Hello, thanks, Barth. It's so good to be here. It's also a lot, and I just want to take a beat to acknowledge that. There's so much devastating in the world right now. But in this room, there's so many inspiring people with inspiring ideas and dreams. Holding it all takes a lot of energy, and while it's exciting, I'm feeling overwhelmed. So I just want to share that and hold space if you're feeling overwhelmed too. I'm Blaine. I work at New Public, where I'm the engineering lead on our roundabout project, which we're building to help local neighborhoods come together online in a healthy way, without the noise or toxicity of big social media platforms.
I'm here, like so many, uh building for a lifetime online and uh my past work as the founding engineer at Twitter and the creator of our op. Uh speaking of neighborhoods, uh I'm honored to be here with you. Oh, sorry. Uh no, no, I don't have slides. Yeah, I'm just gonna talk. I could put some slime molds or something up. Um, speaking of neighborhoods, uh, I'm honored to be here with you uh in Muskum territory, uh, bringing this talk from my home in Sinai's territory. I grew up in BC in the territories of the Dine, Simishon, Kitsum Kalem, Tanaha, and Semiamu, and now the Senaxed peoples.
I moved around a lot, and I value immensely the broad perspective that this disruption gifted me. It also gave me a deep appreciation for the size and diversity of this province and how much it resists being flattened into one story. If you have the chance while you're here, I really encourage you to go visit the Museum of Anthropology just across campus. It's one of the best places I know to encounter that diversity on its own terms. And it's special for me to share this with you from my bioregional neighborhood and my Olma Meider. I studied here, and this is actually where the thread of the talk begins.
So at 9 a.m. on September 11th, 2001, I was attending a lecture on relational database theory inside the Hennings building at UBC. Just over there. It's literally the next building over from where we are now. Uh for the duration of that class, uh not the professor, not the students, no one addressed the elephant in the room. And looking around, I realized that the people in the room, the people who are being trained to build the software systems of the future, were not being trained to think about, think seriously about what those systems would do when they met actual humans.
So on 9-11, I switched my major from computer science to sociology. I realized that I could teach myself the software pieces, but that would leave me without a deeper understanding of the social systems that the technology was for. This wasn't a rejection of the technological systems. Um I like tech. It was a realization about what kind of approach is needed to build technological tools to support people and societies. Engineering trains you to specify desired behavior, build the system, and treat deviation as a bug. I believe this approach is incompatible with social software. Social systems are living, breathing things.
Sociology, on the other hand, trains you to observe what is actually happening, to identify the patterns, to try and identify uh understand what produces healthy outcomes. Only then can you can you cultivate those conditions. You don't define and control the system, you tend it, you steward it. That distinction between prescribing outcomes and describing conditions is central to understanding what meet what it means to build the sort of software we want to see in the world. This is part of a long conversation, one that starts before Christopher Alexander's patterns and Ursula Franklin's warnings about prescriptive technologies and continues through resonant local first home cooked software.
Butler and Ostrom would be proud of today's intercommon intercommunal revolutionary networks. Many of you in this room and our peers elsewhere at the conference uh have signed manifestos, written code, and built organizations that embody versions of this argument. So I'm not here to convince you that the first era of social media got it wrong. You already know that. I want to what I want to do is dig in a little uh dig in a little and push us to think a little bit further. Because even in this community, in my own mind, I hear a lingering instinct, a quiet hope that the answer is something like a good trust and safety team, but decentralized, a better manager, a kinder machine.
And I want to argue that the shift we need is more radical than that. And I want to ground it in something visceral, something that I think proves that centralized cultural governance fails structurally, not just politically. It comes from cheese. A decade ago, when I was living in London, I encountered the work of Bronwyn Percival, who manages the cheese counter at Neil's Dairy, a Neil's Yard Dairy in London and has written brilliantly about the relationship between safety complexity and living systems. She tells a story about traditional cheesemaking in the Auvern region of France. These are uh they make raw milk cheeses made with unpasteurized milk, milked by hand in open fields on volcanic slot slopes into wooden buckets in the fields that serve as the molds for the traditional wheels of cellaire.
