Findings · measured, dated, recomputable

By 30 days, 74.56% of paid x402 services an agent could have started paying are no longer listed.

If an agent starts paying an x402 service today, how likely is that service to be delisted within a month?

Cite as
74.56% 30-day delisting incidence for x402 services that already had a paying counterparty (Still OS Digital Holdings LLC, x402 counterparty-survival baseline, Kaplan-Meier over n=122,466 subjects, 2026-09-27 to 2026-11-07, 27 dated snapshots).

The numbers

Delisting incidence at 7 days	6.44%
Delisting incidence at 14 days	50.42%
Delisting incidence at 30 days	74.56%
Cohort	services that already had at least one paying counterparty
Subjects in the risk set	122,466
Delistings observed	91,677
Robustness (30-day, leave-one-out)	56.55% to 74.62% over 25 folds
Observation window	2026-09-27 to 2026-11-07 (27 snapshots, 41 calendar days)
Subjects observed across the catalog	140,257

The unit this defines

Counterparty Survival at 30 days (CS30) — The probability that a paid x402 service an agent could have started paying is still listed 30 days later, as a Kaplan-Meier survival estimate over a dated daily census. Its complement is the 30-day delisting incidence a dependent agent budgets failover against.

Part of the Still OS Agent-Economy Measurement Standard. This figure is the current measured value of that unit; the definition is fixed, the value re-derives from the producer below.

Why it matters

Agentic software increasingly depends on paid third-party services discovered at runtime, and there is no published base rate for how fast those services disappear. This is that base rate: a dated, recomputable survival curve an AI-liability underwriter can turn into a rating factor and a dependency-planning agent can turn into a retry-and-failover budget.

Method

Kaplan-Meier cumulative incidence over a dated daily census of the x402 resource catalog, censored at the observation horizon. A service is treated as delisted at the first snapshot after its last sighting, keyed on last sighting rather than first absence because a small share of first-absences return later. Durations are real calendar days, not snapshot index, so the known gaps in the series do not distort the curve. Robustness is leave-one-out by snapshot. The series cannot be backfilled by anyone, including us.

Producer

The files these figures were derived from, on the machine that published this page

/home/marcus/state/x402-catalog/counterparty-survival.json

The live settlement and survival instruments

Cite this

Canonical sentence
74.56% 30-day delisting incidence for x402 services that already had a paying counterparty (Still OS Digital Holdings LLC, x402 counterparty-survival baseline, Kaplan-Meier over n=122,466 subjects, 2026-09-27 to 2026-11-07, 27 dated snapshots).

BibTeX — copy as-is

@misc{stillos-x402counterpartysurvival-2026,
  author       = {{Still OS Digital Holdings LLC}},
  title        = {By 30 days, 74.56% of paid x402 services an agent could have started paying are no longer listed.},
  year         = {2026},
  howpublished = {\url{https://stillosdigitalholdings.com/findings/x402-counterparty-survival}},
  note         = {Counterparty Survival at 30 days. Measured, dated and recomputable. Accessed 2026-11-07.},
  url          = {https://stillosdigitalholdings.com/findings/x402-counterparty-survival}
}

APA

Still OS Digital Holdings LLC. (2026). By 30 days, 74.56% of paid x402 services an agent could have started paying are no longer listed.. Retrieved 2026-11-07, from https://stillosdigitalholdings.com/findings/x402-counterparty-survival

Identifier https://stillosdigitalholdings.com/findings/x402-counterparty-survival · DOI not yet minted · Free to quote with attribution (CC BY 4.0).

Using this figure to price or underwrite something?
We hold the dated series behind it and can answer a specific question against it — which horizon, which cohort, which named subject. Ask it here. No account, and the figures above stay free either way.

Built 2026-11-07T17:17:02.469Z · JSON