When you buy an electric blanket, you want it to be cheap to run (i.e. cheap electricity) but also you want to get the one that’s most expensive to run (i.e. highest wattage, since they’re 100% efficient at converting electricity into heat).
(Not entirely true because some of them might generate heat in suboptimal places, and there’s a limit to how much heat you’re interested in them generating, but close enough.)
When you choose a password hash function, you want one that’s slow to run for security, but then you want to choose an implementation of it that runs as fast as you can.
You’re pessimizing/optimizing within an optimized/pessimized domain, or something?
And if you’re writing a program or a blog post, succinctness is generally considered a virtue but if you go too far with that then you’re not just “removing wasted lines/words”, you’re also sacrificing features/speed/examples/clarity/something. This feels like the same kind of thing, just less obviously so?
Sounds like an analogue to the observation that optimal decisions are often on a boundary and maximizing something. Extremism in the pursuit of optimality is no vice! For example, bang-bang control: the optimal strategy is to either go all out or do nothing at all. If you are going to make your password hard to guess at all, you want to make it so hard that an adversary can’t guess it; if you want to turn on your blanket at all, then like a thermostat, you want it turned on to heat up as fast as possible and get back to the preferred temperature.
I don’t know if the blog post or program one would be an example of that… Some people argue that the middle has been hollowed out, and you should be aiming for eg. either the 10s Tiktok video or the 5 hour podcast, and nothing in between. I don’t think that’s really true—at least, I see plenty of medium-sized works still doing well in the attention market. And when it comes to programs, there are tons of programs which are well-used which are not either ultra-minimalist or the last word in kitchen-sink plumbing gadgetry, usually a whole Pareto frontier of them.
Hm, doesn’t feel like the same thing to me; I predict that most other examples of that observation would not have the property that I’m pointing at with the blanket and the password hashing.
When you buy an electric blanket, you want it to be cheap to run (i.e. cheap electricity) but also you want to get the one that’s most expensive to run (i.e. highest wattage, since they’re 100% efficient at converting electricity into heat).
(Not entirely true because some of them might generate heat in suboptimal places, and there’s a limit to how much heat you’re interested in them generating, but close enough.)
When you choose a password hash function, you want one that’s slow to run for security, but then you want to choose an implementation of it that runs as fast as you can.
You’re pessimizing/optimizing within an optimized/pessimized domain, or something?
And if you’re writing a program or a blog post, succinctness is generally considered a virtue but if you go too far with that then you’re not just “removing wasted lines/words”, you’re also sacrificing features/speed/examples/clarity/something. This feels like the same kind of thing, just less obviously so?
Sounds like an analogue to the observation that optimal decisions are often on a boundary and maximizing something. Extremism in the pursuit of optimality is no vice! For example, bang-bang control: the optimal strategy is to either go all out or do nothing at all. If you are going to make your password hard to guess at all, you want to make it so hard that an adversary can’t guess it; if you want to turn on your blanket at all, then like a thermostat, you want it turned on to heat up as fast as possible and get back to the preferred temperature.
I don’t know if the blog post or program one would be an example of that… Some people argue that the middle has been hollowed out, and you should be aiming for eg. either the 10s Tiktok video or the 5 hour podcast, and nothing in between. I don’t think that’s really true—at least, I see plenty of medium-sized works still doing well in the attention market. And when it comes to programs, there are tons of programs which are well-used which are not either ultra-minimalist or the last word in kitchen-sink plumbing gadgetry, usually a whole Pareto frontier of them.
Hm, doesn’t feel like the same thing to me; I predict that most other examples of that observation would not have the property that I’m pointing at with the blanket and the password hashing.