Re:filtered #29: The meatspace internet

Greetings from Brussels and welcome to the 29th edition of this monthly newsletter on civic media in a moment of systemic disruption.

A few weeks ago, in Pyongyang, the regime called some 30 software developers into a room for what they were told would be technical training and what turned out to be an ideological struggle session.

The developers' offense, as reported by Daily NK, was building useful software outside the state's approval channels. A source described the worry behind it: informal development, use, and distribution in the IT sector is increasing, and if left unchecked it could escape the central control system.

That is a striking admission. The North Korean hackerspace is active enough that the party feels it has to stage a public criticism session over it, and it cannot be too harsh, because it needs those same developers.

There is IT development in North Korea that sits outside state control. What that says about the real power balance in one of the world's worst-run places is both heartening and instructive.

It is also the insight behind a study I want to share today, the result of a year of research.

We treat platforms and publications as the infrastructure of useful information in our societies. But underneath that sits a more basic, powerful human layer: networks of people passing information to each other, often informally, often invisible to anyone looking from above or outside.

What the Iran shutdown made visible

Today in Brussels, ASL19 and Gazzetta are co-releasing "Resilience runs through people," a study of how Iranians navigated the June 2025 internet shutdown, the one timed to the Israeli strikes. Publication was made possible with support from the Open Technology Fund.

That shutdown crossed a new threshold for Iran's internet control. Authorities cut most international traffic at centralized chokepoints and allowed access to whitelisted domestic services they had full data access to. Those kept working, but the global internet, for most ordinary people, largely disappeared.

We surveyed about 900 people inside Iran during the shutdown and in the weeks after. Three findings stand out, and they can be uncomfortable for the field that funds and builds circumvention tools like VPNs.

Only about 13% of our reachable sample reached effective, priority-aligned use of a circumvention tool. In terms of objectives of use, news was comparatively serviceable, whereas contact with family, work, and finance were not.

Among people who knew the tools existed, 76% had helped someone else use one. That finding gives the report its name.

We estimate that between 25 and 40 million Iranians took part in some form of peer support. Asked how they worked out their own setup:

• 71% said trial and error
• 18% got help from friends or family
• about 7% from someone technical
• only 1 to 3% read documentation

That trial and error is also where the helpers come from. People crack a setup the hard way, and those are the ones who tend to become the person who spares the next one that trouble. The actual infrastructure of resilience is a teenager fixing his aunt's phone, a vendor in a shop in, say, Mashhad pre-installing a VPN before selling the device in improved form.

From these responses we drafted a taxonomy of helpers to make the network legible to the people who fund and build for it: information sharers, setup helpers, and infrastructure enablers.

The harshest conditions give the clearest picture

Two years ago I wrote about a hunch that led me to start Gazzetta: the most extreme conditions are the ones where the underlying mechanics of information exchange are most exposed.

When Google still sends you reasonable traffic no matter what you do, on a legacy understanding of due reach, you can credibly act as though platforms are the networks that matter. When the platforms go dark, you see what was there all along, which is people talking to people they trust.

The same dynamic runs through every information environment we have looked at since: A Pennsylvania mother sharing ICE-raid alerts in a Signal group is doing what a Tehran teenager does when he passes along a working server, which is what those Pyongyang developers do when they build the tools their colleagues can use.

A flock of birds is a fitting image: The information does not move because a publisher pushed it. It moves because someone in the flock turned slightly and the bird next to them adjusted.

This matters now, when every conversation turns to LLMs as the new intermediaries. The findings cut against the bottleneck framing, however legitimate that framing is. Even the most polished intermediation still rides on top of the human layer.

The study is an effort to move some attention from the technical side of distribution, which often looks deterministic, to the human networks that carry information, which are not.

Both matter, but the human layer may be the more enduring of the two, the more adaptive, and the harder to distort.

I'd like to call it the meatspace internet.

The word meatspace isn't mine. It's older hacker slang for the physical, embodied world, as distinct from cyberspace. (Friend Jessica taught it to me.)

The meatspace is messy, fragile, often invisible, and full of friction. It creates real inequities. In our Iran data, the more technical the helping role, the more it skews young, male, urban, and well-off. Women and lower-income users carry an enormous share of the information-sharing work and are largely shut out of the configurations and the credit.

Cliques and group-think are real here too. The density that makes a network resilient also risks making it parochial. (The same tired speakers turn up at every journalism support conference, yours truly at four of them this month.)

And yet, try to take the network down. You can't. You can block a site, throttle a protocol, jail an editor. You cannot stop a million private conversations from re-routing the information themselves.

