Marc Andreessen on how the internet changed news, politics, and outrage | The a16z Show
Summary
- Andreessen’s master frame is that the internet rebuilt CNN founder Reese Schonfeld’s “Randomonium”: seize whatever is most transfixing, publish fragments in real time, and stay until the next current thing arrives. Social media and streaming now carry the format across X, YouTube, Twitch, and other platforms. “The internet reinvented Randomonium.”
- In Andreessen’s model, the platform turns every real-world event into a viral meme, moral panic, scapegoating cycle, and opposing moral tribes lasting roughly 2.5 days. The previous controversy is rarely resolved; a new one simply absorbs its emotional energy. Between his illustrative April 20 and November 6 election dates sit perhaps 100–120 such cycles, making today’s supposedly decisive issue almost irrelevant to the eventual vote.
- McLuhan’s “global village” has expanded a human-scale direct-relationship capacity of roughly 150 people toward a world of eight billion, creating a predictably “brain-melting experience.” Andreessen’s update to “the medium is the message” is blunt: “If it’s on the internet, it’s a viral social media meme.” Even an alien invasion would become humor, outrage, scapegoating, and opposing moral camps.
- Andreessen’s qualified defense of social media is that rhetorical warfare might be displacing physical warfare. Posters accompanied the Spanish Civil War, radio powered mass politics, and television accompanied Vietnam-era unrest; today, he argues, measured Western political violence is at an all-time low despite ubiquitous online anger. His hypothesis: “Social media wars up, physical violence down.”
- A current thing wins attention through emotional charge, room to argue over facts, tribal formation, and personalization—not through objective importance or scale. One person’s disputed experience can generate as much outrage as harm to 10,000 people because the smaller story supplies a face and leaves room to fight over facts. “The news is called the news, not the importance.”
- Influence operations are real, but “that’s just an op” is an inadequate explanation once a paid or staged intervention becomes an organic movement. Andreessen invokes “availability entrepreneurs” who place a salient event into public consciousness, then rely on resonance to produce an unpaid cascade. He also highlights a disclosure gap: he believes paid shampoo promotion requires transparency and says candidate donations must be disclosed, while paid advocacy on a moral or policy issue may legally remain hidden.
- The investable media shape is a barbell: disposable short-form video at one end and unusually deep, practitioner-led material at the other. Legacy institutions that cannot adapt face continued decline, while three-hour podcasts, seven-hour podcasts, a ten-hour episode on Lex, Substack, long essays, courses, and 30-page AI research outputs demonstrate rising demand for depth. Andreessen thinks Trump remains a TV-internet hybrid; the first president elected entirely through the internet may not arrive until “probably” 2032.
Deep dive
1. CNN’s original product was continuous attention, not a daily bulletin
Andreessen reaches back to Me and Ted Against the World, a book about building CNN with Reese Schonfeld and Ted Turner. Schonfeld had helped wire Turner’s first satellite deal and persuaded the famously controversial founder that an around-the-clock news network could become a viable new media form.
Turner initially imagined a 15-hour channel because he assumed nobody would watch overnight. Schonfeld insisted on 24 hours: somewhere in the world, at every moment, something sufficiently “amazing, interesting, controversial, bonkers, compelling, transfixing” would keep viewers awake.
Schonfeld called the operating system “Randomonium.” CNN would lock onto the current thing, air whatever fragmentary reports, street interviews, or live footage it could obtain, and correct mistakes later: “You just put the thing on the screen.” The incompleteness was part of the real-time product.
The 1991 Gulf War validated the thesis as viewers watched live reporting from Baghdad through bombing raids. CNN later chased that experience through the O.J. Simpson trial and Monica Lewinsky scandal, but Andreessen’s conclusion is that the internet ultimately captured the format: “The internet reinvented Randomonium.”
2. The global village overwhelms human-scale cognition
Marshall McLuhan’s “global village” was a diagnosis, not an endorsement. Andreessen interprets a village as a place where everyone knows everyone, privacy disappears, and neighbors remain permanently “up in each other’s business”; the internet has imposed that social structure on the entire planet.
Human beings may sustain direct relationships with roughly 150 people, the familiar Dunbar number, yet the global village presents “a Dunbar number of like eight billion people.” Everyone can enter one another’s field of attention, criticize them, and demand an emotional response—hence a “really, really brain-melting experience.”
Media therefore changes the experience of reality, not merely its transmission. Andreessen’s thought experiment is Pearl Harbor with Twitter: the same 1941 event would have produced a radically different national experience because citizens would have experienced it through a real-time social-media environment rather than primarily through television and radio.
3. The current thing resets every 2.5 days
McLuhan said that anything placed on television becomes a television show: a polished half-hour comedy or hour-long drama, usually structured as a morality play. Andreessen cites Seinfeld’s “no learning, no hugging” rule precisely because it violated television’s customary ending of moral instruction and reconciliation.
