So, I really don’t like the scientific publishing ecosystem. I used to find the idea deeply inspiring and faintly aspirational, and then I had to actually interact with it, and came away with some confused blend of angry and profoundly disappointed.
I think we can do better, and the good news is that a lot of the heavy lifting has already been done: Octopus reimagines scientific publishing in a way that’s modular and open. ATProto lets you build social networks on top of the open protocol that powers Bluesky. And I kind of think that if we bring those two ideas together, we might just be able to fix everything.
So last week, I got far too excited after an ATProto meetup and built a thing: Octosphere, a web app that bridges your Octopus publications to the ATProtocol “Atmosphere,” turning your research into distributed, permanent, social records.
It’s a bit of an experiment, and I’m not sure how useful it actually is, but I think it might help share a vision of science the way I’d like to see it: collaborative, distributed, resilient, and approachable — a big digital public square where we’re all invited, where new ideas are always welcome, but robustly challenged. (It was built with a lot of vibes and will absolutely fall over quite soon, but I hope it gets you thinking about how we can do science better.)
So what’s wrong with science now?
The idea of science is deeply compelling to me — everywhere, we should all be thinking and sharing new ideas (or hypotheses), which everyone should be able to test, and share their reactions — creating a global feedback loop of idea, theory, test, and findings, that constantly generates new, evaluated knowledge. Peer review is continuous, built in, and welcome.
That’s the idea. The reality is miles away, and I put the blame firmly at the door of one specific culprit: the goddamn paper. I hate the academic paper format. Most papers are long, static, stuffy exercises in self-indulgence, written fully in the knowledge of their own intellectual snobbishness, and my god we should do better. (This varies by field—medicine is leaner, machine learning has largely moved to open preprints—but the core dysfunction remains.)
Once upon a time, papers made sense. If I was a physician in 1803, sharing my ideas with a colleague contemplating some grand new theory, of course I’d write a long paper — we may very well not have read the same books or have the same base knowledge, and frankly, it’s going to take a while for my damn letter to reach them anyway. Once I’ve finished that paper, it’s published, and we can all move on.
But it still absolutely doesn’t fix the core problems, and we can do so much better!
- A paper is static. You finish it, you file it away, and it’s done. Science is and should be iterative — you work, you develop, you listen to feedback and change. But a paper is a one-time thing, a PDF you finish and file and move on.
- A paper goes in a journal: it’s written for other academics, by them, with bespoke structures and guidelines. But nobody reads an actual journal these days (if they ever did).
- We can collaborate so much better now! Do you want to post a reply or critique to an academic paper? Well, you’d better write another paper! My god people, scientists started sharing ideas on Usenet like… 5 decades ago now? We’ve seen whole eras of social media live, grow up and die. We can do better than this.
And I’m not even going to get started on the ridiculous scam that is publishing costs — other people have criticised that at length, and done far better than I will.
Octopus is a web publishing platform funded by the British Government that aimed to fix a lot of this when it launched. Anybody can publish, and you don’t need to share a whole paper: publications are broken down into the components of the scientific method — from hypothesis to analysis to results and beyond. Anybody can review them too: if you disagree with a publication, just post a peer review directly on Octopus.
Octopus is great. But it’s also centralised, and doesn’t feel hugely approachable. Science should be (say it quietly) a shared social network. If only we had a tool to build something like that?
So how does ATProto help, and why Octosphere?
ATProto is the decentralised social media protocol behind Bluesky, and helpfully, is designed to address pretty much this problem — if you want to build a social network for books, or movies, or events, or, you know, research, that is decentralised and built collectively rather than relying on infrastructure owned by a single rich man, ATProto will help you do it.
I’m very far from an ATProto expert, so I’m not going to go too deep into it, but in essence: each user has a PDS (Personal Data Server), which hosts their identity and their content. For most people, that’s Bluesky, but you can just as well host your own PDS. The magic is that once you’ve got a PDS, it becomes part of the network — your content is shared across a network of relays, so everyone can see everything, without relying on any one person.
I really like ActivityPub, but it doesn’t give you as much out of the box—you’d need to handle authentication, data storage, and more yourself. You could absolutely build this on the fediverse, but for a quick “spin up a social network and see what happens” experiment, ATProto just worked.
And the really cool thing? That data can take any shape at all. In the ATProto world, the shape of your data is defined by a Lexicon — think of it as a schema that describes what a piece of content looks like, what fields it has, and how other applications should interact with it. It acts as data structure, API schema, and validation all at once. If you want to store a blog post, you define a Lexicon for blog posts. If you want to store a scientific publication, you define one for that. You can read my lexicon for Octopus publications here.
So the cool thing about building on ATProto is how little you need to worry about: you define a lexicon, and help users post your content to the Atmosphere… and that’s (mostly) it. ATProto acts as backend, CRUD, and authentication, and you just need to visualise the results.
And that’s exactly what Octosphere does. You sign in with your ORCID — the universal researcher identifier — which lets Octosphere find your Octopus publications. It then creates ATProto records for each of them, storing the metadata (title, abstract, type, and link back to the original) as structured data on the ATProtocol network. Optionally, it can cross-post them to your Bluesky feed, so your followers see your work alongside everything else in their timeline.
So what’s cool about this?
A few things, actually.
First, your publications become eternal and distributed. They’re not locked behind a paywall, not sitting on a single server, and not dependent on any one organisation keeping the lights on. If Octopus goes down tomorrow, the ATProto records survive across the network’s relays. That’s a meaningful resilience improvement over the status quo, where decades of research can vanish if a publisher changes its business model or a university cancels a subscription.
Second, it puts science where people increasingly are. A growing number of researchers have moved to Bluesky, especially as other platforms have become less hospitable. Cross-posting publications to that feed means your work reaches people in spaces they’re already checking daily, rather than sitting in a database that only gets visited for very specific searches. It lowers the barrier between “doing science” and “sharing science” to almost nothing.
Third (and this is the cool bit) it’s a proof of concept for something much bigger. Octosphere is just one app reading one type of record. But because it’s built on ATProto, anyone else can build on top of the same data. Someone could build a better visualisation layer, or a recommendation engine that surfaces related work across disciplines, or a peer review tool where critiques are threaded directly to the publication record they’re responding to — persistent, attributable, and impossible to quietly ignore. The data is open, the protocol is open, and the ecosystem can grow without anyone’s permission.
I’m not sure whether this is the future of science. But I love the idea of research as something decentralised, distributed, and genuinely collaborative (which, honestly, is what science was always supposed to be). The tools to make it real are here. Octosphere is a small experiment in that direction: a bit rough, probably temporary, but hopefully enough to get you thinking about how we could do this better.
I hope you give it a shot!