AI

Practical AI for app makers: on-device models, LLM features, and AI-assisted development.

A Privacy Checklist for Shipping AI Features in a Small App

Adding an LLM feature means user text leaves the device and can end up in places you never planned for. This is a practical checklist for small teams: map the data flow, send less, redact before sending, control logs, and tell users plainly what happens.

Keeping LLM Feature Costs Predictable in a Small App

An LLM-backed feature can cost almost nothing in testing and then become your biggest bill after launch. Here is how to measure cost per action, cut wasted tokens, cache what repeats, and set budgets so one heavy user can't sink your margins.

Build a Small Eval Set Before You Ship an LLM Feature

If your only test for an LLM feature is trying a few prompts by hand, you can't tell whether a change made it better or worse. Here is how a solo developer can build a small eval set, score it cheaply, and use it to decide when a feature is ready to ship.

Episode

00:00:00 00:00:00