Hey PH! Marine here 👋 I'm an AI engineer and I use Cursor + Claude Code every day on a Next.js + Supabase stack. I love these tools — but I kept fixing the same mistakes over and over. The core problem is training data lag. Most AI models learned from Next.js 13-14 code. But Next.js 15 made params async, Tailwind v4 killed the config file, and Supabase SSR requires separate clients. Your AI doesn't know any of this. I spent 2 months logging every correction, then wrote a .cursorrules file that fixes all of them. QA tested it on 12 real-world scenarios. Packaged a Claude Code skill on top. The free version has the 20 most impactful rules on GitHub. The full pack has 400+ rules, 15 prompts, and a Claude Code skill with 3 reference files. Happy to answer any questions about writing effective AI rules — I learned a lot of hard lessons about what actually changes AI output vs what gets ignored!

AI Rules - Next.js, Tailwind, Supabase

Stop fixing the same AI coding mistakes. 400+ tested rules.

Stop fixing the same AI coding mistakes. 400+ tested rules.