Yet another article that reinforces my opinion that people that unleash agentic LLM’s on their codebases with zero oversight cannot be trusted; and that LLM’s can be used responsibly but it takes a lot more restraint and deliberate cognitive involvement than most people can handle.

I think I take issue with the idea that it’s the “chat” interface that makes people shut off their brain and stay in that groove. I feel that the answer is more the very nature of LLMs being one of our attempts to get AI to pass the turing test. Continuing the conversation is such a core part of what an LLM is beyond the UI. If it made you think, you wouldn’t believe it to be a human.