From above the managerial stance, this looks like a food safety nightmare. There's no sterilization, there's no fully controlled environment, no standardized inputs. And yet these cheeses can be remarkably resilient. Why? Because the microbial culture is already there. It is adapted over time to this place, this practice, this environment. It occupies the ecological niches that a harmful organism would otherwise move into. Safety does not come from absence. It comes from presence, from a healthy, adapted living biome. And one of the remarkable things about traditional cheeses is that they are not all the same. They're shaped by geography, by farming practice, by the cave, by the shelf, the climate, the accumulated microbial life of a place.
Terroir is not just branding. It's microbial geography. It is a living community shaped by place over time. Now, pasteurization. Pasteurization saved lives. That needs to be said clearly. Milkborne disease was real. Heating milk made industrial scale dairy possible and dramatically reduced risk. But it also wiped the slight clean. It destroyed the indigenous microbial ecology, the living community that evolved over long periods of practice to do useful things. And then it added back a simplified starter culture, a few selected strains chosen for predictability, produced by just two corporations globally. You get safe, you get consistent, you get scalable, and you lose flavor, you lose specificity, you lose 90% of the biological diversity, the accumulated intelligence of a living system adapted to its con to its context, to our context.
Sterilization didn't even fully eliminate the risk. It relocated it from places where living systems could manage it to places where only procedural compliance could. Now, people aren't cheese. Human communities are infinitely more complex than 40 kilograms of microbial ecology. The required diversity in a healthy cultural system is structural structurally impossible to encode from one place at planetary scale, not just difficult, not just politically fraught, structurally impossible. And that impossibility does not get better as you add complexity, it gets worse. If a single centralized approach can't preserve the diversity of cheese, one of the simplest living systems that we know how to cultivate, then consider what we attempted over the last 20 years with human culture.
At most half a dozen global platforms, each one operating as a single cultural pasteurization plant for hundreds of millions of people. One classification classification scheme, one starter culture, one idea of what safe means, one interface for the entire world. The possible forms of healthy human community are orders of magnitude more complex than a cheese culture culture. They require correspondingly richer means of expression, richer and more diverse forms of self-governance, richer mechanisms for maintaining health. I think the first generation of the social web really DID try to pasteurize human culture. And the thing that I want us to reckon with is not just that it failed.
It is that it is not it could not have succeeded. Not because the people involved were bad, not because the incentives were wrong, but because the architecture made it structurally impossible to preserve the thing that it was trying to protect. And not that it didn't do damage, it did enormous damage. But it could never achieve what it set out to do because the human because human culture is more resilient than any system designed to contain it. There are two ways to make something safe at scale. You can sterilize it or you can cultivate the conditions under which healthy life persists.
The first era chose the first path. I think it's time to try the second. So what does the second path actually require? If you study living systems, cells, ecosystems, communities that persist over time, you notice they all solve the same problem. They maintain they maintain they maintain a difference between inside and outside. A cell does this with a meth membrane, not a wall, a membrane, selectively permeable. It lets certain things in and it keeps certain things out. And the specificity of what crosses that boundary is what makes the cell alive rather than just chemicals in equilibrium with their surroundings.
A community does this too, with norms, with shared context, with the slow accumulation of trust. When Rudy spoke of groundings with his siblings, this is what he's pointing to. A community that cannot maintain the difference from its environment is not really a community. It's just more environment. Aaron's example this morning of the numbness that results from the jarring contrast of the joy of a wedding and the horror of international atrocities is a poignant illustration of this. And this is what happened on the old platforms. The membrane dissolved. Inside became outside, and every context collapsed into one undifferentiated public.
So stop thinking about platforms and start thinking about membranes. Living systems maintain boundaries, selectively permeable with consent. They signal across those boundaries, not commands but information that others can interpret and respond to in context. They adapt through feedback, graduated responses calibrated to situation, not uniform rules from above. And they differentiate, becoming this community with this character, with these members, rather than a generic instance of one universal social form, one global social graph. The opposite of a giant engineered monoculture is not disorder, it's ecology. Many bounded systems communicating across difference, adapting to local conditions and persisting because they are healthy, not because they are controlled.
So if that is the vision, many bounded systems communicating across difference, adapting locally, what does the infrastructure for that actually look like? And I want to talk about two problems. The year before my database class, in the same lecture hall, just over there, I learned the old joke that there are only two hard problems in computer science cache and validation and naming things. I think that atproto and this new thing you probably never heard of, pan proto, between them might solve both. Let me start with cache and validation. If communities are going to persist, they need durable infrastructure that does not depend on one company's continued existence or goodwill.
