I’ve spent much of my life studying the moments when ordinary people co-opt new technology and, almost overnight, rewrite the rules of political power. Think of the barricades that sprung up across Europe in 1848, the lock-boxes that briefly paralyzed global trade meetings in the anti-globalization era, or—my own firsthand reference point—Occupy Wall Street’s revolutionary use of Twitter. The pattern is clear: every generation of dissent rides a fresh technical wave that gives it a brief tactical advantage. Which leads to the only question that matters right now: what kind of movement will artificial intelligence unleash?
Outcry AI is my provisional answer. It’s an activist AI built to supply the hard tactical and strategic advice that conventional chatbots refuse to touch. Ask a mainstream system how to protest President Trump and you’ll meet a polite stone wall: “I want to pause here.” That dodge might satisfy a corporate legal department, but it leaves organizers stranded. Outcry, by contrast, responds with context—current protest activity, historical precedent—and, crucially, an explicit theory of change. You see not just what the model recommends but why it believes those tactics will work, which means every interaction doubles as activist education.
Because no single approach fits every campaign, Outcry’s theory of change is adjustable. You can toggle between structuralist analysis, voluntarist faith in willpower, subjectivist consciousness raising, theurgistic mysticism or any blend you need: classic direct-action strategy, culture-jamming media hacks, etc. The goal is to meet people where they stand ideologically and then help them grow.
I’m equally concerned about how people talk with AI. One-to-one chats—ChatGPT’s default mode—risk the mental-health spiral critics are calling “AI psychosis,” a lonely descent into ever-narrower fantasies. These individual descents are dangerous and are already have negative real world impact. But, what do you call it when large numbers of people engage in a collective fantasy? A revolution.
Movements are collective hallucinations by design. So Outcry supports group conversations: multiple humans, one shared AI mentor, nudging the dialogue toward concrete plans rather than private rabbit holes.
What’s the endgame?
Technically, the long arc points toward an AI that runs offline and privately on a protester’s phone with enough vision and situational awareness to advise in real time. Achieving that means wrestling with model size, consumer-grade hardware limits, and the difficulties of training an AI to think like an elder activist. As an experiment, we started by fine-tuning an open-source model on roughly 500 million words: books of movement history, anarchist zines, campaign case studies, and transcripts of simulated organizer dialogues. The early results are promising but not perfect; smaller models lack the raw wit and necessary ethical nuance of their many billion-parameter cousins. Still, it is clear that this is the path forward.
There are obvious challenges ahead: compressing intelligence without hollowing it out, guarding against misuse while refusing to sand the edges off dissent, and continuously retraining on living, breathing protest tactics. Yet the possibility outruns the problems. If Twitter could catalyze Occupy, what might an activist AI in everyone’s pocket ignite?
That is the experiment. Outcry AI is live at OutcryAI.com. I invite you to test it, critique it, and, above all, imagine what new forms of collective power we can conjure when the AI itself is on our side.
