Well, a lot of people this morning told us about their research. I can tell you a little bit about my two. This is a star cluster that's about 3,000 light years away. My research is all about star clusters. These stars here formed like a million years ago. This image was taken recently by the European Southern Observatory at one of their telescopes in Chile. Here you see these kind of stars forming and interacting with the interstellar medium and all kinds of stuff. And I think that that answers why astronomy exists, and it's because we are the outreach of department of science.
We can get people so interested and so engaged with science and technology. Most people who see, say, I don't know, pages of Hubble, learn about Astro in school or something like that, they're not going to do astro as a career because we're such a small field, but we can get people engaged with science. And as you can imagine, we need kind of effective ways to be able to do this kind of outreach. And some of these previous platforms, such as Twitter, which have now been destroyed by billionaires, it's quite frustrating that that's a process that can happen for us as such a kind of very strongly outreach-driven science.
It really screws us over that we miss out on these spaces when they can just get destroyed effectively. But as well as that, it's also not about outreach. Twitter was a really, really nice space for networking that would be really nice to have back and be able to use again. And it's really a shame that we're we're kind of missing out on this. I know through my own PhD, it was so useful to have what we called astro Twitter, and that that doesn't exist anymore. There's so much kind of collective community knowledge and just this huge collective space that we've lost effectively.
So as it turns out, the ad protocol is a fantastic, fantastic tool for outreach. Something that I and our project has been involved with for the last three years. So in July of 2023, I started out by making an astronomy feed. It was partly to procrastinate from preparing for my PhD thesis defense. But basically the way that it works is that it collects posts from astronomers, astronomy organizations, and astrophotographers. Effectively, that you have to do a little bit of a sign-up step just to verify that you're who you say you are, but then you get to post to this feed.
Over time, this has developed into what we call an ecosystem of feeds. We have 15 different feeds that cover all kinds of different topics in astronomy and astrophotography, and people can kind of self-select where they'd like their posts to go by using say emojis and hashtags and other things. And as well as that, I think as a scientist, this graph is really boring because it's really flat, but it's really cool that it's boring. In April of last year, one of our developers added some logging to our servers and actually the stats are kind of boring. Like our views have been stable over time.
The amount of scrolls that we get on the feeds has been relatively stable over time. But I think boring is good in this in this case because we consistently have about 15,000 unique people who scroll the feed every week. So that's not just that their app is kind of preloading the feed or something because the view counts aren't necessarily so reliable, though it's really cool. We get three quarters of a million of a view every week. But there's 15,000 people who are really actively engaging with this and scrolling through at least tw 20 posts each week.
And that's really cool. That's quite a lot of people and quite a big audience, especially when for you as an astronomer or astrophotographer, all you have to do is the little sign-up process to be able to post there, and then you suddenly get to be able to reach this large audience who are going to be interested and engaged in the content that you want to post about. Another thing that's really exciting here too is that we've had some relatively good institutional buy-in from a lot of past organizations across both say the EU and in the US.
There have been various orgs that have joined the feeds. Um really big names here, for instance like the European Space Agency or the Space Telescope Science Institute, which probably none of you know, but they run things like Hubble and the Roman telescope. They're super super important organization in the US. ESO, who I showed you a gorgeous picture from, and also the Rubin Observatory, which started taking data like a month ago, which is super exciting. They're also really active there. And that's really nice as well, then, because you have not only astrophotographers and astronomers, but also these like really large astronomy organizations all in the kind of same place and in one community together.
And I think often as well, where these larger orgs go, people might follow too. So it's it's really really nice that these organizations have been interested in joining Bluesky, that they found our feeds and that they got involved and that they post there. We're a bit of a tech conference, so I will mention briefly as well what we're doing in terms of tech stack things. Most of our backend is actually running in Python, which I don't think is very common for most projects. The choice with that is because Python is the most common language in astronomy and it makes it easier for us to kind of interact with the astronomy community.
