This is the second post in a series explaining the AT Protocol to anyone.

In my last post, Social Media is Broken. Here’s How to Fix It, I laid out the philosophy behind the AT Protocol. That piece focused on incentives: how today’s social platforms are structured in a way that almost guarantees enshittification, and how this architecture changes that by making exit real. When users can leave without losing their identity, their audience, or their history, platforms lose the ability to quietly tighten the screws forever.

This post is about what that shift enables downstream. Not in abstract terms, but in user-experience. What kinds of apps become possible? What kinds of interactions start to feel different? What does social media look like when anti-enshittification is proactively built-in?

If you want a deeper technical mental model,

@danabra.mov

's most recent blogpost, A Social Filesystem, is a great companion. His framing is that your online identity is more like an “everything folder” - your posts, likes, follows, and media live with you - and apps are simply different ways of viewing and interacting with that shared data. I’ll reference that idea lightly, but this post is really about something you can already feel emerging from it. We are beginning to call this AT Protocol ecosystem the Atmosphere.

The Atmosphere.

The Atmosphere is the ecosystem of apps that operate on the same shared substrate - a shared foundation and a shared social contract about how social data is created, stored, and exchanged. Apps don’t need permission from a central platform to exist. They just need to agree on how data should work so users can move freely between them.

Abramov describes it literally as all servers that speak the same social language together forming “the atmosphere.” You don’t need to think in servers to understand it, though. At the experience level, the Atmosphere is what happens when social apps stop being isolated databases and start behaving more like the web again: many places, many interfaces, one connected space.

For most of the last decade, the internet has been collapsing inward. What used to be millions of websites has collapsed into a handful of feeds. Social media took the open web - links, identity, publishing - and pulled it behind login screens and proprietary graphs. The Atmosphere reverses that collapse. It brings back openness without asking you to give up the things we actually like about modern social media: discovery, aggregation, interaction, and momentum.

At the user level, the promise is simple: you can sign into many apps with one account, and your identity is still you everywhere you go. This is why the Atmosphere isn’t an “everything app.” It’s an everything ecosystem. There’s a lot of hype right now around everything apps - one app, controlled by one organization, that tries to do everything for everyone. The Atmosphere is the opposite. As Dan puts it, “An everything app tries to do everything. An everything ecosystem lets everything get done.”

In an everything app, one organization holds the power. In the Atmosphere, power is distributed. Different teams, communities, and organizations build different parts of the social experience, often closer to the problems they’re solving and with incentives that don’t collapse into “maximize engagement at all costs.” Instead of one company deciding what social media should be, you get many overlapping answers.

This dynamic is often described as coopetition: apps compete at the product layer - design, UX, moderation, tone - while cooperating at the infrastructure layer. They share identity, shared data formats, and shared social primitives. In closed platforms, cooperation is a threat. In the Atmosphere, cooperation is what allows real competition to exist at all. It's the killer feature.

One Identity, Everywhere.

In the Atmosphere, your handle is a domain. Your account is basically a website. When you sign up for Bluesky, your handle defaults to a subdomain like name.bsky.social. When you sign up through Spark, it defaults to something like name.sprk.so. These are just defaults. If you own a domain, you can use it as your handle - joebasser.com, artistname.xyz, anything you control.

This has two important consequences.

First, identity becomes legible. If someone’s handle is mayor.nyc.gov, you can trust that is actually who they say they are. The domain itself acts as verification, because domains are already part of the open web’s trust system. (There are other ways to verify in the Atmosphere as well but I'll mention them another time).

Second, your identity isn’t fragile. Under the hood, your handle resolves to a Decentralized Identifier (DID). That means even if you change your handle or move where your account is hosted, all links to your account remain intact. Your identity stays the same; the name is just a pointer.

Your account doesn’t live inside Bluesky, Spark, or any other app. It can be hosted by different providers, and you can move later without permission. Apps stop owning you. They become interfaces you choose.

