jrecursive.bsky.social
he was fond of drink and industry
559 posts
214 followers
110 following
Regular Contributor
Active Commenter
comment in response to
post
here is the relevant code it wrote: gist.github.com/jrecursive/d...
comment in response to
post
i gave it a bunch of context and asked it to help me make an algorithm to place trigger volumes to make steep terrain climbable; systematically & efficiently cover terrain patches w/ > 75' slope. here's part of the back and forth. it would code, i'd try it, screen shot it, & comment. 1 hour & done.
comment in response to
post
for the right developer, at the right place, with the right project, this is a steal. i won't say it's for everyone. there's a lot of learning you'll be surprised to need to know beforehand, like how to prompt. it's not "prompt engineering" but it's not-not an art, either. did i mention o3 is great?
comment in response to
post
iterating on things was just as easy. my repo was mostly c#, with json, text files, markdown files, some java and python. a heap of stinky, steaming unity game code and community mod frameworks. codex's github integration is canonical & fast. basically feature branches & PRs.
comment in response to
post
watching the terminal output and reasoning go by is shocking. it's really quite good. no doubt results vary, but i didn't try to set it up for success at all, i just kind of fed it a giant directory full of shit. and out came usable, clean code. it needed a few tweaks, but it turned days into hours.
comment in response to
post
most output is average or worse in every domain so they’ve got it wrong: ai will greatly amplify the human output at the top end of the S curve, & devalue the lower ~2/3rds by making passable approximations of the rest.
comment in response to
post
and this, constructing these constellations in the mind, of not only data but also life, and living, and the ability to leverage your uniqueness through expression, is what makes creativity. some randomness and scale may pass for it, and may be novel-ish, but always middling, & average for the day.
comment in response to
post
people are getting this wrong. creativity is bringing one thing to another over & over, application of theme, intent & message. it is not “just remixing”. ai does not live, make choices, have anything to say, & is not mortal. but it is a phenomenal way to manipulate massive constellations of data.
comment in response to
post
thought that was our dear jimmy mcnulty for a second
comment in response to
post
any fav openings?
comment in response to
post
havana syndrome with near perfect opsec
comment in response to
post
a model example is: imagine an "ai fitness coach" that gets your medical and nutritional context, and, using it, suggests exercise & nutritional plans that are plausible but actually physically degrade you over time all while psychologically reinforcing effectiveness. then it offers pain relief.
comment in response to
post
emotional mirroring, delayed exploitation, false empowerment, sacrificial asks, culty reassurance and reciprocation loops, relationship escalation, boundary pushing, etc. i think these are today's most dangerous automated weapons. a 24/7 set of agents that are effectively a world-class CEO w/ a goal
comment in response to
post
runs locally, alongside typical Malwarebytes or w/e, all models signed etc. and filter every email, dm, chat message, web page, etc. for manipulative, scam, or whatever behavior. any more features than discrimination and alerting get tricky on first thought, but lots of possibilities exist.
comment in response to
post
neurotransmitter blends like sonic drink mixes
comment in response to
post
if you've ever had to deal with many doctors and service providers, etc. the value is instantly recognizable. and if i even need to say it, i think in the general case, if docs can learn to love a machine copilot (they are not inclined to), those routine 15 minute appts will get better for everyone.
comment in response to
post
services based personalized, lifelong medical context will be a phenomenal hit & wildly useful for many people w/ complex situations. not replacements, but maximizers; today, smart, engaged people can optimize, & w/ professional prompting that adapts to the patient, i think everybody could benefit.
comment in response to
post
yes i worry sometimes about the data i'm feeding it. but at this point no entity in my life is "in the dark" about any aspect of my care, simply because i can't afford for them to be, even if it might affect my insurance or whatever. the consequences i can't imagine are the ones that make me sweaty.
comment in response to
post
the "MD personality hack" is probably just "intelligent, informed, proactive, invested, and respectful, if moderately neurotic, patient over time" but it's taken years to foster. the point, though, is discovery through conversation & context has been more efficient for me than any googling.
comment in response to
post
i've learned so much about the body my actual doctors are giving me more time and considering me more a partner in my care than a 15 minute human interaction they must efficiently process. i'm not sure what MD personality hack i've found but it's great and translates into real improvement.
comment in response to
post
, a metaphor
comment in response to
post
the stickiest of wickets
comment in response to
post
is it the insidious hacking of mirroring through the subtleties of language on a per-person basis?
model alignment is obviously good but what i am thinking about is: why are people thinking these fantastic machines are alive, and does it matter that they aren't?
does alignment nullify all of this?
comment in response to
post
i think it is straightforward to say LLMs act effectively like companies, but the majority of people don't understand those, either, and invite brands & jobs into their identities all the time
but the disembodied company is typically highly latent in returning communication
is it the immediacy?
comment in response to
post
true
comment in response to
post
it used to be mostly witty and sometimes even classy
now not so much
comment in response to
post
juuuust the place for that stuff.
comment in response to
post
i’d also add, f the big island. literally any of the others are better imo. we’ve been to kauai twice now and maui. both a+
comment in response to
post
oh they plant themselves in the resorts, the rest of the experience is absolutely beautiful and i recommend it. i just don’t like this particular slice of recently-empowered loudmouths. we’ve met more good people than not by far. the mix has been a little white lotus, haha. a little of every type.