Why you yearn for craft in the age of AI
Lessons from ancient Greek
If I had written this with AI vs coming up with the words and research on my own, would you judge this post any differently? It would certainly sound different. I’d also feel differently about which I publish into the world. There’s a different feeling that comes from reading a hand-written letter vs an AI-generated email, for both the receiver and the sender. There’s a different feeling to writing software “by hand” vs with AI’s assistance (although I would say that orchestrating agents and engineering context is closer to the craft of writing software manually). It’s akin to the difference between jewelry that’s mass produced and perfect vs handmade and imperfect. And it’s not just about the result being artisanal vs generic, it’s about intention and purpose. With the help from some ancient Greek, I’ll articulate those differences that you may have been feeling.
ποίημα
poíēma — “a thing made; a work”
From the verb ποιέω (poiéō), to make, do, produce. Root of the English word poem.
When the Greeks said ποίημα (poíēma), they meant something made by a maker who was genuinely present to the making. Not just mechanically producing, but actively negotiating with their material.
Think of a metalsmith at the forge. The metal resists, and it may do unexpected things. It warps when you expect it to hold, and holds when you expect it to give. The smith must constantly read it… its temperature, its grain, its memory of previous stress. Every decision is a response to what the material is doing right now. The finished object carries all of that. It is, in the language of semiotics, an index of its maker: a sign physically caused by the thing it represents.
This is what makes a poíēma a poíēma. Not that it’s handmade in some precious sense or that it’s small-scale or slow. After all, an armorer producing hundreds of chainmail links was mass production. A jeweler using a hydraulic press is using industrial force. Neither disqualifies the work from poíēma because the maker is still in negotiation with the material and the outcome.
ἐμπειρία
empeiría — “experience”
The accumulated knowledge of having done something many times, of having learned through particulars. Root of the English word empirical.
τέχνη
téchnē — “craft, art, skilled knowing”
Not just technique, but knowledge that lives in the body through repeated encounter with resistance.
Aristotle distinguished between someone who merely knows a rule and someone who has ἐμπειρία (empeiría), who has been changed by repeated encounters with the thing itself. The experienced physician doesn’t just know that a drug works; they know how it works in this kind of patient, at this age, in this season. That knowledge was earned not only through years of study, but through time in the field, mistakes, and attention to patients.
τέχνη (téchnē) is what empeiría becomes when it matures into genuine understanding, when you know your material well enough to make nuanced judgments, to improvise, to recognize when the rules don’t apply.
What mass AI generation often enables is output without the maker passing through empeiría at all. The tool has the accumulated experience of millions of human makers and makes it available without the passage through resistance that builds understanding. It’s why genAI content can feel hollow. I’m not saying all generative AI feels like this (after all, I have struggled through lessons on how to wield AI effectively), but it may explain why people who have spent decades refining their craft hands-on feel this way.
εἴδωλον
eídōlon — “image, phantom, likeness”
A copy that has lost its connection to the original. Root of the English word idol.
Plato’s critique in the Republic was that representational art, like a painting of a bed, is an εἴδωλον (eídōlon): a mere copy of a physical object, which is itself a copy of the true Form. The eídōlon severs the indexical link because it’s depicting an object without possessing actual knowledge of what they are depicting.
Generated content is the quintessence of this. It’s trained on millions of human encounters and produces content without any experience of its own. It is statistically shaped human expression with the humans removed.
τέλος
télos — “end, purpose, that toward which a thing aims”
For Aristotle, you cannot evaluate a made thing without understanding its τέλος. A well-made sword in service of murder is not the same as a well-made sword in service of defense.
For me, the tension between poíēma and eídōlon ultimately comes from its purpose or intent. Most AI slop isn’t just hollow because there is no maker present, it’s hollow because its τέλος (télos) is usually engagement, content volume, or the appearance of productivity. And it’s why I’d said in this LinkedIn post, “One thing that will stop someone from making a quick buck in this AI gold rush isn’t a lack of ideas or skill or grit. It’s ethics.”
The eídōlon isn’t just imitation; it’s often aimless or aimed at something shallow.
A poíēma requires that you were actually changed by the poiéō. The noun requires the verb. The made thing requires that the making passed through you and left a mark. To want poíēma over eídōlon is to insist that making is a form of being present to your material, to your audience, to the world.
If you have been feeling something you couldn’t name, like a dissatisfaction with generated content or a longing for a skilled craft, I hope this post helped give words to what you’ve been feeling!