Interesting take as usual. Utilization doesn't matter, throughput is what you need to keep in mind.

And this is why you likely need to optimize your data pipelines at some point. There are plenty of levers available.

I didn't know about the Lindy effect, this is an interesting point. Obviously I have a different setup (Plasma has been around longer than XMonad after all) but the overall advice is good.