I think the math is pretty clear.

Today AI is 100% automating the software engineering market. We all know it and it is not evenly distributed with some fields being more impacted than others.

However, what is not mentioned anywhere is for long this will continue and what shape it will take. It is a trickier question.

There are two factors in my mind: cost and volume.

On the cost side, AI coding agents could become unfordable. What happens then? Back to engineering jobs? Hard to tell. It might impact seriously the bottom line of many companies while jobs remain lost forever.

On the scale side, the more code is written the more complex it becomes - that is a fact. Who is going to look after this code? More AI? That will make it more costly. Or is that most software is dead on arrival - then why build it?

So it is not clear cut. The investment is there but how to turn the investment in meaningful long-term ROI is not.

Even on the highly subsidies plans most software written today has marginal ROI. Building landing pages and features that nobody asked for is not a good execution plan and frankly I have done this mistake myself.

The last bit I would like to flag is that if software becomes almost free then why build at all? Why would companies even bother building anything at all?