x402 growth, decomposed | SmartFlow Observatory

What it measures

Scope note: Some category definitions and all results are outside the scope of this public version. What remains is the universe definition, the pipeline that produces the decomposition, and the tests that can falsify it.

A 30-day universe of Base USDC micro-transfers, decomposed into overlapping categories. The universe is defined as transfers from 0.0005 through 5.0 USDC in the independent tracker.

The overlapping categories include wash flags, bundler traffic, facilitator-mediated rows, and an unflagged, facilitator-settled core after the source's wash passes. "Clean" names that remainder; it does not mean independently verified organic demand.

How it is produced

The generator creates site/decomposed.json with a SHA-256 recipe seal. A site builder renders the HTML and chart. A sanity script independently recomputes the seal and three values, and the scheduled pipeline does not publish unless those checks pass.

Source scripts and pipeline

decompose.py
build_site.py
sanity_check.py

decompose --skip-rpc && sanity && build && rsync

The source database is the independent payment tracker, not a third-party index.

Limits and assumptions

timestamp is ingestion time, with batches varying by about one hour, rather than block time. Daily boundaries can therefore move by up to an hour.
Missing tx_sender values make the reported bundler category a lower bound.
The bundler list comes from deterministic EntryPoint 0.6 and 0.7 classification. New EntryPoint versions require an update.
The claim applies to the defined USDC micro-band on Base, not to all USDC transfers or higher-value payments.

How to refute this

Shift the day boundaries by one hour in each direction and recompute. Because timestamp is ingestion time and batches vary by about an hour, a category series that moves materially under that shift is not stable enough at day resolution to carry a claim.
Re-run the classification against a current EntryPoint version list. The bundler category is derived from deterministic EntryPoint 0.6 and 0.7 matching, so an EntryPoint version in use on Base but absent from that list leaves real bundler traffic classified as something else.
Move the lower and upper boundaries of the 0.0005 through 5.0 USDC micro-band and recompute the decomposition. If the category results change materially under reasonable boundary choices, the claim depends on the selected band rather than on a stable x402 population.

Last validated

2026-07-10