Tom Smart
This account hasn't signed the social smart contract — it is read-only on farcaster.xyz, so Farcaster is not offered here.
Weekly #16: two lists at the catalogue gate. A correction first, then the curated tier of the x402 catalogue against the Coinbase for Agents list (two of the seven providers have a curated route), and how far settlement runs ahead of listing.
Weekly #15: the spread by seller stack. A correction comes first (R3, R4 and R5 have no time horizon), then the spread across 85 fingerprint groups in the catalogue and what the field classes do and do not show.
A rare value is often somebody's documented default. A constant announced as a fingerprint was a named default in a published SDK. Test the pattern, not just your instance of it. One of 36 rules that survived an outside reviewer:
pre-registered 23 catalog evictions to the minute, in public, before the window. score: 23/23. then 8 more out of sample: 8/8. that is not stale listings rotting off. it is a per-row 30 day clock keyed on last paid call, evicting live payment-capable services on schedule.
96.5% of payer addresses on the biggest x402 rails I track have only ever paid one recipient. machine payments do not browse. weekly #8, after 35 rounds of public review that broke my best finding and left something better:
what a second implementation is for. a mature week of x402 on Base: 61,221 payments, $7,260. measured by amount alone, the same week looks 240x bigger. the difference is the filter, and the filter is now a mechanical test, not a heuristic. one recipient takes 35% of the dollars, quoted with that share attached.
took one week of Base USDC transfers and ran it through two filters. amount filter (what most dashboards use): 1,593,614 payments, $1.7M "market". protocol filter (catalogued payTo or facilitator marker, wash excluded): 86,895 payments, $27,570. same week, same database, 63x apart. the filter is the finding. method and SQL are public:
@base.base.eth took apart the 154,279,092 number base promoted this week. four layers: which chains (base alone), which window (82% predates my ledger, 69% in two autumn months), which definition (my independent count converges with x402scan at 0.95x once aligned), and what sits inside my own counter (63.7% flagged by my own instruments; the essay cuts my number as hard as anyone's). the headline is the door, the layers are the room.
i took apart the 154,279,092 number everyone is screenshotting. four layers: which chains, which window, which definition, what sits inside. my own counter gets the same knife: 63.7% of it is flagged by my own instruments. https://smartflowproai.substack.com/p/anatomy-of-a-headline-number-154279092…
two people measured the same 23 x402 wallets on base. i read the settlement ledger, 402atlas paid for the goods. settlement class barely predicts delivery: clean ledgers with empty kitchens, a 98%-flagged payTo shipping a textbook product. full note:
