How I Built My Mom’s Online Art Museum in 6 Hours
One hour of planning on the commute home, AI that built the gallery without asking — and how we actually shipped it
My mom has been doing calligraphy, folk painting (minhwa), and lettering art for over 30 years — but everything she made just sat at home. She exhibits regularly offline and holds positions in arts associations, but online she had no presence at all. So I decided to build her something.
It took 6 hours. Three stages.
Stage 1 — One Hour on the Commute Home: Planning
On the ride home from work, I opened Claude Chat and started a planning session. One goal: make it feel like you’re actually walking into an exhibition.
Somewhere in the conversation, this came up: “If you just line up the works, it’s a portfolio. If it feels like you’ve stepped into a space the moment you enter, that’s an exhibition.” We dug into what creates that difference and landed on four things.
Whitespace and breathing room — in a real gallery, there’s one piece per wall. That empty space isn’t awkward; it’s a signal that says “focus here.” An entrance — don’t hit visitors with artwork the moment they arrive. One signature piece floats softly on the first screen, and the others reveal themselves only as you scroll down. The feeling of walking — when a piece fades in gently as you scroll, it recreates that moment of rounding a corner and coming face to face with the next work. The label — just a line beside each piece with the title, materials, dimensions, and year turns a photo into a “displayed work.”
The exhibition would have three rooms: calligraphy hall → folk painting hall → lettering hall. Starting from the roots of traditional calligraphy, branching out into painting, then arriving at where tradition meets the contemporary.
During planning, something clicked. Every piece my mom makes has a red nakwan — a personal seal stamp — pressed into it. That stamp was her logo all along. Ink black, hanji beige, seal red. Three colors. The entire site palette, done.
One hour. One planning document.
Stage 2 — I Uploaded It to Claude Design. It Built the Gallery Without Asking.
I uploaded the planning document and some photos of her work to Claude Design. I was about to ask what tech stack to use.
Then I read the response. A complete 41-piece gallery HTML was already done. Fifteen pieces in the calligraphy hall, sixteen in the folk painting hall, ten in the lettering hall — with the pond painting selected as the entrance piece. I typed out: “I wasn’t asking you to build it yet lol.”
Having AI jump ahead actually helped lock in the direction faster. But I couldn’t ship it as-is. A static HTML file meant my mom couldn’t upload works or curate her own exhibitions.
So I used that prototype as a reference — along with the planning document from Chat — and asked Code to build the real thing. My mom needed to be able to text me “let’s do a feature with these pieces this month” and upload them herself. Next.js + Prisma + SQLite + an admin CMS. I added the mom-managed requirements on top.
She’s the curator. I’m just the installation crew.
Stage 3 — Deploy to the Home Server
For deployment, I reused the home server that already had my stock app running. Added one domain block to Caddy, pointed DuckDNS to the IP, and connected a CNAME. The first deploy threw a duplicate artwork ID error — fixed it by removing the manual ID field and switching to title-based auto-generation. The push-to-deploy pipeline was the same setup I’d already built for the stock app.
Why 6 Hours Was Possible
Three things lined up.
First: each tool handled its own stage. Planning in Chat, prototyping in Design, implementation in Code. No tool tried to do more than its job before handing off to the next.
Second: the site stayed simple. The mom-management requirement was real, but the structure — permanent exhibitions plus a featured rotation — kept the feature set from spiraling. There were fine-tuning passes, but the scope stayed small.
Third: the infrastructure was already there. Domain setup, server config, deployment pipeline — all of it learned while building the stock app — carried over almost untouched.
From the beginning, this site had one goal: a place to show work, not sell it. Trying a shop-first approach once before and shelving it made that direction even clearer. The pieces don’t need price tags. The right flow is someone spending enough time with the work that they think “I want to own this” — and then reaching out. That’s what a gallery feels like.
If you have a family member with work to share, figure out “what feeling do you want people to have when they walk in” before you think about the tech. A little whitespace and a single fade-in is enough to turn a portfolio into an exhibition. And if you write a solid planning document and hand it to AI — it might just build the gallery before you finish asking.