TL;DR: I built an algorithm that finds accidental haikus in old books. Some of them are weirdly good. This made me question who the real author is — the code, the original writer, or the reader.
The Discovery#
I found Dimitri Rataud’s Haïku Marinière — he blacks out pages from old books, keeps a few words, calls it poetry.
My reaction: isn’t this just cherry-picking with extra steps?
The Algorithmic Mirror#
That question bugged me enough to build GutenKu in 2023. I wrote code that scans 70,000+ Project Gutenberg books looking for word sequences that fit the 5-7-5 pattern. No writing. No intent. Just extraction.
If Rataud’s a fraud because he didn’t write the words, GutenKu is worse — it doesn’t even know what it’s selecting.
And yet. Some of these found haikus hit hard. They work. I didn’t expect that.
The Reader as Author#
Barthes argued that meaning comes from readers, not writers. If code produces words with no intention and people still find something in them — maybe poetry was never about the writing at all.
Maybe it’s about the reading. We decide to find meaning. The text is just raw material.
A few questions:
- If a poem moves you, does it matter how it was made?
- Does AI write, or do we project meaning onto noise?
- What’s the real difference between human curation and algorithmic selection?
GutenGuess: The Game#
I also made GutenGuess — you see a haiku, guess which book it came from. Six tries. Daily puzzle.
Under the Hood#
No LLMs. Genetic algorithms, NLP scoring, neural networks. Code is open source.
Perhaps poetry was never in the words themselves, but in the space between text and reader.