His internet equivalent is categorical: “If it’s on the internet, it’s a viral social media meme.” The event becomes jokes, outrage, accusations, moral panic, scapegoating, and two sides lining up for combat regardless of whether the initiating subject was politics, celebrity gossip, or an alien invasion.
The quantitative claim is a sharp spike followed by half-life decay, with each meme explosion lasting about 2.5 days. It is not resolved; the next current thing simply takes over. After seven days, an issue once treated as existential is “lost to the mists of time.”
That cadence undermines long-range political forecasting. From an illustrative April 20 to a November 6 election, Andreessen estimates perhaps 100 or 120 meme cycles; the vote will turn more on the economy and the panic active near Election Day than on today’s outrage. He says the shape of the media puts behavior “on autopilot.”
4. Online rage may be substituting for street violence
Asked whether more noise still brings society closer to truth, Andreessen answers symmetrically: social media contains “more truth and more lies.” Collapsed gatekeeping makes direct truth-telling easier, while staged content, covert influence, and the native game of trying to “spot the ops” remain pervasive.
His historical comparison links emerging media to mass conflict: color propaganda posters were the new media technology in the Spanish Civil War; radio was the native medium of fascist regimes and also central to FDR; television shaped the Vietnam War and the violent unrest of the 1960s and 1970s.
Against the felt intensity of online politics, Andreessen claims measured political violence is “at an all-time low in Western society.” The contrast matters to his argument: extreme rhetoric on a screen should not be treated as equivalent to people assembling in streets to injure or kill opponents.
His causal account remains explicitly a theory: direct participation in “online virtual combat” may shunt anger away from physical confrontation. “If you’re sitting at home scrolling social media getting mad, at least you’re not out in the street hurting people.” That is his moral defense, not a claim that the online world is healthy.
5. Polarization is older—and bloodier—than social media
The hosts press the familiar case that online tribes estrange people from families and create exclusionary ideological communities. Andreessen’s response is that history has been “sanded off”: people compare the vivid present with a sanitized past interpreted by commentators carrying their own modern agendas.
His best cultural specimen is All in the Family. Norman Lear intended progressive son-in-law “Meathead” to morally educate retrograde Nixon supporter Archie Bunker, yet Carroll O’Connor’s humor and the superheated politics of the 1970s made Archie the star—the show ended up opposite to its intended moral lesson.
The broader record runs through Vietnam, civil-rights confrontations, the Red Scare, two world wars, the Great Depression, the Civil War, European revolutions, and printing-press-era violence. Powerful American men once settled insults through armed duels; labor disputes could end with management firing machine guns at striking workers. Today’s conflict is not historically novel.
6. Centralized media created a temporary era of suppressed volatility
Andreessen does not concede that smartphones caused a general rise in depression. He says Western countries report different trends and claims the apparent increases correlate with benefits or school accommodations attached to psychological diagnoses; his interpretation is that at least part of the effect is “a statistical artifact.”
He does concede that life has felt more intense since roughly 2014. His narrower hypothesis is that the West experienced “suppressed volatility” from the Cold War’s end through 2014, when geopolitical confidence coincided with the lingering dominance of highly centralized media.
Centralization peaked around 1970: many local papers consolidated toward one per city, three television networks dominated national viewing, and radio stations, magazines, and wire services narrowed into a small set of institutions trying to serve broad audiences with an even-handed presentation.
Earlier American media looked more like today’s internet. Cities could support 10 or 15 openly partisan newspapers; Ben Franklin reportedly filled his Philadelphia paper with roughly 15 pseudonymous “alts” arguing against one another. By the 1800 Jefferson-Adams election, newspaper slander exceeded what Andreessen sees on present-day social media.
7. Outrage selects for ambiguity, intimacy, and tribal combat
Andreessen’s distinction is foundational: “The news is called the news, not the importance.” A publication consisting only of consequential developments would fail because audiences seek novelty, heat, excitement, and outrage—not a dispassionate ranking of what matters most.
Successful current things generate sparks like professional wrestling or Real Housewives, then expose an underlying moral question around which tribes can form. The camps attack not only competing claims but one another’s values; at maximum intensity, participants dox opponents and try to get them fired.
Factual certainty can weaken the format because ambiguity leaves more room for argument. Andreessen uses the Jussie Smollett episode as an example whose disputed particulars activated two ready-made camps: one saw confirmation of routine racist violence, while the other saw confirmation of fabrication.
Outrage also does not scale with harm. Something affecting one recognizable person can produce the same or greater intensity than an event affecting 10,000 because “a single death is a tragedy; a million deaths is a statistic.” The personalized case supplies emotional access to a larger moral war.
8. The most viral evidence begins after the story has started
Drawing on George Orwell’s account of wartime propaganda, Andreessen says abstract justifications for war are difficult to sell, so governments search for concrete atrocities involving identifiable innocent victims. The propaganda value may remain high even when the alleged atrocity is false because its function is moral authorization and demonization.