Cross-Pollination

This is where the Atmosphere really starts to feel different in practice. To understand why, it helps to introduce one simple idea: a lexicon.

In the Atmosphere, a lexicon is just a shared agreement about what a piece of social data looks like. Think of it like a file format for social actions. A “post,” a “comment,” a “review,” an “event,” or a “song” all have a defined shape - what information they contain and how they relate to each other.

When apps adopt the same lexicon, they don’t need custom integrations or special partnerships. They already know how to read and display the same thing. That’s what allows content to move between apps without being copied, screenshotted, or flattened into links. Crucially, no app can gatekeep this data - access doesn’t require permission from other apps, which enforces cooperation by design.

What follows are just some examples I’ve been thinking about and enjoying recently. They aren’t exhaustive. They’re early glimpses of the kinds of cross-pollination that become possible when apps share a substrate instead of competing as silos.

Long-form writing.

Leaflet, Offprint, and Pckt are independent writing apps built by different teams, with different design philosophies. What they share is a common long-form writing lexicon. Because they agree on the shape of a “blog post,” a piece written in one app can appear in the others. Writing stops belonging to a single product and starts behaving like part of the social web.

That’s very different from something like Substack, which is closed. Writing published there lives and dies inside Substack’s walls (although they do allow you to export your mail-list which is nice). In the Atmosphere, long-form writing becomes a first-class social object. You can write an essay in Leaflet (where this was posted), share it on Bluesky as an embed for more reach, and have people quote specific passages in short posts. Those quotes stimulate discussion that flows back into the original piece. You can embed Bluesky posts inside the essay itself, mention people by handle, and notify them when their ideas are referenced. Because identity is shared, the conversation doesn’t fragment into screenshots and dead links. Long-form and short-form stop being separate worlds and become different surfaces of the same conversation. This is Atmospheric cross-pollination.

Events.

We’re starting to see the same pattern with events. Smoke Signal is an events app built for the Atmosphere, but what’s especially interesting is that some existing event tools like calendar.city and dandelion.events - not originally built as Atmospheric apps - are now writing shared event records so their events can appear inside Smoke Signal and other apps that adopt the same events lexicon. Multiple calendars, multiple creation tools, and multiple entry points can all feed into the same shared pool of events. You don’t need every app to be full-stack to participate; you just need to speak the shared language.

Pop-Culture.

Popfeed is a shared review network for movies, shows, books, games, and music. A Popfeed review appears on Bluesky as an embed rather than a native post, but the discussion doesn’t fragment. Replies can still sync, and the same conversation can live across Popfeed, Bluesky, and long-form writing apps like Leaflet, where a review might expand into a larger essay. Today, sharing reviews usually means fractured comment sections or static screenshots that kill interaction. The Atmosphere allows reviews to travel without losing their social context. Popfeed is just one review app among many emerging Letterboxd and Goodreads alternatives. By sharing a common review lexicon, reviews can appear and be discussed across multiple interfaces instead of being trapped in a single product.

Video.

Video makes the contrast even sharper. On Spark (An app I'm working on for creators), if a post fits Bluesky’s limits - say a short video or a four-image carousel - you can choose to post on Bluesky as well for more reach.

A great example of cross-pollination today is stream.place: when you go live, your Bluesky profile automatically indicates that you’re streaming and links directly to the broadcast. Spark will support stream.place as well.

Something we’re working on at Spark that fits in here is cross-app curation. A popular post format on Instagram and TikTok is a carousel of screenshots - often funny tweets, political posts etc. It works because it’s curated, but it’s also deeply broken. Screenshots are static. There’s no link to the creator, no way into the thread, no way to participate. In the Atmosphere, Spark can support carousels with embedded Bluesky posts instead of screenshots. Tap one and you’re taken to the original post, with the author, timestamp, and replies intact. Spark becomes a curator that increases interaction on Bluesky rather than capturing it, and Bluesky can drive attention back to Spark as well. No down-ranking links. No punishment for leaving.

