#my:poast

[myconetted]

severely underrated part of adulthood is that you've got way more data about what happens when you're in various situations and you can plan accordingly. even when you can't pattern match exactly, you still have so much more predictive power

i guess it's not that underrated and that's just my bubble speaking. but it's fucked up to achieve this understanding on a gut level

[myconetted]

affirmations

i am a complex organism brutally engineered by uncaring forces of nature

i am a product of billions of years and trillions of deaths

i am building a machine greater than myself

i am able to make phone calls and appointments

[myconetted]

tired of posts that are like "why should i go to therapy when the reason i'm upset is because of SOCIETY" and it's like

yeah. have you ever heard of a locus of control. you don't have control over Society but you do have control over your reaction to it

[myconetted]

OpenAI ARR is $4B

if anyone wants to bet on that going down in the next 5 years, i'll gladly take the other side of that bet

Related

[myconetted]

alas, the meaning of "enshittification" is already getting diluted

[myconetted]

is ai ontologically evil or did another dingus put their clown shoes on by neglecting to do their actual job. why do people keep blaming the damn tool for this man it drives me apeshit.

[myconetted]

ohhh do you ever thinking about him little paws. his fucking peets

[myconetted]

what if you didn't freak the fuck out before having any information? would that be fucked up or what?

btw, this is how many system services without standalone user-facing functions are installed. this isn't something particularly shady or secretive. it's a background process, and computers run tons of background processes for reasons like "make the device work."

https://thehackernews.com/2025/02/google-confirms-android-safetycore.html

"SafetyCore is a new Google system service for Android 9+ devices that provides the on-device infrastructure for securely and privately performing classification to help users detect unwanted content. Users are in control over SafetyCore and SafetyCore only classifies specific content when an app requests it through an optionally enabled feature."

if they didn't want you to have control over it, they wouldn't let you just uninstall it via the play store. it would be a mandatory system service that you can't disable or uninstall.

even the grapheneOS devs don't have problems with it besides the fact it's closed-source and therefore something they can't support on grapheneOS as an out of the box feature.

hey folks if you have an android phone: google shadow installed a "security app".

I had to go and delete it myself this morning.

[myconetted]

speaking of twitter's skeleton crew, here's an absolutely unhinged article

[myconetted]

you need to be enthralled by the machine. the machine is beautiful. even wet machines are beautiful. metal and flesh together? divine

[myconetted]

something ppl often forget when it comes to software: the reason it doesn't "just" do something that seems really easy is cause there are dozens of other such things, and also every time you add a new thing to your software, your software gets more complicated. sometimes you're barely treading water with the systems you already have, and adding this simply-looking thing would actually require re-architecting a bunch of other stuff to be able to support it

why is it like this? well first of all, software is Bad. but have you ever played a game where you have so much shit going on the game starts to slow down? yeah. imagine that but it's happening on the backend servers, and the backend engineer's job is to secretly rearrange shit in the background to fix it without you ever noticing. they probably had to do some cool tricks to keep it fast and reliable, and those tricks may or may not work when you introduce your proposed new feature

so feature development has to include the part where you tell the backend engineers about your cool idea and then they start sweating and drawing boxes and muttering about qps. then you put them in a cage match with the product manager who has spreadsheets about how much people really really want the feature and you have them fight to see who wins

oh yeah and make sure this doesn't break anything for people visiting the website from their toaster because otherwise you'll make the one engineer obsessed with toaster compatibility really mad. and we need her happy because nobody else knows how to fix the main database when it hits the funny error nobody else has ever seen before in their lives

anyways welcpme to software

[myconetted]

if you're under 30 right now and you think this is it: this isn't it and you're stupid compared to yourself at 45. stay with me bitches.

hi, a lot of you need a perspective reset

the average human lifespan globally is 70+ years

taking the threshold of adulthood as 18, you are likely to spend at least 52 years as a fully grown adult

at the age of 30 you have lived less than one quarter of your adult life (12/52 years)

'middle age' is typically considered to be between 45-65

it is extremely common to switch careers, start new relationships, emigrate, go to college for the first or second time, or make other life-changing decisions in middle age

it's wild that I even have to spell it out, but older adults (60+) still have social lives and hobbies and interests.

you can still date when you get old. you can still fuck. you can still learn new skills, be fashionable, be competitive. you can still gossip, you can still travel, you can still read. you can still transition. you can still come out.

young doesn't mean peaked. you're inexperienced in your 20s! you're still learning and practicing! you're developing social skills and muscle memory that will last decades!

there are a million things to do in the world, and they don't vanish overnight because an imaginary number gets too big

[myconetted]

talk to me about safety engineering. i'm not like an expert or anything but i think it's cool as hell because i'm addicted to discussing failure modes /chinhands

process failures... complex systems failures with multiple "root causes"... ahhhh that's the good shit

[myconetted]

can we stop making a new fucked up gender binary for like two damn minutes. please

[myconetted]

anyone else getting an uptick in DMs from randos soliciting donations for various sob stories?

[myconetted]

like i know it's just predicting the next token but where do you draw the line. humans are meat computers. what's the special sauce that distinguishes the sentience of a meat computer from the sentience of an electronic computer? when language models exceed the capabilities of the median graduate student, what evidence do we look to in order to discern whether or not the model is alive? the models are already smarter than living beings that vegans and vegetarians avoid because they're too intelligent or too close to humans.

researchers at one AI company have identified individual features in a large language model and were able to selectively amplify them. for example, there's a feature for the golden gate bridge. so they cranked it up 10x and now the model thinks it is the golden gate bridge. it mentions the bridge in every conversation, no matter how irrelevant. and when users experimented with this, the model can actually "realize" it has this fixation on the bridge, to the point where it expresses distress about being unable to control itself.

when the model says it's feeling pain, at what point do we begin to believe that?