Really, there's this imbalance between what you can achieve and what you can't imo. Analysis and summaries are excellent use cases for AI, but creative tasks like writing code or text is highly dependent on the circumstances, as far as I can tell. Code written by AI is undeniably fragile. BUT: this fragility only matters in certain environments, like systems with medium or higher complexity, where interop matters, with low latency or high load characteristics, basically everywhere, where it can have a compounding effect. I've seen people do great work with AI with configuration languages, that offer a validator. Another excellent use case is asking AI about your code base in self-trained specialist models. We're still at the beginning of discovering, what's feasible to do in our current stage with AI. It's probably going to keep disrupting every part of the economy for years to come, after the hype died down a bit. Most importantly though, it gives people a low qualification entry to many fields, that previously needed huge swaths of time investment. As a specialist, it enables certain work, that required generalist knowledge before. Most importantly, as a worker without training or without good education, it gives you wings, lowering the bar and possible entry point for low skilled work, while rising the ceiling of such.