We are living through a Gutenberg-level transformation, yet most people cannot tell you what happened during Gutenberg in the fifteenth century. Most know the broad history: the Renaissance, the Protestant Reformation, the European Enlightenment. But the best analysis was written by Victor Hugo in Book Five of Notre-Dame de Paris. Published in 1831, more than a century before Marshall McLuhan was born, Hugo’s entire analysis can be summarized in just six words: “The book was to kill the edifice.”
What Hugo describes is how writing displaced the cathedral as the dominant institution of Europe. Prior to Gutenberg, writing was expensive and scarce. As a result, few could read or had access to writing. The church was the interface between text and masses. If you wanted to read an important document, like the Bible, you needed a priest to translate it from Latin into the common tongue and interpret it for you.
After the printing press thought became “more imperishable than ever.” Hugo writes that:
“It is volatile, irresistible, indestructible. It is mingled with the air. In the days of architecture it made a mountain of itself, and took powerful possession of a century and a place. Now it converts itself into a flock of birds, scatters itself to the four winds, and occupies all points of air and space at once.”
The printing press took a previously scarce resource and distributed it to the masses. In doing so, it took power from the cathedral, and by extension the church, and shared it with everyone. Everyone had access to text and could interpret it for themselves. From this, the Renaissance, the Protestant Reformation, and European Enlightenment naturally followed. Now, AI does the same.
Gutenberg’s transformation was not peaceful. Much of Europe’s population died in religious wars as the most powerful institution in the world tried to hold on to its monopoly on knowledge. Yet at the same time, fortunes were made. Art and science flourished. The concept of human rights rose out of this redistribution of knowledge. In historical hindsight, what the population experienced as brutal change was the growing pains of human civilization advancing. Understanding this Gutenberg-level transformation could be the difference between dying in war or ascending as one of the greats of history. Maybe both, if you’re Galileo.
What is the cathedral of our time? When I asked various AIs this question, the most common answer was the document. AI allows you to chat with documents. Now, instead of reading books you can ask an AI to read them for you and apply them to your specific use-case. Many already use AI this way, myself included. However, this answer is simplistic, obvious, and wrong.
The internet already killed the book. Long before LLMs existed, we had search engines, Wikipedia, and social media to answer any information we wanted to find. AI accelerates this change by allowing you to more intelligently search documents, but that is a continuation of the internet revolution, not a new use. Plus, the document still matters as source material for these searches. If anything, the document has more power when presented by an AI as truth without attribution, than when it arrived as a separate webpage where one could evaluate the source.
What AI changes is how you interact with the document. Previously, gathering and acting on knowledge were two different tasks. Just because you read a legal article did not mean you could draft a contract. Reading your labs did not mean you could accurately prescribe treatment, having an idea for an app did not mean you could program it, knowing what style you wanted did not mean you could paint a picture, compose a song, or direct a film. You needed a professional, a lawyer, a doctor, a programmer, an artist, etc. to bridge the gap between your intention and your skill.
AI closes that gap. Now, if you can accurately describe your legal situation, medical data, app idea, or artistic project, AI can create for you. Whereas the internet democratized knowledge, AI democratizes skill. AI doesn’t kill the document, it transforms it into action. AI is agentic. It redistributes agency.
If AI redistributes skill and agency, what cathedral previously captured those? It used to be if you wanted to draft a contract, you needed a law firm. If you wanted a medical diagnosis, you needed a medical practice. If you wanted software, you needed a development team. Films required a film studio, research required universities, journalism required newsrooms, marketing campaigns required advertising agencies — we could go down the list, but all of these have one thing in common: they are institutions.
The institution was the cathedral that previously housed skill. Since complex tasks required skill beyond the average individual, the institution was the intermediary between desire and the knowledge, skill, and coordination to accomplish that task. Even if one was skilled in a particular task, they still needed other skilled professionals for the areas where they were not skilled. Now, they don’t.
This is not hypothetical or theoretical. I am doing this right now. I have personally used AI for research, investigative journalism, music, short films, animation, comics, art, websites, apps, editing, and analyzing legal and medical documents. I am not an anomaly. Among my friends interested in AI, I have below average skills.
Consider what it would cost to do just one of those projects through an institution. It would bankrupt me. Most would require at least a six-month commitment. The time and money required would force me to deliberate over which project to pick. Most would never be made. With AI, it takes less time to make an idea than I would spend debating which idea to create without AI.
