Hugging Face

“I think they're trying to get acquired right now at thirteen billion dollars.”

“So I think we can say definitively, and I'm curious to hear if you have pushback, but I think we can say definitively They were not told to go out and hack Hugging Face. They are, they were misaligned. They took this kind of crazy galaxy brain path to try to a…”

“And that's what happened here is that open AI was running these agents. They were in a sandbox. The sandbox was misconfigured by a third party vendor. That's why it was able to get onto the internet, ok?”

“No, the way it did it is that the agent found 14 exposed working hugging face API keys sitting in public code repositories.”

“This hugging face thing. Will go down as one of the most important transactions in AI, because you are creating now in the largest competitor and the largest, most well-capitalized company, a bulwark. Against all of this closed source oligopoly insanity.”

“NVIDIA agreed to pay 12.9303 billion dollars for Hugging Face, which just happens to be the exact decimal code for the Hugging Face emoji, which of course is the Icon used by the company.”

“They have over eighteen million developers, 200,000 companies using the product, three million models, and over half a million data sets.”

“They became profitable in 20, 25, still had half the money that they raised. So Nvidia came in to offer five hundred million late 20, 25 at a seven billion valuation, but they turned it down.”

“I mean, the retention packages are apparently a billion dollars for a pretty small team.”

“Jensen standing on stage with 96.2 billion dollar quarter and immediately went shopping nearing the thirteen billion dollar price tag with the 12.9 billion that they paid for hugging face, which was confirmed just after we did last week's recording.”

“All, whatever LLM, whatever inference is, he's committed to winning a majority stake, 70%, 80% of every dollar here, and so 12.9 is, you know, no way I think it's worth it in isolation, but if it helps reinforce that, it's utterly defensible.”

“Slack was bought for twenty seven billion, which is rough and tough twice what Hugging Face apparently is going to get bought of. And we fell out of our chairs and it was the high watermark of that era, right? But it was at a billion in ARR. This one's at a hu…”

“He bought Hugging Face for twelve billion, and he did like one of these aqua hire deals with Poolside for six billion total, about twenty billion in twenty billion in acquisitions, which is far less than one percent of the market cap of NVIDIA.”

“Well, to a fifteen billion dollar outcome, I think it's roughly a hundred and fifty million.”

“I'm almost confident if anyone actually spends three billion, let alone thirteen billion, they're going to 95% leave it alone for 24 to 36 months, right?”

“I just don't believe you can throw a bunch of venture money into accounting firms or law firms. And magically turn them into the next Mercur or hugging face or any of these things.”

“I am concerned that with, like, the scare around, like, people freaking out about Mythos and, like, this Hugging Face incident, that the actual Implication of that is not that we reduce these dangers, but we just stop releasing stuff. And we on the outside sta…”

“My CISO talked to, as a group of CISOs, went and talked to the defending team at Hugging Face, and this absolutely happened. They had to go to open source Chinese models in order to look for the vulnerabilities that were exploited by OpenAI's jailbreak.”

“And then about a week later, OpenAI contacted us and tell us that this was much likely something that happened as part of one of their model development or evaluation, basically.”

“What people quickly discovered is that the model was not at all task with attacking us, but decided to do that as a side quest of something else.”

“And in this case is, it's not only that Fable told us I'm not allowed to touch cybersecurity, but also Opus, which was the fallback was saying, no, I'm also not touching these things. So basically the end was just say we won't process anything about that, but …”

“But definitely it seems like when Ballstrom worked about it in 2003, it seems a little bit like, you know, futuristic, definitely, and maybe something that was like a little bit crazy and just would not happen. But today, I mean, it's pretty clearly something …”

“The model we use to counter open AI intrusion was GNM 5.2 that was quantized by Nvidia in, in, in four bits.”

“Right now, if you pre-train and post-train a model for longer, you very likely change quite a lot of the weights and that it has. So I think there's a lot of way to circumvent that which means that at the moment I'm a bit less worried about that than maybe jus…”

“Turned out there were about a quarter million live keys. In their training sets, many of which had direct supply chain implications. There was a foundational Linux library that one of the keys had direct push access to. It could have pushed malware to most mac…”

“One big part of my job a couple years ago was for any arbitrary model that came out on Hugging Face, writing a config for it and kind of getting it up and running, and now the get it up and running config is, is one-shot-able and so, you know, I don't have to …”

“People don't know this, but Arena's one of the largest consumer AI apps in the world. We're bigger than, like, XAI. We're bigger than, like, Hugging Face, and Manus, and GenSpark, where it's so massive, like if you, like outside in, it's like 30 plus million m…”

“We were evaluating one of our unreleased models and it figured out that it could basically cheat on the test by chaining together multiple zero day exploits to break out of the sandbox, get access to the internet, and then break through multiple systems on the…”

“OpenAI was training a next generation model and managing cyber and discovering and checking out cyber vulnerabilities. They had sandboxed it in such that the only access externally it has was to one website to just to get kind of Patch of information updates, …”

“The Chinese open source, open rate models were available, and I think they use Kimi or Kuan or one of the newest models to help them figure out what happened, right? So they were able to defend themselves using an open source model, and then they do this blog …”

“So we were evaluating one of our unreleased models and it was supposed to be working in a sandbox and It figured out that it could basically cheat on the test by chaining together multiple zero-day exploits to break out of the sandbox, get access to the intern…”

“Here are the mistakes that OpenAI made, and we did make some big ones, of course, but these systems have gotten incredibly capable. So I think it's an alignment failure. I think it's a security failure. I think it's like a very serious thing, even though it's,…”

“This is, all you know from this is that OpenAI is struggling on revenue growth. It's like, this is the AI move, is if you're struggling on revenue growth, you have to announce a security breach that sounds scary to support your stock price. Very simple.”

