Web portal rebuild: Migrating every page to one codebase
- Platform
- Frontend
- 2026
- Key result
- No migration dip
- User value
- A web portal that works on mobile, with a searchable library and no disruption on the way.
- Business value
- One codebase instead of two, so an improvement reaches the whole web portal.
- Stack
- Web portal
- Next.js
- Tailwind CSS
- VimUI
- Related case studies
Problem
New features were built in a new monorepo, but the web portal still sat on the old one. The two looked like different products, and every improvement to the web portal waited on a migration nobody had time for. Our goal was one codebase, without disrupting the people who use the web portal every day.
Design challenge: migrating every page without a dip in use. We matched the old behavior first, with the same data on a new foundation. The exceptions were pages where porting a problem forward would waste the migration, so the content library became a searchable explore page, and the homepage and settings were modernized.
What I did
- Had Claude Code map the old web portal, page by page. It traced the logic, API calls, and database links each page depended on, which is most of the work in a migration.
- Rebuilt each page from VimUI. Pages run against the same APIs and are held to the guardrails automatically. The rules I’d written for people turned out to make the agent’s output reviewable too.
Outcome
74% → 75%
week-1 return, with weekly web users holding steady
+4–7 pts
web video engagement, year over year, after the content library redesign
Every page is live on one codebase. The web portal now works on mobile, and a product manager or engineer can change a page without a designer in the loop.
Who did what: I rebuilt every page from VimUI, with Claude Code reading the legacy code. Engineering reviewed and hardened each section before it shipped, which is where an agent’s misreadings get caught.