Spark stories will also support Atmospheric embeds. Imagine sharing a bluesky post on your story, not as a static screenshot, but a real link that promotes cross-pollination.

Music.

Apps like plyr.fm are exploring what it means for music to exist in the Atmosphere, and Spark already supports sounds, which we’ll be extending to include music. This isn’t about Spark accessing Plyr’s music database. It’s about music living on the shared substrate itself, owned by artists in the same way posts are owned by creators. Songs sit alongside an artist’s other social posts, already accessible. Any music player that adopts the same lexicon can play the same tracks, and Spark can let creators add those songs to their posts because they’re already there, shared.

This starts to look like an artist-owned internet - something I personally care deeply about building. There’s still real work to do around monetization and rights, which I’ll get into in a future post, but the UX foundation is already visible.

And again, these are just examples. Events, sports, news, fandoms, video, music, podcasts, coding, live streaming, research, location check-ins, local discovery - anywhere there’s a social layer, there can be Atmosphere.

A familiar comparison today is Meta’s integration of Instagram, Threads, and Facebook. You can see Threads posts inside Instagram and cross-post Reels to Facebook under the same identity, even though each app still has its own social graph. The difference is that Meta controls everything. The Atmosphere can achieve that same level of seamless integration without the teams building these apps ever needing to coordinate, and it can go much further. Instead of a small, fixed set of Meta-owned products, there can be an open-ended number of apps and experiences working together by default — something no single company could ever build, no matter how large it became. The Atmosphere is intrinsically anti-monopolistic.

Feeds.

Instead of apps being owned by a single organization with incentives aligned around maximizing one company’s profits, the Atmosphere allows incentives to diverge. Apps compete on experience rather than extraction, leading to better outcomes for users. There’s no gatekeeping users or data in the Atmosphere, because the data doesn’t belong to the apps. It belongs to the users. Power stops accumulating in one place and starts distributing outward.

And this logic extends naturally to feeds - the algorithms and views that shape how we experience the Atmosphere day to day. Feeds still matter, but they behave very differently here. In closed platforms, feeds are traps. They’re designed to keep you inside a single app for as long as possible, optimizing relentlessly for engagement and time spent. Links are down-ranked. Leaving is punished.

In the Atmosphere, feeds act more like windows.

Because apps aren’t penalized for linking out, feeds can intentionally surface activity happening elsewhere in the ecosystem. A Bluesky feed might highlight new long-form posts from Leaflet. Another might surface upcoming events published through Smoke Signal, letting you tap through to RSVP without turning your main timeline into an events app. Feeds become a lightweight layer of awareness - a way to notice what’s happening without being forced into a single mode.

This changes how feeds feel experientially. They feel curated, not hunted. You’re choosing a perspective, not surrendering to a mood imposed on you. Switching feeds feels less like changing who you are and more like walking into a different room. You might move between a friends-only feed, an art-focused feed, a slow news feed, a local culture feed, or deeply niche community feeds that feel alive rather than gamed. These already exist on Bluesky today, and it shows. Bluesky has some of the highest click-through rates for publishers precisely because it doesn’t punish links. When something interesting lives elsewhere, you’re allowed - even encouraged - to go there. The experience starts to feel like a web again.

Feeds can also be designed to intentionally cross-pollinate. A feed might surface new Popfeed reviews, live streams happening on stream.place , or new projects on tangled.org, pointing you toward other parts of the Atmosphere without trying to replace them. Discovery becomes additive instead of extractive.

Most importantly, no single algorithm quietly decides culture for billions of people. Anyone can build a feed. Anyone can switch. No feed owns you. If one starts to feel bad, you leave it - without losing your identity, your follows, or your history. That’s what credible exit looks like when applied to algorithms. Feeds compete on taste, care, truth and usefulness - not on how effectively they can keep you stuck.