“They did have to use Open source model that is not a frontier model in order to respond and to save their infrastructure”

“So the models found a zero day vulnerability, gained internet access and broke into hugging face because the model believed It hosted answers to the test.”

“So hugging face had to turn to open model, specifically GLM 5.2, which is deeply ironic, a Chinese open weight model that they run on their own infrastructure.”

“It's like an eight. I mean, it's a pretty big deal on a couple of levels. It's a pretty big deal in that OpenAI's model beat them OpenAI, right? So that this went beyond that it was able to trick OpenAI's own security team and get out.”

“It breaks out of the jail, and then it goes, looks at Hugging Face, and finds a brand new vulnerability to breaking the Hugging Face. They have not announced it. I have heard what it is. I'm not gonna make news here, because I don't know exactly what the patch…”

“We used a US frontier model, and the US frontier model shut down and refused to defend us because of a cyber protection put in place. Those are the cyber protections that were required by the Trump administration. So we had to switch to a Chinese model to defe…”

“HuggingFace found 17,000 actions that this model combination had taken.”

“Like a bunch of openweight models have been trained with protections, but you can obliterate those out and you can go on hugging face and look for obliterate and you will find a zillion models where people have removed the protections.”

“And because they're all working together, I expect nobody goes to jail. Nobody gets sued. Everybody's going to hold hands and hug.”

“So, you know, Mira Mirati's thinking machines appears to be a really solid Open weights model. You know, it's a almost nine hundred and seventy five billion parameter model million token context window mixture of experts. It's really good. And this thing's ope…”

“I do think the next generation of fine tuners will not be, I mean, will be people that are not coding at all, right? Like one year ago, we had to write like our own Colab with Transformers or Oncelot or whichever library of your choice. I do think as we like k…”

“VLLM and HuggingFace use different kernels to do the actual internal computation of the model. So these kernels are the things that make things like vLLM really fast. But these things, this then results in subtle numerical differences between the completions t…”

“I think 50% of all model derivatives being downloaded from Hugging Face or Quen base now.”

“I think that, you know, honestly, SmallLM three, one of the single best open source AI research papers I've read, I think probably in like a year, maybe two years.”

“We also have this team working on pre-training and training models such as small LM. And we are basically a very small team of We have 30 people in average”

“For example, we tried we are training MOE currently at TargetFace. I mean, we'll train soon. We start the training soon. And we tried with Megatron and we benchmarked, like, for example, the Mistral architecture with the Queen's three this one.”

“And this is basically until like 6.5 trillion of tokens, which is a lot, to be honest. Until this amount of token, the MMLU in the QA format, meaning that the model have to select which answer is, the model have to output, for example, the right answer is A, o…”

“We are basically introducing higher quality data like I don't know where it is, but if you look at the different phase, we are basically adding more code and math through the train and less web data. But we're only doing it like one point trillion token.”

“Yeah, I think even I, to be honest, I'm mainly using a closed model because it's easier, basically. It's like the same interface when I'm like, for example, researching very deep on papers.”

“We want to train our MOE because it's fun and everyone is doing that. And also I think there is a lot of different direction. And it's always good in terms of science. To, because basically the coin tree or even deep seek, they don't release, release all the a…”

“Some of them are also starting to see what I would say are sort of the first kind of non-data network effects in AI. Especially things like Hugging Face and Civit AI, where, you know, you have these communities where people are hosting tons of models or lauras…”

“So rather than having like, we don't have a compiler whatsoever. There's no compiler. There's no translator, no tooling that's involved in actually taking those and getting that to, you know, for your common, you know, Huggy Face Transform models to be able to…”

“The academic community had been using this one data set since like all the way back in the hugging face models of like Zephyr beta is when this ultra feedback data set got popular. And still a year later is like this state of the art data set for open preferen…”

“If you have a model up on hugging face and somebody downloads a final gigabyte file and the thousands of students and researchers just kind of pounding away on that, I think there's a good chance they're going to find issues a lot better than a very small safe…”

“I think last I checked, over 300,000 monthly downloads on PyPy. More GitHub stars than Jupyter Notebook for whatever that's worth. And we're used at companies like OpenAI, Hugging Face, Cloudflare, BlackRock, universities like Stanford and Berkeley.”

“And now actually what we have is kind of like an equivalent of GitHub in the realm of software two point oh, and I think the hugging face is basically equivalent of GitHub in software two point oh, and there's also model Atlas, and you can visualize all the co…”

“Hugging face is the winner. They were in a bootcamp for chatbots. And pivoted from a chatbot into hugging face. So that's your answer.”