Hey Product Hunt! FitVerdict started with a simple problem — my daughters would ask "does this look okay?" before heading out, and the answer was always subjective. I wanted something more useful than a family opinion. But building it became a research project into AI agent patterns I work with in enterprise AI — vision tokenization economics, edge-layer enforcement, privacy-preserving observability, and stateless multi-agent coordination. The consumer app is the surface. The architecture underneath is what I was really exploring. A few things that surprised me building this: → A canvas pipeline that downscales photos to 1568px before the API call cuts token cost 8× with zero quality loss for outfit analysis. Understanding this changed how I model vision AI costs entirely. → You can have full production observability — latency, token burn, error rates, verdict distribution — without storing a single photo. Privacy and telemetry don't have to be in tension. → The battle code system (peer outfit comparison) runs as two completely independent AI agents with zero shared state — coordinated by a base64 token. No session management, no database joins. 33 days since launch. Zero errors in the last 3 weeks. ~$0.009 per scan in infrastructure cost. Full technical breakdown here: https://x.com/_vasan_/status/208... Would love your feedback — especially on the verdict quality and the sharing experience. What would make you actually, use this before heading out? Built in public. Shipping in the open. ⚖️

FitVerdict — Know before you go

AI judges your outfit before you leave the house

AI judges your outfit before you leave the house