MojalandWhy It Does Not Reach 日本語English

MOJALAND / 03

Why It Does Not Reach

How should Bluesky, and the Atmosphere beneath it, be explained so that it lands? In Japanese the question keeps getting asked. Expose the jargon or hide it. Reach for a metaphor or stay precise.

I think the cause is not on the explaining side. This is the reasoning, in order. Unlike the other pieces here, it takes precision over plainness.

1

Usability is necessary and not sufficient

There is a line of argument that says: first beat the incumbent on usability. Half of that is right. A product worse at the central job loses whatever its architecture.

But "beat them on usability and you win" does not follow. Several products won on usability and died anyway.

Windows Phone. Metro was rated above iOS and Android by some at the time. Gone.

webOS. Its card multitasking was later copied by nearly everyone. Gone.

BlackBerry 10. The gesture model reviewed well. Gone.

What the three had in common was no apps. Not operability.

The winners did not win on it either. Windows overtook Mac on hardware breadth, price and application count. Android grew on carrier distribution and price tiers, not on being easier than iOS.

Platform wars are decided by developers, not users. People go where the apps are; developers go where the people are. A product that cannot enter that loop falls however well it operates.

2

What the Switch proved was exclusivity, and addition

A machine comprehensively outclassed on hardware beat one that was not. Two reasons.

First, work you could not get anywhere else. What does the work is not quality but exclusivity. If the same game runs on both, nobody buys a second box.

Second, a use the competitor structurally could not follow. Portability. You cannot match it while remaining a living-room console.

And a third that gets overlooked. Nintendo never once asked anyone to switch. Almost nobody sold a PS4 to buy a Switch. Enormous numbers of households kept both. It won by addition.

3

Not moving is the correct calculation, not ignorance

Moving somewhere new means the cost lands first and the benefit later. And the benefit depends on whether other people also come. Whoever pays first loses.

Everyone runs that calculation, so nobody moves first. This does not happen because people fail to understand. It happens because they do. Explaining better does not change the arithmetic.

What is being compared is not two products. It is "where I am now" against "where I would be, plus the cost of getting there." The current side already contains friends, history, muscle memory, followers. The challenger starts at zero on all of it. Being better is not enough; you have to be better by the size of the move.

And for a social product the substance of that cost is other people, which you cannot pay by yourself. The exit is sized for one person; relationships only move as a group. That is why "you can always leave" does so little work.

4

This argument is being had only by people who crossed

Everyone discussing how to explain it is already on this side. They are the people who did not need the explanation, or who got past it. The reasons of those who did not cross are not lying around inside a gathering of those who did.

So refining "how do we explain it" from the inside works on a skewed sample. Framing it as an explanation problem already fixes the cause as explanation.

The person writing this crossed too, and cannot escape that. So below, instead of opinion, things that can be counted.

5 — MEASUREMENT

The utilities are missing

In Japan, social media is where train delays, weather warnings, disaster notices and government announcements are found. That is not entertainment. It is infrastructure. Operability does not substitute for it.

The Atmosphere has a mechanism for institutions to identify themselves: your own domain becomes your handle. No verification to buy from an operator, and impersonation is harder. That is a concrete benefit for an organisation.

So how many Japanese institutions are here?

This section was badly wrong once. The first pass ran 86 major domains through resolveHandle and reported 8 hits, 4 of 16 in news, and zero broadcasters, railways and government. A reader pointed out the error; on recount, the news, broadcast and university figures were wrong.

Three defects. One, only top-level domains were tried. Institutions identify at the department or product level (news.tv-asahi.co.jp, dpri.kyoto-u.ac.jp, photo.bsky.nikkei.com). Two, non-domain handles were not counted. Hokkaido Shimbun, Chunichi Shimbun and Nishinippon Shimbun all operate officially under ordinary handles. Three, the choice of fields was arbitrary: regional papers, trade press, publishers and fact-checking outfits were never in scope.

What follows is the corrected picture.

fieldwhat can be confirmed now
Newsnational, regional, broadcast, wire and trade press: 41
Universities and institutes (.ac.jp)10, identifying at department level
Publishers8
National government (.go.jp)0
Local government (.lg.jp)0
Railway operators0 (as far as could be checked)

The news, .ac.jp, publisher, .go.jp and .lg.jp figures come from lists maintained by @aiueo.ooo (last verified 20 August 2026). Activity was checked here, account by account.

