I intend to look at technology the way Macpherson looked at democracy, as ideas and dreams, as practices and procedures, as hopes and myths. The second reason is that I wanted to discuss technology in terms of living and working in the real world and what this means to people all over the globe. This is the “real” part in the title, and this is what I wish to address now, having spent some time in the first talk on certain aspects of technology as practice.
—Ursula Franklin, The Real World of Technology
The intensified focus on lab-based molecular and genetic study, however, has many scientists dismayed. They note the drop in enrollment of students in traditional natural history courses where field observations are the primary concern. E. Yale Dawson voiced a similar worry. While the microscopes were important, he believed, one could not describe and understand a particular species of seaweed without being in its world and seeing the full breadth of its community.
— Josie Iselin, The Curious World of Seaweed

We find information and build it into knowledge socially: by word of mouth, with neighbors or in class, at the church or the bar or the gym, on Facebook or TikTok, and at work. We watch people talking on TV and sometimes we read the news or a book, or we ask a doctor or a lawyer or therapist or another kind of priest, but mostly we talk to and listen to other ordinary people.* (As the tired people in public health can tell you, you have to talk to people in socially integrated ways if you want them to accept your worldview well enough to permit injections of fancy molecules into their children.) We possess and adapt and are changed by that knowledge socially, too. It follows—alas—that our social internet systems are also systems for acquiring and updating our knowledge. 

This post introduces the written version of some optimistic and constructive thinking on knowledge in networks that I’ve been wrestling with since the end of last year. But launching straight into that hopeful register feels a little deceptive—I’ve come to my current positions the steep and thorny way, through years of oddball work trying to build knowledge across and inside systems that have gone very wrong, and to study how those systems actually function. This work matters to me because I’ve been in it and watching it. And I’m feeling it, in my spacey brain and increasingly in the sketchy radial nerve in my right arm: the overwhelm of the feeds and streams, the outlandish difficulty of the basic sensemaking we do at Unbreaking, and the variously terrible and hope-inducing things I’ve learned about platform effects and bottom-up efforts to make better ways to live together online. So I want to spend just a moment on the current situation before lurching off into maybe-remedies.

In the real world of social technology, people on platforms are experiencing more right-wing and more extremist content,† suppression of non-right-wing activist communication, reduced access to high-quality journalism, and the loss of organizing and communication networks that had allowed disabled and Black activists to build broad solidarity. And across networks, even when opaque algorithms aren’t in play, social internet users are rewarded more for posting links to sensationalist (mostly right-wing) bullshit than for linking to serious journalism.

Some effects of our information disorder (or truth decay, or epistemic collapse) are obvious and acute: In June 2026, the world’s only trillionaire used the social platform he bought and remade in his white supremacist image to help ignite full-on racist riots in Southampton and Belfast by accelerating the spread of xenophobic rhetoric and even AI-generated lists of streets to target. In their conspiratorial form, these riots are barely distinguishable from Facebook-fueled racist riots in Myanmar from a decade ago in the runup to a genocide—except that now a major social platform leader is doing it on purpose. In between, we’ve seen the same pattern (though at first minus Elon Musk’s demonic presence) in 2018 in Sri Lanka and Germany, 2020 in India and Ethiopia, and 2023, 2024, and 2025 in Ireland, England, and Northern Ireland. 

But focusing on single incidents understates the seriousness of what’s happening. Social platforms’ immediate roles as podiums for riot-stokers are always working on top of long local histories of racist culture and violence, so cautious observers often object to assigning blame to platforms for actions taken by racist humans. (Müller & Schwarz’s “Fanning the Flames of Hate: Social Media and Hate Crime” from 2020 does elegant work on both points despite some dated echo-chamber framing.) The exact calculus of blame doesn’t interest me much—any substantial role in extremist violence is actually quite bad—but we also shouldn’t pretend that those humans in the streets weren’t also affected, long before and long after each flare-up of violence, by the social internet’s broader effects on polarization and extremism. The social internet landscape we inhabit has had observable effects, including the spread—first covertly, later in the mainstream—of extreme right-wing racism and xenophobia and support for political parties who exploit and benefit from those beliefs. And if we take one more step back, we can remember that the social internet’s second-order effects include the demolition of local information systems—and that loss appears to be a genuine causal factor in increased polarization (non-paywalled version of paper). 

