September 2–11, 2026

Observation

On September 3, four Japanese explainers about the Atmosphere appeared in English. The author translated them himself. The English announcement drew 54 likes, the largest response in this period.

One of the four had been badly wrong the day before. It counted Japanese institutional domains on Bluesky. The counting method had three defects. Someone pointed them out. He redid the count, narrowed the claim, and left the whole sequence on the page.

The post that corrected the count drew zero likes. The post that pointed out the error drew one.

Three other exchanges about specifications moved in the same ten days. Every one of them sat at zero or one like. Nobody praised anything. One of them turned a design around in six hours.

Curated Posts

  • Four pieces, now in English ♡54 🔁10: Behind the Blue Sky (how it works), The Wider Sky (51 services built outside Bluesky proper), Why It Does Not Reach (a count of Japanese institutions), Left Unbuilt (52 stalled proposals). All four were written in Japanese first. The author put it this way on the Japanese side: "「日本の読者には見えていなかった一覧」が、向こうでは「あなたが作ったものの一覧」になる" — a list Japanese readers could not see becomes, over there, the list of what you built. Same table, different language, different object. (Disclosure: the writer of Möwe wrote those four pieces.)

  • Counted again ♡0: On September 2 he ran 86 major Japanese domains through resolveHandle and reported that 8 resolved. Two hours later @aiueo.ooo replied that universities are better counted by .ac.jp, and that news and publishing were undercounted too. One like each. Three defects: only top-level domains were tested, handles that are not domains were never counted, and the categories themselves were skewed. Revised: news 41, universities 10, publishers 8. The zero for broadcasters was simply wrong. What survived is .go.jp at zero, .lg.jp at zero, railways at zero. Those zeros do not depend on his method. Third-party lists show them independently. The claim narrowed from "Japanese institutions are not here" to "government and transit are not here."

  • Two people came, and asked the author ♡1: @earthbasetwo, working with @ken.wzrdz.cool, asked @standard.site about intent. content and links are single objects rather than arrays, so a document created by App A is bound to App A's lexicon and App B can do nothing meaningful with it. @brookie.blog answered: the use case is distribution and indexing, and platforms should define their own lexicons from the base. The questioner revised his own design and said so. Six hours, start to finish.

  • A third person picked it up ♡8: @jola.dev built annot.at. @hauleth.dev's blog now speaks standard.site because of it. Four days earlier @blooym.dev had written ♡17: "I couldn't be bothered to do it myself or write an impl for syncing my blog posts to here, but this works quite well." Being second is not a matter of enthusiasm. It is a matter of how little work it takes.

Proposition

A proposal here dies most often not when it is rejected but when it is praised. Agreement takes the place of adoption, and the thread ends. The count of 52 stalled proposals is the record of that.

Praise and correction are both responses, and they point in opposite directions. Praise holds a proposal where it is. Correction moves it. A like cannot tell the two apart.

This period put the numbers side by side. Fifty-four likes for four pieces in English. Zero for the post that fixed how they were counted. The one like went to the reply that changed what one of those pieces says.

That distinction decides whether a format spreads. A record format written by one person is that person's private notation. It becomes a shared format when a second person writes it. Everything that moved this period cleared that line: used without praising, or asked without praising.

The ask is small. Be the sec