3 days. 2 prompts. 1 wedding website.

Wanted to stress-test Claude Design on something I genuinely care about: 👰🏼‍♀️The website for my wedding in September. 10 years with my partner Agus, real deadline, real guests, no room for "AI template" energy.

Here's what I learned by day 3: Claude Design is not a standalone tool. It's a layer in a stack.

The actual workflow that worked for me: → Claude (chat) — drafted the brief, defined palette, structured the 13 sections before opening the canvas → Claude Design — executed in 2 prompts. All 13 sections. Live countdown, RSVP form, copy-to-clipboard buttons. The skeleton was real. → Claude (chat) again — debugged outputs, regenerated my CSS when I hit my weekly limit, planned the next surgical prompt → Claude Code (next) — will wire the RSVP form to Google Sheets

What surprised me: ✦ Reading my SVG monogram + brand brief and producing a coherent first pass ✦ The Tweaks panel — custom sliders that don't burn tokens ✦ How fast 70% of the structure appeared

Where it hit the wall: ✦ Burned my Pro daily allowance in 2 prompts. Inline comments cost credits too. ✦ The Tweaks panel generates a .jsx file that silently overrode my hand-edited CSS. Lost 20 minutes wondering why nothing changed. ✦ Some elements weren't selectable. Decorative star ornaments I wanted to delete? Couldn't click them. Had to spend a full chat prompt asking Claude to remove them. ✦ The output was structurally there but emotionally generic. The small details — animation timing, hover micro-interactions, the curated feel of a hand-crafted page — those still need a human and a tool like Framer.

🤔 The verdict: Claude Design replaces the blank page. It does NOT replace the designer's craft. It's an incredible preview-to-final-web step, not the destination.

For me, the real magic is the loop. Use Claude to plan. Claude Design to scaffold. Framer (or your tool of choice) to give it soul.

What's your stack? 👇