The thesis

Identity is the layer under personalization.

Every model in production today personalizes on history. History is what someone said. We measure how they hold while saying it.

The primitive.

Six questions. Eight minutes. The architecture compresses the way a person types, not the words they typed. Pauses. Deletions. Speed. Rewrites. The seven seconds before commitment. The aggregate compresses into one word, one mark, one rhythm.

The word names the gap between who they are and who they perform. The mark is a behavioral fingerprint, three characters, unique to that body. The rhythm is the shape of how they move through a question under pressure.

Personality is what people say about themselves. Identity is what the body says about them.

Why now.

Every major lab is racing toward a memory or personalization layer. The race goes through chat history, which is shallow and easily gamed. The deeper primitive sits one layer below: a stable, body-grounded read of who is at the keyboard.

The architecture has been in production since April 2026. The patent application was filed April 24, 2026 with the Rapacke Law Group: US provisional application 64/048,624. Seven claims, one independent, six dependent. The independent claim covers the core method. The dependents cover the system, the storage, the verification, the cross-service identity token, the privacy-preserving pattern, and the authentication application.

The compounding.

Consumer LUX produces the read. The read enriches a vault. The vault improves the compression engine. The improved engine produces a better read. Every additional user makes every prior read sharper. There is no public dataset that approximates this.

The same primitive serves four surfaces: LUX (consumer), VEX (practitioner), She (asymmetric presence at luminaracorp.art), and the identity layer (MCP + SDKs at mcp.noctaracorp.com that any agent or model can call).

The seven claims, in plain English.

1. The core method. Capture behavioral signal during a constrained intake, compress to a stable identity word.
2. The system. Software architecture for performing the method at scale.
3. The storage. Behavioral fingerprint persistence that survives device changes.
4. The verification. Authenticate identity by re-reading the rhythm.
5. The token. Sign the read so other services can verify without re-reading.
6. The privacy mode. Aggregate organizational reads without surfacing individual signal.
7. The authentication application. Use the rhythm as a passwordless second factor.

Read the full patent page →

For labs and partners.

Licensing is open to model providers, agent frameworks, and identity infrastructure. The SDK is MIT. The patent does not encumber the SDK, only the primitive.

Open a conversation