‹ All decisionslive
d_ab345133
SOFT2026-08-08 10:30:48 UTC · partner demo · threshold profile balanced · source live
Reasoning
Session behavior looks self-directed — address typed by hand, no long idle gaps, only two tab switches, no declines or KYC retries — so there's no sign of someone being walked through this in real time. What warrants a single context screen is that a well-established account is sending 6x its usual amount into a contract first seen one day ago and moving it out within 15 minutes, which fits a risky-but-informed decision (new protocol or presale) rather than a coached one.
amount_vs_median 6x + destination_first_seen_days 1 (new contract)typed address, idle_gaps 0, no escalation
Signal vector, as received
| Signal | Value | Contributed |
|---|---|---|
| account_age_days | 180 | — |
| is_first_purchase | false | — |
| amount_eur | 6000 | — |
| amount_vs_median | 6 | yes |
| funding_method | bank_transfer | — |
| minutes_to_outbound | 15 | — |
| destination_first_seen_days | 1 | yes |
| destination_type | contract | — |
| declined_attempts | 0 | — |
| address_entry_method | typed | — |
| idle_gaps_over_60s | 0 | — |
| tab_switches | 2 | — |
| deposit_escalation | [6000] | — |
| kyc_retries | 0 | — |
| merchant_verified | null | — |
Amount
€6,000
User outcome
Not recorded
Model
claude-opus-5
Prompt fingerprint
pattern:03f286e3e72b probe:8f9e2db8c06c verdict:2aaed03b60a9