I vibe coded an AI App, Here's what I learned building it.

14 days to build. Voice input, AI extraction, relationship intelligence layer , the whole thing.

The idea: walk out of any important conversation, vent for 60 seconds, and let AI pull out everything that matters. Prices quoted. Commitments made. The personal detail they dropped in passing. All of it structured and surfaced before the next conversation.

Here's what surprised me building this:

Voice input is a completely different product than text input

People say things out loud they would never type. The hesitation in how they describe a deal. The frustration that slips in. The detail they mention casually that changes everything. Text is edited. Voice is raw. The intelligence difference is significant.

The AI prompt is the actual product

I spent more time on the GPT prompts than anything else. Getting the AI to extract a specific price and commitment from a messy voice rant , and file it correctly every time , that's the hard problem. Not the UI. Not the architecture.

Dogfooding before any user sees it is underrated

30 days using it myself before showing anyone. Found things no user interview would have caught. Changed what I lead with entirely.

Stack if anyone's curious: Next.js, Supabase edge functions, Whisper for transcription, GPT-4o for extraction. Deployed on Cloudflare.