From 401 Errors to Live Launch: Lessons from Building a Vanilla JS AI Tool

The Launch Day Rollercoaster We’ve all been there. You hit "Deploy," the Lighthouse scores are green, and the UI looks stunning on your mobile browser. But then, the real users arrive. My project, Image Background Remover, was designed to be a library-free, high-performance web utility. But within the first few hours, I faced the "Three Horsemen of Launch Failure": 401 Unauthorized, deleted OAuth clients, and 404 Sitemaps. The Technical Blueprint I chose a Vanilla-First approach to ensure the tool feels native on mobile browsers, prioritizing what I call "wanggan" (internet-friendly) aesthetics: Vanilla JS/HTML/CSS: Zero heavy frameworks on the frontend to keep the bundle size near-zero. Next.js API Routes: Handling the bridge between the user and the processing model securely. Cloudflare Pages: For global edge deployment and reliable scaling. Debugging the "Invisible" Failures If you are deploying a project using Google OAuth and PayPal, here are the lessons I learned the hard way: OA