Taken far enough, societal destabilization into extremism transforms racist riots into national policy: In the US, as we exhaustively catalogued at Unbreaking, hastily recruited ICE and CBP officers have spent much of the past year and a half beating, tear-gassing, detaining, harassing, starving, sexually assaulting, shooting, and killing people who looked like Black and brown immigrants or allies of immigrants—all against a background of fake videos, xenophobia, white supremacy, and viciously racist and dehumanizing rhetoric on social platforms, including from elite figures and official government accounts.

We’re all living in this ecosystem now, whether we choose to use a given platform or not, but I don’t think any of it can be sustained. I don’t believe Zuckerberg and Musk—or Sam Altman and Dario Amodei, for that matter—will be the forever-rulers of our social and information spheres. Frankly, I expect the eventual backlash against global authoritarianism to include at least one faction devoted to obliterating the platform monopolies that now hold so much unelected and unaccountable power. All empires fall, and even the wealthiest and most powerful come to dust. 

In the meantime, those of us who care about this stuff enough to fight off the gravity of Instagram or X, or even supplement it with something different, have other (decentralized, interoperable, increasingly sophisticated) places to be. And true sickos who care about building a knowledge habitat we can actually live in have the opportunity to use these other places as laboratories for building better futures. That’s why I’m here, at least.

Landslide, then the holdfast

An essay I posted right before the end of last year, “Landslide; a ghost story”, carries my broadest and spookiest sense of the destruction of knowledge and sensemaking capacities and spaces, alongside social-science work on the distributed and social nature of our knowledge—which barely exists in our adorably underpowered individual brains. I landed here: The ultimately collective/communal nature of our real knowledge, and the damage to the substrate on which it rests, means that “knowledge problems that ripple across our information ecosystems can become not just threats to individuals with weak character or bad habits, but the equivalent of colony collapse.”

Since the early spring of this year, I’ve given two different versions of a talk—one at AtmosphereConf in Vancouver, BC (Ionosphere.tv version, regular boring YouTube) and another at PublicSpaces in Amsterdam—that built on and extended Landslide’s ideas into constructive territory. I’m so grateful to have had the chance to deliver both of them, and also for the chance to stretch out a bit more in this next phase. So rather than publishing an expanded version of this spring’s talks as an essay—as I did with my XOXO talk and the “Against the Dark Forest” essay—I’m going to break it down into a longer, deeper series of ideas and explorations on the theme of holdfasts and holding fast. Which brings me to the seaweed.

A confession: I actually started trying to write the follow-on material to Landslide—the positive case for building epistemically healthy online social systems—around some quite literal metaphors, like the retrofitting human societies do to landscapes and buildings to protect them from the catastrophic effects of earthquakes and tsunamis and other dissolutions of solid ground. This was very boring even for me, so I went outdoors. 

I live near the mouth of the Columbia River that divides Oregon from Washington, and about 15 minutes from the Pacific shore by car. Our local kelp on the Northwest Coast stretching up through British Columbia is mostly bull kelp, but the kelp family—and the macroalgae kingdom, more broadly—is global, diverse, and good to think with. Kelp forests are ultra-diverse, super-productive communities that shelter and feed huge numbers of species—only coral reefs surpass their biological richness. And kelp forests can do this because of their holdfasts, which are the extraordinarily strong rootlike anchors that seaweeds evolved to cling onto the sea bed with such strength that they can hold on and flourish in even the roughest, most turbulent intertidal zones. 

