Notes on building AI apps that ship.
Generating a runnable route of exactly 42 km
The routing algorithm behind Trevola — Haversine projection, point-in-polygon validation, OSRM, and an optimisation loop that converges on the exact distance you asked for.
Read →
Anatomy of a production-only auth failure
Sign-in worked in every dev build and failed for 50,000 real users. How I traced a Google Sign-In outage to its OAuth root cause — and how to never ship this bug.
Read →
Riverpod architecture for MVPs that scale
A clean structure so your v2 isn't a costly rewrite — the exact layering, error handling and folder conventions I use on every founder project.
Read →
Testing Flutter apps that talk to LLMs
Your output is non-deterministic by design. Test the contract, not the words — mocking strategies, schema validation, and what golden tests are actually for here.
Read →
From idea to App Store, end to end
Scoping, designing, building and shipping an MVP to both stores — the fastest realistic path to launch, from someone who has done it repeatedly.
Read →
Supabase vs Firebase: which for your Flutter MVP?
A practical breakdown of when to reach for each — based on shipping real products with both, not benchmarks.
Read →
Getting users onto the latest build — without begging
"Are you sure it updated?" — solving the update-adoption problem on Android with the In-App Update API and on iOS with an appcast feed.
Read →
Trust beats talent when talent doesn't care
Clients don't just hire skills — they hire trust. What a burned client taught me about how freelance work is actually won and kept.
Read →
The AI bubble talk — and what I tell junior developers
Expectations and returns are out of sync, and some people have stopped learning because "AI will do everything." That mindset worries me more than any correction.
Read →