dvc

yeah I've definitely heard of that and therefore know that this is a significant finding

an ai company is basically just a having way more compute than you company

Flávio, these are all really good ideas

did you guys think this would make him seem less cool somehow or what

It would be cool if there was like some kind of nonprofit organization dedicated to developing open AI

internal Anthropic researchers working with an unreleased version of Claude have reportedly discovered a fifth Noble Truth

Constrained optimization is my business and business is maxxing.

speaking as somebody who's been in/around the field for a long time this is it right here

The threat to the frontier lab’s business model is building a cheaper mousetrap, we have to keep access to Chinese models or we’re going to get fucked.

This is going to be a tremendous victory for hideously bad governance strategies that should be illegal, I hope everybody’s ready for that.

The world needs a high-quality, truly open language model. I wrote about why it’s necessary and why it’s easier to build than you might think.

holy freaking shitballs. matt yglesias just posted something stupid somewhere else on the internet?!

Good to see CoPilot finally pay off for them

Luddites want Anthropic to fail because they hate AI. I want Anthropic to fail because they keep the most cutting edge cybersecurity tools to themselves and think user agency leads to “harms” We are not the same

*MICROSOFT TO OFFER BUYOUTS TO 60,000 WORKERS IN LATEST COST CUT 27% of its workforce!!!!!

*ANTHROPIC FILES PUBLICLY FOR US IPO www.sec.gov/Archives/edg...

*ANTHROPIC WARNS OF CATASTROPHIC, EXISTENTIAL RISKS TO HUMANITY

Anthropic cumulative operating loss since inception is only $12 billion and at their current run rate operating margin they will completely recoup that loss by Q1 27 even if revenue is flat. So much for “these companies will never make money” lmao

Per their IPO docs, Anthropic estimates that between 5% and 7% of its revenue is from breaches of its terms (distillation, banned ML/AI work, etc).

ONE BRAZILLION YEARS LULAG

my Portuguese isn't great but I think I can work this out

It's insane that the campaign against El-Sayed is basically that he has some secret relationship with the people that the president's family has a secret relationship with.

In his acceptance speech, Lula says he will work to regulate digital platforms in Brazil. “Digital crimes must be dealt with like real-life crimes.” This comes after Meta directly interfered with the Brazilian elections, favouring Flávio Bolsonaro.

“PARECE QUE DEFENDER SUA DEMOCRACIA DE DENTRO PARA FORA DEPOIS DE UMA TENTATIVA DE GOLPE TE DÁ PODER DURADOURO”

I'm no fan of Lula, but it's hard to express how big a blow his shock win is for Trumpworld. There's a reason they were meddling to support Flavio Bolsonaro. It's not just the graft for Trump allies, and favorable resource access. It's also the political & military support for Trump's LatAm agenda.

RIO DE JANEIRO (AP) — Brazil President Luiz Inácio Lula da Silva wins fourth term, according to electoral court.

Lula:

Read: Serbia's long-serving strongman leader conceded defeat to a diverse, student-led opposition movement that spent two years protesting his government

*Jack Donaghy voice* That’s Kimi, we count those

At first glance this may seem like an AI bit But ten years ago, she would simply ask the nanny or housekeeper that actually spends time with her kid for those bullet points, so this goes back to being a political economy bit

cool, cool

when someone's linkedin bio says they're a "Founder" it's the quickest way to identify them as a cold blooded sociopath who would slit your throat for your pocket change

what makes claudish different from human text? we're getting real psycholinguistic research! idea is that models are optimizing for information density rather than human understanding. consequences flow from that. i think i buy it. arxiv.org/abs/2610.29162

I'm sorry but banning non-English speakers because machine translation is unclean is more fascist than anything in the AI space except maybe Palantir's XAI enterprise account

Genre is not about differences in the fundamental forces that make the plot go. Genre is about putting on different costumes while the story explores whatever

Arctic circle train has some legs tbh

i have 75 agents praying to send this guy to hell, effectively cancelling out his prayers