> but let's assume that we get AI to a level it can easily translate a grunt from a domain expert to a working + fast code.
That sounds equivalent to saying "what if we develop true AGI?" and the short answer is that I don't think we can know what the effect will be. ChatGPT isn't AGI and it isn't remotely clear what you have to do to push it there.
Is it going to be practical for the sorts of companies that pay developers now to license AGI? If not, the sort of AI that can do development might be a high end luxury.
Any sort of general AGI developer would effectively be a consultant and we pay people to manage consultants all the time. Would this hypothetical AGI require specialized tending? If so, we might just switch to being AI Developers instead of Java/C#/Python/etc. Developers.
Does this AGI developer require architectural guidance? Maybe we are all software architects presiding over squads of AI programmers.
Is AGI so good and so cheap that the market for information workers dries up? If not, maybe we take our problem solving skills to another area. If so, the economy will change so drastically I don't think it's possible to know what the outcome will be.
All in all, I guess I would say to worry less. You can't foresee or control the outcome and you don't have a better plan. So be agile, wait and see. Act when you see opportunity.