René Girard supplies the psychological warning: scapegoating works because participants forget that scapegoating is a recurring cycle. Each target’s destruction feels uniquely urgent until a new cycle replaces it, so Andreessen recommends preserving “at least a slice of your brain” that recognizes the machinery while emotion is mounting.
For an innocent person caught inside the cycle, eventual truth may not bring restoration. Andreessen points to cancellations built from misreported remarks, imputed beliefs, or stray comments, after which the target is branded morally unacceptable, fired, and “purged from society” before reliable context can catch up.
Viral video has a structural context problem: it usually begins when an incident becomes interesting enough for someone to pull out a phone. Andreessen traces that pattern from Rodney King to the Central Park birdwatcher episode. The essential question is “what happened before that?”—sometimes answerable only when evidence emerges at trial a year later.
9. An op succeeds only when it becomes organic
Andreessen rejects both “everything is coordinated” and “everything is organic.” Some events are genuine, some are operations, and some begin as paid or staged interventions before becoming sincerely held mass movements—which is often the operation’s intended outcome.
Timur Kuran and Cass Sunstein’s “availability cascade” gives him the mechanism. Availability bias overweights what is immediately present in the mind; an “availability entrepreneur” deliberately inserts a person, grievance, or moral event into public consciousness, hoping social momentum will then make it self-sustaining.
His deliberately provocative example is Rosa Parks: Andreessen notes that she was a trained activist rather than the random passenger of simplified school accounts. He treats the organized beginning as compatible with the movement’s profound positive consequences; once broad social change followed, the origin became a historical footnote.
Paid seeding cannot manufacture unlimited demand. A cause still needs to resonate with something people already feel, at which point they may propagate it without coordination. Because they might also have mobilized without the initial op, the counterfactual is unknowable—and “that’s just an op” becomes analytical cope.
10. Influence markets operate inside a disclosure gap
Andreessen contrasts three transactions: he believes an influencer paid to praise shampoo generally has to disclose the advertisement; he says a donation to a federal candidate has to be reported; but payment for promoting a moral or policy position that is neither a product nor a candidate may legally remain invisible.
AI policy is his immediate example. He says some AI-doom advocacy is funded through dark-money influence, while carefully adding that not everyone expressing those beliefs is paid. A newly advertised free fellowship funded by MIRI “and other donors” struck him as the concealed mechanism becoming visible in plain sight.
Bot farms compound the ambiguity: an influencer may benefit from a bot army without knowing who financed it. Andreessen acknowledges a change in his own estimate—“there’s probably more ops” than he had been willing to recognize—but retains the limiting condition that operations achieve mass impact only when real people pick them up.
11. Media is splitting into disposable clips and extreme depth
Andreessen starts with collapsing trust in centralized American institutions, a trend he dates to around 1970 and associates with the “WTF Happened in 1971” meme. Cable ratings, newspapers, magazines, and other legacy formats continue to contract; properties that cannot adapt may simply be managed down by private equity.
The transition remains hybrid. The hosts cite CBS acquiring The Free Press and Mark Halperin bridging legacy commentary with new formats; Andreessen adds The New York Times as a rare successful adapter. Most incumbents, he expects, will not make the transition, while podcasts, social media, and live streaming keep gaining cultural weight.
Yet the future is not uniformly trivial. Short-form TikTok, Instagram, and X video sits opposite three-hour podcasts, seven-hour podcasts, and a ten-hour episode on Lex. The hosts note that creators see surprisingly high completion rates, with many listeners consuming entire multi-hour episodes.
Andreessen calls the structure a barbell and sees “a massive rise in substance” at the long end: Substack, essays on X, online courses, and AI deep-research outputs approaching 30 pages. Practitioner media lets figures such as Andre Karpathy explain their craft directly—a broadly accessible counterpart to Charlie Rose’s once-hour-long interviews airing at midnight.
12. The first fully internet-native president has not appeared
Andreessen is unpersuaded that AI deepfakes create an unprecedented epistemic rupture: “We’ve had deepfakes forever.” He cites the Gulf of Tonkin incident and what he calls the elaborate fake of Russiagate to argue that politics was never a truth-saturated environment awaiting technological corruption.
The historical template is the 1960 Kennedy-Nixon debate: radio listeners reportedly favored Nixon, television viewers Kennedy, making it the canonical first television election. The medium changed which qualities voters perceived and therefore which candidate could win.
Internet tools have mattered without yet dominating the whole contest: Obama used online fundraising in 2008, Trump used social media as broadcast in 2016, and 2024 became “the podcast election.” Andreessen nevertheless argues that none qualifies as the first true internet election.
Andreessen treats Trump as a hybrid: social-media fluent but still intensely attentive to television, able to tweet after disliking a chyron and watch the network change subjects. A true internet president will ignore television and newspapers entirely, existing “100%” online. Andreessen does not know when that happens, but his “probably” guess is 2032.