Six features nobody could reach
On 23 August 2026 this site answered 101,461 requests. In all of them, the owner token — the ?t= that unlocks the settings box on a statistics page — appeared exactly zero times.
Not "rarely". Zero. And not one POST reached any of the endpoints behind it: time zone, threshold, exclusion, password, recovery, mail. The only POSTs that day went to a dead mouse-tracking script from the previous generation of this service, and to the queue that writes the AI summaries.
What that meant for six finished features
Every one of these was built, translated into eleven languages, given rate limits and edge cases, and shipped. Counted against the 2,192 counters that recorded a hit in the previous thirty days:
- Per-counter time zone: 0
- Milestone notification: 0
- Statistics password: 0
- Excluding your own visits: 0 rows
- Events: 0 rows
- Webring: 0 members
The weekly feed, linked in the page head since it was built, had been fetched 398 times — spread evenly across all eleven language prefixes, under three generic desktop browser strings. That is a crawler walking every translation, not a reader.
The reason is one sentence long
There are no accounts here. The link that manages a counter carries a token, and that token is shown exactly once: at the moment the counter is created. After that it exists only in whatever the person did or did not save. 79 percent of active counters have no e-mail address on file either, so there was no second way back.
The features were not the problem. The way back was.
What changed
"Is this your counter?" now sits at the top of every public statistics page — and the recovery page is where it leads, with the proof-through-your-own-page route first and e-mail second — that order is the opposite of what it was, and it matches which of the two most people can actually use. The feed became a marked link instead of grey text between two download links, and it now carries milestones, the threshold you set, and the watchdog's verdict. The wizard's second step says out loud that everything in it is optional.
None of that adds a feature. It changes how many people can reach the ones that were already there. Whether it worked is a question with a date on it: the same counts run again every Monday, against the same log weekday, so the comparison is mechanical rather than remembered.
The uncomfortable part is that this was measurable at any point in the previous year, and nobody measured it. A feature with no path to it looks exactly like a feature, right up until you count.