TypeError

@fubarchitect.com

exists. old enough to be tired. young enough for the wired. not safe for anyone

we are going to invent something more useful and polyglot oriented than github shortly here and that will help. approximately an index of function-shapes, which can then acquire names so both humans and models can converse at an appropriate level of abstraction. this works today for ex: gof patterns

✧⁺⋆🌘 Dream Rotator NPC Alex 🌒⋆.˚₊✩@lastnpcalex.agency · 17h ago

ya i don't think people have thought about this fully solutions to problems becoming more bespoke, and private, isolated, and illegible to others does seem to imply problems that show up at scale (beyond the level where any individual with an agent could solve) will be harder to solve, not easier

ok i don't know what was happening those first few days but i have certainly relearned 'desire for a cigarette' and now it's just, in everything we're gonna make it but i might have to pick up diet coke, i finally get it. i'd do sparkling water but my kids drink that up

the thing we will have difficulty translating into other domains is unfortunately not the world modeling but the infinite parallelism; programming isn't uniquely llm shaped, but programming and llms are gradienty and parcelable in a way that sheep shearing is not

Citizen Platano 🇵🇷@daniloc.xyz · 5h ago

an oversimplification you see a lot: “programming is an LLM shaped problem” this is often extended into the claim that it’s a UNIQUELY LLM shaped problem space, and other problems aren’t so tractable and I think the truth is much more interesting: programming was pre-plumbed for agents

i think "this is what you voted for" should entitle anyone it was spoken to one clean shot to the mouth of the speaker, or at least upside the head

PSA there was an X bug recently that reset DM settings. go check your settings. I did my own research about this. also Bluesky is a daycare for adults and if your DMs got reset over there that's what you deserve for being there.

this is a running theme, and something @aparker.io tried to put me on maybe 5 minutes before i was ready for it; agents mostly free you to aim your compulsive 'i should ask someone' behaviours at a respondent who wont give you a politically convenient lie

TypeError@fubarchitect.com · yesterday

danluu called the first thing normalization of deviance, and at first i thought agentic dev would make it worse, but it mostly just unleashes me to try not-doing-the-dumb-seeming-thing much faster

something interesting is that in a software company if you ask "hey why do we do this seemingly stupid thing?" a bunch of humans will evade telling you that no one knew any better by saying "oh, person-who-left did that, for very-smart reasons" leaving you to disprove that

like i want one to consider how totally worth it mars + lakes + earth women would be. perfectly cromulent place to build a life but that's like, just the appetizer with earth. there's so much more and almost all of it i good, especially if you stay the fuck out of the deep ocean