I’ve decided to start a new sort of vibe project — something at the boundary between creative writing, language play, and AI collaboration. It’s called The Brolm, and it’s a literary-interactive web experiment exploring how meaning can emerge from invented words through layered storytelling.

The Brolm begins with a root story written entirely in nonce words: invented content words with no pre-assigned meaning. The grammatical skeleton is English — determiners, prepositions, conjunctions, auxiliaries — but every noun, verb, adjective, and adverb is a pure invention. The result is a text that is syntactically valid but semantically opaque: the reader can feel the shape of a narrative without understanding a single content word.

Each nonce word is a portal. Clicking it opens a sub-story that contextually illuminates the concept behind that word — not through definitions, but through situated narrative exposure. Sub-stories have slightly lower nonce density: a few real English content words begin to appear, giving the reader partial semantic footholds. These sub-stories themselves contain nonce words, each clickable, leading to further sub-stories — recursively, until the deepest layers become mostly intelligible English.

An important aspect of the process: when a nonce word is first written into a story, neither the author nor the model knows what it means. It’s just a sound with a part of speech. Only later, when that word is chosen for expansion, does its meaning begin to take shape — emerging retroactively from the syntactic and narrative context in which it was originally used. This keeps the creative process honest: the semantic world grows organically, one story at a time, without being planned in advance. It also mirrors the reader’s own experience of gradually building understanding through context.

The project grows incrementally. Each week, new story branches are added, expanding frontier words and deepening the tree. Returning readers discover new paths and, gradually, begin to understand the root story.

The web application offers several ways to explore the growing world:

  • Read — the main reading experience, navigating stories by clicking nonce words, with breadcrumb trails and contextual banners.
  • Map — an interactive tree visualization of the entire story graph, with nodes colored by state and a searchable lexicon sidebar.
  • Atlas — a concordance of every nonce word: its part of speech, morphological forms, and every story where it appears.
  • Stories — a sortable index of all stories with metadata: depth, density, word count, expansion status.
  • Stats — a data portrait tracking lexicon composition, tree shape, linguistic density, and cumulative growth over time.

The root story and its lexicon were conceived and authored in conversation with Claude, and each weekly expansion is co-written the same way — with meaning co-discovered, not pre-determined. The web application was built with Claude Code.

You can explore The Brolm at the following link: The Brolm — Nonce Stories.