Hi, I'm Ivan, building Kopai, an AI reliability layer, from Zagreb

I'm Ivan Vuković, co-founder and CEO of Kopai, which I describe as an AI reliability layer. I'm also a partner at Reactor Studio, a Zagreb studio for agile software development, where I help teams build the right product fast. My background is site reliability engineering, and that shapes how I think about AI systems: they should be dependable, not just impressive.

I build Kopai with Claude and Claude Code, on a stack that includes Vercel, Supabase and ElevenLabs. Working this closely with Claude has changed how I build, and I'm still figuring out what good reliability practice looks like when an AI is a daily teammate.

What I'd love from this community: feedback from anyone who has run AI features in production. What broke first for you, and how did you catch it?