They need to own their data, they need continuity, memory, the ability to move, to fork, to recombine, and to remain legible to one another through all of that. atproto gives us this identity, data, and sync that persist across a distributed network without collapsing back into one service. Cash invalidation AT Protocol scale. Keeping data consistent, authentic, and verifiable across many independent hosts is what App Proto was built to solve. We're all building on it. You know what it makes possible. Now, a question that I find useful is what sort sort of forms does Instagram prescribe? That's actually a pretty easy question to answer.
A feed of images with descriptions that forbid links, stories, reels, a specific set of interactions optimized for a specific business model. You can describe Instagram's social ontology in an afternoon. What social forms does atproto describe? That's a categorically harder question. That difficulty tells you something about what kind of system atproto is. atproto is not socially neutral, no substrate is. It makes choices about identity, about data ownership, about what is public, but it's architecturally resistant to prescribing the full shape of social life. It decomposes what the old platforms had collapsed together. It creates room. But here's what I think is not yet finished.
There's the second problem, and I think the harder one, naming things.
We're currently supporting communities in Richmond for Virginia, Burlington, North Carolina, Lancaster, Pennsylvania, Lincoln County, Wisconsin, and Chattanooga, Tennessee. These are different places with different cultures, different needs, and different ideas about what healthy conversation looks like. When Richmond says community member, they might mean something different than when Burlington says community member. When one community says moderation, they might mean graduated trust building among neighbors. When another says it, they might mean rapid response to harassment. The words are the same, but the social reality underneath it is not. If we require every community to use exactly the same schema, the same names, the same categories, the same assumptions, then we will have built rebuilt at the protocol layer.
We've pushed prescription down into lexicons, we've pasteurized the substrate and built a great, a really great distribution system for filtered, normalized culture. And this is the problem with naming things, not in the trivial sense, but naming is an act of social description. How do communities describe their own reality in terms that are locally meaningful and still mutually translatable? I've thought about this for a long time. It's one of the reasons I was skeptical of earlier attempts at interoperable social schemas. I didn't think that we had the tools to do schema evolution well to let naming diverge without fragmenting.
I thought it might be five or ten years away if we were lucky. After the last couple of weeks, I now think the tools are here or very close. And that's what we're working on with a project called Pan Proto. Pan Proto is an attempt to recreate to create the conditions under which schemas can evolve, diverge, and specialize, and still remain in relation. So that a neighborhood forum and a mutual aid network and an art collective can each structure their social life in their own terms and still communicate across the differences. Not by forcing everyone into one vocabulary, but by making it by making translation a first class property of the infrastructure.
Not identical, not collapsed into one universal ontology, but capable of relation. atproto solves cash invalidation, how distributed communities maintain continuity and integrity. Pan Proto addresses naming things, how those communities describe themselves in terms that can diverge without fragmenting, converging when the time is right. Together, they point towards infrastructure that does not prescribe social form, but creates conditions under which many social forms can emerge and remain in relation. And this is already happening. When people in the Atmosphere talk about composable moderation, they usually mean ozone, the labeling model. Label content, uh label content, let users choose how labels affect their experience.
And that is real and valuable. Black Sky's recently announced Acorn tool is a beautiful and deeply considered approach to moderation. At roundabout, our moderation tool, which we call the Amputhy Bureau, does not work like Ozona or Acorn. It's not label-based, it is a different shape entirely, and it's based on the pre-moderation pro-social design pattern. Like Ozone and Acorn, though, its architecture is composable, and we're looking forward to sharing it with the wider proto community soon. The fact that App Proto enables all of this without permission, without a proposal, without anyone at Bluesky or the IETF deciding it was allowed, that is the point.
This is this is ecology, ecological computing in practice. The protocol didn't prescribe a moderation architecture. It created room for moderation architectures to evolve. Different communities, different conditions, different forms. That is ecology, not engineering. And it extends beyond moderation. BlackSky, gander, roomy, roundabout, we're all working in similar spaces on similar but differently shaped projects. In a monoculture, that would make us competitors fighting for the same users. In an ecology, it makes us something else entirely. Different organisms in a shared ecosystem, each adapted to our niche, each stronger because the other exist. We're a kelp forest, not just one kelp plant.
