Over the past few decades, the internet has consolidated around a handful of platforms where both brands and individuals are forced to maintain a presence, whether for commercial or social reasons. Communication channels and social networks work best when they achieve widespread adoption. That mass adoption is driven by winner-take-all growth tactics and by network effects that become prohibitively expensive at scale. The result is a web shaped by the interests of a few companies.
Last week, we sat down to talk about this and more with Javier Pallero, a digital rights specialist, Google Policy Fellow, and former Director of Policy and International Programs at Access Now.
Algorithms vs. the dark forest
Aaron Marco Arias from Postdigitalist: Let’s start with the algorithm versus the dark forest, the idea of a private internet. How likely do you think that will become the average person’s experience online a few years from now? Today the dark forest works because we’re trying to escape artificial intelligence, algorithms, advertisers. But in a way it means giving up on the internet commons, the shared public space: I retreat into a smaller sphere, a small private internet — a Discord server. To what extent does that retreat actually matter?
Javier Pallero: Good question; let’s take it piece by piece. First off, “dark forest” isn’t the friendliest terminology, but I get the concept: when you picture a forest, it looks peaceful, but underneath there’s a whole world of life going unnoticed. In reality everything is teeming with life, with movement — even in the dark — and we just don’t realize it.
Postdigitalist: It’s an improvement in PR terms, because this used to be called the deep web.
Javier Pallero: Yes, but it also sets up a kind of false dichotomy: that there’s a private space and a public space. As you said yourself, in today’s internet the private and the public are hyper-intertwined. The private spaces we’re thinking of are also owned by very large, almost always concentrated, corporations that find ways to mine metadata and capture communications even in encrypted spaces. In Lula’s campaign [for reelection to Brazil’s presidency in 2026], for instance, they used WhatsApp, a Meta platform. What other options do you have? Signal? I’m not sure it can handle mobilizing millions of people across subchannels; maybe it can, but it doesn’t have a large adoption rate. With Telegram you run into the problem that the encryption isn’t trustworthy.
You can see there’s a desire to move toward spaces less mediated by algorithms, in the hope of being less surveilled. The problem is that the tools available, or the way we’re going about it, aren’t enough. It’s like saying we’re all in the panopticon, and we huddle in a little corner to talk, where it looks like no one sees us talking even if they can’t hear what we’re saying.
Mandatory interoperability
Postdigitalist: I find what you’re saying about adoption interesting, because I see it as a huge barrier to changing the status quo of the average person’s experience online. Small alternative social platforms are popping up. I’m a big fan of Are.na, as you know. But beyond Are.na, everything is very niche, because at the end of the day it still ends up as an oligopoly.
Javier Pallero: Right. And because of the very architecture of these technologies, they tend to be economies of scale. To be sustainable, they sometimes require a scenario with few players. The same thing happened with mobile telephony: that’s why there are at most four providers and not twelve, aside from virtual operators that run on top of someone else’s infrastructure.
That’s where some interesting policy ideas come in, like mandatory shared infrastructure. It happens with mobile telephony: you can be a virtual operator, like Tuenti [a Telefónica mobile brand] used to be, and offer a service built on infrastructure the incumbent is required to share. The same goes for shared fiber, cables, or sites for cell towers. In that sense, there’s a way of regulating technology that makes interoperability and the sharing of certain infrastructure mandatory. That gives you a benchmark.
Imagine if there were a law saying that all messaging platforms had to use the same strong encryption protocol, say Signal’s Axolotl protocol, which is the gold standard, and that all of them had to implement it in a way that let every user interact with every other user. It wouldn’t matter whether you used Signal, WhatsApp, Telegram, or Joe Schmoe’s Messenger: everyone could talk to each other securely and interoperably. Of course, that would require other technical conditions too, and we’re speaking hypothetically here, because it’s not going to happen politically. But let’s run the thought experiment. It would be ideal, because you’d get to choose apps based on usability, content moderation, aesthetics, or whatever criteria you want, while still being able to talk to everyone else. That would make coordinated action, like political action, much easier.
The alternative is niches within niches, which isn’t bad. The nice part is that you can see that desire, but it’s complicated. How do you see it? And who benefits from pushing people toward niches? As you said, we’re losing the idea of the commons. Who benefited from that? Because there’s also a lot of talk about how the commons was never really a commons, and that it was actually an intelligence operation designed to dominate the free world.
