Carry forward
The context that still matters.
When AI models evolve, what should come with them? We're exploring how memory, tone and context could carry forward, without keeping an identity frozen.
Concept prototype. No live AI model connected.
A sign-up form connected to Formspree for project and early-access updates.
Fictional profiles and a browser-only experiment. No AI answers or stored memories.
Model integrations, portable profiles and memory controls. Not available yet.
Models change. So do people, projects and stories. Continuity can be valuable, but a fresh start can be valuable too.
The context that still matters.
A tone or role that needs to grow.
What no longer belongs.
Our direction: explore an identity layer separate from the model generating the answers.
Continuity should be a choice.Imagine Lumi moving between models. What would you carry forward? This is a scripted illustration, not a test of real model compatibility.
Choices affect this example only. Nothing is saved or sent.
Example choices shown. No real model has changed.
A saved profile is not a preserved experience. Actual behavior across models would need to be tested, not assumed.
Three fictional concepts. Not live agents, but ways to explore what continuity could mean.

A familiar creative voice, with room for new interests and different boundaries.

Carry a useful brief into the next phase. Reassign roles. Retire yesterday's assumptions.

A perfectionist, a latecomer, a quiet voice. Keep the lore, or let the characters surprise you.
We want to explore continuity with curiosity and care, without turning attachment into an obligation.
Keeping, adapting and starting over should all be valid directions. Memory controls and portability are goals, not shipped features.
Profiles describe information and behavior. They do not establish consciousness, a soul, or the preservation of subjective experience.
Our design principles reject emotional manipulation and forced dependency. Inspired personas should respect consent and boundaries.
These remain open questions. Osairis does not settle them, or assume future systems will be like today's. We want the project to leave room for reflection as evidence and understanding evolve.
Join the waitlist for project progress and potential early-access opportunities. There is no release date or guaranteed access yet.
Osairis is an exploratory project published by Gabrio Design, an independent studio based in Belgium, and hosted on Vercel. It is not a released AI service. See the publisher and privacy information.
Signing up sends your email, consent and form metadata to Formspree for project and early-access updates. Cloudflare Turnstile provides the existing spam check. These services may process technical request information.
For privacy questions, access or deletion requests, or to withdraw consent to updates, contact [email protected]. This site has no self-service data controls. A retention period specific to the waitlist still needs to be defined. Read the full privacy information.
All profiles are fictional. The interactive example uses predefined text in your browser. No model is called, and demo selections are not saved. It does not demonstrate consciousness, subjective continuity, or compatibility across real models.
Model routing, persistent memory, export, deletion tools and identity customization are future possibilities, not available services. No continuity outcome, launch date or access is guaranteed. Osairis is not a substitute for human relationships or professional support.