Technical articles on performance engineering, automation, DevOps, observability and AI-assisted testing.
Four names people use as synonyms that aren't. What each load test type actually measures, which one matches which goal, and what happened to a real retailer that only ran one of the four.
expect(response).toBe(x) stops working the moment a model generates the response. We break down Anthropic's evals framework — tasks, trials, graders, pass@k — and how we apply it to test generative AI features responsibly.
Streaming breaks the assumptions of a traditional load test: you're not measuring an endpoint, you're measuring playback experience under thousands of concurrent viewers. What these tests are and why almost nobody runs them well.
Real adoption data on AI in QA — what it already delegates with confidence, where it still breaks things, and why the gap between piloting and production is 2026's real problem.
A Grafana dashboard tells you something is wrong. Distributed tracing and structured logs are what tell you why — here's how the three pillars fit together.
Why k6 has become our default choice for load testing, and how to structure your first script around a real user journey instead of a single endpoint.