News is not empty. Asahi Shimbun, Nikkei, Sankei, TV Asahi News, Hokkaido Shimbun, Chunichi Shimbun, Nishinippon Shimbun, Okinawa Times, 47NEWS, The Japan Times, HuffPost Japan, WIRED Japan, the Japan Fact-check Center. Regional papers and trade press are thick on the ground. Universities are here too, at department level: Kyoto University's Disaster Prevention Research Institute, the University of Tokyo's Information Technology Center, Kyoto's Center for Southeast Asian Studies.

What is empty is government and transport. The independently maintained .go.jp and .lg.jp lists both stand at zero. No meteorological agency, no municipality, no railway. The one domain where social media functions most like infrastructure in Japan is the one that is missing.

What remains is the difference in scale

News is present in number. The scale is two orders out.

The New York Times1,344,559
NPR1,008,223
The Guardian789,554
Reuters347,443
The Japan Times53,495
Nikkei20,909
Asahi Shimbun8,857
TV Asahi News660

Follower counts, 3 September 2026. Blue is Western, clay is Japanese.

This is not "the West is all here" either. Neither NASA nor CERN nor the USGS identifies by domain. What is present is news, and some research bodies (ESA, NOAA, Nature, Science).

1,344,559The New York Times
Asahi Shimbun
Kyoto Univ. Disaster Prevention
government · municipal · rail

6 — MEASUREMENT

Unofficial bots are filling the gap

Search the names of the institutions that are not here and something else comes back.

NHK — "NHK News (bot)", "NHK News RSS relay (unofficial)"

Japan Meteorological Agency — "JMA disaster weather information Bot (unofficial)"

Yomiuri Shimbun — "Yomiuri Shimbun (unofficial bot)"

JR East, Tokyo Metro, Tokyo Metropolitan Government, the Prime Minister's Office — nothing

Where the official account does not come, volunteers are re-broadcasting the RSS. The infrastructure is not absent so much as substituted.

Which is ironic given the design. There is a mechanism for proving your own name, and the institutions that most need proving are the ones not using it. The space they left is being filled by people who cannot prove anything.

7

Still, one thing can only be built here

Put briefly, what people want from a social network is this. A place where you learn what you need to know and do not feel bad.

Those two do not normally stand together. Raise coverage and the unpleasant comes in with it. Raise filtering and the wanted goes out with it. Lift one and the other drops.

While one company does both, there is no escaping that trade. And because revenue rides on volume of reaction, feeling bad does not land on the loss side of the ledger. It is structurally unfixable.

The Atmosphere can split it in two. The relay carries everything. Judgement is a separate layer, and the reader picks it. Put coverage and filtering in different places and both can rise at once.

And this cannot be copied. Copying it destroys the copier's revenue. The same shape as a console maker unable to follow a handheld without ceasing to be a console maker.

One account opening other companies' apps is the same case. X can clone any individual app, but it cannot open its own accounts to other companies. The moment it does, it stops being X.

Neither asset adds a line to a feature list, so neither is visible in a feature comparison. They are, so far, the only two exclusive assets here.

8

So what should be built

Turn all of the above over and the work falls out.

Do not ask anyone to switch. Asking hands them an arithmetic they cannot win. Not "come over" but "the account you already have opens these too." Arrange it so no cost lands first.

Explain the benefit of identifying to government and transport. News and universities are already here; what is empty is ministries, municipalities and railways. The domain becomes the name, verification does not have to be bought, impersonation drops. And nothing has to be abandoned. Post to both. For an organisation that issues disaster information, impersonation defence is the core of the job. I have never seen this written out in Japanese.

Grow the judgement layer. Japanese labelers exist. But they are mostly playful ones, fortunes and badges, and the layer that takes on "what gets hidden" has not grown. When something goes wrong, the complaint travels to the operating company rather than into building an alternative. A design that allows substitution, with the default left as the only implementation.

None of the three is closed by a choice of vocabulary. Which is to say: build them and they close.

METHOD

How this was checked, and its limits

Handle existence was decided by passing a domain to com.atproto.identity.resolveHandle and seeing whether a DID came back. Activity and follower counts come from app.bsky.actor.getProfile and app.bsky.feed.getAuthorFeed. Anyone can reproduce it.

The first 86 were mine to choose, and were not a survey. And the choosing did in fact distort the result: only top-level domains were tried, non-domain handles were not counted, and the fields were skewed. The corrected set is not a survey either. The only claim that is now certain is that .go.jp and .lg.jp are at zero.

Name searches only looked at the top few results. Official accounts may have been missed. If you find one, say so. Kyoto University's Disaster Prevention Research Institute, and the 41 news accounts, both entered this table afterwards. Without the correction it would still be wrong. This table can still move.

Figures are as of 3 September 2026. The Atmosphere moves fast. In six months these will be different numbers.