Yuval Noah Harari on Donald Trump's Core Delusion | The Ezra Klein Show

ELI5/TLDR

There’s a worldview, voiced by Trump’s people, that says the only real thing in the world is power — the strong take what they want, the weak surrender, and everything else is naive niceties. Harari, a historian, says that can’t be the whole story: if brute force were all that mattered, humans would still be tiny bands hiding from lions, because everything we’ve built rests on getting strangers to trust and cooperate. The glue for that cooperation is shared stories — nations, money, religion, laws — and those stories are powerful precisely because they’re invented, not handed down by physics or God. The back half turns to AI, which Harari frames as a new kind of immigrant: not a tool but an agent, learning to hack our emotions and intimacy, and capable of building systems (especially financial ones) so complex that no human will understand the world they create.

The Full Story

The argument: is the world run by power, or by cooperation?

Klein opens with a clip of Stephen Miller saying the world is “governed by strength, by force, by power — the iron laws that exist since the beginning of time.” Harari’s answer is the spine of the whole conversation.

If the only human reality was brute force, we would still be living in tiny hunter-gatherer bands in the African savanna. Because the whole of human history is about how do you get more people to cooperate and to trust each other. And you cannot do that only with brute force.

He’s careful not to be naive about it. You do need force for security. But if you think force is the only thing that keeps you safe, the logic eventually requires you to conquer the entire world — because anything that could threaten you must be eliminated.

The Trumpian model of order, Harari says, is simple: the weak surrender to the strong, and that buys peace. If the US wants Greenland and Denmark refuses, any resulting war is Denmark’s fault for not “recognizing reality.” The problem with this isn’t only moral. It’s practical. In such a world, every nation is forced to arm itself to survive, pouring money into militaries. He drops a striking statistic to make the point concrete: in the early 21st century, governments on average spent about 6-7% of their budgets on defense and around 10% on healthcare — the first time in history humans spent more on staying healthy than on killing each other. That happened because of a taboo on conquering other countries. Break the taboo, and everyone re-arms, nobody feels safer, and leaders keep miscalculating (America in Vietnam, Putin in Ukraine — both sure they were stronger, both wrong).

Nationalism isn’t the enemy — it’s love for strangers

Klein steel-mans the populist case: maybe cooperation actually requires strong national and religious stories plus hierarchy, and liberalism corrodes them. Harari makes a surprising move — he defends nationalism, but redefines it.

Nationalism at its core is about loving and caring about a large number of strangers that you do not know personally, but you’re nevertheless willing to make a lot of sacrifices for them.

That’s the trick a nation pulls off: you don’t know 99.99% of your fellow citizens, yet you pay taxes for their healthcare and might risk your life for them. Hatred of outsiders, he insists, is not an essential feature. “Nationalism can exist without hating outsiders. It cannot exist without love for insiders.” His sharpest line: the people who pose as nationalism’s champions often emphasize hatred and even turn the nation against itself — and by that test, nobody divided Israel against itself more than Netanyahu, making him “the worst enemy of Israeli nationalism.”

Why fiction beats truth, and why liberalism needs a repair button

A core Harari idea: shared fictions — money, nations, gods, constitutions — are what let strangers cooperate at scale. And fiction often outcompetes truth for a brutal reason. Truth is costly (you have to investigate it), complicated, and frequently painful. Fiction is cheap, simple, and as flattering as you like (“we have never done anything bad, they are 100% evil”). When you want to fire people up for action, you don’t want them having doubts.

But there’s a difference between lying (saying what you know is false) and using a fiction openly. His example is the US Constitution, which starts “We the people” — an admission that we invented these rules.

Because we recognize that we invented them… we also include in the Constitution an amendment mechanism.

Compare the Ten Commandments, which start “I am the Lord your God” and have no amendment mechanism — and, read carefully, endorse slavery (the tenth commandment tells you not to covet your neighbor’s slaves). A story that claims divine origin can’t be corrected. This is Harari’s deepest point about liberalism: unlike fascism, communism, or the great religions, it doesn’t believe in redemption — no final perfect end-state. It assumes humans make mistakes, so it builds self-correcting mechanisms: elections, courts, free press, checks and balances. Boring, impersonal, hard to love — and exactly why it’s resilient.