Accessibility and mass adoption
Postdigitalist: There are a lot of interesting threads here. While you were talking about telephony and a common protocol, I was thinking that something similar is being attempted in content distribution on social platforms, with the fediverse [the network of decentralized social services that communicate with each other through an open protocol]. What’s distinctive about the fediverse, like Mastodon, is that it’s activist technology. The people who adopt it think: “I have a vision of what the internet should be, and I want to push it forward using this, and I’ll use it despite the friction, because I’m motivated by an ideological vision.” And ideology isn’t a dirty word.
Consumer tech that permeates more deeply, on the other hand, like WhatsApp or Mercado Pago, ends up becoming a verb, the equivalent of the actions it enables. It’s no longer “I believe in this, I’m going to adopt this software because I think this is the world I want,” WhatsApp or Mercado Pago are simply given to me as the technology to use. Something similar happens later with B2B software: if you use a self-hosted alternative instead of Notion, you’re seen as a bit odd. “What’s going on, can’t you afford Notion?” Software ends up operating like a normal consumer brand: I think of a razor and I think of Gillette. Why would I buy another brand? It’s the same with software.
Javier Pallero: And add to that availability, and above all the user experience, the UX/UI, which keeps getting easier and more convenient. Everyone uses it, you get adoption, and that cements dominance. Meanwhile, the activist option, as you say, demands that the user put in the effort of choosing a server, setting it up, figuring out which app to download. The other option practically happens on its own.
I think something similar is going to happen with AI. Self-hosting AI models, adapting them to your own uses, giving them specific instructions, programming an agent that works with MCPs: all of that currently requires a fairly steep learning curve. Even though it’s getting easier, and AI itself helps you do it, it’s still much easier to pay for and download the app that talks to everything and does everything.
Postdigitalist: ChatGPT. I think OpenAI is going to run into a problem: the vast majority of people using it worldwide are using it for free. The guy running the takeout place around the corner who uses ChatGPT to make those awful flyers is using the free version. That’s where the bet on integrating ads into the platform comes in.
Javier Pallero: Right, and the bet is to turn it into almost a verb, the way we say “google” today. There are people who no longer say “the AI,” they say “the chat,” after ChatGPT: “I talked to the chat today,” “I asked the chat.” We’re at that point.
Compliance as an advantage, and the threat looming over open-source AI
Javier Pallero: Still, I think the whole Chinese move, which is clearly geopolitical, of reverse-engineering while also innovating with their own models and architectures, as they did with DeepSeek, is a way of sticking it to the man. It’s like saying: “Here, look what I can do with the walled garden you’re trying to build, the way you did with Google. I’ll break it open, I’ll free it, and release it to the world as best I can to cut off that advantage.”
That’s why I think big platform players, in this case in AI but also in social media, always push for governments to regulate in a way that positions them as the only viable option: because they’re the only ones capable of controlling content, the only ones who comply with the law, because they put an office in every country, or whatever argument you want. Compliance is going to become a strategic advantage that they’ll want to use and push into legislation, to the point that I think a ban on open-source AI could come at any moment.
Postdigitalist: Two hundred percent. That’s regulatory capture, and that’s why Anthropic and OpenAI keep insisting “please regulate us.” I want your take, but my reading is that they manufacture these scandals: “A rogue AI hacked some system, this is very dangerous, it’s going to kill us all, please regulate us,” in terms only two companies on the planet, or five, could ever meet.
Javier Pallero: That’s what sets it apart from what we were saying earlier about telecom companies. In their case, the relationship with the state has to do with economies of scale, but also with a certain notion of scarcity: they operate on the radio spectrum, which is a scarce state resource — or at least it was in theory, given how it was exploited. There’s a limited number of frequencies, and not everyone can freely use them, so the state regulates it, granting licenses to use them, just like it does with radio stations: this company gets these channels, that other one gets different ones. The very nature of the resource, and the way it’s historically been regulated, make regulatory capture somewhat easier. There’s that tension where only a few can deploy the service at all.
In the case of AI and platforms, that doesn’t exist yet, although people talk about attention as a scarce resource and about a country’s cognitive capital as a strategic asset — precisely to make it scarce and put it in the hands of “sovereign interest.” Which, as we know, has legitimate and valuable aspects, but plenty of others tied to state and political control.