AI does not provide better expertise or quality than the best professionals yet. However, it provides sufficient capability at a low enough price to bridge the gap institutions exist to solve. The printing press did not produce better illuminated manuscripts than the monks did. It produced good enough books, fast enough, at a scale that made the monastery irrelevant.
AI removes the gap between intention and action. Many of these projects don’t go far, but most readers didn’t become Galileo either. The point is that without AI they would not exist at all. Before the printing press, you went to the cathedral and the priest could read you the good book. After the printing press, you could read the book yourself. Before AI, you went to the institution and it could do things for you. After AI, you can do those things yourself. When agency is dispersed like literacy, the cathedral is bound to crumble. In short: AI kills the institution.
Hugo’s history reveals the institution as an anomaly. Print freed thought, but it could not free action. You could read a legal statute, but you could not practice law. You could read a medical text, but you could not treat patients. Knowledge escaped the cathedral, but got re-centralized into professional guilds, credentialing bodies, and bureaucratic hierarchies that controlled who could turn knowledge into action. The institution was a regression, a re-consolidation of power after the book had scattered it. For five centuries, the cathedral rebuilt itself in a new material. AI is the technology that finishes what Gutenberg started. Now, knowledge can act.
Given that Europe was plunged in religious wars when the priests tried to hold on to power, it is worth considering how the current cathedral owners will react to being displaced by AI. In Gutenberg’s time, the church held a monopoly on knowledge. In our time, powerful institutions held a monopoly on agency. They could fund films, marketing campaigns, legal teams, media outlets, investigative journalism, academic writing, and software products. You could not — until now.
Imagine you were Pope in 1455. If you knew what we know now about the power of the printing press, what would you have done? A church simply interested in maintaining power would hang Gutenberg, smash every press, and drive humanity back into illiteracy. At minimum, they would nationalize printing and make it so only those licensed by the clergy could own and operate it.
Instead, the church funded the technology. The church was Gutenberg’s first client. Their money grew the printing industry. They called printing a “divine art” and assumed it would strengthen their position by allowing them to print more Bibles. It did not occur to them that this also meant anyone could print Bibles, interpret those Bibles for themselves, or print books which went against their interpretation of the Bible. They just thought they could replace their scribes.
Institutions adopting AI are making a similar error. AI growth is funded by enterprise clients. CEOs of these institutions speak about AI with the same euphoria bishops spoke about the printing press. They are pushing AI adoption with the assumption it will make money, cut costs, and replace employees, not realizing that their employees can use the same technology to replace the institutions that employed them.
Censorship of the printing press did not begin until thirty years after the technology was introduced. When it began, it was local, reactive, and aimed at specific books, rather than the technology itself. Cologne University was granted censorship powers in 1479. Würzburg enacted book censorship in 1482. The Archbishop of Mainz followed in 1485. By the time these orders were written, printing presses had spread to over two hundred European cities.
AI opposition is following a similar trajectory as individual rights holders and licensing cartels sue to protect their local province. Disney, Universal, Warner Bros., the New York Times, and various other publishers have all sued AI companies for copyright violations. Illinois, Nevada, and Utah banned AI from providing therapy. New York is moving to ban AI from giving medical or legal advice. At best, these slow AI. None of them stop it.
If history continues the same, we know what comes next: Martin Luther. A single man disrupts the dominant institution by printing at scale in the common language. He uses a new medium called the pamphlet, cheap short-form “slop” enabled by new technology, to flood Europe with his content. The transition killed millions and established the modern world. By the time the church published Index Librorum Prohibitorum, a formal list of forbidden books in 1559, thirteen years after Luther’s death, it was too late.
For those looking to regulate AI, it might already be too late as well. Open source makes every laptop a printing press. Even if AI were controlled now, the modern cathedral has an enemy the Catholic church never had: China. China has strategic reasons to offer the West unregulated open-source models. American tech companies have not been able to prevent their models from leaking. The modern Luther already has local compute.
Modern priests would respond to this analysis that institutions do more than provide agency. They confer legitimacy, trust, and authority. One could say the same of the Catholic church in Medieval times. The same critique applies in both cases. The church selling indulgences is what prompted Martin Luther to write his 95 Theses. Modern corruption drives the desire for parallel institutions. The Catholic church had lost legitimacy before Luther published. Likewise, modern institutions had a “legitimacy crisis” before AI. AI and the printing press are the technologies that allow response to the crisis.
