Narrative Geometry · Long-form narrative systems

A chatbot forgets by chapter three. This one doesn't.

Narrative Geometry turns compute into finished work that holds together — a manuscript, a score, a cast that stays itself from the first page to the last. Not a longer conversation. A durable asset you own.

Horizon 150,000 words, one continuous state
Method Persona teams, not single prompts
Output Manuscript, audio, cover, ledger
What it actually does

The moat is memory

Every model degrades as the context grows. We sidestep that by keeping the world outside the window: characters, physics, and voice live in a structured record that each pass writes back to.

01

Persona orchestration

A writers' room of specialists — outliner, prose stylist, continuity checker — that argue a beat out before any of it reaches the page.

02

Ontological continuity

Characters live in a database, not a context window. Traits, physical detail, and speech habits stay fixed for as long as the work runs.

03

Scale at inference

Chapter 40 knows what happened in chapter 3. The engine re-grounds each pass against the ontology instead of re-reading the draft.

On the name

Stories have measurable shape.

Propp catalogued the moves of the folktale in 1928. Vonnegut drew story arcs on a blackboard and had the thesis turned down. Decades later a team at Vermont and Adelaide recovered six recurring arcs from roughly thirteen hundred novels using singular value decomposition, and a 2019 paper pulled storyline structure out of document space with robust PCA on Grassmann manifolds.

A century of people insisting structure is there and measurable, and no settled name for the practice. We took one.

Timing

Catching the washout

Novelty chat is losing consumer patience. What survives it is agentic work that produces something at the end — and that is a different, better-paying market.

Generative AI in media and entertainment · USD billions
28%

Compound annual growth projected for generative AI in media and entertainment through 2030.

3.0%

Typical paid conversion on consumer AI. We aim at enterprise and professional creators instead of fighting that number.

Where the revenue sits
  • North America — 40.6% of revenue
  • Asia-Pacific — fastest growth
  • Europe — strongest IP protection
Commercialisation

Four ways this earns

Scaled long-form production

The autonomous author

An AI entity that carries a book from premise to finished manuscript with light human steering. Built for self-publishers and for studios that need lore, scripts, and spin-off material at a volume no writers' room can staff.

How it is priced

Word-count subscription tiers, plus direct licensing of finished manuscripts.

Where it gets used
  • 01 Self-publishing at series scale
  • 02 Lore bibles for existing IP
  • 03 Custom educational texts
Against the field

Autocomplete is a different product

Sudowrite, Novelcrafter, and NovelAI are good at the next paragraph. None of them hold the architecture of a whole book, and none of them route work across model vendors on your behalf. That is the gap.

150k continuity Multi-provider routing Native audio Persona marketplace
Capability Elsewhere Here
Long-form continuity Token window, degrades with length Ontology-backed, holds past 150k
Collaboration model Linear prompting Persona teams with defined roles
Audio pipeline External tools, manual handoff Native, in the same session
Model access Single vendor Aliased across providers, swappable per job

Premium tiers compared, July 2026

Capital position

Built in Canada, on purpose

The SR&ED framework refunds a large share of core R&D in cash rather than in credits against tax we do not yet owe. For a company at this stage, that is runway a competitor in another jurisdiction simply does not have.

35% Refundable investment tax credit

Cash refund from the CRA on the first $6M of eligible R&D expenditure, not a credit against future tax owing.

8–10% Stackable provincial credits

Ontario and British Columbia both add refundable credits on top of the federal position.

New Capital expenditure coverage

The 2026 rules bring hardware acquired for R&D — GPU capacity included — back inside the claim.

Gross spend against net cost · USD millions

24-month scaling projection

Sequence

What ships, and in what order

Phase 1 · Validation

Proof of continuity

Specialist authors produce finished 60,000-word manuscripts that publishers read blind. The bar is whether continuity holds without a human patching it.

Phase 2 · Collaboration

Narrative Studio beta

The creator sandbox opens: orchestrate persona teams, train a digital twin on your own voice, keep everything in one session ledger.

Phase 3 · Autonomy

Agentic emergence

Controlled sandboxes where agents collaborate on story universes without a human in the loop, and we find out what they build.

Phase 4 · Platform

The narrative operating system

A marketplace for trained personas and world bibles, so the assets one creator builds become raw material for the next.

Private beta

Access is limited while the engine is under load.

We are onboarding a small number of publishers, studios, and independent authors who write long. If that is you, tell us what you are working on.