A Critical Welcome to ‘AI for Radicals’
Comrades, the French Left Has Given Us a Gift
Red Mole Editorial
Espaces Marx has done something useful. No, better than useful: they have produced AI for Radicals, a thirty-page guide that treats artificial intelligence as what it actually is for the revolutionary movement: a tool. Not a messiah. Not an existential threat requiring hand-wringing think-pieces. A tool, like a duplicating machine or a megaphone, that can either serve capital or serve us depending on who wields it and how.
The guide, available in English and French, walks activists through practical AI use: from writing mobilization leaflets to converting articles into video scripts, from running AI locally on your own computer to the technical art of “prompt engineering.” It is accessible, concrete, and refreshingly unsentimental. It deserves a wide readership on the international left.
But here at Red Mole, we have been wrestling with these questions for most of 2025. Our own internal (shhhhh - top secret - but available to Fourth Internationalists upon request) guidance document, “The Dialectics of Digital Agitprop,” addresses a problem the Espaces Marx comrades touch upon but do not fully resolve: AI does not merely assist revolutionary writing. Left unchecked, it liquidates it.
Where the French Comrades Get It Right
AI for Radicals echoes several core insights we have developed at Red Mole. The convergence is not accidental; it reflects the material reality of what happens when revolutionary organizers actually try to use these tools.
1. The Machine Cannot Replace Collective Intelligence
Espaces Marx insists: “AI is not here to replace collective intelligence or militant work, but to save time, simplify technical tasks, and amplify the impact of our actions.” They warn that AI “doesn’t ‘understand’ in the human sense” and “can fabricate information.” Our own guidance document makes the identical point through Marxist terminology: AI produces “probabilistic averaging” that tends toward “bourgeois objectivity” and “liberal centrist drift.” The French comrades frame it practically; we frame it theoretically. Both arrive at the same conclusion: the machine offers raw material, not finished analysis.
2. Political Viewpoint Must Be Explicitly Injected
One of the strongest sections of AI for Radicals addresses what we call the “Equidistant Fallacy.” The French comrades observe that “your AI conversation partner will often be as consensual as possible” and that without specifying a political angle, you will receive “roughly the same answer you would get from a journalist working for a mainstream or public media outlet: the blame is shared, there are good solutions on both sides.”
They recommend using explicit political markers: “comrade,” “nationalization,” “self-organized collectives.” This maps directly onto our “Vanguard Lexicon” approach: replacing AI’s vague language (”middle class,” “working together”) with the precise terms of our tradition (”petty-bourgeoisie,” “united front,” “regroupment”). Both documents recognize that revolutionary politics must be forced into the output. The machine will not volunteer it.
3. Verification Is Non-Negotiable
The Espaces Marx guide dedicates substantial space to fact-checking AI output: requesting sources, testing links, cross-referencing claims. They warn that “even plausible elements may be slightly or entirely false” and that AI “may invent or distort information.” Our guidance document frames this as the problem of AI “hallucinations” emerging from its RLHF training, but the practical instruction is identical: never trust AI claims without independent verification. The machine lies with confidence. That confidence must be treated as irrelevant.
Three Lessons the French Guide Misses
AI for Radicals is an excellent introductory text. But it is precisely that: introductory. For activists who want to produce revolutionary content that can actually organize rather than merely inform, three additional lessons from our internal work deserve attention.
Lesson One: The Rhythm of Revolutionary Prose
AI produces text with uniform sentence length, typically fifteen to twenty-five words in standard subject-verb-object structure. This induces passivity. Revolutionary writing requires what we call “the Trotskyist Breath”: deliberate variation between staccato interruptions and cascading hypotactic constructions that build momentum through subordinate clauses, holding the main verb until the tension becomes unbearable.
Compare an AI draft: “The failure of the constitutional reform movement demonstrates that relying on parliamentary procedures is often ineffective during times of crisis.” Now the humanized version: “They relied on the constitution. It failed. It always fails.”
The French guide acknowledges that AI text may need “reworking” but does not identify what specifically makes AI prose deadening. The answer is rhythmic uniformity. Break it.
Lesson Two: The Activist Anecdote as Authenticity Marker
AI has no body. It has no memory. It has never stood in a drafty community centre hall watching the clock tick past 9 PM while a comrade from the platform reads a forty-minute statement on the situation in Bolivia. It has never sold papers on the high street, watching people walk past with supermarket bags lighter than they were a month ago.
These “synthetic memories” or “activist lore” must be grafted onto AI text by human editors. They ground abstract theory in lived reality. They signal to the reader: “I am one of you. I know the fatigue. I know the struggle.” AI cannot produce this texture because it has never experienced it.
The Espaces Marx guide recommends “collective proofreading” but does not specify what that proofreading should add beyond factual correction. The answer: sensory detail, geographic specificity, the material weight of organizing experience.
Lesson Three: Deliberate Imperfection as Human Signature
AI text is perfectly proofread and syntactically balanced. This perfection is a tell. Human writing contains cognitive imperfections that signal a mind at work: mid-sentence self-corrections (”The policy is a disaster. No, worse than a disaster: it is structurally necessary for British capitalism”), fragments (”But we knew better. Always had”), rhetorical questions with contemptuous answers (”And the TUC’s response? A press release”).
These are not errors. They are authenticity markers. They mimic the dialectical process of thinking in real-time, the passionate run-on that emerges when argument overflows syntax. AI detection tools are increasingly sophisticated; deliberate imperfection is not merely aesthetic but practical camouflage.
The French guide notes that AI may use “awkward references” when targeting young audiences. This understates the problem. The issue is not occasional awkwardness but pervasive smoothness. Revolutionary writing must be rough. It must feel like ink on fingers, not pixels on glass.
A Tool in the Hands of our Class
Hugo Pompougnac’s introduction to AI for Radicals correctly identifies the core contradiction: “The ease with which we use 21st-century tools will lend us credibility, give confidence to our supporters, and, at the same time, discourage our opponents.” This is the materialist position. The technology exists. Capital uses it. We must use it better.
But “better” does not mean uncritically. AI is trained on the corpus of bourgeois internet discourse. It defaults to consensus, to the “on one hand, on the other hand” structure that is death to revolutionary argument. Every AI draft is a draft written by a centrist. The human editor’s task is not proofreading but radicalization: forcing the machine’s output to speak the language of the struggle.
The result, done correctly, is what our guidance document calls a “Cyborg” publication: powered by the speed of silicon but guided by the iron will of the Bolshevik tradition. It does not merely inform readers. It organizes them.
Download AI for Radicals from Espaces Marx. Read it. Use it. But remember: the machine is the servant, not the master. The final word, the strategic line, the rhythm of the polemic: these remain ours. Always have.
Resources: