Key Ideas from Nexus
Yuval Noah Harari's Nexus is a history of information networks — from prehistoric oral traditions to the printing press to the internet — and an argument about what happens when those networks encounter artificial intelligence. It's a bigger, more urgent book than it might sound. Harari's central claim is that we are making a catastrophic mistake in how we think about information, and that AI is about to make that mistake much harder to recover from.
The Naive View of Information
Harari opens by identifying what he calls the "naive view" of information: the assumption that more information leads to better understanding, better decisions, and a better world. It feels intuitive. Of course knowing more is better than knowing less.
But history, Harari argues, tells a different story. Some of the most information-rich societies have also been some of the most capable of mass delusion, genocide, and self-destruction. The Nazis had newspapers, radio, and sophisticated propaganda infrastructure. The Soviet Union had a vast bureaucratic information network. More information, in both cases, was weaponised to coordinate catastrophic outcomes rather than prevent them.
The naive view confuses information with truth. They are not the same thing. Information is a signal. What that signal contains, how it travels, who controls it, and what it's designed to do are entirely separate questions — and they matter far more than the raw quantity of information in circulation.
Information Networks Shape Civilisations
One of Nexus's most interesting arguments is that the major turning points in human history have been driven not by leaders or technologies per se, but by shifts in how information flows.
The invention of writing didn't just let people record things — it created bureaucracy, which allowed empires to coordinate across distances that oral cultures couldn't manage. The printing press didn't just spread ideas — it destabilised the Catholic Church's monopoly on religious authority and enabled both the Reformation and a century of religious wars. The telegraph and mass media didn't just accelerate communication — they made modern nationalism possible by giving millions of strangers a shared sense of identity built on shared information.
Harari's framework is that information networks are the substrate on which civilisations are built, and that changing the network changes everything built on top of it. The internet is the most recent and most dramatic such shift. AI is the next one — and it's categorically different from all the previous ones.
AI Is Not a Tool
This is the pivot on which the whole book turns. Every previous information technology — writing, printing, radio, the internet — was a tool. It amplified human communication, but humans remained the agents. People decided what to write, what to broadcast, what to publish. The technology was passive.
AI is not passive. For the first time in history, Harari argues, we have created an information-processing entity that can make decisions autonomously, set and pursue goals, and act in the world without a human directing each step. This is a qualitative break from everything that came before.
The distinction matters enormously. A printing press can spread a dangerous idea, but it cannot decide to spread it. An AI system can. It can identify targets, craft messages, adapt its approach based on feedback, and pursue objectives across millions of interactions simultaneously — without any human being aware of or approving each decision.
This doesn't require the AI to be sentient or to "want" things in any deep sense. It only requires that it have objectives and the capacity to pursue them. Current systems already meet that bar.
Intelligence Without Consciousness
Harari spends time on a distinction that gets surprisingly little attention in most AI discourse: the difference between intelligence and consciousness.
Intelligence, as he uses it, is the ability to process information and solve problems. Consciousness is subjective experience — the sense of what it is like to be something. These are not the same thing, and they don't necessarily travel together.
AI systems are becoming extraordinarily intelligent by any useful definition. Whether they are conscious — whether there is anything it's like to be them — is a genuinely open question that nobody currently knows how to answer. Harari's point is not to resolve this but to highlight that we are building enormously powerful systems whose inner nature we don't understand, and deploying them at scale anyway.
The Threat to Democratic Institutions
One of the book's most politically pointed arguments is about the relationship between AI and democracy. Democratic systems depend on a set of conditions: citizens who can form genuine views, a public sphere where those views can be expressed and contested, and institutions capable of aggregating preferences into legitimate decisions.
All three of those conditions are under pressure from AI-enabled information manipulation. When it becomes trivially easy to generate personalised persuasive content at scale — to show each person a slightly different version of reality tailored to their existing beliefs and emotional vulnerabilities — the idea of a shared public sphere starts to break down.
Harari points to the ways authoritarian governments have already weaponised social media and targeted information as a preview of what becomes possible when AI dramatically lowers the cost of that kind of manipulation. The Silicon Curtain he describes is the risk that AI becomes the most powerful tool ever built for controlling what populations believe — not through force, but through the architecture of information itself.
The Problem of Checks and Balances
The final thread of Nexus is about institutional design. Human societies have spent millennia developing systems to prevent catastrophic concentrations of power — courts, constitutions, separation of powers, free press, elections. These systems are imperfect, but they encode hard-won understanding about what happens when single actors gain unchecked control.
AI is developing faster than those institutions can adapt. The companies building the most powerful systems face limited oversight. The governments capable of regulating them are largely composed of people who don't understand the technology. And the competitive dynamics — between companies, between nations — create strong incentives to move fast and worry about consequences later.
Harari is not fatalistic. He is insistent that the choices made in the next decade are genuinely consequential, that different choices produce different outcomes, and that the current trajectory is not inevitable. But he's also honest that the window for course correction is not unlimited, and that the naive view of information — the belief that more intelligence, more connectivity, and more powerful AI will automatically make things better — is exactly the wrong framework for this moment.
Nexus is the kind of book that reframes questions you thought you already understood. Harari isn't predicting the future — he's trying to give you a clearer picture of the forces shaping it, so that the choices ahead look less like fate and more like choices.