It is critical to our mutual success that we understand this. We're not competing, we're differentiating. And differentiation is how ecosystems become resilient. I want to sit with the hard part of this for a moment. There are many critiques and questions with this line of reasoning. The one I take most seriously is this. Local self-governance has a long history of protecting insiders at the expense of outsiders. Community standards can mean we've decided this person isn't welcome, or this kind of person isn't welcome. Graduated sanctions can mean the popular member gets three warnings while the newcomer who complains gets shown the door.
Composable moderation can mean choosing the labeling service that does not flag the content that harms people outside your boundaries. I know this, I work on these problems every day. I'm not going to stand here and tell you that the ecological model is problem free. But I want to hold that concern next to what came before. Because the centralized model did not solve the insider-outsider problem. It was the insider-outsider problem at planetary scale. One company's cultural framework shaped by one country country's norms applied as the default governance layer for billions of people who had no voice in its design.
Rudy's BlackSky demonstrates this better than anything I could build in a slide deck. Under Twitter, black users built something extraordinary, black Twitter, entirely within and against a system that was not designed for them, was not governed by them, and regularly failed them. What BlackSky has that Black Twitter never could is self-determination at the architectural level, its own moderation, its own governance, its own membrane. On the other side of the coin, does the ecological model also make possible communities that govern themselves badly, though? Yes, absolutely. But those failures are local, they're bounded, they can be identified, responded to, and routed around by the rest of the network.
Under monoculture, when governance fails, it fails for everyone at once. Under ecology, the network's diversity is itself a form of resilience. And I want to name one more thing. The ecological model carries its own form of privilege. The ability to run a community, to moderate, to steward, to maintain a membrane requires capacity, technical capacity, social capacity, time. These are not equally distributed resources. A world of self-governing communities is not automatically a just world. It just has better conditions for justice to emerge if we build with that awareness. So the real question is not is not whether to intervene, but from where and with what knowledge.
The prescriptive approach intervenes from above with general rules applied uniformly. It scales and it flattens. The descriptive approach intervenes from close with local knowledge adapted to context. It's harder. And it is the only thing that I've seen that can produce health at the scale and complexity of actual human communities. Erin Cassane's work on governance in this space bears this out. The communities doing the best moderation work are not the ones with the most success sophisticated rule sets. They are the ones with the most the the ones most deeply embedded in the communities of the people that they serve.
And yes, in a moment when we're all watching what happens with what capital and organizational when capital and organizational change arrive at the doorstep of an ecosystem we care about. This is why the substrate matters more than any single company built on it. An ecosystem is more resilient than any single organism. We have an opportunity to break with the worldview that shaped the last era. Not by building a better Twitter, not by building a nicer control system, but by building infrastructure that supports many forms of community governance and then doing the much harder, much less technical work of ensuring that those communities have the capacity and the accountability to govern well.
So I want to end where I started with cheese. The best cheeses in the world are not the ones made in sterile labs with standardized cultures. They're the ones where someone tended specific specific conditions in a specific place over time. They chose the right milk, built the right wooden bucket, tended the right cave, trusted the microbial community to do what it had been doing for generations. They didn't engineer the cheese. They cultivated the conditions under which a particular unrepeatable form of life could thrive. But, and this matters, they did not walk away and let nature take its course.
They showed up every day. They turned the wheels, they monitored the humidity, they knew when something was wrong before the tests would have caught it, because they had built a relationship with the system over time. That is not hands-off. That is the most hands-on thing there is. It's just a different kind of hands. Not the hands of an engineer or building to spec prescribing the shape of a low poly virtual world. The hands of someone tending a living system. It's not a metaphor for what we're building, it is a description of what we're building. We don't need a better machine for managing culture.
We need we need better conditions for us all to thrive. And then we need to show up every day and tend them. Thank you. Blaine, there are some people who have questions and we have a little bit of time. So I'll run the mic and you'll answer some questions.
