What stops working in testing when LLM arrives

▤ Summary

This article explains how large language models change the way QA teams think about testing. It shows that traditional exact-match assertions often fail when outputs become probabilistic instead of fixed. It also highlights new risks such as prompt injection, adversarial behavior, and the need for red teaming. The piece argues for risk-based testing, LLM-as-a-judge workflows, and CI/CD evals to handle AI systems more effectively. It closes by pointing readers to a course and code example for learning LLM testing practices.

▦ Linked Companies