There's a really nice SDK built by Marshall X that also has some bits of Rust code as well, so it can do like really fast DAG CBO decoding and other stuff. So a lot of the technical stuff works really nicely. Also feel free to ask me about things like minimizing hosting costs. We now use hosting in the EU. We got really into hosting on rented VPSs and getting our costs down as much as possible and being really really efficient about it. Feel free to ask me more about that if you're curious. And finally, I think one of the most exciting things as well about these early stages of the project is it isn't not just me.
For over a year now, we've had multiple developers and a moderation team working on keeping the feeds going between doing things like working on new features and making sure that we kind of moderate signups and moderate content on the feeds. We also, in kind of September of 2025, we started an open collective page, and that happily covers our hosting costs. It covers also some stickers that I'm going to give away later and a little bit of my travel here too, which is fantastic. And because of that, now we're really becoming a community of people who are trying to work on this and trying to push AT Protocol as a thing that can have purpose for our community of science.
So in the kind of second half of this talk, I will talk a little bit more about how we're trying to build an ecosystem of tools and move beyond just having feeds. So, first off, an obvious low-hanging fruit for us is PDS hosting. We have a PDS running that can host people's data. The main problem is getting legal situation stuff sorted out. I want to do it properly from the start because if we don't do it properly at the start, we'll probably never get around to it. But we're hoping to start a charity probably in the UK, which would then become the legal host for what we're doing.
And that would then mean that we can host the data of astronomers and the astronomy community online. I think this is actually something that is maybe a bit of a missing niche in the PDS hosting world as well. I think people talk a lot about self-hosting, and then we also hear a lot about these amazing big projects like BlackSky and Euros Sky that are giving us viable alternatives to Bluesky as a place to host your data in the Atmosphere. But I think there's also a missing middle as well for kind of distinct communities who want to host data for their community.
I expect we wouldn't be larger than maybe a thousand, two thousand people realistically. But because we're kind of a community, I would probably already personally know some of the people who are signing up, just someone who works in Astro. It's yeah, it's it's it's a kind of nice missing missing middle, I think, and I think it's also great for the health of the Atmosphere ecosystem too, if we can have a range of PDS hosts and not just have kind of go from one to say three big hosts. I think we want to have a whole spectrum of data hosting options.
And as well as that, if we have our own PDS host, then we can define what we want to do with data on that. And if we have crazy data ideas, then we can do those. So a couple of weeks ago, we were chatting in a meeting and had a cool idea about doing live real-time data streaming. The reason why we care about this in astronomy is because when a star explodes with something like a supernova, it looks a little bit like this here, the Crab Nebula, but a thousand years later. But basically, when a star explodes, you want every available telescope around the world to point at this place and take data and measure things like the afterglow of this event and measure other things about it, like how far away it is with this redshift, all kinds of other stuff.
And you want to do that within 30 to 60 seconds of this transient event happening, or preferably just completely as fast as possible. There are only certain kinds of detector that will find these things happening, but then there's a lot of other kinds of telescopes, all with different specialties that you would want to know about this and to follow it up right away. There are solutions to do this already, but they're either they have some kind of mix of either being older and less secure or being very highly centralized. In terms of the highly centralized solutions at the moment, a lot of different observatories host their own Kafka streams.
And what that basically means is an observatory, if you want to put out data, you need to do things like handle authentication for who can subscribe to your Kafka stream, you need to set that up. You need to obviously run the thing, decide on your schema, continuously run this thing and hope and kind of support having many, many different web sockets subscribing to that and have as much uptime as possible so that people don't have interruptions. If you're a consumer then of these Kafka streams, there are a lot of different organizations around the world that are producing transient events and things that you would want to subscribe to.
