Most AI experiences begin with the intelligence. Which model is fastest? Which provider reasons best? How large is the context window? Those questions matter, but they place the most replaceable part of the system at the center.
TheoSphere starts somewhere else. It is being built as a user-owned memory and continuity appliance: a persistent home for approved source material, history, decisions, projects, relationships, and the context that gives them meaning.
The relationship should not reset because a provider changed its model.
Memory outside the model
The model is an instrument. It may run locally or, under an explicit policy, in the cloud. It can improve, fail, become too expensive, or be replaced. The memory remains anchored to infrastructure the owner controls.
That does not mean pretending no information can ever leave the appliance. When an external intelligence is used, an approved context slice necessarily travels. The honest goal is an inspectable policy and transaction record showing what moved, where, and why.
What exists today
TheoSphere is a working private prototype, not a shipping consumer appliance. The present system has demonstrated bounded retrieval, a locally grounded response with citations, and a deterministic control boundary that exposes a sanitized trace.
The customer configuration, onboarding, support model, recovery guarantees, price, and release date remain in development. We will label those boundaries rather than market the roadmap as finished product.