Reason why it's super hard for juniors at the moment. For "agents" to produce good results, the operator must have the experience to direct it in the right direction and properly validate the output.
I've had some weird bug investigations where it has suggested wrong fixes confidently multiple times before arriving at the actual fix. Only with extra details from manual testing did we arrive at the real bug. The ~5 first suggested fixes would've hidden the root cause even more.
I've had some weird bug investigations where it has suggested wrong fixes confidently multiple times before arriving at the actual fix. Only with extra details from manual testing did we arrive at the real bug. The ~5 first suggested fixes would've hidden the root cause even more.