The problem
Every task app assumes you will stop, open it, choose a list, type a title and set a date. People do not. The thought arrives while driving, mid-meeting or half asleep, and by the time there is a free hand it is gone.
What it does
You speak. It works out what you meant, turns it into something structured — a reminder, a step in a plan, an action with a time attached — and files it where it belongs. In Arabic or English, including the mix of both that people here actually speak.
Why it was hard
- Arabic speech recognition is meaningfully behind English, and dialect makes it worse. Getting it usable meant work at every layer, not a model swap.
- Intent is ambiguous by nature. “Call Ahmed about the quote tomorrow” is a task, a person, a subject and a time, and getting one of them wrong makes the whole capture untrustworthy.
- Trust is binary. A capture tool that loses one thing in twenty stops being used, because you can never again assume it caught something.
My role
Founder. Product direction, architecture, and the parts of the build that decide whether it feels reliable.