@MyWoolyMastadon@toot.community @ronanmcd@mastodon.green yes that's a way that can work. The problem is more that the LLM does not "know" what to test for. So it can guess some edge cases but not much more.