The wet dream of AGI ... LLMs can't reason. Period. This is just a "guess game". I write code with these tools every day and I can tell you that it's not worth it. If you would write bigger software systems with just the AI's output would be complete crap (unmaintainable). It invents software functions and arguments that never existed (and apologieses later).
Training these models needs so much power (for mostly stupid questions from the users they could have answered with a simple research), it's ridiculous.
And what do you think if there's more and more of this AI generated s.. stuff—the models will learn on their self-produced crap. It's a bubble (like self-driving cars) and it will burst if the core methodology of AI does not change fundamentally.