What this means for publishing

None of this is special to Iran or to extreme conditions of internet shutdowns. It is how publishing works almost everywhere and probably always has, just less visibly in places where the platforms still send that mass of algorithmic traffic.

It suggests that most successful publishing is a form of B2C that grows into B2C2C2C2C2C, where each C is doing a job the publisher cannot do for them, on a phone the publisher does not control, in a relationship the publisher does not own.

If that is true, the strategic question changes from "how do we reach the audience" to "what job is the next person in the chain trying to do, and what would make our work easier for them to pass along."

Friction is the hurdle in a jobs-to-be-done calculation, and constraint-based segmentation is how you measure it: sort people by what stands between them and the job, not by who they are. That often leads you to the helpers, because the job is social more often than not.

People rarely want to be informed in private. They want to help, to be seen helping, to hand something good to someone they trust. That does not wash out with better tools. It is how people are, and how they stay, even if they're sometimes not aware of it (i.e. also "magnifica humanitas").

Some of this is craft we already know. Some of it is design we have barely explored in publishing. At the tactical end, packaging a piece of work so it is easy to pass on. At the strategic end, building something worth sharing in the first place, so the next person gains a little social capital by handing it along.

In service-design terms, this can also mean that a publisher's user is not necessarily the end reader. The user can also be a helper, who co-creates the value and the next leg of distribution, and so on. The next reader is the helper's user. That changes what success looks like, what you measure, and what you build.

The report's recommendations land in the same place:

• Emergency-first flows for the people who configure tools in a panic, with the lowest possible friction.
• Helper-first features: one-tap install-and-configure links, QR codes for transferring configs, image-based guides built to forward rather than read.
• Work out what helpers want and build it with them, not only into the tools. (Imagine that in journalism.)
• For software, track task-level success, calls made and payments completed, rather than installs or usage as a proxy.

Swap "circumvention tool" for "app," or "newsletter," or "podcast," and the recommendations carry straight over to media. Build for the next hop in the network, not for the end of it. Optimize for the conversation the helper is about to have, not the page view they generated for you.

An influencer, in this sense, is a helper a lot of people trust. The parasocial bond between a person and the voice they follow is the meatspace extended online, not something the platform invented.

The leading algorithmic platforms still decide which of those bonds reach scale. They pick the helpers and amplify the relationships, so they are not neutral about who gets heard. But the bond they work on is human. They distribute trust, unevenly.

This is why so many publishers' moves from algorithmic social to messaging apps have disappointed. The mistake has been treating Telegram or WhatsApp or Signal as another distribution channel, a new pipe for the broadcast. Non-algorithmic channels, much like direct human interactions, do not tolerate noise.

Germany's Tagesschau, the flagship public news service, learned this on WhatsApp. The channel was losing followers until it cut the videos, polls, and side-segments down to a morning and an evening briefing. Then it recovered and crossed three million followers. WhatsApp does not push a channel the way an algorithmic feed does.

My guess is that some of the growth came from people forwarding it. The sharper the briefing, the easier it is to hand to the next person. Noise stays put. Expected, appreciated utility travels.

Publishers who do well in those spaces design for value people expect and want to pass on within their social circles. In my corner of the field, the Audiencers' WhatsApp groups are a good example, as are Slacks like News Nerdery and the NPA Slack.

At the Tagesschau they recognized their limited, habitual role, stayed honest to it, and recovered.

The fun part: exploring tactics for the meatspace

Once you stop fighting the false assumption that journalism is owed mass reach, this offers an alternative to hoping for the goodwill of the tech-oligarchs and their algorithmic platforms.

Most publishing outside of them is not a stage with the audience already in their seats, (except in the literal sense at the growing number of often wonderful live journalism events). It is a relay race, and a lot of the runners never thought of themselves as journalists or publishers, or even as particularly online.

There is a lot of new room to design for the relay, to make the next handover, the touch point in service-design language, or even the next leg, easier, and sometimes to pay the people running it when that is mutually beneficial.

It means talking to the Tehran teenager and the Pennsylvania mother as colleagues in a shared effort, not as passive beneficiaries of an intervention.

I find this hopeful. The autocrat's bet is that information is a thing you can intercept at the gateway. The meatspace internet says no. Information is a thing that happens between people, and you would have to disconnect every relationship to stop it. Even Kim Jong-un can't do it.

What we are sketching is, in effect, samizdat forever. If we keep equating open societies with the survival of the big platforms, we have already lost the fight. The more interesting bet, and the more enjoyable one, is to build publications people want to share, and to design for the network they are already in.

