Got it — here’s a short, clean version you can paste right into Devpost or read to judges:


Inspiration

I wanted to make AI visual — not just text-based. What if it could draw what you imagine?


What it does

Memento lets you describe what you want, and the AI draws it on the canvas. Say “draw three boxes connected by arrows,” and it does it instantly.

How we built it

Built with Next.js, tldraw.dev, Vercel AI SDK, and Supabase. The AI parses natural language into drawing instructions and renders shapes live.

Challenges

File parsing took too long, so I dropped it. Couldn’t finish SMS integration in time. Doing it solo meant cutting scope fast.

Accomplishments

Got a working AI agent that actually draws from text. Fully functional and clean.

What we learned

Keep it simple, focus on one core loop, and polish it.

What’s next for Memento

Add image parsing, SMS input, and smarter multi-shape layouts.

Built With

  • nextjs
  • supabase
  • tldraw
  • vercel
Share this project:

Updates