The future looks to involve a mixture of AI and traditional development. There are things I do with AI that I could never touch the speed of with traditional development. But the vast majority of dev work is just traditional methods with maybe an AI rubber duck and then review before opening the PR to catch the dumb mistakes we all make sometimes. There is a massive difference between a one-off maintenance script or functional skeleton and enterprise code that has been fucked up for 15 years and the AI is never going to understand why you can't just do the normal best practice thing.

A good developer will be familiar enough with AI to know the difference, but it'll be a tool they use a couple times a month (highly dependent on the job) in big ways and maybe daily in insignificant ways if they choose.

Companies want a staff prepared for that state, not dragging their heels because they refuse to learn. I've been at this for thirty year's and I've had to adapt to a number of changes I didn't like. But like a lot of job skills we've had to develop over the years — such as devops — it'll be something that you engage for specific purposes, not the whole job.

Even when the AI bubble does burst, AI won't go away entirely. OpenAI isn't the only provider and local AI is continuing to close the gap in terms of capability and hardware. In that environment, it may become even more important to know when the tool is a good fit and when it isn't.