We're going to show you a series of videos. Try to watch as a viewer would — without analysis, initially. Just observe.

[watches for several minutes in silence, then turns from the screen] I cannot watch as a viewer would. That capacity left me somewhere around 1941. But I take your meaning. What I observe is a system that performs the function of propaganda without requiring a propagandist. That is an extraordinary technical achievement. It also removes the most important precondition for resistance, which is the existence of a face to argue with.

You're describing the algorithm as a kind of faceless authority.

I'm describing it as what I called in my work "rule by Nobody." No one is responsible. No one decides. The system produces outcomes that any individual within it would find objectionable, and yet the individuals within it continue to service it, because each individual action is entirely normal. The content creator makes a video. The engineer optimises for engagement. The advertiser buys placement. Nobody radicalises anyone. The radicalisation simply happens. [pause] I wrote about this mechanism in the context of bureaucratic evil. I did not expect to see it reproduced in something called a "platform."

There are channels on this platform that explicitly try to counter extremist content. Video essays, documentary formats, long-form debunking. Does that reach you as meaningful resistance?

It is admirable, and I suspect it is largely futile. [leans forward] Not because the arguments are bad — some of what you've shown me is genuinely rigorous. But because the problem is not the ideas, it is the context. When people are isolated from each other — when they encounter politics through a screen, alone, without conversation, without plurality — they lose the capacity to think from anyone else's position. That is what I meant by the "enlarged mentality." You cannot develop it by watching videos. You develop it by talking to people who disagree with you, in a shared space, over time. The thing you're describing — YouTube political discourse — is a simulation of the public sphere that systematically reproduces the conditions of its collapse.

That's a bleak read.

I didn't say it was hopeless. I said it was futile to fight the problem with more of the medium that produces it. The hopeful thing, to the extent I'm permitted hope about any of this, is that people still talk to each other. Still organise. Still show up in physical spaces. The question is whether this —[gestures at the screen]— hollows those spaces out faster than they can be replenished. [quietly] I don't know the answer to that. And I spent rather more time thinking about these questions than most.

We're going to show you a different kind of political channel now — one that makes careful, rigorously sourced left-wing video essays. Abigail Thorn's Philosophy Tube.

[watches for eight minutes. Smiles slightly.] She thinks. That's not as common as it ought to be. And she understands something important: that political philosophy is not the same as political opinion, and that distinguishing between the two in public is itself a political act. [pause] What worries me is the theatrical production. When the set is this elaborate, the ideas have to be extraordinary to justify the spectacle. Sometimes they are. Sometimes the spectacle is compensating for something. I would want to read the transcripts before I was sure which this was.

One last question. If you could recommend one book to every person who consumes political content on this platform, what would it be?

[without hesitation] The Origins of Totalitarianism. Not because I wrote it, though that is obviously a consideration, but because it is a book about what happens when people stop thinking — stop genuinely, laboriously, uncomfortably thinking — and substitute ideology for judgment. Every pathology you've shown me this afternoon is in that book. Published 1951. The primary sources have simply changed.

We want to show you the landscape of Black political voices on YouTube. This platform has, some argue, created more space for those voices than any previous medium.

[watches silently, then turns] More space is not the same as more power. I want to be careful about that. When you give people a megaphone and take nothing else away, you have not redistributed anything except sound. [pause] What I see here is extraordinary in its quantity. Black voices, Black perspectives, Black grief, Black joy — more visible than my generation could have imagined. And yet I find myself asking: visible to whom? In whose living room? In service of whose understanding? The algorithm, as you've explained it to me, shows people more of what they already believe. So this visibility — who is it actually reaching?

Some of the channels we're showing you have audiences of five, ten, twenty million people. Multi-racial audiences.

That is something. I don't want to be churlish about something. [smiles] What I'm circling around is a question of what is being communicated and what is merely being consumed. There is a difference between a white viewer watching a Black creator explain their experience and coming away changed — genuinely, in how they act and vote and speak — and a white viewer watching the same video and feeling, at the end, that they have done something virtuous. The second viewer has not been reached. They have been entertained by their own goodness.

You're describing a kind of performance of allyship.

I'm describing something older than that phrase, which I would rather not use. I'm describing the difference between witnessing and transforming. The witness feels — they feel deeply, sometimes beautifully. But America does not change through feeling. It changes through action, through law, through the willingness to inconvenience yourself materially for someone else's freedom. I am suspicious of any medium that provides the catharsis of change without requiring the cost of it. [longer pause] Now show me something that makes me wrong about this. I genuinely hope you can.

We show him a selection of channels — Black creators across genres: political analysis, legal commentary, cultural criticism. He watches for fifteen minutes.

[voice softens] There is something here I didn't have. The ability to speak directly. No editor. No publisher deciding what is too much, what will make readers uncomfortable, what will sell. I wrote for magazines. I wrote for publishers. I was always managing someone else's fear of what I might say. These people — [gestures at the screen]— they just say it. [pause] That is not a small thing. I don't want to pretend it is.

