Pingi — AI Companion App

What I Built

Pingi is an AI companion app with its own personality, voice, and memory. It remembers past conversations, has emotional states, and feels like talking to a friend — not a chatbot.

Tools Used

• -Claude Code — Built the entire app from terminal prompts
• -Next.js — Frontend and API routes
• -Supabase — Auth, database, conversation memory
• -AI APIs — Personality engine, voice synthesis

How I Vibe Coded It

Started with a simple prompt: "build me an AI companion that remembers who I am." Claude Code scaffolded the full Next.js app, set up Supabase for conversation persistence, and wired up the AI personality layer.

The hardest part was getting the personality right — I spent most of my time iterating on the system prompt, not the code. That's the vibe coding difference: your taste and ideas matter more than technical skill.

Key Lessons

• -Start with the feeling, not the features. I described how I wanted conversations to FEEL, and Claude Code figured out the architecture.
• -Supabase makes persistence easy. Adding memory to an AI app went from "complex backend problem" to "ask Claude Code to add a conversations table."
• -Iterate on prompts, not code. 80% of the work was refining what I wanted, not debugging.