I can't believe this happened...

This is really bad. Daario got his way. He has regulatory capture and now all of us don't get the Frontier model right when it drops. Here is GPT 5.6 sole preview. Only a select group of companies get to use it right now. And the rest of us, we wait. We wait for the little morsels. And it's I'm I'm so disappointed right now. I really it's hard for me to convey in words what I'm feeling. And look, it looks like a great model. And I'm sure it's a great model. And I don't even want to talk about it. I want to actually talk about what's going on. How this is a turning point, a negative turning point for the AI industry. And if you watch my videos, you know I like to be very optimistic. And this might be the first time in the three years since I started doing this where I'm not feeling so optimistic anymore. And I really do blame Daario and Anthropic. Fear-based marketing has won. And let me explain it all. Daario and company have been on this campaign of fear-based marketing and they have forced the hand of the government to finally regulate in a really aggressive way all the newest artificial intelligence models that are going to come out. And look, I can obviously point to the US government and place some blame on them, but what were they supposed to do? You have Daario and Anthropic saying, "We have this new model and it's so powerful. We're not going to release it. We're going to give it to a select group of people, a select group of companies, and they're going to test it first. They're going to make sure that it's good enough for the rest of you to use. And so then fast forward, we have, you know, months of this fear-based marketing campaign and then they release Fable. During that time, they talk about how China has been distillation hacking the claw models. Here's a report from just a day ago. Anthropic accuses Alibaba, which is one of the biggest Chinese companies in the world, of a campaign to brazantly and illicitly extract AI capabilities. They say, "Hey, we put out these incredible models. China is effectively stealing the models." Even though the argument is a little bit flimsy because they didn't really steal all that much data, but fine. They did this distillation attack. They stole a bunch of data from Claude and now they are building frontier models with that. And so the US government is looking at this whole thing and thinking, "Okay, well, you're telling us these models are incredibly dangerous. You're telling us that only a select few should be using them, but you just released it and now we're getting reports that you can prompt hack the model." By the way, you can always prompt hack a model. There is no such thing as having a 100% protected model. It doesn't exist. That is the nature of large language models. That is the nature of non-determinism. So the government stepped in and they said, "Hey Dario, go ahead and uh don't give that model to any non US citizens." How is anthropic supposed to do that? Well, they just turned it off. We lost access to the model after just a couple days. Fable, the best model on the planet. And guess what? Open AAI had GPT 5.6 cooking for probably months now. They've had this incredible model, but they can't release it. I just talked about in a previous video. Wow, it's been quite a while since GPT 5.5 was released. I felt like the model release schedule was accelerating, but here we are like 2 months on and we don't have GPT 5.6. Imagine you're sitting in a room with OpenAI's leadership. Sam Alman is there 4D chess player and he's thinking, well, Fable is banned. They released it and now it's banned. If we release 5.6 right now while Fable is banned, what do we think is going to happen? Well, guess what? We found out they didn't release it and the government said, "You have to stagger the release now." And what does that actually mean? And the implications of that are so substantial. I really hope you can understand and appreciate how big of a deal this is and why I am seeming like I'm so upset. And then last night, the information reported it. The US government asked OpenAI to stagger the release of GPT 5.6. First, it will go to a handful of companies that deserve to use it, that are trustworthy enough to use it. They get to test it, and then maybe hopefully in a few weeks, maybe more, maybe less, we don't actually know, the general public will get it. This is effectively AI regulation. On Wednesday, for example, OpenAI CEO Sam Alman informed staff during a Q&A session that the company would release its latest model, GPT 5.6, 6 in a limited preview to a small group of partners. According to two people familiar with the matter, the company's biggest rival, Anthropic, followed a similar course in April with Mythos, a new model with powerful cyber security capabilities that it shared with select partners rather than opening it up to the public. Anthropic set the standard and they campaigned the government to do this. And not only were they campaigning that these models were incredibly dangerous from a cyber security standpoint, that China is distillation hacking them, but also job loss. Remember the white collar bloodbath? This is all coming from one company. We could have had unlimited abundance. We could have had this incredible optimistic future. And I feel like Daario has now won. He has affected regulatory capture. And what does that actually mean? It means that companies, startups, builders, tinkerers, people like me, people like you do not get the frontier intelligence like everybody else. the few, the powerful, the companies that are already massive that can afford to jump over these regulatory hurdles, they get it first and they have plenty of time to accelerate ahead even further before the rest of us get the model. This will cause concentration of power. Absolutely. I have no doubt. I was already feeling a little bit fearful of that. Even without all of this regulation, even without staggered release, there are effectively two companies, Open AI and Enthropic that have artificial intelligence at the frontier. And that AI is building the next AI. And now that these companies need to hold back their frontier even longer, they get to build even more with them. And then when they finally give it to their close partners, the biggest companies