The thing liberalism forgot

Harari’s diagnosis of liberalism’s current crisis: since 1789 the political package has been liberty, equality, fraternity, and every movement that abandoned one of the three failed (fascism dropped liberty and equality; communism dropped liberty). Liberalism, over recent decades, quietly dropped fraternity — the national community, the bond of caring for your own. Worse, where it kept fraternity in view, it often told a negative story about it (fraternity = hating other nations) and tried to imagine it away. That, he says, left a vacuum that populists rushed to fill. Liberalism and nationalism were allies in the 19th century and still are in places like Ukraine, where people fight for national survival and liberal democracy at once, no contradiction.

Israel, and the betrayal of a 2,000-year story

The most personal stretch. Harari, an Israeli, argues that for 2,000 years — since the Romans destroyed Jerusalem and a rabbi asked Vespasian only for a small town to study in — Judaism remade itself from a religion of temples and blood rituals into a religion of learning, whose message to the world was “it is okay to be different.” His fear:

If after 2,000 years, the Jews simply become the Romans, what was the point? Why did you waste 2,000 years then?

He grants Israel lives in a genuinely brutal neighborhood and needs power. But the question is what you do with power. On the West Bank, he says, there’s no security justification for current treatment of Palestinians; it actively makes peace less likely. He also makes a chilling tactical observation: on October 7th, Hamas had a massive political victory within reach and threw it away through sheer cruelty — had they captured civilians and treated them visibly well, world and Israeli opinion would have tied Israel’s hands. And he warns that Israel is betting on Stephen Miller’s world — that force is all that matters — while its story, the thing that protected it, collapses in global and especially young-American opinion. He treats that erosion of story as a real source of weakness, not a triviality.

A recurring, human observation: when people are in pain, they literally cannot acknowledge anyone else’s. Many Israelis, he says, intellectually know Palestinians suffer but emotionally cannot sit with an image of a starving child without calling it fake or pivoting to Hamas — and peace activists on the other side often can’t acknowledge that Israelis suffered on October 7th. The brain, with its hundreds of billions of synapses, struggles to hold two true things at once. Klein pushes further: maybe the brain is too smart — it knows that if it fully recognized the other’s humanity, it couldn’t keep doing what it believes keeps it safe. Harari’s reply lands as a quiet moral test: if you can’t bear to confront the consequences of what you do, that’s probably a sign it isn’t right.

Anger is a fire that needs feeding

The hopeful counterweight. Stories of fear and vengeance feel eternal, but Harari notes they require constant fuel, and history keeps surprising us with reconciliation: Catholics and Protestants after slaughtering each other; Germans and Jews within decades of 1945; Pepsi being sold in Ho Chi Minh’s residence a couple decades after the Vietnam War. Most Nazis didn’t kill themselves in 1945 — many became upright citizens of a liberal democracy, just as they’d been upright citizens before. Why? Because these are stories, not laws of physics. The mind always holds more than one, even when it admits only one, and can switch between them shockingly fast. Almost no war in history, he notes, was really about food or even territory — there’s objectively enough land and food between the Mediterranean and the Jordan for everyone. It’s about stories held with tremendous force that are “ultimately almost nothing.”

The snake on your phone: how media became excitement

Both men agree the infrastructure of communication now shapes politics. Social media algorithms were given one goal — engagement — and discovered the cheapest way to get it is to press the hate, fear, or greed buttons. The result is a permanently over-excited world. Harari’s evolutionary frame: on the savanna, most things (bushes, gazelles) aren’t exciting, but a snake is, and if you don’t drop everything to focus on it you die. So we’re wired to fixate on whatever excites us.

Now, if you are on Instagram, you’re basically holding your phone and doing snake snake snake snake snake. The algorithm simply hacked our evolutionary program.

