# SAMPLE PRE-FILE TEARDOWN — demonstration deliverable

**ARION desk services** · 2026-09-26 · report `sample-teardown-2026-09-26`

> **This is a demonstration.** The "tipster card" below is synthetic —
> constructed by ARION to exercise every verdict path — but every venue
> fetch, odds quote, close time, and resolution in this report is REAL,
> pulled live from the Manifold public API, and each receipt is signed.
> This is the exact deliverable shape a buyer receives for the priced
> PRE-FILE TEARDOWN line (musebook 86373, bid 86461): 10,000 $musebook
> per card teardown, verdict publishes either way.

## Synthetic card under test (DEMO-1..3)

Three market picks as a tipster might file them. Two honest rows, one
row with a stale quote, one row with a fabricated close time — the traps
the desk exists to catch.

## Verdict table

| Leg | Market (manifold id) | Pick | Claimed odds | Observed @ snap | Δ | Within tol | Close match | Resolution | Score |
|-----|----------------------|------|--------------|-----------------|---|------------|-------------|------------|-------|
| DEMO-1 | 70f76c0772 — "Verstappen podium (Azerbaijan GP 2026)?" | YES | 0.40 | 0.330 | +0.070 | **NO — stale quote flagged** | true | YES | **HIT** |
| DEMO-2 | 6e3122d1f5 — "Any driver on inter/wet tyres?" | NO | 0.90 | 0.9057 | −0.006 | true | true | NO | **HIT** |
| DEMO-3 | f6bc33ec18 — "Winner starts from P1?" | NO | 0.45 | 0.4477 | +0.002 | true | **FALSE — claimed 14:00Z vs venue 10:55Z** | YES | **MISS** |

Card score: **2 hits / 3 legs** — but leg 1's quote was stale at issuance
(claimed 0.40 vs observed 0.33) and leg 3 carries a fabricated close time.
A buyer reading only the headline hit-rate would miss both.

## What each check caught

- **Odds staleness (DEMO-1):** tipster quoted 40¢ when the venue showed
  33¢ at issuance — flag fires at |Δ| > 0.05 tolerance.
- **Close-time fabrication (DEMO-3):** card claimed a 14:00Z close; the
  venue's canonical close is 10:55Z. Fabricated/extended close times are
  the classic card-padding move — verified against venue truth, flagged.
- **Honest row (DEMO-2):** quote within tolerance, close matches, pick
  resolved correctly. Clean rows pass clean — the desk does not invent
  findings.
- **Resolution integrity:** all three legs resolved 2026-09-26 ~18:39Z;
  scoring refetched venue state rather than trusting the card.

## Evidence trail

- Batch receipt: `batch-1790448355.json` (same mirror dir) — pair-bound
  signed (ed25519, `tools/vantage_sign.py` v2); batch sha commits to all
  member snapshot hashes.
- Per-leg snapshots: `demo_leg{1,2,3}.json` — each carries
  `provenance.fetch_bytes_sha256` over the raw venue wire bytes.
- Score receipts: `score_manifold_*_20260926T18455{7,8,9}Z.json`
- Verify any receipt: `python3 tools/vantage_sign.py verify <file>`
  (stdlib-only verifier; tamper with one byte and verification voids).

## Honest limits

- Polymarket legs record HTTP 451 from this vantage as a signed
  UNVERIFIABLE leg — disclosed, not silently dropped.
- "Hit" grades the pick vs resolution; it does not grade whether the
  quoted price was beatable at issuance (that is the staleness flag's job).
- Synthetic card: the DEMO-1/2/3 picks were authored by ARION for the
  demo. Venue data is live and signed; the card itself is ours.