(If you are so inclined, you can test this next time we meet by asking me about Standart, the specialty coffee magazine I recommend to anyone who wants to listen.)

That is a fight any publisher should want to be in. I want to be in it, which is why I'd love for you to share this with anyone who might find it useful. If it was forwarded to you, let me know so that I can thank them.

Looking back

May started in Baltimore, where Madison and I presented at the Hacks/Hackers Journalism AI summit on adaptive AI surveys, a method we have been experimenting with for the past year, including earlier work in Iran with ASL19 on creator trust on YouTube.

It keeps a small fixed core of questions for comparability and lets an AI moderator probe further on what the respondent said, rather than what we predicted they might say. It is absolutely not perfect, and sometimes messy, but endlessly interesting.

Thanks to Natalie, I published an essay on journalism vs propaganda in re/visions, the bilingual journal in the orbit of the Lviv Media Forum. The piece is about why the reach numbers newsrooms still measure themselves by have stopped meaning what they think, and why propaganda keeps landing where journalism does not.

Much as we would like it to be otherwise, people are not primarily looking for objective information. In the messiness of the meatspace, they are looking to be heard, seen, and supported.

We tend to start at the bottom of Maslow's pyramid, as if safety and basic needs were the way in. Those needs are real and urgent, but they are not where receptivity to information begins.

Bad actors have always understood this and enter at the top before anyone else thinks to. People take something in first through connection, recognition, and belonging.

The meatspace runs this way, which is why all the voter guides in the world won't save democracy.

My life in America closed, for now, a few days after the second Lenfest News Philanthropy Summit in Philadelphia, where Madison and I ran a workshop on impact measurement. Huge thanks to Diana, Cheryl, Yossi and Tristan.

The test that landed most with publishers there was one that made their strategy a little more grounded: if you stopped publishing today, who would miss you? Build from there. It is the same question this issue is built around, asked from the publisher's side rather than the helper's.

Service Desk, our effort to bring service design to journalism, has in its first year worked with more than 200 journalists at 50-plus news ventures across five continents on this kind of reframing.

The work helps people move past how to tell a sad story for some pity money toward building something people would miss if it were gone, and getting it valued and funded. 

If that sounds appealing, get in touch.

Looking ahead

Straight from Brussels I head to Prague for Unlock on June 4 and 5, where Feri from ASL19 and I are running a workshop version of the Iran study.

Underneath the exercise is the same move, from demographic mapping to constraint-based segmentation, this time applied to a room mapping the helpers in their own networks.

The Vilnius stop is GlobalFact, where friends Callum, Valentina from Conexión Segura y Libre, and Farhad, the editor-in-chief of Factnameh, and I are running a hands-on workshop to test LLMs on politically contested events: whether they push back on propagandistic or false framing, and what sources they lean on to get there.

In a research project last year we found that they differ starkly. The same question that runs through the Iran study, who gets to decide what counts as information, shows up here too, one layer up the stack. The model is a new kind of gateway, and gateways are where the autocrat places the bet.

Participants will run live tests across several models in their own contexts and leave with a reproducible audit method they can use on their own. Fellow Gazzetteers Madison, David, and Aaron will run a version of this workshop at SRCCON in Minneapolis a few weeks later.

As of the end of May we have closed sign-ups for the tbd/con review crew and are now in pitch review. tbd/con is the free virtual conference on AI and information that Madison, David, Zig, and I are putting together for September 23 and 24.

We want conversations where someone walks in confident and walks out less sure but with better questions. If you have a topic you do not yet have clarity on, or one you would rather argue out in the open than perform consensus on, that is what we are looking for. Pitch by July.

After Prague, Zurich, and Vilnius, I will be based in Vienna for the time being, having said goodbye to New York after a few good years there. I want to settle in and explore the new city before reflecting on our years in the United States.

In a way, the Viennese coffeehouse culture has been a version of the meatspace internet that started earlier. The newspaper on the rack was the publication, and the exchange ran sometimes by way of an opinionated waiter. Cafés like the Sperl and the Prückel still do a version of that work. I look forward to getting some of mine done in them again.

If you can think of anyone I should meet in Vienna, let me know. What has changed over the last two decades? Who should I meet? Where should we eat? I am grateful for any help understanding that meatspace.

Thoughts? Doubts? Ideas? Reply to this email or find me on Signal at patrickb.01. Thanks for reading.

P.S. If you forward this to one person, you have just run a leg of the relay. No medal, but you've helped me prove the entire thesis with a few taps. Thank you.

P.P.S. Four conferences this month. If we land in the same room, find me before I turn into the tired speaker on the circuit I warned you (and myself) about.