> In the other hand maybe all these prices drive folks to program like we used to, conscious of the hardware limitations, without extra slots to rescue from bad programming.

I don’t understand why this comes up on all of these topics.

Dire need will compel a naked woman to learn to weave her own clothes, but there is no weaving machine for just making more efficient software. (American) Software developers are expensive, now I guess compute will either be expensive, more centralized, or have much more demand from competing interests (vibe coders who are implementing their own bespoke software or abandonware-for-one), and compute-for-code (GenAI) is a whole emergent engineering problem.

Then we hunker down and listen to or read a piece by Muratori, I guess, just practice some of those principles? But what seems to keep happening is that we get stuck by constraints that go beyond just writing more efficient code as a solo contributor on a solo project. That you can predict the efficiency of the code by the application domain suggests that there is a whole big system (of people and processes) above our heads that is not simple for any person or group of people to untangle.