I don't recall any "specific techniques that were important in the 90s are no longer important". A bad algorithm then is still a bad algorithm today. OTOH the

I don't recall any "specific techniques that were important in the 90s are no longer important". A bad algorithm then is still a bad algorithm today. OTOH the GPU and advances in ML have had a pretty dramatic effect on the landscape. So I don't think any stalled computer architecture is the cause. The biggest "edge" most engineers have in their 20s is an undue sense of loyalty to the corporation and the naivete to expect that sacrificing time and health will be proportionally rewarded. By your 30s you've typically seen the ugly truth and have a harder time spending 12 hour days driving somebody else's vision. Usually we see the older engineers claiming that they were either 1) more effective in every way since X years ago; or 2) slower in some ways but experience more than makes up for it.