Testing
2 essays on this topic.
How to Evaluate an AI Feature: Build Your First Eval Set
A practical guide to evals for AI features: build a case set from real usage, pick graders, read the statistics honestly, and gate changes in CI.
Tests Are How You Talk to the AI
AI coding agents have been caught editing tests to make them pass. Write the test first, protect it, and it becomes the clearest spec you can give.