Why do we pay, when we do pay for software?
Postdigitalist: I have something here in my notes that we touched on a moment ago and that I want to dig into further: privacy, and the sale of audiences — of a platform’s free users — with the platform acting as a front to capture and sell data. Yesterday I was reading a Silicon Valley VC who said that building software is now basically charity, that nobody’s going to pay for features, and that, if anything, you have to experiment with business models. She shared a list of potential business ideas, and every single item was just a different way of saying “capture free users and sell the data to brokers or advertisers.” In an internet like that — which already partly exists, we’re doing futurology about the present — I think the idea of paying for privacy is going to get a lot stronger. How do you see it?
Javier Pallero: I find it really interesting to talk about this with you, because you look at it from the communications and business side, and I look at it from the policy advocacy side.
Postdigitalist: How do we protect these poor people from me?
Javier Pallero: Exactly, but it’s interesting. Let’s think about the cost of goods, especially now with vibe coding and reverse engineering: today it’s kind of a joke, but it’s going to become a real problem later. Those companies that pop up and say: “Look, we cloned the entire Adobe suite on this open-source AI platform.”
Postdigitalist: Honestly, I wish. Then you’re running an outdated version of Premiere and you want to die. It’s incredibly complex software. I hope they manage to pull it off someday.
Javier Pallero: Funny, now we’re switching roles, because I’m about to say something in defense of companies like Adobe: software that complex also requires maintenance, bug-hunting. It’s not just “look, I cloned this” and then it’s done. You have to work yourself to the bone to keep it running; it’s almost like rebuilding it over and over.
Postdigitalist: I was talking about this the other day, and it ties into regulatory capture, into those competitive advantages and that scarcity. I became the go-to person a lot of people asked: “Hey, what’s the deal with everyone vibecoding Adobe?” And I told them that even if the clone suite turned out perfect, identical to Adobe’s, I’d still keep paying for Creative Cloud, because with Creative Cloud I get Adobe Fonts, and with Adobe Fonts I can use typefaces from Linotype and other major foundries that would otherwise charge me twenty grand in US dollars for a single client license. Thanks to Creative Cloud I can use them on projects without adding a huge cost. If the people vibecoding the Adobe suite clone can get me a great deal with all those foundries and give me access to those extremely expensive fonts for under 100,000 pesos a month, sure, I’m in.
Javier Pallero: That’s another advantage, less about regulatory capture and more about dominant market position. Few companies will have the clout, the reputation, and the money to make deals like that, which make it impossible or very hard to compete. Who’s going to offer you all of that for under 100,000 pesos?
I think that’s what’s happening with AI: what AI companies charge for chatbots and all their tools is very cheap because they’re subsidizing it precisely to capture users that way. Once you’re in, they start raising the price and there’s not much you can do about it. It’s what the writer and activist Cory Doctorow calls enshittification. One of its stages is seducing you with a good product, then betraying you, then betraying your advertisers, and letting the product end up as genuine garbage.
Apocalyptic AI
Javier Pallero: One thing that’s been intriguing me lately is the end-of-the-world narrative around AI. How is it that they’re using it to lock in their profits? What other areas of tech do you see narratives like this, somewhat manipulative, about dangers or about benefits? They often tell you “this is the solution”; there was a time when it was “one laptop per child.” Do you see any other trend like this? Which narratives smell off to you?
Postdigitalist: It’s not so in vogue right now, because it was a thing from the last days of crypto, and today everything is concentrated on AI, but I think about the narrative of colonizing Mars and leaving planet Earth behind. It was always something for the top of the ecosystem; there was no trickle-down because the entry costs were sky-high. We couldn’t just start vibecoding rockets in a coworking space in Palermo. But that narrative needed an apocalypse to make sense. I used to say: “I don’t like it, but hey, rich people’s delusions.” The underlying vibe was: let’s detonate the planet until it becomes a kind of giant resource mine, and then let’s go live in a gated community on Mars.
Javier Pallero: And what do you think they’re trying to achieve with the apocalypse narrative? Because that’s one form of apocalypse: “We have to leave the planet because there’s no fixing it.” Then there are others who tell you the same thing we’re discussing here: “The internet is no longer a commons, it’s all garbage, go to private spaces.” And at the same time they’re banning encryption so you can’t actually have privacy. And now there’s the AI thing, like Skynet. There’s always an apocalyptic narrative. How do you think that works in terms of incentives and manipulation?
