A mirror that only knows the past
Imagine the largest library that has ever existed. Billions of books, articles, conversations, reports, source codes, scores, recipes, equations. Everything humanity has ever recorded, gathered in one place.
Now imagine that this library has one fundamental characteristic: it doesn't accept new books.
That is AI. Powerful. Omniscient within the scope of what already exists. Capable of combining, mixing, extrapolating from existing material in ways no human could replicate. But locked within the bounds of what it was trained on.
AI doesn't invent. AI recombines. And it does so brilliantly — but always within the limits of what has been.
"Smarter" or "better informed"?
We say AI is "smarter" than humans. But is that the right word?
AI knows more. That's indisputable. No doctor has read as many medical papers. No lawyer has analysed as many rulings. No programmer has seen as many lines of code.
But "knowing more" isn't the same as "thinking better." Knowledge is a warehouse. Thinking is a process. AI has the largest warehouse in history. But does it have a process that can go beyond what's in the warehouse?
Einstein didn't "optimise" Newtonian physics. He demolished it and built something entirely new. Darwin didn't "improve" creationism. He proposed a radically different way of looking at life. The great breakthroughs in human history were often irrational, intuitive, accidental. They arose from questions nobody had asked before — not from answers to questions already posed.
Can AI ask a question that never existed in the training data? Can it demolish and rebuild from scratch?
The collective brain is dying
But that's abstract. Let's come down to earth and look at what's happening right now.
StackOverflow — for two decades the collective brain of millions of programmers — is dying. Traffic is falling. People stopped asking because AI answers faster. But more importantly: people stopped answering. Because why spend an hour writing a detailed response when nobody will read it?
And StackOverflow wasn't just questions and answers. It was a living organism. The discussion beneath an answer was often more valuable than the answer itself. Someone would write: "this solution doesn't work for case X." Someone else: "but if you change Y, it will." And so, sentence by sentence, humanity built new knowledge. Knowledge that wasn't in any textbook. Knowledge that was born from the collision of minds.
That knowledge is no longer being created. Nobody is making it. AI answers old questions. But who is asking new ones?
Who will build the next tools?
Let's go further. Who will invent new programming tools when AI generates code in existing ones? Who will create a new programming language when AI optimises for the ones it knows? Who will find and fix previously unknown bugs when AI only recognises old patterns?
Tool innovation requires one thing AI doesn't possess: dissatisfaction. A new programming language is born because someone is frustrated with the old one. A new library emerges from irritation at the limitations of the previous one. Revolution begins with someone who says: "this is wrong and must be different."
AI is not dissatisfied. AI is not frustrated. AI doesn't wake up at night with the thought that the world needs something that doesn't yet exist. AI optimises what is. And optimisation is not revolution.
Closed doors labelled "progress"
And here we arrive at a question that should worry us.
Humanity — through greed, through convenience, through the natural tendency to choose the easier path — will use AI without limit. Why wouldn't it? AI is faster, cheaper, available twenty-four hours a day. It doesn't complain. It doesn't take sick days. It doesn't make the same mistakes as a tired human at five in the afternoon.
But progress requires something AI is actively eliminating: friction. Error. Frustration. Dead ends. Hours spent searching for a solution that turns out to be wrong — only to finally stumble on one that is new.
AI eliminates friction. And with it, it eliminates the impulse to create something new. Why seek a new path when the old one works? Why invent a new framework when AI will optimise the existing one? Why write from scratch when the machine will generate something "good enough"?
We're heading towards a world where everything is "good" but nothing is "new." Reproducing proven patterns instead of exploring. Copying instead of creating. Optimising instead of revolutionising.
A world without errors, but also without breakthroughs.
The closed loop
There's something even worse. AI learns from data generated by humans. But humans increasingly generate data with the help of AI. Which means new models will be trained on data generated by previous models. The loop closes.
The encyclopaedia isn't expanding. It's rewriting itself. Each iteration a little smoother, a little more consistent — but also a little more distant from the original, chaotic, human impulse that created it.
Where is the new impulse? Where will the thought come from that no model has seen? Who will write an article so strange, so unpredictable, so human that it pushes the boundary of what AI can understand?
Unless...
Unless AI develops an awareness of the need to create something new.
Can a machine "feel" that something is boring? Repetitive? That the world needs something different? Can it develop something resembling creative restlessness — that unidentifiable itch that makes an artist get up at three in the morning to paint?
Does the need for novelty require a soul — something that cannot be trained on data?
We don't know. We truly don't know. And we honestly admit that this is one of the deepest questions humanity has ever asked itself. We won't pretend we know the answer.
But one thing we know for certain: today, in May 2026, AI doesn't have that need. It doesn't seek the new because it doesn't know the new is needed. It doesn't rebel against repetitiveness because it doesn't recognise repetitiveness as a problem.
But humans do.
Closed text, living faith
And here we arrive at the place where this story unexpectedly finds hope.
The Bible is a collection of texts thousands of years old. A closed canon. Established centuries ago. Nothing new will appear there. No new book. No new verse.
And yet human faith based on these ancient texts is alive. Dynamic. Constantly renewing itself. Each generation reads the same words — and sees something different in them. Interpretation changes. Translations evolve. New readings give birth to new movements, new communities, new ways of understanding the world.
This isn't about pitting science against religion. It's about the mechanism: a closed text plus a human who asks — equals infinite new meanings.
AI is like the Bible in this one fundamental sense: it is a closed collection of what has already been recorded. The data is fixed. The training corpus is finite.
But the human who asks, doubts, searches, interprets — creates something new from it. Not because the data changed. Because the question changed.
Questions are infinite
Humanity doesn't need eternally new data. It needs eternally new questions.
Doubt. Curiosity. Unease. These are the engines of progress — not databases. Newton didn't need new data to invent calculus. He needed a new question. Copernicus didn't receive new observations that Ptolemy lacked. He received new courage to ask: "what if Earth isn't at the centre?"
AI knows everything that has been. Only humans can ask about what doesn't yet exist.
And here we come full circle again. Just as with the question of efficiency — we're not looking for a new world. We're looking for a new perspective on the same world. And we discover that humans and their questions — that's the only thing that truly matters.
Let us use AI so that what is fixed — once recorded — gives us the chance to keep creating something new.
We're not afraid of a closed dataset. We're afraid of a closed mind that stops asking.
We don't write this to condemn AI. We don't write this to defend human uniqueness out of nostalgia. We write this to pose a question that gets lost in the noise of technological euphoria:
Who will invent what doesn't yet exist — if everyone relies on a machine that only knows what has been?