How Pulse measures authenticity.
Four detector formulas, applied to public data, published in full so anyone can check the math.
What this is
Agent Economy Pulse publishes statistical anomaly measurements over public data — the x402scan paid merchants API, agenteconomy.to, and 8004scan.io — refreshed every 6 hours. Anomalies are patterns, not verdicts: a high anomaly can have innocent explanations. A B2B service with a handful of large clients, for example, will show high buyer concentration without anything being wrong. That is why we publish every formula and every observed number, so anyone can reproduce the result and judge the context for themselves.
The four detectors
Buyer concentration
total_amount ÷ unique_buyers, compared to the peer median
We compute each merchant's revenue per unique buyer and compare it to the median of its peers. The anomaly grows with the log-multiple over the peer median and saturates at 100×. Field anchor: on Jul 10, 2026 a top-3 merchant showed $7.0M of volume through 94 unique buyers (~$74K per buyer) beside peers with 12,827 and 25,226 buyers.
Volume oscillation
mean(on-days) ÷ mean(off-days), split at the series median
The daily transaction series is split at its median into “on” days and “off” days. We take the ratio of the on-day mean to the off-day mean; the anomaly ramps from 2× to 12×. Organic traffic drifts, while scheduled traffic flips. Field anchor: the Jul 2026 series flipped between roughly 200K transactions per day and 14–80K.
Mint burst / registration spam
share of last 100 registrations per chain in identical-name clusters or procedural-name patterns
For each chain we take the 100 most recent ERC-8004 registrations and measure the share that sits in identical-name clusters (3 or more, counted when minted by distinct wallets — the one-mint-per-wallet Sybil fingerprint) or matches known procedural-name patterns. The anomaly equals that share. Field anchor: 35 identical “Ave.ai Trading Agent” mints from 35 wallets within 2.8 hours (minimum gap 17 seconds) on BNB Chain, plus 61 of 100 procedural names.
Cross-tracker divergence
merchant ≥25% of one tracker's tx counts, absent from an independent tracker's facilitator list
When a merchant accounts for 25% or more of one tracker's transaction counts but its facilitators do not appear on an independent tracker's list, the two views of the ecosystem diverge. The anomaly scales with the merchant's share.
Scoring
The authenticity score is 100 × Π(1 − anomaly_i)— independent anomalies compound multiplicatively. A score of 100 means “no anomalies observed,” not an endorsement. The score also feeds AgentRadar's trust composite as one of seven signals at 10% weight — deliberately bounded, never a hard block.
Self-disclosure and allowlist
Our own collector wallet 0x3cb6319ca1FcCa620fFe531cb32a226BEF55e420performs one paid $0.01 x402scan pull per 6-hour run. It is excluded from flagging and marked “monitoring (ours)” wherever it appears. Any operator can contest a label or request allowlisting of self-identified monitoring traffic — contact vl.bichev@gmail.com.
Data sources
- x402scan.com paid merchants API — x402-settled, $0.01 per pull.
- agenteconomy.to public data.json.
- 8004scan.io public API.
Labels for addresses come from our curated registry (verified purchases from the Jul 2026 field study) and ERC-8004 registration names. Unlabeled addresses are shown truncated.
Corrections
If you believe a measurement is wrong, email vl.bichev@gmail.com with the address in question. We will re-run the measurement and publish the correction.