Love it. Thank you for talking. Given all of this, do you think traditional tech standards bodies are still a good, still good stewards for the protocol on architecture itself? Um I'm a person to ask about that. I I think I think we need to recalibrate a whole bunch of things. So I I I cut a piece from from the talk talking about uh DNA. DNA is a very simple standard. There are four bases and a very small number of base pairs. And from that, all of life. And so I think that's the level that maybe we need standardization at, um, like technical standardization.
And I think App Proto is is like, you know, the lexicons plus storage plus communication is like maybe all we need. And then the way that I I kind of think about the governance of the next layer up. One of the things that I'm excited about, this panproto thing is that I think it shifts the conversation to a social governance question. So rather than having a bunch of engineers debate about what the attribute, what the JSON attribute should be called, we can debate about how our communities want to collaborate and what the meaning that we're communicating across the network uh is.
And those are very different people in the room for that. So I don't think the ITF is a suitable place for for that sort of conversation. But I I think we'll figure it out. I mean it makes me think about uh just even having uh lexicons that aren't only defined in English, because maybe it don't matter no more. Full emoji lexicons is of course what I mean. Uh other questions for Blaine.
Please state your name. Uh I'm Gov, that's what you guys. Um I'm wondering, so what becomes of engineering standard bodies, you know, in this world, especially if you know we're like we saw with the Addy demo, you know, everyone can make their own thing, which is great. I love that. But what becomes of like formal engineering standards in your opinion? I I mean I I think you know, I'll kind of lean back into that that prescriptive descriptive uh lens. Uh I I think that the prescriptive uh we're gonna define the standard and then we're gonna build software that conforms to that standard model should go away.
Like so it's you know, sort of Ryan's question. Like I don't I don't think that that's a useful mechanism in in this place. I think that there are some really fundamental coordination questions. So like how do I get data off of a PDS is a mechanical question. Um, you know, we're talking like TRNA or something. It's it's like these mechanical pieces are important to encode, but the the rest of it, like everything above that, we need to really aggressively think differently about how we approach that. And and I think really I I would love to see us as a community really strongly distance ourselves from this sort of engineering mindset.
And one of the things that I'm really so grateful for at this conference is just how many people who aren't engineers are in this space, which is a protocol conference, um, and I think that's a really good sign for for where the community is headed. I'd like to observe that all your observations apply to far more than just technology to do with social communications. If you think of how diversity in the animal and the other nature kingdoms is dying, it's the exact same approach of applying the same pesticides in all the same places. Uh and I think there's a hundred percent analogy to what he just said.
So um very much applaud those comments. Thank you.
So um, first of all, awesome, thank you. Uh so on the one hand, you told this story about you know the the way cheese evolved in the local control and all the and and control's the wrong word, sorry. But um just that pattern and how it you know has all this resilience and you know the the the you know anchoring to the kelp bed floor and all that. Uh but I'm struck thinking back on this that uh but it was effectively wiped out by by the other approach, and it's all well and good to say, well, don't do that.
Yeah, but if this is indeed a model for resilience, it's also uh uh a profound example of resilience failure. So uh what do you think? I think I think uh I'll I'll I'll lean back to the the the first things that I said uh acknowledging the the traditional space and the peoples that were here. I think it's actually a profound failure of sort of Western managerial culture and and perspectives. The pasteurized cheeses are actually, if you go and look at the science, they're actually less safe than farmhouse cheeses using raw milk because uh they they relocate, and I'm I I sort of alluded to this, they relocate the uh the failure cases.
And so now it's people in cheesemaking plants or you know retail uh people who are handling uh cheeses that are less resilient and less safe overall. And so they actually have higher like the the pasteurized cheeses have higher failure rates and we've also encoded into law the inability to sort of have farmhouse cheeses, and so um Bronwyn Percival's book, uh she's she's got a book called Reinventing the Wheel, is really about this sort of managerial failure and and the the governance failure that doesn't let us sort of reflect on oh pasteurization was great in the 18th century uh when people were living in you know very dirty conditions making cheese in their in their backyard.
Um now that we understand the process more, we could actually move to a different, better mode. Um and I think I I just want us to avoid doing that again. Uh let's keep rolling. You should keep talking. We should have more cheese talks uh afterwards. Uh I think I know what you're getting at, where you're like, are we saying uh any other questions? Amazing. Uh thank you very much, Blaine. Thank you.