The End of the Database Era: How AI Is Freeing Us from 60 Years of Data Obsession

For sixty years we have lived under the dictatorship of the database. Every aspect of our digital existence has been shaped, deformed, and forced into the rigid cages of tables, columns, and relationships. The tax form that forces us to choose between categories that don't represent us, the movie ticket that requires data no one will ever use, the creation of catalogs that transform the complexity of the real world into cold numerical lists, the online purchase that obliges us to navigate through endless filters and forms. Everything, absolutely everything, has been guided by databases.

Until today, when you conceive of a computer system, you start with the database. It's the DB that formally, rigidly, and relentlessly dictates what kind of information will be handled and what will not. And we have adapted, taming ourselves like well-trained dogs. If you forget a piece of data during the design phase, all hell breaks loose—months of work to modify the systems, yes, because every piece of data is correlated with many other pieces of data, with verbose and often contradictory rules. Just think of the rules underlying the creation of a payslip, to give one example.

The Obsession with Perfect Data

This approach has pushed humanity to live a formal reality. We are all obsessed with data. When we go to the city hall for an errand, or when we fill out an online or company form, the data is in command. If a piece of data is missing... we're screwed. The system freezes, the clerk shakes their head, the process stops. In the same way, data is what matters when we consume information: search results ordered by algorithmic relevance, products filtered by predefined characteristics, personalized content based on demographic profiles.

But do our reality, nature, and physical laws work like that? Absolutely not. The real world is an approximate world. We humans forget a lot of things, we have to repeat ourselves, roads are interrupted, they change, and signage is often imprecise, biology and our DNA are not perfect. We are faced with a restaurant menu and often certain dishes are sold out, but the menu was still complete, and in the end, we somehow have dinner and take home a good memory. We order sparkling water and get still water. We often deliberately tell lies. Real life is an imperfect life, and yet it goes on, it works, it adapts.

Despite this, we have been led to believe that the perfect life is a data-centric life. Marriages pop like popcorn, but of course we must have 4 witnesses, rigorously registered in some registry. To what end? What are they for? Yes, yes, there is a reason they exist, but out of millions of marriages, how many times is this reason actually put into practice? And yet marriages fall apart, and nothing is done to improve this trend, they fall apart with the witnesses... how ridiculous we have become.

I once renewed my passport. When I went to pick it up, I was asked for a document to be able to retrieve it. Data sometimes becomes absurd in itself: the passport is supposed to certify who I am; if we doubt this, to the point that I need another document, its primary function fails.

The Absurdity of the Current System

Let's think about how an eCommerce site is built today: you have a DB, you dedicate 20 to 60 minutes to upload every single product into this database, taking images from the supplier's catalog, color, sizes, etc., then you mark the quantities, and you put it online. The customer has to browse, using filters for size, color, etc., through hundreds of pages of products, then put them in a cart and fill out a long and super annoying form.

Years of human labor spent taking data from various sources, organizing it in a database, and then forcing the customer to follow that rigid structure to be able to make a purchase. But it gets worse: each of those sources—the supplier's catalog, the price list, the Excel sheet with sizes—in turn represents months of human labor. Someone had to take information from yet other sources and organize it into databases. It's an endless chain: databases that feed other databases, which in turn feed still more databases.

To make all these systems communicate with each other, we invented a new category of workers: "middleware" developers, those who build digital bridges between databases. The result? We saved hours of employees paid 22 euros an hour by hiring IT professionals paid 110 euros an hour. A perhaps convenient operation, but one that articulates the world even more towards technical and less humanistic figures. The author of this article is one of these software developers, how we enjoy passing ourselves off in society as alchemists of processes that, in hindsight, are now old.

The World Is Changing

We can no longer afford to pay either 22 euros an hour or 110 euros an hour. Databases have served their purpose. Data exists in the world whether someone crams it into a DB or not. There's no need to take a PDF catalog, three Excel sheets, and a WeTransfer and organize them into a complex database. We just need to know that they are there and where they are. Just as there's no need to prepare complex filters and screens to let a lady choose which top goes with her new handbag: the data is there, you just need to ask.

We are witnessing AI agents that, like us, are imprecise, forgetful, not formally deterministic, and yet they are exponentially accelerating our productivity.

The Vector Revolution (and the Answer to Stubborn Programmers)

A programmer might object that it's not true that DBs are disappearing, they are simply transforming: we will have fewer relational DBs (tables of columns and rows) and many more vector DBs. A vector database is a sequence of numbers (typically between -1 and +1) that represents, in a non-precise but approximate way, a concept. The AI sees things, transforms them into vectors of numbers, in an approximate way, then when it has to respond to a request, it also transforms the question into vectors of numbers, using the same rules, then it searches the vector DB for which previously stored vectors are similar and re-transforms them into a meaning for the user (imagine it as looking for the closest dots in a space).

To my programming colleague who might criticize my proclamation that DBs are finally disappearing, I respond bluntly. The relational databases (imagine them as Excel sheets) that have governed the world are, at their core, also numbers, like the vector ones, but with two peculiar characteristics that make them hateful and now obsolete.

The first characteristic is that if the data, in a relational DB, becomes a number (the Bit, 0 or 1), this transformation is only syntactic; it is not loaded with any real added value. I could read 10001110 and 11001110 which differ by only one number, but they could mean totally different things. The semantics (the meaning) are in the code written by the programmer and in their mind. This is why a programming error causes enormous disasters.

In the vector database, on the other hand, two almost similar numbers indicate very similar semantic concepts: 0.0012667 and 0.0012567 indicate that we are talking about almost the same thing.

The second fact is that the numbers in vector DBs are memory in transformation with a changing semantic and are not readable by a human; they are processing material. They do not have a meaning in themselves, as Jean-Paul Sartre would say, but rather a meaning for those who must use them (for themselves). But we humans, who use the information, do not need precise data, but useful data that can be used in a context not necessarily thought, imagined, and budgeted for in advance. This is why those ugly little numbers between -1 and 1, which tell us nothing when read, are so powerful in practice.

Towards a More Human World

We don't need 4 wedding witnesses if there is a priest or a mayor and a full room and a royal banquet. When someone eventually needs to access a testimony, there will be a wide choice. Making 90 people wait in the sun because we have to put the town of birth of a witness on a registry that no one will ever read is a total idiocy.

For this purpose, I recommend reading A Confederacy of Dunces by the American writer John Kennedy Toole: we have created a world without geometry or theology, a world where form has killed substance, where the container has become more important than the content.

Artificial intelligence is bringing us back to a more human, more natural paradigm, more similar to how the world really works. A paradigm where approximation is not a flaw to be corrected, but a characteristic to be embraced. Where information flows freely, without being imprisoned in rigid schemes, where we can finally go back to communicating as human beings instead of as database terminals.

The reign of databases is over. It's time to go back to being human.