I'm not saying anything that hasn't been said a thousand times before. But I find it's evident when I'm getting it to do something I consider myself good at. And that's what's worrying to me. I work in DevOps and there are a couple of tools I'm really good at. If I were trusting the output all my configuration would be outdated and set up like a blog example with all the issues and shortcuts one takes in them (and I see that in the PRs that I get from the team members that rely on claud heavily). But if you didn't know the tool it would look fine. So when I code with the agent, it all looks really good, but I must be missing things right? For scripts that have no impact if they fail, I llm the shit out of that.