The Train Station
First, a short stretch of the mind muscles.
Imagine the AT protocol like a train station. New people come in all the time, jumping off the train onto the Platform, which is where they get their start. This is where they create an account, take in the scenery, and (depending on the Platform) are either calmly led to the concourse by good signage or are hurried along by the anxious crowd. The vast majority of people currently step off the train onto Bluesky. Once you get your bearings you notice that people aren't really using the exits. In fact, you don't see any exits at all. You're still on the Platform.
To your left you see a scrum where someone had apparently called another person's jean jacket "forgettable" in passing. About half of the crowd around them is jeering one way, the other juxtaposed. To your right you see someone who has found a large wooden box to stand on. He seems to be yelling something very obvious that everyone agrees with. People raise their fists in solidarity as they pass anyway. On the concourse outside the Platform, past the glass windows, there are a few construction workers, adding stores, meeting rooms, maybe even a bar. There are screens all around you playing every news channel you've ever heard of. It's hard to hear each TV over all the shouting unless you step up close. You can barely move. Why isn't anyone getting off the Platform??
The Problem
You probably understand where I'm going with this: Bluesky has been good for the AT protocol as a Platform to onboard users, but Bluesky being basically the only big Platform that people know on atproto, with no easy "exits", is actively hurting the protocol.
The Ugly Face of the Protocol
Bluesky is the face of the AT protocol, but it’s an ugly one for most people. It’s not as much a pleasant experience anymore as it is a nostalgic (feels like Twitter circa-2012!) or political one (I’m doing X, but not on X, and that matters!). Bluesky as a platform excels the most at providing users a way to collectively react to real-time events and zeitgeists. That's what bsky.app should be for! The problem is that on a platform like Bluesky you're absolutely inundated with information, often more than you know what to do with. The very next thing that's natural (and what we've been trained, by social media) to do is to earnestly discuss. We're millions of people being shoved into a small space with TVs that never turn off feeding us endless information and nowhere else to go. Of course this isn't pleasant, and of course we're going to want to talk about what we're seeing!
The Medium
This experience on Bluesky has even made some communities give up on it. Chanda Prescod-Weinstein laments about it in this thread:
Prescod-Weinstein talks primarily of a content imbalance whereas I'm primarily talking about a format imbalance, but I think they're related. She makes this connection herself in her thread when talking about how Bluesky doesn't allow videos longer than 100mb, which prevents its use as a true "creative outlet" and keeps people using Instagram for that creative purpose. It's also true that the format Bluesky provides (short, reactive, speedy) is also the format that political discourse thrives in and that cultural conversation often doesn't. There's a good reason for the phrase "the medium is the message", though I think it could be specified further: the limits of the medium will limit the message.
Custom Feeds are a great feature of Bluesky, but over time they have perverted the content on the platform in a couple of ways. First, Custom Feeds have gotten faster. When Custom Feeds were starting out the amount of content on the platform was manageable for a single person to check periodically. But as the daily bandwidth of the firehose exploded the amount of content rushing behind our screens, just waiting to be viewed, is staggering.
These changes in feeds have corresponding effects on content. When feeds get faster users find that their posts have less time and space in which to land. That might mean more posts get created that will get a quick like or easy comment and less posts that take time to sit with. The former is often content that makes us feel momentarily buffeted in some way by the current of culture but does nothing to connect us to others; content that emphasizes or analyzes culture and places us in its context is often the latter.
I believe that the culture of the protocol is being defined by people's experience on Bluesky, that experience is no longer a good one for many people, and their experience is very often tainted by toxic moments of "discussion". Essentially, the problem right now is that there's too many people on Bluesky with nowhere to go to take a second and have a chat.
The Solution(s)
The solution, however, is less obvious. I don't think the best path forward is for Bluesky PBLLC to devote more time to making Bluesky feature-complete, or to start building (or funding) many different kinds of apps that fulfill tons of use cases on the protocol. The developer community is doing a good job of bootstrapping the latter already and it's showing the vast potential of atproto. I think instead, as stewards of the protocol they need to follow the natural progression of the network and release the pressure where it builds up. This means building or funding the development of a preponderance of apps that solve the current largest problem facing the protocol. That problem right now is too many people on Bluesky with nowhere to go to have a relaxed conversation. In a year the most pressing issue might be that the protocol isn't meeting the AI moment well enough, or that it needs governance to be more intrinsic and easier to build programmatically. But right now I think it's clear that we simply need to release the pressure in the network that's making bsky.app an unpleasant place to hang out.
I think that means that the solution is for Bluesky PBLLC to both actively guide and fund the development of discussion apps on atproto, along with eventually linking out to one or many from the Bluesky app to encourage discussion there.
A Forcing Factor for Better Tooling
Above and beyond the inherent problem of the medium, the labeling and moderation systems within the AT protocol are both very sophisticated and utterly unsatisfactory. A concerted push to move earnest discussion off of Bluesky and onto apps that are more designed for it would reduce the overall moderation load for Bluesky PBC and could be an excellent opportunity to mature its labeling/composable moderation architecture goals. Ozone, the labeling and moderation tool for atproto, is right now technically app-agnostic but practically Bluesky-preferent. For just one example, if you try to use it to display a record that's not Bluesky-specific, it shows JSON instead of displaying the record natively. The ideal solution in my mind would be for Ozone to refactor RecordCard.tsx so that it defaults to be record-agnostic, not Bluesky-specific, and you can include your own components that match your specific lexicons in your Ozone fork that you deploy (Bluesky's repo would have Bluesky's types of course, but anyone who forks would be able to easily insert their own).
Labeler adoption has been thin thus far because the tools available and Bluesky's moderation have been "good enough" so far. But discussion platforms would have strong incentives to develop their own labelers with community-specific logic, which would mature and stress-test the labeler architecture in ways that haven't happened yet. It's true that discussion apps might be the hardest single medium to make moderation work for, but that might be the best reason to tackle it early. If atproto's moderation layer can't support discussion apps, it suggests the protocol has a real ceiling on the kinds of communities it can host.
Wrapping Up
It’s becoming more clear every day that Bluesky as a platform is not a sufficient arena for discussion. I think the reason is because we’ve crammed millions of users into relatively small feeds, where most of us see the same general things but can’t talk to the person next to us because of the noise. Discussion apps are not the only meaningful alternative experience that could be built on atproto. They are simply the most urgent. Addressing this gap would relieve real tension in the network, improve the day-to-day experience for users, and make Bluesky a healthier place to create culture.