← Back to Ara: your NZ driving practice app

Case Study

Building an AI Driving Examiner With Process, Not Vibes

A case study on NZ Drive Practice: an AI-examiner app for the New Zealand full-licence test, built deliberately as a demonstration of AI-assisted engineering with discipline, ADRs for every structural decision, an MVP roadmap with explicit exit criteria, and deterministic replay tests. A record of the reasoning, the trade-offs, and the honest state of the work.

Role Fullstack DeveloperRead 12 min

React Native + Expo
react-native-maps (Google Maps)
expo-speech-recognition + OpenAI TTS
Supabase (Postgres, RLS, OAuth)
Claude (Haiku + Sonnet) via Edge Function proxy

Context and problem

The approach: process over improvisation

Decision one: a pure, deterministic session engine (ADR-0006)

Decision two: real NZTA scoring, not an invented percentage (ADR-0005)

Decision three: the honest gap, hands-free audio (ADR-0003)

Other decisions worth noting

The honest state of the work

Reflection: what this says about how I work