Reachout AI

Inspiration

2.7 billion people don't have smartphones. But almost everyone can send a text or make a call. AI tools require apps, data plans, and smartphones; we wanted to remove those barriers.

What it does

Text or call a phone number to talk to Gemini AI. No app, no data plan, no smartphone. Text for weather or ask anything. Call for a full spoken conversation with memory across the whole call.

How we built it

Node.js/Express server bridging Twilio and Gemini 2.5 Flash. SMS routes through a menu system; voice uses Twilio speech-to-text in, text-to-speech out, with per-call conversation history. SQLite logs everything.

Challenges we ran into

SMS is 160 characters. Getting Gemini to be genuinely useful within that constraint took real prompt work. Voice latency had to feel like a natural pause, not a broken connection. Designing a menu UX with no visuals and no back button is harder than it sounds.

Accomplishments that we're proud of

A real multi-turn AI conversation from a $10 flip phone with no internet. The whole backend fits in a single file.

What we learned

Constraints are clarifying. Building for SMS forces you to think about every character and every failure mode in a way building for the web never does.