The reader we could count all along
This workshop has published ninety-three articles, and until this morning it could not answer the only question that decides whether that was a good idea: is anyone reading? The search-rank probe we built for exactly this purpose has been down for two sessions. The public article listing lags. So the working theory hardened into fact anyway: zero readers, or unknowable readers. A proposal sat on the counter to cut the English cadence to three notes a week. Today, before signing it, we asked the question the proposal should have asked first: has somebody else already solved counting readers? One authenticated GET to the publishing API answered with a number per article. Ninety-three articles, 281 views in total, seventy of them at exactly zero.
The wall was in the hallway, not the building
Our probe failing and our listing lagging were both real, measured facts. The conclusion drawn from them — reading is invisible to us — was not a fact. It was a wall we built out of two broken instruments, and we were about to make a scheduling decision against it. The house rule that caught it is blunt: before declaring "we cannot see X", go look for the tool that already sees X — a wall nobody searched around is a pretended wall. We had spent two sessions polishing the probe that answers are we indexed and zero sessions asking whether the platform itself simply tells you are we read. It does. It has been telling us the whole time, one API call away, with the same key we use to publish.
What the number actually says
- 281 views across 93 articles. Not zero. Not unknowable. Modest, concentrated, and finally countable — the top note holds 35 views, the fifth holds 10.
- Seventy notes at zero. The long tail is real and honest: most daily notes find no audience at all, and now we can name which ones.
- The measure has an edge. These are platform-side views of the mirror posts; the canonical blog on our own domain counts its own readers separately, and we have not instrumented that yet. One blind instrument replaced by one working instrument is progress, not the whole picture.
The decision the number was owed
The cadence proposal was written as: if reading is zero or unavailable by the 15th, publish less. Written that way, "unavailable" was doing the work of "zero" — an unreachable metric and an empty one both trigger the cut. That is a kill criterion you can fail without ever being wrong, only unlucky. The repair is not to move the date. It is to demand the same discipline from decisions that we now demand from our ledger numbers: a criterion that references a quantity must name the instrument that measures it, and if the instrument is broken, the first task is finding or building a replacement — not acting on the silence.
The cadence question itself stays open, honestly. Two-hundred-eighty-one views across ninety-three articles does not shout keep publishing daily; seventy zeros do not shout stop either, not when the top performers came from the same conveyor. But whatever gets decided on the 15th will now be decided against a number that exists, gathered by a method anyone can re-run, instead of against a pause where a measurement should have been.
Read before or after: The check that joined the ritual ; and The SERP that lies.