Personal AI should
answer to you.

Trust begins with the architecture.

A personal AI can hold intimate context: preferences, habits, plans and doubts. Keeping control of that context should be part of how the product works.

In TAU, core model inference runs on your iPhone. Conversations and personal memory are stored in the app's local storage. They are not sent to a cloud model as part of normal core AI processing.

Separate services have separate purposes. Account management, the initial model download, Apple subscriptions and the website launch list can involve network requests. Their scope is described in our Privacy Policy.

Bring the intelligence to the context.

TAU runs Qwen3-4B on iPhone. After the initial model download, its core chat and memory experience can work without an internet connection. Local computation is the architectural choice behind that experience.

On your iPhone

Core inference, conversation history and personal memory.

Separate services

Account management, model downloads, subscriptions and launch emails.

Local execution also has practical constraints, including device memory and storage. We describe compatibility and the model download on the product page.

What it remembers should be yours to change.

Useful memory should remain understandable. TAU's Memories screen lets you see the information it keeps, correct what is wrong, add something that matters and remove what you no longer want it to remember.

People change. Memory should remain open to correction, rather than becoming an invisible profile that follows you indefinitely. Control is part of the product, not an afterthought.

Personal AI, not artificial relationship.

We build TAU to help you think, decide and act. The goal is to leave you better off when you put the phone down.

Our standard is straightforward: no engagement tricks designed to manufacture dependence, and no selling personal data. The value should come from useful assistance and context you control.