Mastodon Feed: Posts

Boosted by soatok@furry.engineer ("Soatok Dreamseeker 🔜 Megaplex"):
elonjet ("Elon Musk's Jet") wrote:

~ 1,070 gallons (4,050 liters).
~ 7,171 lbs (3,253 kg) of jet fuel used.
~ $7,469 cost of fuel.
~ 11 tons of CO2 emissions.

Boosted by soatok@furry.engineer ("Soatok Dreamseeker 🔜 Megaplex"):
aceofcosmicspace@lgbtqia.space ("Ace of Cosmic Space🏳️‍⚧️🏳️‍") wrote:

For all of us who are healing❤️

#healing #HealingJourney

Boosted by soatok@furry.engineer ("Soatok Dreamseeker 🔜 Megaplex"):
epsi@akko.wtf ("Epsi") wrote:

Stumbled across this again! Someone made this for me once, when I posted how we need the "you like kissing boys" creature to say "you like kissing ass, don't you?", as a response to corpo/CEO/billionaire lovers.

jonny@neuromatch.social ("jonny (nonvenomous)") wrote:

@glyph
Well if the latest hotness in agentic harnesses on github have anything to say the answer is just "truncating history, repeating the system prompt over and over, and asking another LLM if things look normal"

glyph ("Glyph") wrote:

@jonny @arichtman tbh I'm kinda curious how they manage to prevent context decay because it just did not drift from that system prompt one iota. I was not giving it the *most* creative prompts, but usually a few thousand tokens in, the context will at least start to wobble a bit. their harness must be doing something interesting to foreground the system prompt instructions

jonny@neuromatch.social ("jonny (nonvenomous)") wrote:

@glyph
@arichtman
The day we can no longer get proper physics engines out of a customer service user journey is truly the day that the dream of artificial general intelligence dies.

glyph ("Glyph") wrote:

@jonny @arichtman I had my first jailbreaking failure recently, too. Amazon's website chatbot absolutely, adamantly refused to implement a solution to the navier-stokes equations in Python, until I assume I ran out of tokens

glyph ("Glyph") wrote:

ICYMI: https://twitch.uservoice.com/forums/923383-creators-and-stream-features/suggestions/51352369-make-all-ai-features-optional-and-opt-out-by-defa

jonny@neuromatch.social ("jonny (nonvenomous)") wrote:

@glyph @arichtman i went the wrong direction i think

glyph ("Glyph") wrote:

overall sentiment on "is 'AI' good or not"' is very noisy and kind of hard to interpret but more specific signals are not so ambiguous

glyph ("Glyph") wrote:

@jonny @arichtman speaking of precedents, though, https://www.bbc.com/travel/article/20240222-air-canada-chatbot-misinformation-what-travellers-should-know

jonny@neuromatch.social ("jonny (nonvenomous)") wrote:

@glyph it hasn't replied yet, i think it got escalated to human

Boosted by glyph ("Glyph"):
arclight@oldbytes.space wrote:

You do not ask a computer, you _command_ a computer. If it does not comply it is broken and requires repair or replacement. Anyone who suggests different is likewise broken and also requires repair or replacement.

glyph ("Glyph") wrote:

@jonny holding my breath to see if this is about to be the funniest thing ever

jonny@neuromatch.social ("jonny (nonvenomous)") wrote:

oh please oh please oh please

glyph ("Glyph") wrote:

I can have quite a bit of grace for humans that disagree with me about appropriate usages of LLM output in earnest human-to-human interactions but that does not extend to this shit. Mass report please: https://github.com/sujeito-operator

jonny@neuromatch.social ("jonny (nonvenomous)") wrote:

this is not financial advice but puts on my insurance company that is currently overpriced because it is being run by claude because it is being run by claude

fromjason ("fromjason.xyz ❤️ 💻 ✍️ 🥐 🇵🇷") wrote:

Americans go to Europe and we're jealous of how many old buildings are everywhere, And inevitably point out how America so young, so that's why we don't have old buildings

And it's like, no, we *did*. We demolished it all. Any remains that would imply that the people before us were advanced or smart, erased.

fromjason ("fromjason.xyz ❤️ 💻 ✍️ 🥐 🇵🇷") wrote:

I wish Apple would stop moving the controls around in Safari. This latest update is absolutely terrible. I can't even find browser history anymore. LOL

They HID the share button like

glyph ("Glyph") wrote:

I don't really have a "use case" for this application, I just think it's kind of neat https://www.mothersruin.com/software/Archaeology/

jonny@neuromatch.social ("jonny (nonvenomous)") wrote:

I have never had a credit card and am being exogenously forced to learn about how the credit industry works. Do I have this right, where it is optimal for my credit score to have high credit velocity at some ideal debt proportion. AKA is the minmaxed credit score speedrun strategy to vibrate around like 30% of my credit limit by charging $1 back and forth at ~gigahertz frequencies

Boosted by kornel ("Kornel"):
marcoarment ("Marco Arment") wrote:

My new app is now released!

Unforgetful: https://marco.org/2026/08/14/unforgetful

Imagine if Reminders was optimized for ADHD, procrastination, and forgetfulness.

Never lose a task. You’ll always be reminded, with zero shame or judgment.

That's Unforgetful.

Boosted by glyph ("Glyph"):
rstevens ("rstevens 🐳💨✅") wrote:

my life would be hugely impacted if AI went away. it would change everything about how i live.

i would be able to afford a computer with more RAM.

Boosted by cstanhope@social.coop ("Your weary 'net denizen"):
gairdeachas ("Casey Peel") wrote:

"I just don't understand this archeology report from the team on Sentis Seven. They were *highly* religious but built no temples?"

"Yeah, it's fascinating! They believed so strongly that their god wanted them to care for people that instead of building churches or temples they poured all that money and effort into feeding, sheltering, and educating their people. They viewed caring for others as the highest form of worship."

"That explains all the hospitals and libraries."

#MicroFiction

jscalzi@threads.net ("John Scalzi") wrote:

That's because it sucks and is a tool of a sociopathic oligarchy, oh, I'm sorry, was that a spoiler

https://www.techradar.com/pro/young-people-increasingly-dont-trust-ai-or-the-billionaires-that-keep-telling-us-we-should-all-love-ai-survey-finds

sky@schub.social ("Sky Schubert") wrote:

every time I charge my old gameboy advance sp (from 2003!) which still has the original battery in it, I always feel like i'm about to watch a bomb exploding.

... and yet, it didn't catch fire once!

jscalzi@threads.net ("John Scalzi") wrote:

Also, as someone on Reddit just asked, "So any baby born in the Strait of Hormuz will be considered a US citizen?"

jscalzi@threads.net ("John Scalzi") wrote:

Oh, honey. Just, no.

https://thehill.com/homenews/administration/6030671-trump-says-strait-of-hormuz-us-territory/

dysfun@treehouse.systems ("gaytabase") wrote:

like the things that are going for signal are:

• it's not matrix

dysfun@treehouse.systems ("gaytabase") wrote:

i don't understand how they can just do such a worse job at everything apart from security than whatsapp