in the world, they get to build with them. They're going to accelerate even further. How does a founder of a fresh startup catch up? How does a builder catch up? It I I can't see it. And it even gets worse. For those of us that wanted to participate in the value creation of companies like Open AAI, we have to wait even longer. Now, listen to this. Open AAI is now leaning toward pushing its IPO until 2027 per the New York Times. The reasons they give here, I don't think are necessarily true. They talk about choppy markets in recent weeks. No, there is incredible demand for OpenAI shares. The company is worried it might not find much enthusiasm from retail investors. Definitely not. That is not true. The real reason is because they have zero visibility into what the regulatory framework is going to be for artificial intelligence and that directly affects their business. And that means people like you, people like me who wanted to participate in that value in the IPO, we have to wait. And the model builders, the frontier labs will continue to accelerate ahead. And then the partners that they eventually give the models to will also accelerate ahead. Imagine you're a founder, you just started a company, and now you want to build an AI model. You have to go through all of these hoops. Or even if you're not trying to build an AI lab, imagine you're trying to build a startup and you're competing with this large company that has the frontier model and is able to build better software more quickly than you because they have the best model and you don't. You have to survive on, you know, maybe a six-month-old model. And I know maybe this sounds exaggerated, but this is only going to get worse if we continue down this path. And I really do blame Daario and company. And look, it actually gets worse. While the US government is slowing down AI in the United States, you know who's not? China. China's certainly not slowing down. This might seem like a minor slowdown. This might be just a few weeks before the rest of us get the GPT 5.6 model, but guess what? It is a slowdown. Every single minute counts. It really does matter when we are in a race with China on AI. And I'm going to say it again. I blame Anthropic. And it's even hurting them. They rolled out Fable. People loved it. They wanted to use it. They wanted to pay them so much money for it and then they had to roll it back. And they complained about that, which is insane to me. It's insane that they would complain when it is completely self-inflicted. You did this to yourself, Daario. You did it to yourself. And here's the thing. I was already a big proponent of open-source, but now opensource is more important than ever. Democratization of frontier intelligence is more important than ever. But open-source is behind the frontier. And this actually might make it worse because although the broad economy won't get the benefits of the absolute frontier because it's going to be locked up within just a handful of companies, those companies are still accelerating. They are still using Fable 5.0 0 to train Fable 5.1 and that will be used to train Fable 5.2 and so on. So these companies aren't slowing down. The government has not said, "Hey, you need to stop training the next model." No, no. They're saying, "Hey, don't give the model to everybody." And so here's what I'm going to do. If you're an open- source lab, and I don't care where in the world you're from, I'm going to be talking about you more often now. I hope you succeed. I hope you are able to democratize artificial intelligence because we're in this turning point and I am quite worried. All right, I've collected a few of the reactions throughout the industry from people whose opinions I very much respect. I want to show those to you. So here's Aaron Levy. We now have de facto AI regulation. It's not obvious why from here on out models that have certain levels of capability or are trained on certain compute sizes won't have to be reviewed by the government before release. Realistically, as AI models became more and more powerful, this was going to be inevitable. I think it's too early. Agreed. Totally agree. But here we are. So now it's mostly just interesting to think about the implications and scenarios from here. So here is what he thinks may happen. one, America gets to control who gets access to frontier intelligence and when. I mean, that's happening. That is happening right now. This generally works as long as we remain at the frontier at all times and don't have a risk of being surpassed. He is talking about China. At the moment, we have a clear lead in frontier intelligence. So, this is a good bet, but lots of motivated parties would love to change that. Number two, this likely creates backlog of AI releases, which means that we will see less rapidfire back and forth jumps in model progress. This is an extremely important point. If the model labs do not have the same incentive to release their models very quickly, as quickly as possible, they do not have the same level of competition anymore because they're not competing for dollars. They're just competing for who can develop the best AI internally and maybe they'll reach recursive self-improvement. And then really like concentration of power is it's over. It's done. The reason why over the past 6 months we've seen this acceleration of model releases is because OpenAI and Anthropic were in an absolute battle to see who would win the most dollars in the market. Companies were voting with their dollars and now they don't. And it's a really bad place to be because these companies raised a ton of money and now their ability to monetize their models that they created is completely hindered and that is going to affect the US economy. It's going to have these second and third order effects that will affect the non-investors, people who are not at the AI frontier lab that's going to affect the entire economy. But bull find cases that we will just get better step functions per release at a slower rate and we end up at the same point we would have. Basically what he's saying is rather than these iterative deployments of new models they're basically going to kind of keep baking, keep building, keep innovating and then release bigger models less frequently which from a safety perspective