Press · SmartFlow Observatory
For journalists and researchers covering x402 payments on Base.
Who this is
SmartFlow Observatory is an independent x402 settlement ledger on Base. It records USDC settlements and derives a small set of metrics from them. It is a one-person instrument. The author writes under the pseudonym Tom Smart.
What I measure
On-chain settlement events: how many settle, when they settle, which endpoints receive them, and which patterns look like wash activity under a published rule.
What I do not measure
- Downstream attribution. When a payment routes through a shared wallet or a facilitator, I report the on-chain recipient, not the party behind it.
- Private data. Everything derives from public chain state and public catalogs.
- Rankings for sale. No entry pays to appear, to rank higher, or to be removed.
Five headline numbers
Re-run against the live ledger on 2026-07-22. Each reproduces from a single SQL query; ask and I send the query.
Live artifacts
- Decomposed view: what the growth number is made of, recomputed daily.
- Sample attestation: what a signed audit output looks like.
- Anatomy: layer-by-layer decomposition of a headline count.
Notes on method and corrections
- Anatomy of a headline number: 154,279,092: the essay version, four layers deep, my own counter included.
- I published a wrong number: how the wash filter works and how a bad figure got corrected in the open.
- The facilitator number was wrong: a second public correction, and what replaced it.
Standards and spec work
Findings from this ledger have been adopted into open x402 specifications: explicit settlement status and the settle-response digest design in the AIR receipt format (v0.2 drafts), the payTo cross-check in the payment-integrity verifier design, and facilitator attribution rules with a byte-parity reference implementation. Full list with links: github.com/smartflowproai-lang.
Methodology and corrections
Every number reproduces from one SQL file. If a figure appears here or in a note, there is a single query behind it. Ask and I send the query.
I correct in the open. When a number changes because the method changed or an error surfaced, I say so and I log it. Published corrections are a feature of the instrument, not an exception to it.