Using LLMs to generate automated tests after writing production code misses the epistemological foundation of testing. Tests work best when you see them fail first, corroborating that they actually test what you intend. LLM-generated tests without human engagement become mere ceremony, providing a false sense of security. More

8m read time From blog.ploeh.dk
Post cover image
Table of contents
How do you know that LLM-generated code works? #Using LLMs to generate tests #Toward epistemological soundness #Cargo-cult testing #Ways forward #Conclusion #
2 Comments

Sort: