In approximately 1998, a fifteen-year-old in suburban Ohio — let's call him xX_FreedomEagle_Xx because that's spiritually accurate — discovered that he could log into EFnet, join #politics, and immediately start screaming his half-formed opinions into a room full of strangers who would scream back with equal ferocity. There were no moderators. There were no community guidelines. There was no algorithm deciding which screams deserved amplification. There was just raw, unfiltered human conflict, running 24 hours a day, powered by dial-up connections and teenage hormones.
Facebook would spend billions of dollars and employ thousands of engineers to accidentally recreate this exact environment twenty years later. EFnet did it for free.
The Anarchist Infrastructure
To understand why EFnet's political channels were uniquely catastrophic, you need to understand what EFnet was and, more importantly, what it refused to be. EFnet — the Eris Free Network — was the oldest and, for most of the 1990s, the largest IRC network in existence. It operated on a philosophical principle that would give a modern Trust and Safety team a full cardiac episode: there was essentially no central authority. Channel operators controlled their own channels. If a channel had no operators, it was a free-for-all. Server admins could ban individual users, but coordination between servers was minimal and contentious.
This was a feature, not a bug, to the people who built it. The internet, in this vision, was a wilderness where competence and aggression determined survival. Moderation was for the weak. If you couldn't handle getting flooded off a channel, maybe you shouldn't be on IRC.
Applying this philosophy to a channel called #politics was like handing a flamethrower to a room full of people who were already arguing about who gets to hold the matches.
The Toolkit of Digital Warfare
What made early IRC political channels genuinely different from a bar argument or a letters-to-the-editor section wasn't the content — humans have been having the same arguments since the invention of agriculture — it was the tools available to participants.
Flood scripts let you overwhelm an opponent's IRC client with so much text that it disconnected. Winning an argument, in this context, was literal: you could physically remove the other person from the conversation. Clone bots let a single user create the impression of a crowd — join the channel with fifteen different nicknames, all agreeing with your position, and suddenly you'd manufactured consensus from nothing. Nick collision attacks could steal someone's identity. DCC exploits could crash their client entirely.
This was debate club with packet weapons. And the people who were best at it weren't necessarily the ones with the most coherent political philosophy — they were the ones who'd spent the most time in #warez learning how to run scripts.
The overlap between the hacker/warez scene and the political channel regulars was significant and underappreciated. Kids who'd spent months learning to automate IRC operations for file sharing purposes discovered that those same skills translated directly into political influence operations. A teenager who could run a coordinated XDCC bot network could absolutely run a coordinated sockpuppet campaign. Same skill set, different application.
The Radicalization Pipeline (Dial-Up Edition)
Here's the mechanism, reconstructed from the accounts of people who lived through it and emerged ideologically transformed in ways they're still processing in their thirties:
Step one: curious teenager discovers IRC, finds political channel, initially lurks. Step two: teenager posts a mild opinion and gets absolutely destroyed by veterans who've been doing this for years. Step three: teenager, now humiliated and furious, decides to learn the rules of the game. Step four: teenager acquires flood scripts, learns to clone, studies the rhetorical patterns of whoever seems to "win" the most arguments. Step five: teenager begins to internalize the worldview of whoever taught them to fight effectively, because humans are social creatures who adopt the values of their communities. Step six: six months later, teenager is a true believer in something they'd never heard of a year ago, defending it with a ferocity that has nothing to do with the ideas themselves and everything to do with identity and belonging.
Facebook's engagement algorithm does this with likes and reshares. EFnet did it with op privileges and flood protection. The psychological mechanism is identical: reward the behavior that keeps people engaged, punish deviation, create in-group identity through conflict with out-groups. Zuckerberg's engineers rediscovered a wheel that #politics had already invented, stress-tested, and broken several times over.
The Characters Who Ran the Show
Every long-running EFnet political channel developed its own ecosystem of regulars, each occupying a distinct ecological niche.
There was the Veteran, who'd been on the channel since 1995 and had developed opinions so baroque and internally consistent that engaging with them was like trying to argue with a conspiracy theory that had achieved sentience. There was the Script Kiddie Ideologue, whose political positions were secondary to their desire to demonstrate technical superiority by flooding anyone who disagreed. There was the Lurker Who Finally Snapped, who'd been watching for months and suddenly unleashed a 3 AM manifesto that read like someone had fed the entire channel's logs into a fever dream generator. There was the Person Who Was Clearly a Fed, or at least everyone assumed so, which created a fascinating layer of paranoia that made every conversation feel vaguely cinematic.
And there was the Bot — the Tcl-scripted, always-online presence that logged everything, delivered canned responses to certain keywords, and occasionally went rogue in ways that were difficult to explain and impossible to predict. Every serious channel regular ran bots. The bots outlasted their owners. Some of them are probably still running on shell accounts somewhere, arguing with each other in empty channels about politics that have since been resolved or replaced by new disasters.
What They Accidentally Built
The researchers who study online radicalization today — and there are now entire university departments dedicated to this — describe a set of conditions that reliably produce ideological extremism: anonymity, absence of moderation, high-conflict environment, in-group identity formation, repeated exposure to extreme positions as a baseline. Every single one of these conditions was present in EFnet's political channels in 1998, operating at scale, with zero funding and zero oversight.
The kids running flood scripts in #politics weren't trying to destabilize democracy. They were trying to win arguments with strangers on the internet, which is arguably the most human motivation imaginable. But the environment they were operating in was a near-perfect radicalization machine, and a meaningful number of them came out the other side holding views they'd never have encountered, let alone adopted, without it.
The darkly funny part — and there is always a darkly funny part — is that the same properties that made EFnet's political channels so effective at manufacturing ideological commitment also made them completely exhausting. Most people burned out. The channel would empty out at 4 AM, the bots arguing with each other in the silence, and by morning the whole cycle would start again with a fresh batch of curious teenagers who'd just discovered that politics was apparently happening on IRC.
It was unsustainable, unscalable, and ultimately self-limiting in ways that modern platforms are not.
Facebook fixed that problem. Nobody should be proud of this.
EFnet still exists. #politics is considerably quieter these days, but the bots are eternal.
