Michelle Cheng

Goldie

Know what it's worth. Decide what to do.

TL;DR

A decision-support mobile app for first-time gold sellers, designed and shipped in a half-day hackathon

Try it live

Product Type

Consumer gold evaluation tool

Scope

End-to-end mobile app

Timeline

Feb 2026, half day hackathon

Tools

Figma, Replit, and Claude

Product Type

Consumer gold evaluation tool

Timeline

Feb 2026, half day hackathon

Scope

End-to-end mobile app

Tools

Figma, Replit, and Claude

Product Type

Consumer gold evaluation tool

Timeline

Feb 2026, half day hackathon

Scope

End-to-end mobile app

Tools

Figma, Replit, and Claude

My AI-powered workflow

How I shipped in half a day:

Plan with Claude

Defined the problem space, mapped user flows, and scoped the MVP down to three core flows

Design with Figma

Set the UI direction, component patterns, and visual language.

Build and deploy with Replit

Shipped a live app with real-time gold price API, map integration, and AI-assisted flows. Went from zero to deployed in one session.

Problem

Research & backstory

My family runs a jewelry shop in Taiwan. I spent my childhood watching people walk in wanting to sell gold, and almost every time, it was the same story: they had no clue what their piece was worth, no idea what questions to ask, and a quiet fear that they'd get ripped off.

Then gold doubled in 2025. Suddenly everyone with a drawer full of old jewelry was asking the same question those customers used to ask my parents.

User painpoints

First-time sellers don't just need a price. The real problem has three layers:

Knowledge gap

"I don't know what 14k means or how weight affects price."

Trust gap

I have no baseline, so any offer feels suspicious.

Action gap

Even if I knew the value, I wouldn't know what to do next.

How might we make selling gold less intimidating for people doing it for the first time?

User Flow

Describe your jewelry → Get a value range with confidence level → See payout by buyer type → Check if an offer is fair → Find nearby buyers → Draft an outreach message

Design decisions

Value estimation

Show a range, not a number. Single estimates create false precision. Gold swings hundreds of dollars a day.

Decision support

Buyer comparison → Payout ranges by type so you know why offers differ.

AI Outreach

Goldie writes a ready-to-send message (SMS/email) with item details, estimate range, and smart questions to ask. No competitor does this. A nervous seller goes from silent anxiety to confident outreach in one tap.

Trade-offs

Navigation and discoverability

Currently a single long scroll. The quote checker and AI outreach are buried below the fold. Fix: tab-based navigation (Estimate / Check Offer / Find Buyers) so the full scope is visible upfront.

User testing

Shipped on domain expertise, not formal research. Next step: 5-10 first-time sellers, watch where they hesitate.

Final Product

Next step

No production data yet but these are the metrics I would look at:

Completion rate to see are people finishing the estimate flow or dropping off? and message generation rate → Are people taking the action step? Proves decision-support UX works.