those kinds of tests are so fragile, human will retain champion of this.. They can use AI for a lot of it but at the end of the day a human will catch way more bugs, and triage them, and then send the proper data upstream to be fixed. It's underated skillz that will be ressurected and adopted by devs that will not want to be left behind. (imagine game development) No more 'elite devs that didn't even bother compiling' throwing code over the fence and then disappearing when it deploys and pretending they can't find or read a logfile 😂

I think agentic bug finding is easy and needed. But I agree that the most important thing we do is to evaluate if the feature and UI makes sense. I am discarding a lot of things the Claude does not because it is wrong or not working, but because it's doesn't make sense to the user.