Possible futures in my mind:

- Junior jobs become scarce. You either jump to senior very fast, or you are dumped out of the industry. The juniors that make it are those special unicorns that somehow learn everything about the job, including how to do it with AI, within a couple of years. There's a little bit of guidance, but mostly it's the kids who have good taste in blogs/books/videos that end up learning it all on their own. Also the kids who have the motivation to keep studying without a syllabus.

- Instead of junior devs, we just have domain experts who are crap at coding. Quants who can write a model in pandas, but when they venture offpiste, they get AI to build them a monstrosity. Working monstrosity, but if you could code, you would cry. This ends up happening in every industry: there's very few coders left anywhere outside of FAANG, everyone just does the modern equivalent of thinking Excel has solved their problem. Balls of spaghetti the size of which the world has never seen are written, hidden in various domains.

- Universities wisen up about how to teach people to use AI. Once upon a time, they used to teach you how to punch holes. Assembler was taught. Systems languages like c++. Java, JS, Lua. Kids who came out of these universities were somewhat ok for industry. Why not AI as well? There are going to be lessons learned roughly this decade that will be useful to teach the kids. What to tell the AI, what not to. How to leverage it to make the most progress.

Perhaps all-of-the-above:

Universities will (still) teach only part of the necessary tradecraft, but that fraction will include some basics for how LLMs can churn out dodgy prototypes. Junior/introductory roles will feature the crap-work of taking excel-esque monstrosities and making them marginally less-terrible.