Having spent the better part of two years bouncing between various social protocols, I thought I’d share my thoughts on the big three contenders for the future of social media: AT Protocol, ActivityPub, and Nostr. Spoiler alert: I have opinions, and they’re probably not what the crypto evangelists want to hear.

AT Protocol: The One That Actually Gets It

Let me be clear from the start: AT Protocol is my preference, and for good reason. The Personal Data Server (PDS) model is genuinely brilliant—it’s what convinced me that decentralised social media could actually work without being an absolute nightmare to manage.

User-Centric Design That Actually Works

The beauty of AT Protocol lies in its user-centric approach. When I run my own PDS, I’m not just hosting a server; I’m hosting my data. If Bluesky decides to go full Twitter circa 2022, I can pack up my entire digital identity and move elsewhere. My followers, my posts, my connections—all of it comes with me because it’s mine.

The protocol’s modular design is refreshing too. Unlike the monolithic approach of traditional platforms, AT Protocol separates concerns cleanly. Want a different algorithm? Use a different AppView. Want better moderation? Choose a different labelling service. It’s like having a social media platform where you can actually swap out the bits you don’t like without rebuilding everything from scratch.

Developer Experience That Doesn’t Make You Cry

I’ve spent considerable time working with AT Protocol’s APIs, and they’re genuinely pleasant to use. The documentation is clear, the responses are consistent, and the authentication flow doesn’t make you want to throw your laptop out the window. When I built my moon phase bot for Bluesky, the most challenging part was getting the UTF-8 encoding right for hashtags—hardly a protocol-level issue.

ActivityPub: The One I Grew Tired Of

ActivityPub held promise, and for a while, I genuinely enjoyed my time on Mastodon. The federation model works, communities can be vibrant, and there’s something satisfying about the distributed nature of it all. But here’s the thing: the server-centric model is fundamentally flawed for user experience.

The Ghost Server Problem

I’ve lost count of how many Mastodon accounts I created over the years. Fifteen? Maybe more? Each time a server would shut down unexpectedly, or the admin would go rogue, or there’d be some federation drama that would effectively strand my account. The worst part isn’t losing the account—it’s the digital ghosts that linger.

These abandoned accounts float around the fediverse like digital debris. Sometimes people would reply to old posts from servers that no longer exist, or I’d stumble across my own defunct profile on some instance’s user directory. It’s unsettling, this permanent record of temporary presences.

Community Toxicity and Tribalism

The Mastodon community itself became increasingly hostile over time, at least in my experience. What started as a refreshing alternative to Twitter’s toxicity gradually developed its own particular flavour of gatekeeping and performative outrage. The constant discourse about federation politics, instance blocking, and who’s allowed to use what software became exhausting. Social media shouldn’t require a PhD in network topology and community management theory.

The handle system made this worse by emphasising tribalism. Your @username@instance.social wasn’t just an address—it was a badge of allegiance. People would judge you based on your server choice before they’d even read your posts. Are you on a “good” instance or a “problematic” one? The technology that was meant to promote diversity ended up creating digital class systems.

This tribalism extends to the broader ecosystem politics as well. The hostility towards Meta’s integration with Threads was particularly telling—whilst I somewhat agree with concerns about corporate influence, the reaction felt disproportionately apocalyptic. Meanwhile, alternative ActivityPub software like Pleroma, Akkoma, or Misskey gets largely shafted by Mastodon’s dominance, with some users from these platforms sarcastically referring to Mastodon users as “Mastocels” in response to the gatekeeping. It’s a mess of competing fiefdoms rather than a cohesive protocol ecosystem.

The Dormant Account

Don’t get me wrong—ActivityPub works as a protocol. But the user experience of constantly wondering if your server will exist next month isn’t sustainable for most people. The technical concept is sound; the practical implementation leaves users at the mercy of volunteer admins and their life circumstances.

I still maintain an account on Mastodon (@ewanc26@mas.to), but it’s largely dormant these days. It sits there like a digital time capsule, a reminder of what could have been if the ecosystem had prioritised user experience over ideological purity.

Nostr: The Alleyway Lurker

And then there’s Nostr. Oh, Nostr.

If AT Protocol is the friendly neighbourhood tech enthusiast and ActivityPub is the well-meaning but chaotic community organiser, then Nostr is absolutely that one person lurking at the mouth of an alley, trying to convince you that this cryptocurrency is different from all the others.

Bitcoin But Social (Unfortunately)

The “bitcoin but social” aspect is precisely what puts me off. I hate cryptocurrency—not just the environmental impact or the speculation, but the entire mindset that everything needs to be financialised. Nostr takes this philosophy and makes it central to the social experience. Want better relay performance? Pay for it. Want persistent data? Pay for it. Want people to see your posts? Well, you get the idea.

The protocol feels like it was designed by people who think the primary problem with social media is that it’s not enough like a financial market. The obsession with Lightning Network integration and zaps and micropayments turns every social interaction into a potential transaction. It’s exhausting.

Evangelical Cryptocurrency Culture

There’s also something deeply weird about Nostr’s community culture. It has this evangelical quality where proponents seem genuinely convinced they’re building the future of human communication, when in practice they’ve recreated Twitter but with more steps and mandatory cryptocurrency enthusiasm. The whole thing feels less like social networking and more like a recruitment scheme for bitcoin maximalists.

I understand the appeal of censorship resistance, truly I do. But Nostr’s approach feels like using a sledgehammer to hang a picture frame. The complexity-to-benefit ratio is absurd for most social media use cases.

The PDS Difference

Coming back to why I prefer AT Protocol: the PDS model solves the core problem that both ActivityPub and Nostr struggle with in different ways. With ActivityPub, your data is hostage to your server admin’s commitment and circumstances. With Nostr, your data is scattered across relays that may or may not keep it depending on their economic incentives.

AT Protocol’s PDS approach means I control my data destiny. If I run my own PDS, I know my posts will persist as long as I want them to. If I choose a hosted PDS, I can migrate if the service quality declines. The protocol treats migration as a first-class feature rather than an emergency procedure.

It’s the difference between renting a flat where the landlord might sell the building without notice (ActivityPub), buying property that requires constant maintenance fees (Nostr), and having a proper lease with reasonable migration rights (AT Protocol).

Looking Forward

None of these protocols are perfect, and the social media landscape will undoubtedly continue evolving. But after bouncing between various implementations and experiencing the practical realities of each approach, AT Protocol feels like the only one that’s actually designed around user needs rather than ideological commitments.

I want my social media protocol to be boring in the same way that email is boring—reliable, portable, and not constantly demanding I think about its underlying mechanics. AT Protocol comes closest to that ideal.

Maybe I’m wrong. Maybe in five years we’ll all be zapping each other satoshis while our data floats between relay nodes in some glorious decentralised utopia. But I doubt it. More likely, we’ll still be debating the merits of different approaches while actual users gravitate toward whatever works most reliably.

For now, that’s AT Protocol and Bluesky. The moon bot posts on schedule, my followers come with me if I migrate, and nobody’s trying to sell me on the revolutionary potential of micropayments for social interactions.

That’s good enough for me.