What to try — hotelsweep

# WHAT TO TRY — Item 6 · Hotel Sweep ("Night Clerk")

**Open:** `index.html` in this folder — any browser, works offline. Then read `SPEC-6.md` (3 minutes) for the full session concept.

1. **Read the ledger top to bottom.** Two real captures from Oct 31 — the Moto and Samsung 60-minute paired run — are laid out as BASELINE vs. SWEEP, with the diff in three buckets: 175 shared · 234 baseline-only · 80 sweep-only · 489 union.
2. **Check the math yourself.** The page prints the identities (175 + 234 = 409 · 175 + 80 = 255 · union 489). They tie — that's the fixture working, not decoration.
3. **Sit with the caveat.** The amber box is the most important part: the gap between the buckets is mostly address-resolution asymmetry, *not* one phone hearing better. If the demo teaches anything, it's that a diff is a question, not a verdict.
4. **Picture the real use.** Arrival capture in a hotel room, second capture before checkout, same spot — passive listening only, diff read at home. That's the whole product; the spec (§2) is four steps.

**Nothing to install, nothing captured.** No hotel session has been run — the page says so on its face. A first real sweep is a separate, owner-run commission whenever you call it.