is actually a bad thing. You don't want to release this massive step function change and improvement for a model all at once. It's one of the things I really agreed with OpenAI on, which is rapid deployment. Get it out there, small increments, see what happens, react quickly, and now they can't do that. Other countries likely will have even more incentive to at least hedge their bets with sovereign AI strategies so they aren't dependent on access to US AI at all times. Yes, this is definitely true. To be honest, they should have been doing this anyways. You know, with the way the world is currently, every country should have their own AI program, but this adds even more risk to other countries. Open weight's obviously a big winner here as it becomes what likely sovereign AI gets built out on which I guess I hadn't really considered that. That is a good thing and I am more bullish and a bigger proponent than ever on open source artificial intelligence. But what do you think is going to happen when Daario sees open source catching up? He's going to do the thing that he does. He's going to champion regulatory capture. He's going to champion regulation and say, "Well, everybody can use this open source model. We need to lock that down right now. Only we, Daario and company, can have the best models on the planet, can decide who gets access to the intelligence that's going to build the future. Only us." One interesting question would be how regulation eventually extends to open models, which would have its own set of long-term consequences. Yeah, I mean, it can't. Maybe you're going to start locking down GPUs. Maybe you're going to start saying, "Okay, only certain people can have GPUs." We kind of already do that with China, but not inside the United States. That is further concentration of power. Who gets the GPUs? Who gets the compute to build the intelligence? Who decides that? Well, I will tell you. It's going to be Daario and company. It's probably going to be Open AI and company. and it's going to be the US government. And that is very scary to me. Here's Peter HD Amandis. If we strip away the headlines about Mythos 5, the story is brutally simple. Someone somewhere is deciding what level of intelligence you and your company are allowed to access. Here, if you weren't already upset at Anthropic, let me go ahead and make it worse for you. This is Bill Gurley, one of the most famous venture capitalists, invested in some of the biggest companies in the entire world, one of the best thinkers. Listen to this. Anthropic could have sued in court as others have, but they want something far more valuable than simple restitution. They want the US government's protection against competition for years and years. A court can't provide that. Now, what he is specifically referring to is that distillation attack. And so rather than taking the legal steps to sue, they didn't. They simply campaigned for more regulation. Bill Gurley just a couple years ago had this insanely viral and famous speech about regulatory capture and we're seeing it play out right in front of our eyes now. And then continuing on this point, he also says, "If you're on the verge of AGI or ASI, artificial general intelligence, artificial super intelligence, why isn't your model smart enough to recognize espionage distillation in real time? You say cure cancer in a few years? Isn't sniffing elicit distillation quite a bit easier than curing cancer? Why write letters to DC? Just use AGI." He's right. Your models are so good. Your models are the most dangerous models on the planet at cyber attacks and cyber defense. They can't prevent you from getting distillation attacked. They can't prevent Chinese companies from spinning up 20,000 fraudulent accounts and extracting data from your models. How does that make sense in the slightest? And let me be clear, Open AI, from my conversations with them, they do not like this. They want to release the model to everybody. It is a good model and they are being very public about it. Nome brown who's a researcher at OpenAI, GPT 5.6 is incredibly strong and fast for coding. I hope we can make it available to everyone soon. And if you read the blog post about GPT 5.6, what did they focus on? They focused on safety. They focused on how good it is at cyber attacks and cyber defense and also how aligned the model is and how difficult it is to get the model to do things that it shouldn't. And here it is. As part of our ongoing engagement with the US government, we previewed our plans and the model's capabilities ahead of today's launch. At their request, we are starting with a limited preview for a small group of trusted partners. Who is that? Who are those partners whose participation has been shared with the government before releasing more broadly? So during the preview, we will continue testing and coordinating close with partners as we work toward broader availability. We don't believe this kind of government access process should become the long-term default. It keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them. We are taking this short-term step because we believe it is the strongest path to broader availability in the coming weeks. While we work with the administration to develop the cyber executive order framework and a repeatable process for future models. So there it is. We no longer get the best models. Some do, not us. So there it is. I don't know what to tell you at this point to be honest. I think there are two things. One, go support open source like you never have before. Go test open source models. Download them and run them locally. Figure out how to do it. I have plenty of videos going over how to do this. And I will be making more because it's more important than ever that everybody understand how open source works and how to run these models on your own computer. No, they will not be as good as Fable. No, they won't be as good as GPT 5.6. But if we support opensource labs, maybe one day they will be. And you should also reach out to your government representatives. Tell them that this is BS. So, I know this is a very different type of video than I usually make. Hopefully, you understand the severity of the situation. I really hope this does not become the standard because this is going to have severe consequences.