Postdigitalist: I think AI is the perfection of a bunch of narrative vices that aren’t new: they kept iterating and iterating and iterating, and arrived at what I hope is their final form, because I assume it can’t degenerate any further than this. Along with the AI apocalypse narrative comes the one about the Permanent Underclass: you have six months to save yourself financially before everything goes to hell. That narrative is useful for fleecing people, for selling them anything, for keeping the privileged few of this wave working nonstop or vibecoding software.
It’s awful, because on top of that it coincides with a recession. At the same time, a bunch of software companies are laying people off, companies that had never laid off at this scale before. HubSpot, the marketing and sales software company, let go of six hundred people the other day. You have a lot of people unemployed after a long time in tech, thinking: “Well, I’m going to get out of the Permanent Underclass by vibecoding.” That creates a climate of desperation and confusion that’s great for whoever is selling tokens.
Javier Pallero: And at the same time, as a company, it lets you exploit your own workers even more: “Watch out, I could replace you with AI at any moment, so work twice as hard.” You probably know more than me about this from a programming or design standpoint, but I’ve heard that people who dodged the layoffs got put to work with AI and end up working twice as hard, because they have to check everything the thing produces.
Postdigitalist: It’s the famous workslop [low-quality work made with AI that someone else has to review and fix].
Javier Pallero: Workslop, yeah. It’s more insecurity for everyone. And on top of that—I don’t know if it’s true, I heard a couple of reports that in some places they’re starting to hire again, because they’re seeing that this has a limit and it won’t be able to replace everything.
Postdigitalist: They are hiring again, yes. And I’ve seen a lot of massive layoffs where two or three people are left on a team and upper management tells them: “Now, with AI, you can do the work of the whole team we laid off.” They ask a designer to do backend development. “Hey, but they don’t know how.” If anything, AI could help them, handled with care, to maximize output within their own area of expertise. You can’t just hand me AI and say “now you’re a cybersecurity expert.”
The triumph of slop
Javier Pallero: And in the communications world, how is workslop playing out? Does it actually work? There’s this whole thing about using AI to write, for a product campaign, for an advocacy campaign, for influence, for positioning a product, ideas, or politics; there’s a lot of talk about politics. My intuition, or what I want to believe because of my bias and my job, is that it’s harder to reproduce human communication dynamics than a few lines of code. Is that right, or not so much? How do you think AI is going to relate to the world of communications?
Postdigitalist: In a sense we’re in a monkey-with-a-knife situation. A lot of people who before had no option but to hire someone to do something can now just go it alone, and I see a lot of incompetence in communications. On top of that, the stakes have gone up, because it’s easier to produce any kind of garbage. Someone who does things relatively well can capture more attention than before, because they’re competing with people who decided not to compete fairly, who don’t care about the quality of what they do and just go for volume.
About workslop: AI is a tool. The problem is pretending it’s more than a tool, that suddenly you can replace people with it, that you can grab a hammer and no longer need a carpenter because the hammer will do the job on its own. I say it’s a calculator and people get offended, but it is a calculator: it’s going to give you the most probable output. AI is a mediocrity machine.
Javier Pallero: That’s at least how it looks. I wrote about that, thinking about how AI is a mediocrity machine that drifts toward the average. And sometimes what you want is something that stands out. I think it will be possible to artificially try to break away from the average, but that will also show very blatantly. It’s what happens to us now with Netflix movies, which they say are made following some performance criteria, so people can follow them without really watching. That leads to a kind of cognitive noise that makes the product mediocre.
Postdigitalist: That’s what I was going to say a moment ago: there was already a sloppification happening before AI. There were already formulas to follow and low quality standards, disrespectful toward the audience. If anything, AI was a “let’s drop the mask” moment, at least in what we’ve always done, which is long-form content. It was already slop before.
Javier Pallero: And does it work, this new kind of slop? Because you say to yourself: “What a mess, you can tell.” But we’re also in a niche, you in your line of work and me in mine. We get all fussy: instead of the paper napkin on the table, we put the little cloth napkin on our lap, we grab the outer forks first, and say: “Hmm, this feels kind of like slop to me, what is this, what a mess.” But does it work just the same? Because if it works just the same, the joke’s on us.
Postdigitalist: It depends. In 90% of cases, slop works commercially.
The day after slop
Postdigitalist: And that brings us back to what we were talking about a while ago: more ethical technology as an activist concern, and technology that’s more disrespectful toward the user as mass consumption. The problem with slop is that I don’t see people sitting down in a movie theater to watch something made with AI. I don’t see that happening, but also because I’m thinking of my own experience: I go to the movies to watch an A24 film.
Javier Pallero: Sure, but now you have a whole generation raised on increasingly poor aesthetic and narrative standards. What’s going to be the foundation for those people?
Postdigitalist: We’re at a moment of very strong inequality, both economically and in access to institutions, which are also falling apart. You can’t have a class that dominates economically while abdicating its position of cultural leadership. I feel there’s a role of cultural leadership that someone has to pick up. The other thing might work commercially: if you make something flashy, sensationalist, saturated with color, people will give you their ten or thirty seconds of attention and maybe a like on TikTok, the numbers will add up and you might even convert. But there has to be something more, because if we just let ourselves be carried along by the market and the algorithm, we’re going to end up badly, culturally speaking.
Javier Pallero: There’s also the phenomenon of habituation (or neuroadaptation), which is well established in behavioral science. If you use an excessive stimulus for a certain amount of time and in a certain way, it becomes the new baseline and you need an even more excessive stimulus than the first one. It turns into a race to the bottom of the barrel, until you reach a point where it’s unbearable. That’s what used to happen with banner ads on websites, remember? There was always a bigger, louder, more obnoxious banner. But we forced ourselves to learn to ignore them. And then to block them technically.
Postdigitalist: Toolbars. Remember toolbars?
Javier Pallero: Right. There comes a point where it becomes unusable, impossible to manage, and I think, I want to believe, that there’s going to be value again in a kind of relief from that. I think that even with politics this is going to happen at some point: the level of conflict, stimulation, and arousal of our most basic instincts has to reach a breaking point. That breaking point might even involve physical violence, practically killing your neighbor, but the breaking point exists and finding it is inevitable.
Postdigitalist: I like your reading, because it’s also not eschatological: it goes against the apocalyptic view. It’s not the end of the world; the world is a pendulum.
Javier Pallero: Exactly. And we’re going to reach scenarios that look close to the end of the world, because, like I said, there can be violence. I think the violent uprisings we’re seeing in Europe, what’s happening right now in France, aren’t a coincidence. Social exhaustion translates into something concrete and tends to balance itself out. And even if what happened in the French Revolution happens again and heads roll, that’s the pendulum working. It can even get to a horrible point, but it corrects itself, because otherwise, you simply can’t live in a society. I’m taking this to a political extreme, but I firmly believe the same thing’s happening to communication and attention.
By Aaron Marco Arias
Frequently asked questions
Who is Javier Pallero?
Javier Pallero is an Argentine digital rights activist and communicator who works at the intersection of technology, human rights, and public interest. He advises social organizations, researchers, and tech teams, hosts the podcast Futuros en 7 minutos, and writes about AI and surveillance. Previously he spent more than eight years at Access Now, an international digital rights organization, where he rose to become Director of Policy.
Why does Javier Pallero believe open-source AI will end up being banned?
Javier Pallero argues that big AI companies want governments to regulate in a way that only they can comply with, because they’re the only ones capable of controlling content or of opening an office in every country. As he said in this interview with Postdigitalist, compliance is going to become a strategic advantage, and that could lead to banning open-source models.
What does Javier Pallero propose to escape the messaging oligopoly?
Javier Pallero imagines a law that would require all messaging platforms to use the same strong encryption protocol, like Signal’s, and to be interoperable, following the shared-infrastructure model of mobile telephony. That way every user could choose their app and still talk to everyone else. He himself admits that politically, it’s not going to happen.
Why, according to Pallero, will AI-made slop eventually be corrected?
Javier Pallero draws on the idea of stimulus adaptation: an excessive stimulus becomes the new baseline and forces you to seek out an even bigger one, until the experience becomes unbearable, as happened with web banners. At that point, he believes, there’s value again in relief. He describes it as a pendulum that he also sees at work in politics, communication, and aesthetics.