Writing
Notes on testing things properly
Practical pieces on automation, quality gates and AI testing — written to be useful to an engineer whether or not they ever hire us. No listicles, no vendor comparisons, no thought leadership.
Why your Playwright suite is flaky, and the three causes behind almost all of it
Flakiness is not random. In practice it comes from three specific causes — shared state, implicit timing assumptions and unstable locators — and each has a concrete fix.
Test the API first: why UI-first automation is the most expensive mistake in QA
Most teams automate the UI first because it looks like testing. It produces the slowest, most fragile and most expensive coverage available. Here is the order that actually works.
How to test an LLM feature when the output changes every time
Conventional automation assumes the same input gives the same output. LLM features break that assumption. Here is the practical shift from assertions to evaluation.
RSS feed — no email signup anywhere on this site.
Start with a free QA Health Check
Answer twelve questions about how your team tests today. You get a maturity score, the three gaps most likely to cause a production incident, and a 30/60/90-day roadmap — whether or not you ever work with us.