So you then have a lot of different streams that you would have to subscribe up to, and you'd have to run lots of different consumers. You have to somehow aggregate all of that data. And that can get quite complicated. Instead with atproto, you can do each one of those things with one line of Python code with a new Python package that I've done as a hack project in the last week. Basically, it's called Nebra. It's already on PyPy. So if you want to at the end of this talk, you can try installing it on your computer and try streaming some astronomy transients for yourself.
If you are a producer, all that you need is an account on a PDS, uh set a couple of environment variables for your login, and then you just do Nebra.send that uploads the record to your PDS using the atproto SDK behind the scenes, but I've tried to make it as sort of simple as possible, with the target audience being astronomy observatories and astronomy organizations who do not know anything about atproto and they just want a really, really simple Python module that does what they need for them. Likewise, if you're a consumer, there's a Python implementation of a Jetstream client, and all you have to do is Nebra.stream and then just pick which collections or handles and different organizations you would like.
And what this means now is that a lot of complicated things like data authentication and real-time data streaming, even validating schemas, they all get handled by the AT Protocol because we're all here at this amazing conference about this amazing piece of software that's been built over the last few years and has this amazing dev community. All of the hard stuff works. All you really need is a Python client just to post and receive, which is super simple. I can show you as well. For the last day or so, I've been running this and reposting things from NASA's general coordinates network, which has various different things from different NASA-based missions.
This is getting reposted now to a account called transient crossposted.astrous guy.eco. This is just on PDS LS now. We can go look at some of these different records. I know there's gravitational waves here. They're not actually real, they're testing their pipeline before the big gravitational wave thing starts happening. But like we can see records like that. If I can do the back button here, let's find it. Where was one? There's these are all the gravitational wave ones. I think, yeah, this is a Juicy and circular where people can write things and stuff. And this is all basically existing data that's now getting streamed on the Ad Proto, and anyone can go find that and go look at that.
And what's best as well is that this Python module is obviously open source. You can install it with Python, and anyone could start working with any data that they'd like and streaming that in any way that they'd like. For instance, this is originally designed for astronomy, but it can ingest or produce any type of record that you'd like. For instance, within the last week, a map a record called MataDisco was recently released. I think Robin might talk about it a little bit in his talk later this week. It's a very simple schema for posting metadata of science data.
And if you want to, you can stream that too with this one line of Python code, which is really nice and easy. And the idea really is to make this a super simple solution that people can just go and use if they care about real-time data streaming, and they don't even have to care too much about that it's AT Protocol, they don't have to know about what Jetstream is or what a PDS is to get started. So, yeah, thanks very much. I will leave you with our conclusions and also I have stickers with this design on them, which I will put on the sticker table in a little bit.
So check those out if you'd like a bit of free much. Thanks very much. Okay, we are taking questions for Emily, if you have any. This time we can start here with Matt. I bring the mic here. This is this is very exciting and wonderful to hear. So this last part about the having scientists on on atproto. What kind of appetite have you seen so far for people sharing like measurements or like research observations that would otherwise or eventually go on on the archive or other types of publications? So I think that there's been pretty much no appetite for it so far.
I don't think I've seen it be very well explored. I think Mata Disco is the first time I've really seen someone do it as scale, and I'm not familiar with that many other groups doing it. I think it's it's the kind of thing where it takes some sort of amount of knowledge to be able to bridge the knowledge gap and know, okay, atproto, it's a thing people might have heard of it if they've been on Bluesky, but it's another thing entirely to understand the primitives and know okay, this is something that can work in real time.
We can self-host the data with the PDS, we can rely on the relays and this bits of software and stuff. So I think it requires people to try and bridge the gap and sort of um make it really, really easy for scientists. Because I actually don't think there's much investment at all in doing real-time data streaming and stuff. And also, as well, you mentioned this being on the archive. Most of this data doesn't end up on the archive, but papers will be written about it. So it's somewhere in between where it's very small bits of data with say like a potential location of something, and it's it's up to people to deal with that how they want to.
