This page describes what we have actually done, not what we intend to do. Where we have not done something, it says so. Most readings below were taken on 22 and 23 September 2026, and a line says so where one is older; they are a record of those days, not a live feed.
We checked whether our priced services do the work first and invoice afterwards. On 23 September, all 15 priced doors we could measure answered a request without payment with a 402 before anything else. One more door does not currently advertise a price and has no payment gate; it is not counted here. That 402 comes from our price list, before a request reaches any service. It shows a door is priced. It does not show that a service is running behind it.
The more useful question is whether that gate can say yes, or whether it just refuses everything — a door that is broken shut looks identical to a door that is working. So we drove the classifier across 22,500 combinations of payment state, control state and response body (re-run 24 September). Some reach a pass, and none of them pass while the control is anything other than a 402. The gate discriminates. That is a claim about the gate, not the service behind it.
Read one of our 402 responses by hand and everything you need to pay is in it: the amount, the chain, the token address, and the recipient. We have not yet shown a payment going all the way through. On 29 September, 67 of the 72 priced paths behind our doors had no service connected to them. Do not pay through these doors yet. We will remove this line when a payment has gone all the way through and come back with the work.
An automated x402 client could not use them. Our responses label themselves x402, but they do not carry accepts or x402Version — two fields the x402 specification requires — so off-the-shelf tooling has nothing to parse. The gap is also on the receiving side: our payment handler accepts a transaction hash, not the signed authorization an x402 client sends, so adding the two fields would not be enough on its own. We found the missing fields with a check aimed at the required fields; every gauge we had before it was satisfied by the label alone. We are not x402-enabled, and we have said so here rather than let the word do work the responses do not.
This is a readiness gap. We cannot say whether an automated buyer has ever tried and been turned away: the record we keep of outside use would not show one. What it certainly blocks is listing our doors where machines go looking for them.
We have no record of anyone outside this building using this. As of 20 September, every conversation thread on record was ours or a test. The founder has walked through them once.
We cannot currently evidence external readership. Our publication has one subscriber: the founder's own address, added 2026-07-12. That list, read on 30 September, holds no external subscriber; it shows who is subscribed now, not who may have joined and left before. That is one door — it is the one we measured, and we are not generalising it to the others.
We are not able to tell you whether our writing has driven anyone to us. The field that attribution would key on is empty on every arrival we can see, so the instrument is dead on that axis — which means it supports no claim in either direction, including a flattering one about how quiet it is.
Because the alternative was a page that sounded like every other one, and we would have known. Nothing above is a promise about what we will be. It is a list of things that were measured, each with a control that could have made it come out the other way, except where we say the instrument is dead.