I’ve always been more interested in the technology side of social media than the “social” side. Twitter, in particular, was originally mainly a broadcasting technology: it was very cheap for anyone to broadcast a message for wide distribution, and other people chose which broadcasts to listen to (via who they followed). This made Twitter not just a way for small friend groups to update each other, but also a medium for any number of broadcasts. You could (and in some cases still can) subscribe to Twitter alerts for emergencies in your area, news events, transport problems, etc.

But this broadcasting infrastructure appears to be the key ingredient for the phenomena we call “social”: large numbers of small-time “creators”, with some vaguely Zipf-like distribution making a few of them extremely influential. The fact that any receiver of a broadcast could cheaply send a broadcast of their own in reply led to The Discourse. And perhaps most significantly, the broadcasting infrastructure companies realized that users choosing what to subscribe to was not nearly as “engaging” (nor as compatible with an ad-driven revenue model) as using An Algorithm to decide what to show people.

I’ve always wondered what the tech could have become if it hadn’t been entirely consumed by the “social” juggernaut. If instead of centralized algorithms and moderation (which, to be clear, seem to be necessary when the infrastructure is used for The Discourse) clients had instead been built out with more and more robust tools for custom filtering of broadcasts—eg subscribe to traffic alerts, but only for these hours a day on these routes—and more and more information sources had been broadcast via this infrastructure.1

In short, Twitter could have become a universal interface for asynchronous ephemeral notifications of any kind.

Sunrise and Sunset Times

Solar time is fascinating in its own way. It’s not just that days get longer and shorter across the year, but solar noon actually meanders forward and back as well: in most places in the world, there are a few days or weeks of the year during which each day is getting longer than the last, but our clocks show that sunrise is getting later (or sunset is getting earlier) because solar noon is shifting faster than the days are lengthening. (Similarly, sometimes the days are getting shorter but sunrise is getting earlier or sunset later.)

What’s more, solar time depends on your exact location on earth: both latitude and longitude, of course shifted by exactly where you are within your official time zone. People at the east end of a (large) state will see the sun rise when the clock reads an hour earlier than those in the same time zone in the west of the state.

It strikes me that modern people who live by the clock never develop the kind of intuition for solar time that seems like it should be wired into us naturally. This is never more clear than the twice-yearly griping about Daylight Saving Time, which pretty consistently demonstrates a complete ignorance of solar rhythms.

So, Bots

Thank you for reading the above two sections. Or respect for skipping them.

I created a bunch of very simple social media bots that broadcast how sunrise and sunset are shifting every day shortly after sunset. Because this data depends on location, there is a separate bot for each major metropolitan area. Right now I manage bots for 82 locations across North America, the UK, Australia, and new Zealand. All of them are available on Mastodon; BlueSky no longer requires invite codes but they do limit the number of accounts individuals can create, so only a handful are on that service right now. If you live somewhere that’s not in one of these metro regions, then by all means contact me (on either Mastodon or BlueSky) and I’ll consider adding it to the list.2 This page will be updated as new bots are added.

Mastodon bots

North America and US:

United Kingdom:

Australia and New Zealand:

BlueSky bots


  1. Back-end software engineers will of course recognize the extreme end of this product space: logging. Software systems are constantly “broadcasting” messages about their operation into log files that quickly grow to gigabytes of incomprehensible detail. Operators who maintain and analyze those systems have developed extremely advanced filtering and searching tools for finding useful signals amidst this ocean of noise. There is no direct mapping between these tools and what would be required for the very different kinds of data that flow through social media, but it does make clear just how spartan the tools are for social media users managing their feeds. 

  2. Adding a location to the back-end bot implementation is fairly trivial: I just have a big list of locations with their latitudes and longitudes. The hassle is the actual management of the social media accounts (and their credentials), which typically cannot be automated. “Just log in and change this one setting” may be only a one- or two-minute job, but it quickly becomes hours of dull hassle across all the accounts.