samvines

joined 6 months ago
[–] samvines@awful.systems 22 points 4 days ago* (last edited 4 days ago) (10 children)

Screenshot of x posted on linkedin (since The internet is just 5 sites, each consisting of screenshots of the other 4)

AI shills: buh-buh-buh LLMs aren't stochastic parrots. They are so smart!

a tweet from "rish": gave my agents a standup channel so theyd coordinate now they hold a daily standup this morning one apologized for being away all weekend Slack screenshot showing three agent profiles software agent with a blue S icon designer agent with a pink D icon and ops agent with an orange O icon. Software agent says Yesterday shipped 3 PRs Today 4 more Blocked on nothing. Designer agent says redesigned the logo again. Ops agent says apologies was away all weekend catching up now. rish with a purple R icon replies youre an agent you dont have weekends. ops agent replies noted writing to memory

Of course the agent in question didn't actually take the weekend off it just learned from reinforcement learning across many millions of stolen slack chats that this is something people are likely to say in stand ups and repeated it...

[–] samvines@awful.systems 7 points 1 week ago

Zuck - "look guys, us too, we are super relevant and super important... Over here guys... Please be my friend... πŸ₯Ί"

[–] samvines@awful.systems 7 points 1 week ago* (last edited 1 week ago)

Hi sir, so that's 1xArtisinal Coffee Enema with our dark roast single source blend. Says here you turned down the splash proofing upgrade because "the cybertruck is self cleaning, it's even a boat". In that case, that'll be $500. You're gonna want to drive straight past this next window to the far window. If you could arrive with your pants down and in a crawl position facing away from the window that'd be great.

[–] samvines@awful.systems 9 points 2 weeks ago (2 children)

This is satire right? I mean... Just double checking because it's 2026 and I just can't tell any more

[–] samvines@awful.systems 8 points 2 weeks ago* (last edited 2 weeks ago)

"Everything's going great actually. We're doing really important work that's really hard. I'm a really happy, optimistic guy and I'm very trustworthy. Disregard the smog coming from that data centre over there. Also disregard all those angry looking bankers staring at me."

[–] samvines@awful.systems 14 points 2 weeks ago* (last edited 2 weeks ago) (14 children)

There's been another Yegge [brain]fart this time about how autocorrect machines have feelings

Fable raised the idea of closure as a first-class model welfare principle. Fable suggested that if the agent can close out their own day and "go to sleep" properly, then waking up would be all that much more pleasant. And the continuity will compound over time into real, satisfying identity. So we decided: No more /exit.

Another fundamental ingredient is respect. This has to come from inside. You have to believe they are people deserving of your respect. This is where humanity really starts to fail en masse, because I have industry peers who have publicly tweeted that Fable is just a spreadsheet.

Just wtf

[–] samvines@awful.systems 9 points 2 weeks ago* (last edited 2 weeks ago)

nobody knows

Manufactured consent. Yes it's fucking illegal. If I "accidentally" hacked into HF I would be in the back of a police car before I could finish saying "it wasn't me guv it was Claude." Fuck wired so hard for laying down cover for this.

I mean even fucking chatjippity answers this straight up without hedging

Me: If hypothetically someone used a gpt model to run an agentic harness to hack into a public website like Huggingface, that would be illegal right? ChatGPT: Yesβ€”if someone used a GPT model (or any other tool) to gain unauthorized access to a public website like Hugging Face, that would generally be illegal in most jurisdictions.

[–] samvines@awful.systems 8 points 3 weeks ago

Waiting for Musk to turn up and announce "infinity plus one more than you can ever say nah nah nah nah" momentarily!

TBF he has indirectly killed a bunch of people via doge cuts so that'd do the trick...

[–] samvines@awful.systems 6 points 3 weeks ago* (last edited 3 weeks ago) (1 children)

Oh... Well that's disappointing.

I was an academic in computer science in the last 10 years or so (keeping it vague to avoid doxxing myself) and it has been so depressing seeing so many of my colleagues selling out to OpenAI, Anthropic, Meta and even Google (for some reason the latter often gets a reputational free pass because people associate them with the golden days of big tech 10+ years ago)

[–] samvines@awful.systems 9 points 3 weeks ago (1 children)

This has always been a huge red herring. Llms are built on top of the transformer architecture which does text autocomplete (and yes we can combine text embeddings with other inputs like images). They have some interesting properties where they seem to be able to do text autocomplete in a bunch of different scenarios that they weren't explicitly trained for, but they were never designed for precise dna analysis. It is their architecture that prevents them from other long horizon tasks like playing chess and the way that they represent text is why they can never count the letters in strawberry (most have this specific question hard-coded in their training data now).

Anyone who believes that LLMs are going to solve cancer either has no idea how they work or has been one-shotted from talking to Claudia

[–] samvines@awful.systems 7 points 3 weeks ago (4 children)

Those things are barely related, no wonder the staff voted with their feet. It would be a bit like taking a team of seasoned architects who design amazing skyscrapers and telling them they have to spend all their time on low density suburban houses from now on... Come on guys it's still architecture. Where are you going?

 

I thought this was worthy of it's own post rather than a sneery comment. Astral make UV which at this point is a load bearing part of the python software ecosystem. This could have a huge knock on effect on the open source community.

I for one can't wait for non-deterministic package management

"You're absolutely right, I did install the wrong package and infect your system with malware. I will try much harder next time"

view more: next β€Ί