Dear Future Me is a reflection notebook. You tell it a thought, a proud moment, or a letter to your future self, and it gives your words back to you later. Before the VoiceOS hackathon it was a website where you typed into forms. At the hackathon we gave it a voice.
The project is built on real data. Dear Future Me surveyed 1,109 students across 26 schools, and 720 students have attended its workshops. I work on it as a full-stack engineering intern, and the goal of the voice app is to take the friction out of journaling for those students.
What it does now
You tap a widget on your iPhone lock screen. The app opens already listening, and you talk. An agent decides what to keep, saves it to your notebook in your own words, and answers out loud. If you wrote a letter to your future self, it comes back months later as a text message.
The code is in the repository.
How it is built
The phone does very little. It records speech and plays audio. Everything else runs on a gateway server:
- The conversation runs through VoiceOS, which generates the replies.
- A FastAPI backend checks every request and writes to Supabase.
- ElevenLabs turns the replies into speech.
- Letters go out by SMS, and only to numbers that have confirmed a code first.
The iPhone app is SwiftUI, with a WebGL orb that shows the state of the conversation, widgets for the lock and home screens, and Siri shortcuts. The same API also serves the website, so the notebook reads the same everywhere.
VoiceOS
VoiceOS is the reason this was possible in a hackathon. The hard part of a voice product is the conversation itself: knowing when the person has finished, when to ask a follow-up, and when to just save the thought and say so. VoiceOS handled that, and we got to spend our time on the product around it.
It is a new tool, and it has stuck with me after the event. It has become part of my own workflow, and it has changed how I work.
What I took from it
Typing into a form makes you edit yourself. Talking does not. For a notebook you will read years later, sounding like yourself is the point, and voice gets you there with far less effort.
Thanks to the VoiceOS team for building the platform and running the hackathon.