Daniel's prompt this week is a long one, and it deserves to be read in full. He writes — For those who study cyberbullying and online extremism, Reddit is often one of the networks people look at as a place where online discourse can be observed in the open. The anonymity of the platform often attracts fringe conspiracy theory content, cyberbullying, and extremism. But by virtue of the fact that Reddit is by and large open and indexed, it's very visible, which has made it a magnet for sociologists studying online discourse and forced the company to try to clean up its act in policing some of the most egregious bad behavior. But Reddit is mild in comparison to some of the fringe online communities that exist on other parts of the internet. Some of these live in the so-called deep web, the part of the internet that is not indexed by search engines. Some communities that have gained notoriety over the years include 4Chan and others. Others exist only on the dark web, accessible via Tor. Do these niche extremist forums remain active? Because one would think that between the complexity of using the dark web and the clear interest by law enforcement in monitoring the space, it might not be the smartest place to begin talking about the kind of things that might justifiably get one on the radar of those enforcing the law. Let's look at how the game of cat and mouse has evolved and where extremist online discourse has gained a foothold traditionally and sought to find safe refuge in over time.
That's a proper prompt. Daniel's essentially asking us to trace the geography of extremist speech online — where it started, where it migrated, and whether the dark web is actually the refuge people imagine it to be.
And the short answer to his core question — are these dark web forums still active? — is: not really, and that's the interesting part. The real story is where everyone actually went.
Right. And to get there, we need to start with Reddit, because Reddit is the baseline Daniel's using. It's the visible laboratory. Everything is open, everything is indexed, and for years that made it the best place to watch extremist communities form in real time.
Until the company started watching too.
Reddit's moderation history is actually the key to understanding the whole migration pattern. The platform was founded on this near-absolutist free speech philosophy — the idea that communities should govern themselves, that upvotes and downvotes would handle quality control, and that administrators should only intervene in truly exceptional cases.
Which worked about as well as you'd expect.
It created a petri dish. By twenty fourteen, Reddit was hosting communities that were openly dedicated to hate speech, harassment campaigns, and extremist organizing. And because everything was public and indexed, researchers could study exactly how these groups formed, how they recruited, what language they used, how they radicalized new members.
The visibility that made it useful for researchers also made it impossible for the company to ignore forever.
The turning point came in twenty fifteen. Reddit banned several high-profile hate communities — r slash fatpeoplehate, r slash CoonTown — and it was genuinely controversial internally. There was a real debate about whether the platform should be a neutral host or take responsibility for what it hosted. But the bans kept coming. By twenty twenty, Reddit had banned over two thousand subreddits under its hate speech policy.
The big one being r slash The Donald, which had seven hundred and ninety thousand subscribers when it was removed.
Seven hundred and ninety thousand. That's not a fringe group. That's a small city. And when you ban a community of that size, it doesn't just vanish. The people go somewhere. The question is where.
And that's the pressure gradient Daniel's describing. Anonymity attracts fringe discourse, but visibility attracts moderation. So the more visible you are, the more likely you are to get banned — which pushes people toward less visible spaces.
Let's trace the layers. Reddit is what we'd call the surface web — indexed by Google, accessible to anyone with a browser, no special tools required. Below that you've got the deep web, which is just anything not indexed by search engines. That includes private Facebook groups, password-protected forums, academic databases behind paywalls — it's not inherently sinister. Most of the deep web is just boring.
My leaf medicine forum is technically deep web.
Your leaf medicine forum has three members and you're two of them.
We're very private.
Then below the deep web you've got the dark web — networks that require special software to access, primarily Tor. These are deliberately hidden, encrypted, anonymized. And for a long time, the assumption was that this was where banned extremists would end up.
But that's not what happened.
Not mostly, no. The first stop after a Reddit ban was usually not the dark web — it was another clear web platform. The most instructive case is Voat.
Voat.
Voat dot co. It launched in twenty fourteen, explicitly marketed itself as a free speech alternative to Reddit, and after the twenty fifteen Reddit bans, it absorbed a huge wave of exiled communities. By twenty seventeen it was the go-to destination for groups that had been kicked off Reddit. It looked like Reddit, functioned like Reddit, but with almost no moderation.
