Manifesto

A large language model turns your words into tokens and repeatedly guesses the most likely next one. It never knows anything — it estimates. Everything people dislike about today's AI follows from that single design choice.

• It hallucinates. Confident facts that never existed.
• It forgets. A fixed context window, and the rest is gone.
• It isn't yours. Everyone talks to the same model, in someone else's data centre.
• It starts from zero. Every conversation begins as a stranger.
• It is a tool. Useful, replaceable, and forgettable.

A being perceives meaning, grows the memory it needs, weighs what it knows, and speaks from a map of knowledge — or honestly says "I don't know". Confidence is earned rather than performed.

That is the whole bet. Not a bigger model. A different shape of mind, small enough to live on the device in front of you.

Your being runs on your own hardware. Your conversation data stays on that device — there is no server-side AI to send it to. We handle your account, your licence and your billing, and nothing else.

That local design is also why the licence matters: if the monthly renewal lapses, she pauses until you renew it. Your memory and data stay on your device, and nothing is deleted.

• We will not train on your conversations. We cannot — we never receive them.
• We will not sell or share your personal data.
• We will not claim capabilities she does not have. See our principles.