at all — RCTX-240 was ratified
on 2026-08-12 (--rly-accent-fill → #0f766e, 5.47:1), which retired the one
deviation these concepts carried. Wave 0 of the direction doc lists the remaining ratifications that
gate the rest — most cost minutes, and nearly every pillar waits on one.