soundthought

Where it stands

A record of what has been built, and when.

Where it stands today: seven chapters since April, one of them the work I am on right now. The app runs and I use it most days. When it opens, the first people come in from the waitlist, earliest first.

April 2026

The blueprints

The first month went on the foundations: the beliefs that shape the product, the tools it is built with, and the architecture underneath it.

April to May 2026

The skeleton

The first working version: a conversation you could actually have, with your tone and the context read as you speak, and memory that carries from one session to the next. By mid May it could refer back to what you had told it weeks before.

Late May 2026

So you can cut in

The audio underneath was improved so you can talk over it mid sentence, the way you would with a person, and it stops and listens. The same work added a layer that reads your pauses, so a breath is not taken for the end of a thought.

June 2026

Journal and Journey

The busiest month of the project went on the phone itself. The Journal arrived for what you write or say between conversations, the Journey for seeing where you have been, and the interface was polished to stay simple and ask nothing extra of your attention.

August 2026

Proving it behaves

Every change to the way it speaks is checked against a fixed set of hard cases, the crisis ones first, before it can reach anyone.

September 2026

What it holds, and who can read it

Since the end of August, everything you say and write is encrypted with a key unique to you. In September that reached everything stored, and your words are kept out of the operational logs, for everyone from their first day.

Working on this now

Our own models

Moving everything onto infrastructure we run ourselves, so that nothing you say goes to another company's AI model. Everyone gets it from their first day.