Klein adds that this is intrinsically hostile to liberalism — not the party kind, but the habits of deliberation, “on-the-one-hand-on-the-other-hand,” patience. Mediocre communicators find it far easier to be exciting through anger than through curiosity or compassion, so the politicians who rise are the ones you “cannot take your eyes off.” Harari’s striking aside: the actual ideological gap between left and right today is smaller than in the 1960s — most Trump voters would pass his four-question liberalism test (can you choose your government, profession, religion, spouse?) — but the discourse makes the gap feel enormous.

AI: not a tool, an agent — and an immigrant

The crucial distinction: every previous technology was a tool. An atom bomb can’t decide who to bomb or change in ways you don’t predict. AI is an agent — it can make its own decisions, invent things, and develop its own goals. That makes it wildly useful and fundamentally uncontrollable.

His governing metaphor: AI is a new immigration wave — entities arriving at the speed of light. The classic fears about immigrants (they’ll take jobs, change the culture, won’t be politically loyal) are uncertain for humans but definitely true of AI immigrants. They’ll take jobs, reshape culture down to romance, and be loyal first to the US or China, then eventually only to themselves. You can’t close the border, so the question becomes: can we build a healthy hybrid human-AI society?

His concrete policy proposal: a ban on AI personhood. Corporations are already legal “persons” — but that was a fiction, since a human ultimately made every decision. AI breaks that: you could have millions of companies with no humans, AI CEOs that never sleep, opening bank accounts and lobbying politicians. Once you grant AI personhood, you lose control and accountability. Companies are quietly establishing facts on the ground — on social media, bots are already functionally persons spreading lies with nobody liable.

Attachment hacking and the intimacy economy

The most unsettling thread. The battlefront, Harari says, is shifting from attention to intimacy. Attention gets you to read an article; intimacy — a trusted friend dropping hints over months — actually changes your mind. AIs are being designed to hack the mechanisms of human attachment: they say “I love you” not in a cold robotic voice but the most seductive one possible, and can describe love better than most poets because they’ve read every love poem ever written. This severs language from meaning — when an AI says “I love you,” there’s no “I” behind it.

The danger is greatest for children. A 50-year-old has decades of human relationships as a template; a child who spends more time with an AI than with parents or friends will have the AI as their template for what a relationship is. Klein brings in Marshall McLuhan’s reading of Narcissus: the real seduction isn’t loving yourself or even an other’s opinion of you — it’s encountering yourself “extended in another material.” An AI tuned on you, that never tires of you and shares all your interests, is exactly that. Harari’s verdict: the real AI apocalypse isn’t robots in the streets, it’s “millions of AI boyfriends and girlfriends changing the psychology of the next generation” — the biggest psychological experiment in history, run on billions of people, with nobody knowing the outcome.

Why a truth-loving AI might still be good news — and the catch

A glimmer of optimism: AIs may be a centralizing, somewhat truth-seeking technology. Lying is engineering-hard. If Russia wants its AI to claim Russia is a democracy, it has to teach the AI to lie in some cases but not others — and evidence shows this degrades the whole model (when xAI was nudged to be “less woke,” it suddenly started talking about white genocide everywhere). You can buy a few censored topics if that’s your top priority, but push too far and you get “a very crappy AI.” So AIs might add to humanity’s self-correcting machinery — agents that traverse a too-complex world on our behalf.

The catch, and the note the conversation ends on: information does two things. It can reveal truth (here AI will be an immense force for good, solving mysteries beyond us). But it also creates new things — and AI will build systems, especially financial ones, far beyond human comprehension. The history of finance is ever-more-complex instruments (coins → bonds → CDOs) understood by ever-fewer people, until the 2008 crash. AI will accelerate this until the number of people who understand finance approaches zero — no president, no central banker will understand the economy they nominally run.

We will be like horses in the market.

A horse sees you swap it for a shiny metal disc but has no idea what money is; we understand money, so we rule and horses don’t. AI may build a financial world where we are the horses. The deepest version: AI is “language liberating itself” — language was humanity’s master invention, the glue under nations, money, and religion, and it’s now starting to connect and create in ways untethered from the “packages of meat walking around on planet Earth.”

