You're geo-pinned to the home you're inspecting. You ask quietly (hold to ask — PTT for public), the agent speaks a one-line verdict and that verdict lands as the big glanceable line up top with a Replay. Below it, the artifacts you'll actually want to scan before walking in: appraisal band, fit-against-brief, the comps, and what to ask inside.
RESEARCH §4: real estate is browsed in public, at an inspection — design for the silent/visual fallback. This is the literal scenario. The verdict is a computed fact (“computed · not the agent's opinion”), every fit ✓ links to its data source, and the agent flags what it doesn't hold (B&P, flood) so you ask the right questions inside.
Two: the fit checklist (each ✓/✗ tagged → guide / listing / transit / census / comps) and “the 7 comps behind fair” with agency-direct source links, caught before the portal.
Real shape: Coogee 3-bed at $3.4m, $8.x k/m² comps, Coogee→Bondi Jct transit, auction-Saturday context, agency-direct early-catch. Faked: static mockup, no live audio/geo, picsum photos, illustrative comp figures.
The pragmatic middle: it keeps the conservative card grammar but reframes the whole screen around one high-stakes scenario rather than a generic call. Less novel than the orb, more opinionated and contextual than the plain call screen — the one most likely to win a real buyer on a Saturday.