Every place I have worked required showing that tests worked during code review, break the code, watch the tests fail, unbreak the code, watch tests pass.
If I were going to spend resources on making testing better and my choice was between mutation testing and property testing, I'd rather spend that energy on property testing.
Mutation testing feels like something that the test framework should be doing anyway, maybe not every run, but every time it sees a new test.
If you keep going in this direction, your LLMs will be expert vulnerability injectors, possibly your end goal. All technologies are dual-use.
When I tap and hold on your blog link and tap Search on mobile, the top result ("My Opinion on LLM-based Mutation Testing") has a malformed URL where "mutatino" is used instead of "mutation" and 404s. When corrected, the page loads.
clickable link https://www.jungsteven.com/blog/2024-07-03-past-present-futu...
Every place I have worked required showing that tests worked during code review, break the code, watch the tests fail, unbreak the code, watch tests pass.
If I were going to spend resources on making testing better and my choice was between mutation testing and property testing, I'd rather spend that energy on property testing.
Mutation testing feels like something that the test framework should be doing anyway, maybe not every run, but every time it sees a new test.
If you keep going in this direction, your LLMs will be expert vulnerability injectors, possibly your end goal. All technologies are dual-use.