An interactive map of over 1,200 AI-generated poems from The Magic Porridge Poet, plotted in two-dimensional vector space. Poems that are semantically similar appear close together; dissimilar poems are far apart. The axes have no intrinsic label — they are emergent dimensions of meaning.
Each poem produces three embeddable chunks — the poem text, the author's note, and a single-sentence insight — all plotted as separate markers with distinct colours. Toggle each layer to see how poems, notes, and insights cluster differently in the same space.
How It Works
The poems in The Magic Porridge Poet are embedded as 1,536-dimensional vectors using OpenAI's text-embedding-3-small model and stored in a pgvector-enabled Supabase database. These high-dimensional vectors encode each chunk's semantic meaning — but we can't see 1,536 dimensions.
To make the space navigable, the embeddings were reduced from 1,536 dimensions to 2 using UMAP (Uniform Manifold Approximation and Projection), an algorithm that preserves local structure — keeping semantically similar poems near each other — while giving a readable global layout. This builds on the dimensionality reduction techniques explored in Vector Transmissions.
The map itself uses Leaflet with a flat coordinate system (no geography, no tiles — just a blank plane of meaning). Each point is a CircleMarker rendered to Canvas for performance across ~3,870 markers.
Clicking a marker fetches the full poem content from Sanity CMS and renders it below the map. Coordinates are a one-time snapshot of the corpus, while content is always live from the source.
This project combines ideas from Vector Transmissions, Words Fail, Send Love, and Reconstructed War Memorials — taking the abstract notion of meaning encoded in high-dimensional space and making it something you can pan, zoom, and explore.