AI accelerates the legitimacy crisis. Like anyone with a terminal illness, the institution will wither for a long time in hospice before being pronounced officially dead. Credentials remain, but deference to them will not. You might be legally required to use a lawyer or doctor for certain cases or prescriptions, when the public realizes the AI in their pocket is superior to a professional in most use-cases, they will view those professions as a barrier to services rather than provider of them.
However, the failure of institutions does not automatically confer legitimacy on their successors. The priest class is correct that AI can be used for false information, just as they were correct when they made the same criticisms of the printing press. One of the first bestsellers of the printing press was Malleus Maleficarum, a witch-hunting manual published in 1486, that reads like a viral shitpost. In one passage, a witch steals men’s penises and keeps them in a nest where they move around and eat oats. When a victim tries to pick a replacement from the nest, he is told he cannot have the biggest one because it belongs to the parish priest. Despite being officially condemned by church, the book contained front matter that made it appear officially endorsed. Tens of thousands of women were tortured and executed over the next two centuries using procedures codified in it. Modern AI deepfakes of politicians are equally false, often just as hilarious, and have the same potential for misuse.
The priest class is right that we still need coordination, accountability, and trust, not just agency. In Medieval times, they would have proposed the church as the source of such solutions. The printing press found superior solutions in the form of the scientific process and peer-review. It created decentralized authority where anyone could check the claims of another, rather than accept them on faith. AI has the potential to generate superior solutions that do not look like the architecture of an institution.
Just as priests are concerned with their position, modern peasants are as well. One of the most common questions professionals have is whether or not AI will take their job. It will. More than that, AI will take what gives you a job: the institution. Your job was never yours. It was a role offered by the institution. Now, that role will be granted by what replaces it. AI will take the entire concept of a job.
Ultimately, no objections matter, because the modern Luther won’t win through argument, but adoption. The previous Gutenberg revolution was about publishing. The AI revolution is about building. This means the modern Luther will be someone who uses AI to replace or decentralize institutions that were previously cathedrals at the center of the world. Media distribution will be a part of this revolution, but action that utilizes the agentic power of AI will drive it.
The most obvious application of this analysis is: become your own institution. Look at where you previously paid institutions for agency, rather than trust, relationship, or connections, and replace them with your own systems. You can start a one-person film studio, newsroom, app developer, small press — whatever you want to be. Many are already doing this, myself included.
While we could end there, history will not. Though it would be good advice to get into the Bible printing business in the 1500s, the printing revolution did not stop at Bibles. Victor Hugo wrote about it three hundred years after the Reformation, with a view as far from those who lived through it as a person from 2300 would be from us. He describes printing as the greatest of liberated arts. He says that “all the life which is leaving architecture comes to it” and “all human thought flows in this channel.” He imagines that “if all the volumes which have issued from the press since Gutenberg’s day were to be piled one upon another, they would fill the space between the earth and the moon.” He concludes that “it is the second tower of Babel of the human race.” Now, that tower of Babel is what we call training data. Through AI, the tower speaks.
The first printing presses made Bibles. Then, a church insider printed his own commentary on the church. Within a century, religious texts were the minority of printed material. Printing became dominated by the novel, a new artform only possible due to Gutenberg’s technology. The novel gave access to the inner world of other people. Historian Lynn Hunt argues in Inventing Human Rights that the rise of the epistolary novel created the capacity for empathy across social boundaries. Once readers could inhabit the inner life of those different from them, the separation between them dissolved. Public torture went from entertainment to unthinkable when audiences could empathize with those punished. The human rights revolution soon followed.
Victor Hugo participated in this movement. Novels of the period incorporated lengthy digressions, interspersing non-fiction commentary chapters in the fictional story. Hugo’s commentary on the printing press appears inside an encyclopedic novel. Notre-Dame de Paris humanizes the outcasts, the deformed Quasimodo and gypsy Esmeralda. It portrays the church as a hollow institution that no longer shows compassion to the kind of characters Christ spent time with and uses the novel to fulfill that mission. In our time, institutions no longer serve their intended purpose. The legal system does not provide justice, the medical system does not heal, and the media does not tell the truth. AI can fulfill the underlying intention of these institutions.
No one living in the 1500s could have predicted the novel. Even if they had, none could have foreseen that it would end public torture and make human rights self-evident. The printing press did not just create a new artform. That artform produced a new society and made intolerable what the previous society accepted as normal. Somewhere in our present there are practices as brutal and routine as public execution was in the 1500s. The AI revolution will not stop at better institutions. It will produce forms we cannot imagine, and those forms will transform society. The ways we organize agency and people will become as different from the institution as the novel is from the cathedral.