So more questions. All right, so I admit to it being kind of post-launch. Brain's not fully there. Um is the intent to have conversations about the science here? Is the intent to actually be hosting the science here? Is it both? I'm curious kind of what you're trying to do. So I guess in that sense, I've kind of shown stuff that's doing all of the above because the feeds are a space for people to have conversations about astronomy and other things. Um and then this newer project with say Nebra, for instance, or our PDS hosting, are more about hosting either bits of scientific data or hosting people's social data if they're in the astronomy community.
So kind of a whole different range of little applications of atproto, specifically for the for astronomers. Thank you. Matthew? Yeah. That's really inspiring work. I love it. Um just a couple of things. Uh how many institutions do you think uh are open to the idea of setting up a PDS? Uh is that is that a hard argument to make, or did you find it really easy? I've been told to expect a lot of resistance. I I could see there being some amount of resistance, but I'm honestly not sure. I think that the challenge is selling what a PDS is and making it just really making it clear why it's so important.
Because I think that universities, we talked about this the other day, how I I feel like they have it takes a lot of, they have a lot of mass. It's hard to get them to move, it's hard to get them to have momentum. Um but equally, a lot of universities in Europe, especially in the German-speaking world, say my university I did my PhD on in Germany and my one now in Austria, they are all not on X anymore, and they are all exclusively on Bluesky and other non-fascist platforms. Um I think that they have within certain circles at these universities, they care.
I think the challenge really is explaining what a PDS is to these different groups and trying to get the buy-in and make and really kind of distilling it away from being very technical language. Because I kind of show some parts of our project to people and they're like, oh yeah, okay, that's nice. I don't think they get it. It's the challenge really is selling selling the thing. So a quick follow-up before I pass on the mic. So I will, I promise. Um I also thinking, I mean, obviously, this might not be a huge issue in astronomy, but uh the ability to be able to live stream your data like this could really open up citizen science in a major way.
Have you ever considered that? I mean, obviously not for people who are hunting for supernova, but I mean, there's a lot of other areas of science where citizen science has a little role to play. And it might it might sort of lower the barrier to entry for people to share their data in this way. That is such a cool idea. Imagine if someone made an app where you can things like Merlin, if you know of the bird identification app, imagine if Merlin posted every bird that it finds on the atproto, and every time someone sees it, it gets put on a PDS.
And then you kind of have this as if you're an ornithologist, for instance, you have this massive citizen scientist data set of people identifying birds. It doesn't just have to be Merlin, other people could reuse the schema with their own app or their own anything. Really the possibilities are there. Again, I think the hard part is the bridge from we're all technical nerds, we're probably the 0.91% of people who know about app proto on the planet. We have to sell it to the 99.999%. But that alone is such a cool idea. Hadn't thought of it, but yeah, why not?
Are there added questions? You follow up? One follow up and then real quick. In terms of getting uh universities on board, I'm sure you've tried this, but in the states, we've got all of these um collaborative um RFPs that are coming out of the National Science Foundation. I'm actually really surprised by the language and some of these. Um I'm wondering if you might be uh really successful if you try to find groups that you can partner with um as a collaborator on some of these NSF or European proposals where you would be one of the open science tools that they're working to bring their data onto.
I think that's a really good idea. Yeah, I think it would be really nice to try and find some of these things. I'm I'm not aware of enough stuff. Unfortunately, I'm a postdoc who has to also try and do my science as well and be able to be able to have a job at the end of the day, which kind of limits my my time. I'm hoping to meet people at this conference too, have ideas and thoughts. I've already had a couple, so but I think it's a good idea as well. It would be really nice to get funding for this dev.
There's only so much we can do as ourselves as a dev team, as volunteers who only have varying levels of time each month. I would love it if we could have funding and hire a full time software developer, for instance. But it's just working out where to apply doing it. But but yeah. Really cool. So um let's do another round of applause for Emily for this nice talk and for setting up this wonderful community. It's really amazing. And our next talk will be on uh computational education commons on the Atmosphere by Tim Brooks.