The guy who gave us QLoRA, says DeepSeek's inference system is on par with OpenAI's stack and better than what Anthropic runs, at least at their scale. Now check the license. Apache 2.0. Until today a production stack at that level existed inside a handful of companies. It was the thing that made their API fast and their margins fat, and none of them ever showed it to anyone. Today any startup, any university, any cloud in any country can download it, read it and run it. It runs on NVIDIA Blackwell and on Huawei Ascend, so it isnt tied to 1 chip vendor either. This is what I mean when I say open weights are about more than downloading a chatbot. DeepSeek is giving away the knowledge that only a few labs had, and that changes who can build an AI company at all. Respect @deepseek_ai

Today we’re releasing Underdog Bark 3.0, the #1 frontier speech model It's a tiny 163 MB download as accurate as NVIDIA’s Parakeet (15x bigger). It beats OpenAI's Whisper large-v3 (57x bigger) It's fast- transcribes 1 min of audio in <0.25s on a Macbook. On recorded meetings (AMI), Bark 3.0 gets 7.82% of words wrong. Parakeet gets 9.42%. OpenAI's Whisper large-v3 gets 13.63% error rate Bark runs on both your Laptop & Phone. It’s free and open source under CC-BY 4.0 Try private on-device meeting notes & dictation, powered by Bark, in Underdog today: underdog.ai/bark

🔴 As Russia is gaslighting the world, framing the humanitarian catastrophe in Oleshky as yet another instance of “Ukrainians bombing themselves”, SM platforms slow down the reports from the ground, as usual. Here is an important exclusive report, hidden from the world.

Radwan Mortada is one of Lebanon’s most prominent investigative journalists. He spent two years conducting an exhaustive investigation into the Israeli pager attacks, tracing the operation’s planning across nine countries. His documentary recently began airing on Al Jazeera. Israel just bombed his house.

Trump on Camp Lejeune water contamination claims: "If we win the House and Senate, I am directing the Department of Justice to settle all remaining legitimate claims as soon as possible ... if we don't win, there's no money."

is this good

Stanford researchers did it again. They made language models 31× faster. Right now, every major AI, from ChatGPT to Claude, writes exactly the same way. It guesses the next word. One token at a time. Left to right. It’s like writing an essay on a typewriter while looking through a straw. It is computationally exhausting and incredibly slow. Some labs tried to fix this with "diffusion models”, generating everything at once and polishing it over and over. But that requires massive compute and endless denoising passes. A team from Stanford just dropped a paper that kills both methods. It’s called "Next-Concept Modeling" (NCM). Instead of predicting the next word, the AI predicts the next concept. It generates text exactly how a professional writer drafts a book. Coarse to fine. First, it generates a high-level conceptual outline of the entire sequence. Then, in parallel, it fills in the precise words and details across the whole document at once. The results are staggering. • Generates text up to 31x faster than current diffusion language models. • Requires drastically fewer computing steps to reach high-quality outputs. • Actually outperforms traditional step-by-step models on complex reasoning tasks like Sudoku. AI is no longer blindly typing word by word, hoping the sentence makes sense when it reaches the period. It is conceptualizing the entire thought before it opens its mouth.

Unaccompanied migrant children, some as young as 5 years old, are appearing in immigration court without attorneys. Now, Bloomberg Law reports that they are being pressured to accept deportation in exchange for $3,000. According to the Acacia Center for Justice, immigration judges have reportedly presented these payments to children, sometimes describing them as one-time offers.

Alina Habba was an advisor to the Attorney General, was interim US Attorney in New Jersey, and all the while was negotiating a massive bribe to get a Venezuelan billionaire a pardon.

A billionaire wanted a pardon. His family gave a $50 million deal to Alina Habba, a lawyer with close ties to President Trump. nyti.ms/4zNq59Z

She got $50 million ! For arranging a pardon!! $50 million. $50 million. $50 million.

😳 Trump loses it and SNAPS at a string of reporters, says Ilhan Omar might "go to jail," calls Somalia "a shit house," and boasts that thanks to a new program he announced today Americans "will make millions of dollars by turning others in" during absolutely deranged Oval Office event Here's a supercut:

Nobel laureate @avalaina has documented Russian war crimes in Ukraine since its full-scale invasion. Her reaction to Trump saying Russia has been “fine”’ there?: “We document intentional murders of civilians every day…I'm sorry to say, but this presidential administration completely lost their ability to distinguish between good and evil.”

Eric Pratt, the republican running for Angie Craig's open seat in MN-02, left the AI restating his prompt on his website. "The President is not popular, so all messaging should be focused on local politics and needs." lmfao

Our largest model yet. 763B parameters, compressed with GSQ-RCO and co-optimized with Strata (by @coldniko) for local inference. On Hugging Face this Friday. Much more is coming. Follow us for future releases.

Jason Alexander says he feared visiting Palestine as a Jew until 10 Palestinians protected him during an Israeli army raid in Ramallah “How much of a risk would I be taking as an American Jew walking into Palestinian cities? And Mohammed said, ‘The greatest dishonor that could happen to a religious Muslim is to allow harm to come to an invited guest’” “And he said, ‘Jason, if they had to, they’ll take a bullet for you.’ And what was so interesting about that trip is, we were having lunch in Ramallah with a bunch of actors and artists at a sort of outdoor restaurant” “And the Israeli army was conducting a raid on a bank that they thought had terrorist ties a couple blocks away. And they must have, to announce themselves, just shot into the air. And I remember 10 of those guys at the table surrounding me. And I went, ‘Oh my God, he wasn’t kidding. They’re not going to let me get hurt’”

a brand new model paradigm! introducing 𝚌𝚕𝚊𝚒𝚛𝚒𝚌: a system one model for extracting, computing, estimating, and predicting values from unstructured inputs. blog.cloudflare.com/clairic-…

It's not hard to work out the playbook here: acquire a widely used model of voting machine, hack it, and then one of three possibilities: – don't disclose the vulnerability until after the voting is done, creating pressure for votes cast using that model to be discarded as insecure; – disclose it just before the vote and demand that the machines not be used, throwing the election into chaos; – use the disclosure as justification for federal forces to physically seize the voting machines, with the same result. Whichever scenario they had in mind, it's possible, maybe even likely, that this represents preparation for events never seen before in a Western democracy. I'd be interested to know what @C_C_Krebs thinks about all this...

According to the treacherous US president @realDonaldTrump —a supporter of murderers and terrorists—the daily Russian slaughter of 🇺🇦 civilians and attacks on 🇺🇦hospitals, schools, bus stops, and shopping centers are perfectly fine.

This is what the current US president supports. Yes, today the world—and the majority of people in the US—no longer doubt that Trump has become a supporter of the most heinous criminals and crimes. For the Trump, crimes are perfectly fine, provided they bring him personal gain.