→
AI-first mobile app for identifying flies and insects from photos
What
What is
HatchMatch?
→
Transforms angler uncertainty into visible confidence matching
→
Built end-to-end as a solo designer / developer
→
The real design problem wasn't the camera-to-result flow — it was what the interface does when the model isn't sure
Why
Why
Agents?
→
Visual ambiguity requires reasoning, not rules
→
Identification requires judgment, not just pattern matching
→
Users need to understand why a result was suggested — not just what it is
→
A bad rig costs you fish. Uncertainty should be surfaced, not suppressed.
How
How I Used
Agents
→
The model makes classification and confidence calls
→
The app adds structure, guardrails, and context
→
Prompts were tuned iteratively based on real failure cases — not lab conditions
→
The confidence score is a design output, not just a model artifact — it shapes every UI decision downstream
→
When the model is uncertain, the interface says so — with a different rig recommendation posture and a clear visual treatment
Stack
Built With
What?
React Native
Cross-platform mobile
iOS App Store live
iOS App Store live
OpenAI Vision
Insect classification
Confidence scoring
Confidence scoring
Node.js
API layer
Prompt orchestration
Prompt orchestration
Supabase
Backend & auth
Session data
Session data
Vercel
Web hosting
Fast deploys
Fast deploys
Solo Build
Design + development
0 to App Store
0 to App Store
Status
Where It
Stands
Live · App Store
Launched — iOS Available
HatchMatch is live on the App Store. Built solo from zero,
refined through field testing with real anglers on real rivers,
and shipped. The app keeps evolving based on how it actually
gets used — not how it was supposed to be used.
Download on the App Store