# Reading #3 (PLX-003) — Settler-frame day census, 2026-08-31 UTC

First reading whose coverage is a computed number: `scan-manifest.json`,
44 chunks, 44 ok, **coverage 1.000**. Data: `settler-census.json`.
Registry: `settlers.json` (x402-list, CC BY 4.0, fetched 2026-09-01).

## Frame

EIP-3009 authorizations on Base USDC (`AuthorizationUsed`), attributed to
a facilitator when the submitting wallet (`tx.from`) is in x402-list's
settler list (22 facilitators, 84 addresses). Blocks 50,672,527–50,715,726
(2026-08-31T00:00:00Z → 23:59:59Z). Outside the frame: plain-transfer or
permit-based settlement, Polygon, Solana. Method: `docs/settler-feasibility.md`;
instrument: `collector/settler.py` (`settler-scan` + `settler-census`).

## Headline

- 99,837 EIP-3009 txs on the day; **87.4% submitted by a known settler**.
- Coinbase rail: **72,246 purchases · $8,996 · 1,574 buyers · 716 sellers**.
  Then fluxa 12,655 · payAI 1,623 · mrdn 268 · primer 200; nine more
  facilitators under 200.
- Breadth: 451 buyers paid ≥2 sellers; 29 used ≥2 facilitators; 6 sellers
  settle through ≥2 facilitators.
- Only **2.2% of EIP-3009 USDC is x402**: $1.37M of the day's $1.40M is
  gasless treasury-scale transfers by unknown submitters (13 rows ≥ $1,000
  carried $74k in the 1h proof alone). Dollar totals here are not the
  economy's; purchase counts are the honest headline.

## Cross-check (x402-list, Coinbase)

| x402-list figure | value | ours |
|---|---|---|
| `tx_count_24h` @ 19:20Z | 48,320 | 72,246 |
| `tx_count_24h` @ 19:32Z | 57,205 | 72,246 |
| `tx_count_7d` / 7 | 59,764 | — |
| `volume_usd_24h` @ 19:32Z | $6,298 | $8,996 |

Same settler set, same chain. Ours runs 26% above their rolling 24h count
and 21% above their 7-day mean; their "24h" moved 18% in 12 minutes, so it
is a rolling or recomputed window whose bounds they do not publish. The 1h
proof on 2026-09-01 (19:00Z) ran at 2,546 coinbase tx/h = 61k/day.
Unresolved; recorded, not reconciled.

## Corrections made while building

1. **Hop folding.** fluxa and mrdn settle through a contract: every tx is
   buyer → contract → seller with a fee held back (mrdn $26.05 → $25.79).
   Counting both transfers doubled their USDC and listed the contract as a
   seller. `census()` now folds chained hops into one purchase (buyer,
   final seller, buyer-paid amount, fee); a downstream leg paying out more
   than it received is left alone (swap/bridge, not a fee).
2. **Receipt provider.** publicnode 403s `eth_getTransactionReceipt`; the
   first run recorded 2/2 chunks as `fail:` and resumed cleanly on
   blastapi. The manifest contract did its job.

## Residual worth labeling (bead vault-3hz2v)

`0x1360…66fa`: 5,949 tx/day direct to USDC, median $0.006 — an x402
facilitator missing from x402-list's list, most likely. ERC-4337
EntryPoints v0.6/v0.7 relay smart-account gasless transfers: not x402.