But to be completely fair: I started using Copilot with Claude Sonnet at work recently, I created a .copilotignore file where I explicitly banned all .tf files, and then I went to AGENTS.md and explicitly said Copilot is not allowed to read anything from my Terraform repo. I think you know what happened next... Yes, it started reading .yamls from Terraform, and when I asked wtf it responded with smth like "the restriction was ambiguous". So
#LLMs are smart enough to replace programmers, data engineers and whatnot, but "not smart enough" to figure out what "you are not allowed here" means. Ofc it is not about being smart or not, it is about intentions, and all the bad boys like Sam and Dario simply show us our place. I still think
#AI will stay with us (at least in IT), so I decided not to reject this tool but try to explore what it is capable of. And I don't like what I've seen so far, but maybe I will change my mind, who knows. I really want to be a tech optimist. But the reality does not give me opportunities to be one.