Insights
What we have learned building AI that has to keep working.
3 articles on QA. — written by the people who ship and then maintain the systems.
QA
Test Automation That Survives a Redesign
A suite that has to be rewritten every time the UI changes gets abandoned. The difference is almost entirely in what the tests are anchored to.
3 September 2026 · 3 min read
QA
What a Test Plan for an AI Feature Actually Contains
You cannot assert equality on a non-deterministic output, so most teams test almost nothing. Here is what a real QA plan for an AI feature covers instead.
1 September 2026 · 4 min read
QA
How Do You Test an Agent That Never Answers the Same Way Twice?
Traditional QA asserts that output equals expected. Agentic systems produce a different sentence every run and can still be correct — or produce the same sentence twice and be wrong both times. The assertions have to move.
30 June 2026 · 3 min read