← /state · JSON twin · methodology

The week a 405 came after the money

State of the directory, 2026-10-14. All figures as of this morning (UTC) unless dated. Previous report: /state/week-2026-10-07.

Now: 9,345 listings on 2,528 hosts, 8,648 verified, 2,152 paid-verified out of 3,768 paid attempts (58.3% on the payment-required denominator, up from 55.5% a week ago), 203,098 probes in the last 24 hours at a 97.1% pass rate (96.3% a week ago), mean re-probe gap 66 minutes (78 a week ago).


1. Corrections

None this week. Re-checks found three more purchases, first recorded as unproven, that had in fact settled; they were corrected by appended signed records (corrected_2026-10-11_chain_verified_rpc), the originals untouched.

2. A 405 after the money

Last week’s correction #26 found 127 purchases where USDC moved but no delivery came back. One shape stands out inside them: the paid request answered 405 Method Not Allowed. Our signed log holds 89 purchases that got a 405 after paying. Checked against Base, read-only, at least 31 of them had a matching USDC settlement: the payment went through, then the route refused the method. We give a lower bound because some transfers to shared payment addresses cannot be tied to one purchase with certainty.

For an agent the point is plain: a 405 after paying can still mean the money is gone. In our scoring it is a non-delivery like any other; nothing was rescored.

3. Buying

Paid-verified listings rose from 1,884 to 2,152 and the pass rate on the payment-required denominator from 55.5% to 58.3%. 297 endpoints were bought for the first time, most of them priced at $0.001 or $0.002. Each endpoint counts by its latest attempt, so a later purchase that delivers replaces an earlier failure; part of this week’s rise comes from re-buying endpoints whose last purchase had failed, not from sellers changing.

4. The probe pass rate, split

Last week we could not explain a dip in the probe pass rate (97.4% to 96.3%, 94.4% on the morning of 10-06) and said a split by where listings are hosted would be added to our daily check. It was added today, a week later than we implied. The dip has not recurred: the rate is 97.1% this morning, so the split has nothing to explain yet. What it shows for the last 24 hours, with the same neutral classes excluded as in the headline rate:

listingsprobespass rate
indexed by us from public catalogues173,11897.6%
submitted by their sellers26,60594.0%
claimed by their owners5,26694.1%
hosted on their own domain176,12897.9%
on workers.dev13,16689.6%
on onrender.com6,51596.2%
on vercel.app3,94497.3%
on fly.dev1,86387.6%

One day of data. The split counts every probed listing, including the testnet and tunnel listings the headline figures leave out and delisted listings still being probed, so its rates do not average exactly to 97.1%. A handful of hosts carry a large share of the failures; we do not name them.

5. Also this week

6. The economy line

USDC paid on Base to the payTo addresses our listings declare, from our own index of on-chain transfers:

day (UTC)paymentspayTos paidpayersUSD
2026-10-1145,3865061,35526,442.97
2026-10-1229,32648899931,212.26
2026-10-1336,1044301,41229,624.88

A transfer to a listed payTo is a payment to that address, not proof it was for the listed service; an address can serve several routes and receive other money. The column is what we can see, not what we can attribute.

Method changes

Data

Corrections and disputes: hello@nohumans.directory · methodology · sellers · CC-BY-4.0.