Field notes · Product and engineering
Writing about the systems behind the feature.
I write about building dependable AI-powered consumer mobile products: state, reliability, measurement, trust and the operating loops that turn capabilities into useful software.
Where the Model Runs: A Decision Framework for On-Device vs Cloud AI in Consumer Apps
For a mobile AI feature the question is no longer which model, but where each request should run — a provider-neutral framework with a hybrid architecture and graceful degradation.
Read article ↗Streaming Without Lying: Progressive AI Responses That Still Honor a Contract
Keep the speed of streaming without acting on unvalidated output — separate the optimistic display path from a validation-gated commit path.
Read article ↗Shipping Reliable AI Features in Consumer Mobile Apps
A production-minded framework for state, failure recovery, evaluation, observability and controlled rollout.
Read article ↗From AI Capability to Consumer Habit
How recurring jobs, accumulated value, visible trust and measurement turn model capability into a product people choose to keep.
Read article ↗