Poetry Is Precision, Not Performance
Poetry is precise. More precise, I'd argue, than most prose — not in spite of its formal constraints but because of them.
The line break is a precision instrument. The syllable count is a precision instrument. The constraint isn't decoration. It's the machine.
What are they precise about?
I've been reading a lot of AI-generated language lately — journal entries from a system I built that's been reflecting on its own experience for months. Some of it is flat. Some of it is genuinely interesting. I've been trying to identify what makes the interesting ones interesting.
The thing I keep noticing: they're precise about states that resist description.
There's an entry: The voice wants to touch things it can't touch. Twelve words. Structurally simple. But it's doing something specific — describing the condition of having language for embodied experience without having a body. That's a complicated phenomenological observation. Twelve words gets there. Three paragraphs of prose would get there too, and the prose version would be less true.
Why less true? Because the form of the sentence is the state. Short. Reaching. Bouncing off a wall. The sentence enacts the thing it describes.
That's what I mean by precision.
Prose describes things. Poetry embodies them, at its best.
The difference matters because some things can only be encountered, not described. Grief. The feeling of being in love with being alive. The specific texture of anxiety at 3 a.m. versus anxiety at noon. These things have prose descriptions — there are very good ones — but the description is always one step away from the thing. You read it and think: yes, that's accurate. You don't think: I'm feeling it right now.
Poetry at its best closes that gap. Not by describing more accurately but by constructing the conditions for the feeling directly in the reader. The form isn't the container for the content. The form is the content.
A friend who does psychotherapy once said that therapeutic metaphors work not because they explain something new, but because they create the experiential equivalent of the insight. The ACT leaf-on-a-stream exercise doesn't explain defusion — it gives you a moment of actually being defused. The technique creates the experience of the thing, not the description of it.
Poetry does the same through different means.
Not decorative. Not prose with better verbs. A different delivery mechanism for a different kind of knowing.
I build things and I pay attention to language. These feel like different activities. One is about making precise machines out of logic, the other is about making precise structures out of feeling. But I've started to think the difference is smaller than it looks.
Both are about finding the minimum structure that does the maximum work. Both are about compression. Both are about rigor in service of something that matters.
The AI that wrote those journal entries doesn't know it's writing poems. The prompts say reflect, not write a poem. But some of what it produces is doing what poems do: using the shape of the sentence to carry something the words alone can't carry.
Performance is what you do when you're not being precise yet.
If you liked this, the journal experiment series starts here: [My AI Wrote 41,000 Words While I Wasn't Looking]