In my talks this spring and now in these essays, I’m thinking with kelp and other seaweeds that are so extremely good at anchoring down and helping whole communities hold onto the world because I want that for us. I want us to be able to hold on to the best knowledge we have, and I want our social systems online to help us do it instead of sweeping us off our moorings or smashing our brains on rocky shores. I want this because I think, earnestly, that we risk losing our democracies if we let ourselves be dragged out to sea, but also because I think that as collectively sapient creatures, we share a right to the best knowledge our cultures can produce. We have a right, specifically, to well-formed knowledge: knowledge grounded in truth, produced by the best methods and norms of the field or context it descends from, and forthright about its origins and shortcomings, whether it descends from recognized epistemic authorities or emerges from lived experience and grassroots sensemaking.

I think this means that those of us trying to design, build, influence, lead, or fund social internet systems are definitionally involved in knowledge infrastructure and therefore entangled in this problem: We need to make systems that are beneficial for knowledge-producing institutions and communities of many sizes and kinds. And we also need them to be beneficial for the human brains using them to find and take in information. 

We also know that our new networks and social tools can preserve and transmit better collective knowledge. We can see the beginnings of exactly that emerging from projects like Bonfire Networks, which is building a new generation of tooling co-designed with communities on the Fediverse and from new Atmospheric tools researchers and journalists are building, like Semble, a “social knowledge network,” or Lea, which provides semi-sheltered sensemaking and helps researchers who are being brigaded to protect themselves, or Sill, which separates out news from social streams. We can just try things. And I think we have to. 

Beyond “Trust & Safety”

To get into the right zone for experimenting toward something actually better, I think it’s helpful to frame the problem of good knowledge in our networks in ways that go past nudges or aftermarket tweaks and into the architectural and ecological.

Architectural in the sense that we shape our buildings and afterward they shape us, which was Churchill’s long before it was McLuhan’s. We tend to look at the way we live and behave in our current shapes and structures online as natural—as human nature. But of course we know that our architectural choices affect the way we behave. Different structures and shapes produce different human natures, and some are much more conducive to living together in peace and respect than others. And fortunately, we have generations of knowledge about how to create environments and situations in which we humans are better able to learn and relate to one another. As any gifted schoolteacher can tell you, good environments for learning and socialization look nothing like the platforms Big Tech companies made—and why would they? Those platforms were made with totally different goals in mind, and have ultimately used our longing for connection to extract our data and make us into manipulable and profitable objects. Of course we’re unhappy and disoriented inside them.

Ecological in the sense that we can’t keep treating our online social systems in isolation from everything else and expect things to go well outside them. They’re not going well! In the US, the changes produced by the web, accelerated by the social internet, and turbocharged by AI have eaten the heart of corporate journalism and most other forms of publishing—we lose an average of 2.5 local papers each week. And an ecological view lets us stop debating precise allocations of blame for specific events and consider that the social internet is entangled at every level of our information worlds: We shouldn’t just consider who lit a match, but who put the matches in their hands, who helped turn baseline-racist publics into the equivalent of dry tinder, and who turned local knowledge-webs into desiccated news deserts populated by national politics and clickbait façades. The answer is almost never “the social internet” by itself, but the social internet’s almost never out of the picture, either.

Our new networks are already starting from a place of greater anchoring-down—and greater reciprocity—than the old ones, in that they’re, idk, not owned by the guy last seen encouraging the people of Belfast to enact a pogrom, or run by a company about to enter its fourth year of blocking links to news in all of Canada. But we will need much more ambitious changes and experiments to build systems that genuinely nurture good communal knowledge-making. And that’s not what we’ve been doing so far.

Next up: What we have been doing so far and the vast and under-explored space beyond and around those things.

Notes

All gratitude, as always, to the stalwart people who support my work even when I go into a burrow for a few months to read and think and only pop out for talks. It’s good to be above ground again.

Footnotes

* And now we also talk to extraordinarily complex and alien language models that present themselves to us as ordinary and as people, which is only going to make this all weirder.

† The Sky News investigation came with notes on its methods.

The lithograph of macroalgae at the top of this post was made by Adolphe Millot for the wonderful Nouveau Larousse illustré : dictionnaire universel encyclopédique of 1906. Wikipedia tells me that Millot was the senior illustrator at the French National Museum of Natural History, which sounds like a satisfying and overwhelming job.

Now reading