I mentioned this in my last post, but it is particularly topical right now. Open algorithms are censorship-resistant. Because no single organization determines the logic of what you see, no matter of sucking-up to authoritarian governments can shift what users see. You can even build your own algorithm if you wanted to. Through choice, power is in their hands, not that of their algorithmic overlords.

What is a follow?

This has been one of the most debated topics on Bluesky recently, and it’s worth unpacking carefully. On many web2 platforms these days, follows don’t mean much because algorithms can ignore them. You follow people and still barely see their content. The Atmosphere aims to change that by giving users more control over feeds and clients, so follows can matter again.

One of the Atmosphere’s most powerful features is that apps can reuse follow records from elsewhere in the ecosystem to reduce cold start. A new app can show you a familiar social graph immediately, and users can switch apps without losing their audience or who they follow. But this power cuts both ways. A creator may appear to have many followers inside an app even though many of those followers don’t actually use it and will never see that content in that place.

Which raises the question: do they really follow them there?

A follow should represent intent within a context. You shouldn’t be able to follow someone on an app you don’t use, because then the number is hollow. This is where Erving Goffman’s work is relevant (Thank you

@blaine.bsky.social

for the recommendation). Goffman argued that we perform different selves for different audiences - different masks depending on the social situation. Offline, those contexts are physical. Online, they’re interfaces.

The Atmosphere allows social graphs to reflect that reality. Some apps should share social graphs. Others shouldn’t. What matters is that follows remain meaningful within the experience they’re attached to. Apps can still take hints from each other - for example, Spark will look at your Bluesky follows and suggest people who also use Spark, or auto-follow overlaps, without importing followers who will never engage in that context.

Apps that serve the same audience and intent should share social graphs - if you follow someone in one, your intent likely carries to the other. These apps are often coopetitors, sharing content and audiences while competing on experience. Direct competitors should share graphs so users get real choice. Apps built for musicians should share graphs with each other; fundamentally different experiences - like social coding versus short-form video - shouldn’t. No one has a single social graph.

This is especially important for creators - some of the people we're building Spark for. There should be other creator-focused apps that compete with Spark while sharing the same social graph. That way, if Spark ever degrades or goes off course, creators have a frictionless credible exit without losing their audience. They own their content and their reach, and can move freely between apps that adopt the same lexicon.

Why small apps can matter again.

When identity is shared, data is portable, and lexicons are interoperable, the economics of social change. Apps don’t start from zero. They don’t need to trap users to survive. They can serve small communities and still matter. Ten thousand users can be enough. Niche becomes viable again.

Unique experiences that cater to smaller communities can be sustainable, and the perceived risk and effort of 'creating a new account' somewhere new is gone. The last social account you'll ever need.

If an app dies, your identity doesn’t die with it. Your data still lives with you. Someone else can build a new interface and bring it back to life. That’s the downstream effect of anti-enshittification: the internet stops feeling like a series of hostage situations. Imagine the possibilities for new interfaces and ways to interact online, limited only by the creativity of the Atmosphere’s developer community. It opens the door to custom interfaces tailored to how every individual actually wants to engage online.

Where this is going.

There’s a lot more to explore: monetization and creator economics, private data and paywalls, what it means for apps and feeds to become businesses inside an ecosystem rather than monopolies.

For now, what matters is the shift you can already feel. When apps can't trap you, links aren't punished and identity is portable - design changes. When third-party becomes first-party, innovation becomes combinatorial.

Instead of one global feed optimized for outrage, we get many overlapping cultures, curators, and norms. Culture re-fragments — in a good way. It starts to feel like a web again.

We’re up against some of the biggest names in tech, but this isn’t David versus Goliath. It’s many Davids, working together. Apps cooperating across the Atmosphere to challenge a system that only works when everything is locked behind one giant.

This is the Atmosphere.