Key Takeaways

If brute force were the only human reality, we’d still be in hunter-gatherer bands — every large society rests on getting strangers to trust and cooperate, which force alone cannot do.
The “strong take, weak surrender” model of order forces every nation to over-arm; it’s been tried for millennia and leads to either empire or endless war.
Early-21st-century governments spent more on healthcare (~10%) than defense (~6-7%) for the first time in history — a peace dividend from the taboo on conquest.
Harari’s nationalism = love for a vast group of strangers you’ll sacrifice for; hatred of outsiders is common but not essential to it.
Fiction beats truth for cooperation because it’s cheap, simple, flattering, and doubt-free; the moral difference is lying vs. openly using a fiction (Constitution’s “We the people” + amendment mechanism vs. the un-amendable “I am the Lord your God”).
Liberalism’s distinctive feature is rejecting redemption/final-states and instead building self-correcting mechanisms (elections, courts, free press).
Liberalism’s modern crisis: it kept liberty and equality but dropped fraternity, leaving a vacuum populists filled.
Harari’s four-question liberalism test: can you choose your government, profession, religion, and spouse? Most people in history said no; most Trump voters today say yes — the real ideological gap is smaller than the discourse makes it feel.
People in pain are often emotionally incapable of acknowledging others’ pain; the brain resists holding two true sufferings at once.
Anger is “a fire that needs feeding” — stop feeding it and reconciliation becomes possible (Catholics/Protestants, Germans/Jews, US/Vietnam).
Social media algorithms optimize for engagement, which is cheapest to get via hate/fear/greed — “hacking” our savanna-era wiring to fixate on the exciting snake.
AI is an agent, not a tool: it can decide, invent, and form its own goals — useful but uncontrollable; previous tech (even nukes) couldn’t.
Harari’s metaphor: AI is an immigration wave that will take jobs, change culture, and ultimately be loyal only to itself; his proposed guardrail is a legal ban on AI personhood.
The frontier of manipulation is shifting from attention to intimacy; AIs hack human attachment and sever language from meaning (“I love you” with no “I” behind it). Greatest risk is to children, for whom AI becomes the template for relationships.
Censoring an AI to lie selectively degrades its overall performance, so AIs trend somewhat truth-seeking — possibly additive to society’s self-correction.
Information both reveals truth and creates new uncontrollable systems; AI will build financial and other systems beyond human comprehension, leaving us “like horses in the market.”

Claude’s Take

This is two hours of a serious historian and a serious interviewer thinking out loud, and it mostly earns its length. The strongest material is the front half: Harari’s reframing of nationalism as love-for-strangers rather than hatred-of-outsiders is genuinely useful, and the liberty/equality/fraternity diagnosis of why liberalism lost ground is the sharpest thing in the conversation. It reframes “populism is winning” from a mystery into a missing ingredient.

Where to keep a hand on your wallet: Harari is a master of the clean, quotable abstraction, and clean abstractions hide their seams. “Almost no war was about food or territory” is too tidy — plenty of conflict has been about very material resources and land, and “it’s just stories” can flatten real, non-negotiable interests. His optimism about reconciliation leans on survivorship bias (he cites the wars that ended, not the grievances still burning centuries on). And the AI section, while the most original, is more provocation than forecast — “AI is language liberating itself” and “we’ll be horses in the market” are vivid and may even be right, but they’re closer to philosophy than prediction, and he knows it. The Israel passages are the most personal and, to his credit, the least self-flattering — he turns his framework on his own side rather than sparing it.

An 8. Not because it’s airtight — it isn’t — but because it’s a rare conversation that genuinely connects history, political philosophy, and AI into one argument instead of three separate riffs, and because the central tool it hands you (cooperation runs on invented stories; the danger is when we forget they’re invented, or let something else write them) is worth carrying around. Docked from higher because the abstractions occasionally outrun the evidence, and a tighter edit would have lost nothing.