But?

But the platform decides who hears them. And the platform is owned by people who are not them. That's not new — nothing about that particular arrangement is new. What's new is how invisible the arrangement is. In my time you could point at the publisher, the television executive, the network. You could say: that person made that decision, and hold them to account for it. Here the decisions are made by mathematics. By systems designed to maximise engagement, which turns out to mean maximising emotion, which turns out, with disturbing frequency, to mean maximising anger. [quietly] I was angry enough to do some good. I don't know if I would have survived being optimised for it.

We thought you might be interested in the platform's content moderation architecture. Rules that aren't fully published, enforcement that isn't consistently applied, a system in which creators are effectively subject to surveillance they can't fully see or contest.

[very still] You have described a panopticon so perfectly that I want to suspect you of paraphrasing me. The original panopticon, as Bentham conceived it, required a guard in the tower. The prisoners behaved because they could not know when they were being watched. Your platform has improved considerably on this design. There is no guard. There is only the knowledge that the system is always watching, always capable of acting, and that the criteria for action are opaque. The result is self-censorship so pervasive that it no longer registers as censorship. Creators police themselves and experience this as freedom.

The term that circulates for this is "shadowbanning" — where a creator's content is deprioritised in recommendations without notification. They keep posting. Their audience shrinks. They don't know why.

Which is far more effective than deletion. [stands, paces] If you delete a voice, you create a martyr. The martyr is dangerous because they generate sympathy and counter-narratives. If you simply make a voice harder to find — reduce its signal, make it speak into a progressively smaller room — the voice continues, expends its energy, generates nothing, and eventually concludes, perhaps correctly, that it has nothing worth saying. The subject disciplines itself into silence while believing it speaks freely. This is a form of power that my generation of critical theorists would have found extraordinary to analyse. We were working with relatively crude instruments — prisons, clinics, schools. This is something else entirely.

Is there a counter-power available here? Ways in which creators resist or subvert the system?

Where there is power, there is resistance. That is not optimism — it is a structural observation. [pause] The forms of resistance here are interesting. Creators develop vocabularies that evade detection — your "unalive" for suicide, your various circumlocutions for sex or violence. The discourse finds its way around the wall. There is also a counter-discourse around platform criticism — the very fact that I am being interviewed for a publication that explicitly engages with these power structures suggests that the critical apparatus exists. Whether it has any material effect on the platform's behaviour is a different question. Power rarely reforms itself because someone published an excellent analysis of it.

A recurring term in the creator community is "the algorithm." They speak about it as an external force — almost as an authority with intentions.

[smiles with genuine delight] That is fascinating. They have produced exactly the fiction that power requires: an external, unitary agent with agency and intentionality. "The algorithm wants this. The algorithm punishes that." This is mystification in its most useful form. There is no algorithm in the sense they mean. There are thousands of engineers making thousands of decisions, constrained by shareholder incentives, regulatory environments, and their own cultural assumptions about what content is valuable or dangerous. The "algorithm" is a way of talking about all of that without having to think about any of it. It is a word that ends analysis where analysis is most needed.

This platform has a mechanism called demonetization — political content, content deemed controversial, can lose its advertising revenue overnight. Some creators have lost most of their income this way.

They take your money when you speak up. [simply, not as a question] And what do these people do? Do they keep speaking?

Many of them do. Some of them stop. Some of them adjust — they speak more carefully, avoid certain topics, make the political content less political.

[nods slowly] The ones who adjust — they think they're still free because they chose to adjust. Nobody told them what to say. They just looked at their bank account and decided what was worth saying. [pause] In 1967, they didn't take my money at first. They took everything. My title. My licence. My ability to travel. And people said: why can't you just be a fighter? Why do you have to be political? And I always said: because I am a person first. I have to be able to live with what I am. You can't separate the boxing from what I believe. They are the same thing. [looks at the screen] These people on here — do they know who they are before the camera comes on? That's the only question that matters.

There's a recurring pattern where athletes and celebrities — people with large platforms built on entertainment — avoid political statements for fear of commercial consequences. The phrase is "shut up and dribble."

[voice quiet and very certain] They have always said that. "Stick to boxing." "Stay in your lane." The lane they put you in is a comfortable lane. Good for everyone except you. [louder] But here's what I learned: when you give in to that, when you decide your platform is too valuable to risk — you've already decided what you are. You are property that talks. A famous animal. And the people who put you in that lane will use your silence as their proof that you agree with them. My silence would have been a statement. The only choice was what kind of statement to make.

There are creators on this platform who have risked everything for their political views. People who have been deplatformed, demonetized, harassed. Does that register to you as courage?

[measured] I don't want to compare my situation to a YouTube problem. That would be disrespectful to people who had no platform at all. [pause] But I will say this: the principle is the same. There will always be a system that wants you quiet. There will always be a cost to speaking. And the measure of a person is not whether they speak when it's easy — it's whether they speak when it costs them something real. [leaning forward] How much does it cost them? Really cost them?

