It will come to this, you're either competent with a tool or not, nothing will change this.
It can be a terrific assistant, but your mileage may vary. I can only use myself as an example, for example I can't remember the last time I got to be this productive, I coded 10+ games (one of them is actually good, which is better than my previous track record) with LLM assisting me. I set up a Linux gaming server with 4 windows specific games, which I had zero chance of doing myself in a little time (I work as an IT Admin, but... I'm kind of an imposter, not very good, but I get the job done), it has solved several legal issues for me.
Maybe one could draw a parallel to that of using Google as an search engine, it is a terrific tool if you know what you are looking for, if you don't then it's terrible.
Same for an LLM:
If you ask an LLM to "code me a game", your game or app will suck. If you scaffold it with lots of details, have a reasonable good idea what you want it to do, describe it well, scaffold it by project managing it (like making it modular so you don't have to get the entire code rewritten every single time, noob mistake, been there), then you can absolutely get terrific results. I've tested them extensively since ChatGPT 3.5 that could hardly get definitions right, with 5.0 and Claude that are so good at it that if you DO proper project management, then you can get really good results.
So mark my words buddy, it will happen.