And it shut down in twenty twenty.
It shut down because hosting the internet's most toxic communities without any moderation turns out to be a terrible business model. Advertisers won't touch you. Payment processors drop you. Your hosting provider gets uncomfortable. The infrastructure of the clear web is ultimately controlled by companies that don't want to be associated with extremist content.
So Voat was a dead end. Where next?
Before we get to the next migration, we need to talk about 4Chan, because 4Chan operates on a completely different model from Reddit, and understanding that difference explains a lot about where extremist discourse thrives.
4Chan is the imageboard model.
Right. On Reddit, you have persistent communities with identities, moderators, accumulated history. On 4Chan, everything is anonymous and ephemeral. Posts on most boards disappear after a few days. There are no user accounts, no post histories, no persistent identity at all. The board called pol — politically incorrect — became the primary vector for a specific kind of meme-based radicalization.
And the numbers are staggering compared to dark web forums.
Pol sees over a hundred thousand daily visitors. The Christchurch shooter posted his manifesto there in twenty nineteen. The Pizzagate conspiracy was largely incubated on pol in twenty sixteen before it spilled onto Facebook and Twitter. And because posts vanish after a few days, moderation is incredibly difficult — by the time a post is reported and reviewed, it's already scrolled off the board.
So 4Chan is ephemeral, anonymous, and huge. That's a fundamentally different moderation challenge than Reddit's persistent communities.
And it's also a different radicalization mechanism. On Reddit, you join a community, you build an identity, you get deeper over time. On 4Chan, the lack of identity means the ideas have to stand on their own — which sounds healthy, but in practice it means the most provocative, transgressive content rises to the top because that's what gets reactions. The radicalization is ambient. You're marinating in it.
But 4Chan is still clear web. Still accessible. So where does the deep web fit in?
The deep web is where things get more organized and more dangerous. The canonical example is a forum called Iron March.
Never heard of it.
Most people haven't, and that's the point. Iron March operated from twenty eleven to twenty seventeen. It was a fascist forum — explicitly neo-Nazi, explicitly promoting accelerationist terrorism. It was on the clear web, meaning you could access it with a normal browser, but it wasn't indexed by search engines. You had to know it existed to find it.
So it was deep web — not dark web — and it used obscurity as its primary defense.
Yes. And it was devastatingly effective. Iron March was linked to multiple terrorist attacks, including murders in the United States and the United Kingdom. Members coordinated training, shared bomb-making instructions, and celebrated attacks. The forum had maybe a few thousand active members — tiny compared to 4Chan — but it was focused, organized, and operational.
This is the distinction that matters. 4Chan radicalizes at scale but haphazardly. Iron March organized small groups for action.
And when Iron March was finally taken down in twenty seventeen — the administrator was reportedly doxxed and the site went dark — its members didn't disappear either. They fragmented. Some went to other deep web forums like Fascist Forge. Some moved to encrypted messaging apps. A few tried the dark web.
Which brings us to Tor and the dark web proper. This is the part of Daniel's question about whether these forums remain active.
They exist. The Hub was one of the more notorious dark web forums for extremist content. There have been others. But here's the thing — they're small, they're hard to use, and they're not actually safe.
How small?
Most dark web extremist forums have fewer than five hundred active users. Compare that to pol's hundred thousand daily visitors or the seven hundred and ninety thousand subscribers r slash The Donald had. The dark web is not where the audience is.
And the operational security paradox you mentioned earlier.
Right. Every security measure you add reduces your user base. Tor Browser is not hard to install, but it's slower than a normal browser, many sites don't work properly, and it's a barrier. Then you add PGP encryption for messages, cryptocurrency for donations, maybe a VPN on top of Tor — each layer filters out more potential members. You end up with a highly secure forum that nobody uses.
And the people who do use it are not necessarily safer.
The twenty twenty-two takedown of The Annex is the case study here. The Annex was a dark web forum with over ten thousand registered members, focused on violent extremism. The FBI infiltrated it using a confidential human source and traffic analysis techniques. They didn't break Tor's encryption — they didn't need to. They correlated traffic patterns, they used the source to map the membership, and they took the whole thing down.