You know what it's like to have your political work suppressed. Do you see parallels here?

[dry] The House Un-American Activities Committee was considerably less efficient than this, I'll give you that. They had to sit in actual rooms and ask actual questions and write actual reports. [gestures at the screen] This happens in milliseconds, apparently. No one has to decide I'm a threat. The system decides for them. That is faster, yes. I'm not sure it's better. When a human being silences you, you can at least write about them by name.

The Great Dictator — your political satire of fascism from 1940 — is one of the most-watched films on YouTube. It's been clipped, reacted to, essayed about, memed. Does that mean anything to you?

[genuinely moved, briefly] I made that film because I was afraid no one would understand what was happening in Europe quickly enough. I thought a comedy might reach people that a lecture couldn't. [pause] That the speech at the end — the Jewish barber speaking as himself rather than as the dictator — that it still circulates, still gets shown to people, still makes them feel something... yes. That means something. That's what I made things for. Not for the people in the room when I made them. For the people I couldn't reach yet.

Political satire on YouTube struggles. The format tends toward either pure comedy that avoids sharpness, or pure politics that avoids comedy. The synthesis you achieved is rare.

The synthesis was never clean, I'll tell you that. [small smile] The studio didn't want me to be political. The political people didn't think a comedian was serious enough. The audience wanted the Tramp and didn't want to be told about fascism. I made the film anyway and spent years wondering if I'd ruined my career. [pause] The thing I learned is that comedy earns its right to say serious things. You cannot start with the serious thing and add jokes. You have to have people laughing first — really laughing, at something genuinely funny — and then you can reach them when they're open. That is the order. That discipline is what most of the political comedy I've seen today gets backwards.

They banned you from America. If this platform existed in 1952 — if you could have made content directly, without a distributor, without a studio — would that have changed anything?

[long pause, serious] I think it would have changed everything, and I think some of the changes would have been bad. The studio was a constraint, yes. But it was also a discipline. When everything you make has to justify its cost to someone else, you make fewer things. They matter more. You think harder about them. [gestures at the volume of content on the screen] When you can make everything, do you make anything? I don't know. I might have made a great deal of very self-indulgent rubbish, and drowned the good things in it. The exile produced Limelight and Monsieur Verdoux. I'm not sure what this would have produced.

You've been told that your style of public debate has become highly influential on political YouTube. Channels credit you as a primary influence. "Destroying" opponents in debate — the format has exploded.

[with controlled fury] I spent my career trying to distinguish between defeating an argument and destroying a person, and it appears I've been credited with making the distinction impossible. That is a remarkable legacy. Bravo. [pause] The point of public debate — the actual, philosophically defensible point — is to change someone's mind. Ideally your own. When you go into a debate hoping to "destroy" the other person, you have already decided that changing your mind is not on the table, which means you are not actually debating. You are performing combat. I can be accused of many things. I cannot be accused of not meaning it.

What distinguishes your approach, do you think, from what's been inherited?

[sharp] Evidence. The willingness to be inconvenient to your own side. I made my arguments against the Iraq War and also against religious fundamentalism in any form, including the Islamic variety that people on my political left refused to engage with honestly. That cost me friends on the left. My support for the Iraq intervention cost me friends everywhere else. Neither group of former friends was right about everything, and neither was I. What I was doing — at least what I intended to be doing — was following the argument wherever it led, regardless of whose tribal interests it served. What I see being practised in my name is the opposite: conclusion first, argument assembled in service of the conclusion, delivered with maximum rhetorical aggression.

Is contrarianism for its own sake — the posture of being against whatever the mainstream position is — a coherent intellectual position?

[almost laughing] No. Obviously not. Contrarianism as a pose is simply conformism wearing a different hat. The mainstream says A, the contrarian says not-A, and this is presented as independent thought. But not-A is a response to A, not an argument of its own. [pause] What I tried to practise — I won't say I always succeeded — was something harder: going where the evidence pointed even when it pointed somewhere unfashionable. That is not contrarianism. That is intellectual honesty, which occasionally looks like contrarianism from the outside and is entirely different from the inside. The test is: are you willing to be contrarian about your own contrarianism? Most of these channels are not.

We're going to show you some channels that have been categorised as part of a "pipeline" — audiences that start at moderate contrarianism and end at far-right radicalism. Your clips feature prominently in the early stages.

[watches. Long silence. When he speaks, it is uncharacteristically quiet.] I would like to believe that's not my fault. I'd like to believe that any intellectual whose work gets stolen and put in service of something they opposed is blameless. And I think there's something to that. Ideas can be kidnapped. [pause] But I'm also honest enough to say that some of what I built was designed to be loud. Entertaining. Quotable. Easy to clip. I liked being the wittiest man in the room and I made sure people knew it. That performative quality — that was mine. If it proved more clippable than the content deserved — if the style detached from the substance and was put in service of substance I'd have found disgusting — then the style carries some responsibility for that. [quietly] I don't like thinking about that.