What does text do?
Text analysis usually groups passages by what they're about — all the fear passages together, all the money passages together. We trained a model on a different signal — which passages appear near each other in a text — and it discovered what passages do: narrative functions, literary traditions, and structural patterns that recur across centuries of English-language writing.
Explore the demo →A chase scene in a Victorian adventure novel and a chase scene in a Russian psychological novel look nothing alike — different words, different settings, different century. But they perform the same work in a story. They sit in the same structural position: tension building before them, resolution or escape after. The words are different; the shape is the same.
Instead of asking “which passages use similar words?” we asked “which passages tend to appear near the same kinds of neighbours?” We trained a small neural network on 373 million of these neighbourhood relationships, extracted from 9,766 Project Gutenberg texts spanning four centuries. The model couldn't memorise them all — it had to compress, and in compressing, it found the patterns that recur across thousands of books.
What emerged were hundreds of structural patterns — from broad modes like “direct confrontation” and “lyrical landscape meditation” to precise registers like “sailor dialect,” “courtroom cross-examination,” and “Darwin-Huxley scientific correspondence.” These aren't topics. They're recurring shapes in how text works — and they appear across authors, genres, and centuries.
The narrative timeline shows a novel's structure as a sequence of colour-coded concepts. Each colour represents a structural pattern. Use the resolution slider to zoom from broad narrative modes (k=50) down to specific registers and traditions (k=2,000). Here you can see where a novel shifts between dialogue, introspection, and action.
The hierarchical view shows all six resolutions simultaneously — from k=50 (broad narrative modes) down to k=2,000 (specific registers). You can see how a single broad category like “natural history exposition” fractures into finer distinctions at higher resolution: “hunting and animal behaviour,” “maritime observation,” “maritime exploration description.” Click any chunk to see its full decomposition.
Click the ✦ button on any cluster to get a plain-language explanation of what structural pattern it captures — and why apparently unrelated passages from different genres and centuries belong together. This feature makes the clusters much faster to interpret without reading through every entry.
View the prompt sent to Claude
You are an expert literary analyst examining clusters discovered by an unsupervised AI model (Predictive Associative Memory) trained on 10,000 Project Gutenberg novels. The model groups text chunks by temporal co-occurrence patterns — passages that serve similar narrative structural functions tend to appear in similar sequential contexts across different novels, regardless of their surface content.
Important: the model knows nothing about themes, topics, or meaning. It only knows what kinds of passages tend to appear before and after each other. Two passages can be structurally identical (same rhetorical pattern, same pacing, same position in a narrative arc) while being about completely different things.
When the cluster label seems wrong for the chunk, this is often the most interesting case. Look for the structural parallel — the rhetorical pattern, pacing, narrative position, or formal technique that connects them despite different surface content. Explain what structural feature the model likely detected.
Explain in 2–3 sentences: (1) What narrative/structural function this chunk serves. (2) Why it belongs in this cluster — what structural feature connects it to the other samples, even if the surface content differs. (3) What's interesting or surprising about the grouping. Be specific about the text. Reference actual phrases. Be concise.
Paste any text — a chapter you're writing, an essay, a speech — and the model will assign each passage to its nearest structural concept, showing you which patterns it detects in your writing.
This is not a topic model. It won't tell you a passage is about war — it tells you a passage is performing the same structural beat as thousands of other passages across hundreds of books, regardless of subject matter. The concept labels are AI-generated descriptions of each cluster's contents. The model produces the groupings; the labels are our attempt to name them.