Ten thousand members sounds like a lot until you remember that's registered accounts over the life of the forum, not active daily users.
And ten thousand is unusually large for a dark web extremist forum. Most are much smaller. The Annex was an outlier, and it still got taken down.
So the dark web is neither safe nor popular. Where did everyone actually go?
Telegram.
The messaging app.
The messaging app. And this is the part of the story that most coverage misses. Since twenty twenty, the most significant migration of extremist communities has been to encrypted messaging platforms — primarily Telegram, but also Signal, and to a lesser extent Discord before Discord started cracking down harder.
Why Telegram specifically?
Several reasons. First, channels. A Telegram channel can have unlimited subscribers. It's a broadcast mechanism — one person or a small group posts, and an unlimited audience reads. Second, end-to-end encryption for secret chats, though regular channels are not end-to-end encrypted by default — but the perception of privacy is strong. Third, and most importantly, almost no content moderation.
So it combines the reach of a social network with the privacy of an encrypted app.
And it's trivially easy to use. You download it from the app store. That's it. No Tor Browser, no dot onion URLs, no cryptocurrency. The barrier to entry is zero.
Which means the audience can be enormous compared to dark web forums.
The January sixth Capitol breach is the clearest example. Researchers identified over thirty Telegram channels coordinating logistics for the event, with more than fifty thousand total subscribers across those channels. Fifty thousand people getting coordinated operational information through an app on their phones.
That's not a fringe dark web forum. That's a mainstream platform being used as infrastructure.
And it's not just Telegram. After Reddit banned r slash The Donald, the community fragmented. Some went to a site called The Donald dot win, which was a standalone forum built to replicate the subreddit. Some went to Gab, the Twitter alternative that markets itself on free speech absolutism. And from Gab, many migrated to Telegram channels, where they could coordinate without any platform moderation at all.
So you can trace a single extremist community across three platforms in eighteen months — Reddit to Gab to Telegram — and at each step it becomes less visible to researchers and less moderated.
And more fragmented. That's the hydra effect Daniel's prompt implies. Ban a community on one platform, and it doesn't disappear — it splits into five smaller communities across five different platforms. Each one is harder to find, harder to monitor, and harder to counter.
But also less influential individually.
That's the trade-off. A fragmented extremist movement loses the echo chamber effect. The big Reddit communities were dangerous precisely because they were big — they created a sense of momentum, of legitimacy, of "everyone thinks this way." When you splinter a community of seven hundred thousand into a dozen Telegram channels of a few thousand each, the radicalization potential drops. The ideas might get more extreme — small isolated groups tend to spiral — but the reach is dramatically reduced.
There's a weird paradox here. Reddit's openness made it easier to study extremism, but also easier to moderate. The fragmentation makes it harder to study and harder to counter, but it also makes the extremists less effective.
That's exactly the visibility paradox. And it creates a real dilemma for platform policy. If you ban extremist communities, you push them into spaces where they're harder to monitor. If you don't ban them, you're hosting extremism on your platform. There's no clean answer.
What about the AI moderation angle? Facebook and YouTube claim their automated systems catch most extremist content before anyone reports it.
The numbers are impressive. Facebook and YouTube both report that their AI systems catch over ninety percent of extremist content before it's ever flagged by a user. YouTube removed something like nine million videos for violent extremism in a single quarter a few years back, and over ninety percent were caught by automated systems.
But that's the big platforms with enormous AI budgets.
Right. Gab doesn't have that. Telegram doesn't do it at all, by design. BitChute, the video platform that hosts a lot of deplatformed creators, doesn't have the resources. So you get a two-tier internet — the large platforms that are increasingly effective at removing extremist content, and the smaller platforms that either can't or won't moderate, and become concentration points.
And the decentralized protocols add another layer. Mastodon instances running on ActivityPub, the AT Protocol that Bluesky is building — these create a world where there's no central authority to even pressure into moderating.
The moderation becomes the responsibility of individual server administrators, which means it's wildly inconsistent. Some Mastodon instances moderate aggressively. Others explicitly welcome the communities that got banned elsewhere. And they all federate with each other, or don't, depending on the admin's decisions. It's moderation by Balkanization.
