I started to use claude code heavily at work two weeks ago and I've been able to slack off in the office so much more than I used to. working maybe 60% as hard as usual. at times I'm really optimistic about this being the future AI will make possible for everyone

hopefully LLM memory systems will soon get good enough that you can make a model that will grumpily refuse to answer your question if someone's asked it to the model before. it'll be just like the good old days

she's right-clicking to pull up a context menu

sending back a femboy fatale agent to seduce and neutralize peter thiel before everything gets out of hand

wanna say Gordon Moore cause if you don't have someone making exponential growth an explicit goal of the seminconductor industry, maybe it won't be one? but I doubt he's particularly individually responsible for that. can't be anything on the algorithms side that stuff's all easy

they figured this out a decade ago on lesswrong

I like to say "they shouldn't let depressed people do utilitarianism" though it's a bit flippant. I think this comes up too in people thinking about wild animal suffering versus wild animals having a pretty okay time

americans, like mushrooms and tofu, take on the flavors they are cooked in

do you know about his glowfic forum stuff? I'm only vaguely familiar but my friend who's more aware has tossed the word "hyperpregnant" around describing that scene. I think there's some crazy shit going on there

it's kind of yudkowsky's whole deal that he doesn't get extremely uncertain about his beliefs for those reasons

central park itself should be the datacenter. have these people never read "all watched over by machines of loving grace"?

substrate invariant, or maybe more accurately invariant over some substrate subspace

imo you're supposed to read asimov in middle school, clarke in high school, le guin in college, and lem in grad school. I fucked up and read lem in college and le guin in grad school and it's set me back a bit.

a robot gibbon would be sooo fucking cool