Nobody set out to build an identity system.
It fell out of failing at something else for long enough to understand why.
Origin
Where the problem came fromThe work began as an attempt to generate a full-length book against a single model API. It fell apart around chapter three. The failure was not sentence quality; the sentences were fine. It was that the person speaking in chapter forty had stopped being the person introduced in chapter one, and prompt engineering did not fix it, because the identity was living in a context window that fills.
Fixing continuity meant moving character state out of the prompt and into a structured record. That decision produced an attribute schema. The schema turned out to be the actual product, and the book turned out to be a test harness for it.
Sequence
What broke, in the order it brokeOne provider, one problem
A readable full-length book against a single model API. Continuity failed at chapter three.
The ontology
Character state moved out of the prompt and into a database. The attribute schema arrives.
Collaborator teams
Single-voice generation gives way to teams — outliner, stylist, continuity checker — each with its own psychology.
First provisional filing
USPTO provisional covering the computational identity architecture. Priority date established.
Separation
The identity layer is pulled out and developed on its own as Persona Protocol. Narrative Geometry becomes its first consumer rather than its container.
Who
Alalonde Corporation, OttawaA deliberately small operation working on a problem larger companies have mostly routed around, because the money in generative AI is currently in short-form output and agent orchestration rather than in whether chapter forty remembers chapter three.
That is a reasonable commercial judgement on their part and a genuine opening on ours. Long-horizon identity is not a feature you bolt onto a chat product. It changes the architecture, and it has to be there from the start.
It is also why the underlying platform routes across multiple providers rather than committing to one. Model quality moves monthly. An abstraction layer that lets the best model for each task do that task is worth more than any single model relationship — and an identity that lives outside the model is what makes that swap survivable.
We say collaborator, not agent. An agent executes a task. A collaborator has a psychology, a voice, and a history — and those are exactly the properties that make long-horizon work hold together. The name is a reminder of what is actually being built.