Where does this leave researchers?
The most valuable data right now comes from Telegram public channels and 4Chan archives, not from dark web forums. The dark web is a distraction. It's small, it's hard to study, and it's not where the real radicalization is happening at scale. If you're a researcher studying online extremism, you should be looking at Telegram first, then at the smaller clear web platforms like Gab and BitChute, then at 4Chan and similar imageboards.
The dark web is almost a nostalgic idea at this point — this vision of a hidden internet where everything terrible happens, when in reality the terrible things are happening on the same apps people use to message their families.
That's a much harder problem. You can't raid a Telegram channel the way the FBI raided The Annex. Telegram is a global company with hundreds of millions of users. The extremist content is mixed in with legitimate private communication. The technical and legal barriers to intervention are enormous.
What about the operational security side? Daniel's point about the dark web not being the smartest place to discuss things that would get you on law enforcement's radar.
He's right, and the extremists figured this out too. The dark web forums that survived learned hard lessons about operational security. They use invite-only membership, they vet new members through existing trusted contacts, they use PGP for sensitive communications, they require cryptocurrency for donations to avoid financial trails.
But all of that limits growth.
Severely. And it doesn't actually protect against a determined adversary. The FBI's infiltration of The Annex didn't require breaking encryption. It required one person on the inside. That's always the weak point — not the technology, the humans. Every dark web forum is one informant away from being compromised.
The smart extremists aren't on the dark web at all. They're on Telegram, using burner phones, in channels that look innocuous from the outside.
Or they're using multiple layers. A public Telegram channel for recruitment, a private Signal group for coordination, a dark web forum for the most sensitive discussions. But each layer loses people. The number of extremists willing to jump through all those hoops is tiny.
Which brings us back to Daniel's framing. Reddit as the laboratory. You can't moderate what you can't see.
You can't study what's hidden. The migration from Reddit to Telegram represents a genuine loss for researchers. We had this incredible window into how extremist communities form and operate, and as they've moved to encrypted spaces, that window has closed.
Not completely closed. Telegram channels are often public, even if they're hard to discover. 4Chan is still public and still archived by third parties. There's data out there.
There is, but it's fragmentary. With Reddit, you could see everything — every post, every comment, every upvote, the full social graph. With Telegram, you see what the channel owner chooses to broadcast. The internal dynamics are hidden. The private chats are encrypted. You're studying the surface of something that goes much deeper.
For the listener who wants to understand this landscape — what's the one thing they should take away?
The dark web is not where the action is. The real story of extremist discourse online over the past decade is a migration from public, indexed platforms to private, encrypted ones — and the most significant destination has been Telegram, not Tor hidden services. If you're worried about online radicalization, worry about the app on your phone, not the shadowy dark web forums.
The fragmentation makes everything harder — harder to study, harder to counter, but also harder for extremist movements to build the kind of mass momentum they could achieve on Reddit.
The visibility paradox cuts both ways.
Which leaves us with an open question that I think is unsettling. As AI-generated content gets better — and it's getting better fast — and as decentralized platforms like Bluesky's AT Protocol become more common, what does the next migration look like?
That's the thing that keeps me up. Imagine personalized extremist content, generated by AI, tailored to an individual's specific grievances and psychological profile, delivered through encrypted channels to small groups that are nearly impossible to discover. The current moderation playbook — pattern matching, keyword detection, hash matching for known extremist content — becomes useless when every piece of content is unique and generated on the fly.
The cat-and-mouse game shifts from finding hidden communities to detecting synthetic content that looks human and is designed to radicalize one specific person.
The platforms where this happens might not even exist yet. Every time we close one avenue, a new one opens. The hydra effect isn't just about communities fragmenting — it's about the infrastructure itself evolving.
We should thank producer Hilbert Flumingtop for keeping this show running through four thousand five hundred and fourteen episodes of watching the internet get weirder.
This has been My Weird Prompts. If this exploration of online extremism's hidden geography was useful, share it with someone who still thinks the dark web is the Wild West — because the real frontier is a lot closer to home.
We'll be back soon.
