The Medieval Shortcut to Genius
I was digging through some old occult history when I stumbled upon a 13th-century manuscript called the Ars Notoria. It wasn't a book of spells to turn lead into gold or summon demons. Instead, it was a manual for the impatient scholar. The premise was simple: if you stared at specific geometric figures (notae) and recited the right prayers, God would infuse your mind with the complete knowledge of the liberal arts. No years of grueling study required. No slow, painful memorization. Just a divine download directly into the cerebral cortex.
It struck me that this is exactly what my Twitter feed looks like today. We've replaced the monastic cell with a browser window and the holy figures with a chat interface, but the underlying desire is identical. We aren't just looking for tools to help us work; we are hunting for a way to bypass the cognitive friction of becoming someone who actually knows things. We want the result without the process.
There is something deeply human about this search for a cheat code. We have always hated how slow our brains are. It took us thousands of years to go from the Ars Notoria to Large Language Models, but the psychological itch hasn't changed a bit. We are still looking for that specific sequence of words—the perfect prompt—that will unlock a hidden door to mastery.
The Ritual of the Prompt
Watching a "Prompt Engineer" work today feels remarkably like watching a medieval alchemist. They iterate on specific phrases, adding a dash of "act as a world-class expert" here and a pinch of "think step-by-step" there, hoping the latent space will yield the gold they're looking for. It is a ritualistic interaction with a black box. We don't really know why "take a deep breath" makes the model perform better on math problems, but we do it anyway because the ritual seems to work.
This makes me wonder if we are developing a new kind of literacy that has nothing to do with understanding the subject matter. If I can prompt a model to write a flawless Python script for a neural network, do I "know" Python? In the medieval view, the answer was yes—the knowledge was infused. But in our world, knowledge has always been structural. It’s built like a house, brick by brick. If you skip the bricks and just manifest the house, what happens when the wind blows?
Photo by Sabine Fischer on Pexels
We are essentially outsourcing the "working memory" phase of learning. When I spend three hours struggling to understand a concept, my brain is physically reconfiguring itself. When I ask an AI to summarize it in three bullet points, my brain stays exactly the same. We are getting the information, but we might be missing the transformation that usually comes with it.
What Happens When the Grimoire Closes?
I’m curious about what this does to our long-term ability to think originally. True expertise isn't just a database of facts; it's the ability to see patterns that don't exist yet. It’s the "gut feeling" a doctor gets when a patient's symptoms don't match the textbook. That intuition is a byproduct of thousands of hours of boredom, repetition, and failure. If we use AI to skip the boredom and the failure, do we also lose the intuition?
Consider the way we navigate cities now. Before GPS, you had to build a mental map. You got lost, you found landmarks, and eventually, you possessed the geography. Now, we follow the blue dot. We get to our destination faster, but we are functionally blind the moment the battery dies. I worry that we are doing this to the landscape of human knowledge. We are following the blue dot through chemistry, law, and art, but we aren't actually living there.
- The Latent Space as a Digital Aether: We treat the weights of a 175-billion parameter model as a mystical realm of infinite truth, much like the celestial spheres of the past.
- The Death of the "Aha!" Moment: Real learning usually involves a sudden click after a period of confusion. AI removes the confusion, which might also be removing the click.
- The Rise of the Curator: We are shifting from being "makers" to "editors." We don't create the thought; we just decide if the AI's thought is good enough to use.
The Alchemy of the Modern Mind
There’s a specific kind of hollow feeling that comes from using an AI to do something you haven't earned the right to do yet. I felt it last week when I used a model to translate a complex Latin phrase. I had the answer in 0.4 seconds. I felt efficient, but I also felt like a bit of a fraud. I hadn't wrestled with the declensions; I hadn't felt the texture of the language. I just had the output.
Is it possible that the "friction" of learning is actually the point? Evolution didn't design us to be efficient data processors. It designed us to be survival machines that learn through struggle. When we remove the struggle, we might be accidentally signaling to our brains that the information isn't actually important enough to keep. We are becoming a species with access to everything and a permanent grasp on nothing.
Maybe the real danger isn't that AI will become smarter than us, but that we will become so reliant on the "infused knowledge" of the digital Ars Notoria that we forget how to build our own intellectual structures. We might become a civilization of prompt-alchemists, standing over a bubbling cauldron of tokens, waiting for a miracle that we no longer understand how to perform ourselves.
What This Actually Means
We are entering an era where the barrier between "not knowing" and "having the answer" has effectively vanished. This is a massive win for productivity, but it’s a potential disaster for cognitive development. If we mistake the speed of retrieval for the depth of understanding, we’re going to end up with a generation of experts who are ten miles wide and one inch deep. They will be able to operate the systems, but they won't be able to repair them if they break.
I think we need to start treating AI like a calculator for the mind—useful, but dangerous if you use it before you learn how to do long division by hand. We have to be intentional about what we choose to struggle with. The goal shouldn't be to eliminate the process of learning, but to use AI to reach higher levels of complexity that were previously out of reach.
The Ars Notoria ultimately failed because staring at drawings doesn't actually teach you geometry. The LLM is different because it actually gives you the answer. That makes it much more useful, but also much more seductive. We have to make sure that in our rush to possess the world's knowledge, we don't lose the very thing that makes that knowledge worth having: the human mind that fought to understand it.
Quick Answers
Is using AI to learn always a bad thing?
No, it's an incredible tutor if you use it to explain how a conclusion was reached rather than just asking for the final answer. The risk is using it as a replacement for thinking instead of a catalyst for it.
What is the main difference between AI and the Ars Notoria?
The medieval ritual was based on faith and didn't actually work, whereas AI provides functional outputs that can deceive us into thinking we've mastered a subject when we've only mastered the interface.
How can I avoid becoming a 'Prompt-Alchemist'?
Force yourself to do the work manually first. Use AI to check your results or break a stalemate, but never let it be the first and last step in your creative or analytical process.
