It's an interesting thought, since you might effect a larger momentary difference than you could achieve any other way, depending on personal predisposition. B

It's an interesting thought, since you might effect a larger momentary difference than you could achieve any other way, depending on personal predisposition. But for one, the total change we can achieve as a society that way is limited--gcc and LLVM are almost optimally efficient already, the total possible saving probably wouldn't exceed 10%; larger changes could be achieved on higher abstraction levels (higher-level languages, full-program analysis, ..), though. And secondly, such efficiency gains tend to be eaten up as they come. The same energy is just spent on either an inefficiency in another place (e.g. saving on human cost instead), or on new things that can be done with it. (Similar to how ~20 years ago the thought was that the office of the future would be paperless, whereas in reality the amount of paper used hasn't decreased last I heard.) This is because computing is used as much as it is affordable[1]. Making it cheaper just affords more of it. E.g. if you make LLVM or gcc faster, it allows for people using Python for longer before moving to Go or Rust or whatever. PS. I guess if you're subscribing to the view that technological progress is good, that's still a positive. But technological progress can also be viewed as being largely the reason why we've got a climate problem, and especially the speed in which we have progressed is why we're running into the problem so hard that it will be difficult to handle. [1] affordable in the sense of energy cost for large organisations, or in the sense of time cost or human cost for small organisations.