The app opens on a populated digest ("Found while you slept: 3 strong fits"), never a blank chat box. Scroll a feed of compact verdict cards — each a mini-appraisal — and act inline: Pass / Comps / Shortlist. The middle ground between the full report (A) and the swipe deck (C).
Every card is a small Appraisal Card: verdict chip + fair-value band + confidence + a provenance link ("4 comps · sources"). The agent's one-line read sits visually softer (purple) beneath the computed band — KNOW vs RATE, even at card scale.
Shows populated cards, a live agent-working note + skeleton ("still appraising 2 more — pulling comps…"), and the over-priced verdict kept visible but demoted. Tap "resolve loading" to watch the skeleton fill.
The middle take. More inventive than a report (it's a proactive, ranked-by-value digest), safer than a swipe deck (no destructive gesture; everything stays scannable and reversible).
Numbers are realistic Eastern-Suburbs placeholders; images via picsum. Verdicts, bands, comp counts & confidence map to live D1 data (sold comps, $/m², SEIFA, catchments).