Imagine that the most important factory in the world is not a machine factory or a oil rig. It's a lab where new artificial intelligence is being built. People used to write the code themselves, search for the mistakes themselves, start experiments themselves. Now, part of this job makes the most artificial intelligence. He's not just answering the chat room, he's helping to write the code for his next version. He's checking someone's code, finding mistakes, launching experiments. And most importantly, he's suggesting where to move on. We often argue whether artificial intelligenceuals of the software, designers, analysts and you will replace. But that's too small a question. The main question is, what happens to the market if companies that create artificial intelligence start to accelerate very much with artificial intelligence? I'll sort three things out in this release in simple language. First, what is happening inside Anthropic, OpenAI and Google DeepMind. Two: why can this very quickly speed up the entire market for drugs, your routine and business. And the third one, perhaps the most important thing, is where self-inflicted is real and what you can believe. And that's not another chat-bot story. It's a story about a factory that starts to collect itself. Anthropic said that in her internal work Claude, it's like ChatGPT, it's not just a chat assistant. I mean, there's a corporation. It's clear that they use chat-bot inside, just like the OpenAI uses GPT chatting for a huge number of processes. And, accordingly, we're at the base of the view that it's their chat assistant or some kind of agent system. Now, the OpenAI recently issued a release and showed a large number of different agents that help them sell, sell or market for analysis. I was also saying that these are stupid, weird boxes. Yeah, it's a case like that two years ago. So, Anthropic writes, yes, and says Claude writes a code that helps to start experiments and, as I said, check errors and in some tasks, takes over most of the performance.
I mean, basically, he's working his own way. If you're looking at the software system now, uh, with artificial intelligence, you can't press the button, you're gonna have software, yeah. There's nowhere to pay that money. So, that's what Anthropic said about their system inside, uh, functioning. It's very important, yes, it's working inside. It is clear that Anthropic itself is not being created and Claude itself is not created, uh, by a model of artificial intelligence. But the fact that most of it is already being built by artificial intelligence is a fact. In fact, you and I can see, uh, this, sometimes in systems interfaces and systems work. I am deeply convinced that a large piece, such as OpenAI, is, uh, all portals, services, that's all about user support services, soupports, all about different lines, control panels. and even the Chat interface. It's written by means of, uh, code or by, uh, conditionally, GPT chat, yes, by their code development services. It can be seen in strange things from the perspective of menu, menu display, business logic. And even if you read their materials, descriptions, help them, how they form. So you can see that's done with artificial intelligence. Uh, and this, look, is a very important aspect. It doesn't mean it's done badly, but it's done in some way. Yeah, it's not a product cut by a certain product team or not a product visually, in the sense of XUI, which carries this kind of imagination, the meaning, the foundation of a certain team. We could see that, for example, in the Airbnb example. I remember very clearly, there, more than ten years ago, when Airbnb product came up, and it was seen that its makers were designers. And now, Max Grigorieev, who was on the canal, he was a technical director there, and he said that the designer approach was to business and software development. There were therefore a great number of problems in terms of technical infrastructure. They had a very cool interface, but it was just a badass.
It is clear that he was frozen at some point of time because such big projects, they stop from development. I hope that doesn't happen to Anthropic and OpenAI after they're public now. I hope very much that this will not happen and that they will not be pressured, and they will not move, so, in companies that make releases very long, yes. Uh, and Google's got some kind of speeding up, but it's unlike many so-called start-ups. It's important to understand that in the slang in the current OpenAI, Anthropic is still a start-up. So, according to Anthropic, as of May this year, more than eighty percent of their code-- which company is pouring into its own, uh, what is called code base, yes, it was written by Claude.
I mean, it was written, uh, automatically artificial intelligence. It's important to understand that again, yes, so that we all understand, it's not eighty percent of the company's work, that's the code that was written and it's in the repository. Uh, again, eighty percent is a lot. And that again confirms my statement that the entire UX/UI is written by Anthropic, well, with a high probability written, yes, Anthropic. Anthropic says that a typical engineer in the second quarter of this year released about eight times more code per day, and more than in 2024. Here. It's not-- it's not a matter of engineering performance, because, yet, uh, code counters are not a quality, and so on. But it's still pretty serious, it's a lot of volume. And of course, I've been in the process of developing over twenty years, creating at least some software products. I see it's relatively developmental. Somewhere less. And the speed that is now is, of course, feno-menal things, and they're uneven. I see how many companies are standing there right now, and I understand how much the problem is to make some update. And there are companies that just roll at space speed, yes, ahead. It's important to understand that, yet, Anthropic, they have their own system. I'm talking about Anthropic, I'm probably gonna make a separate issue because Anthropic originally made a software application rate. And so to put this issue on a certain direction, and of course they made a phenomenal bet because they didn't make a bet on the software, they made a stake in creating their own soft. And what Ilon Mack does Macrohard again proves a complete-fit-- that's the whole link, all, all that's going on.
What Sergei Breen said to him is that he believes in artificial intelligence more than in mankind. Ah, what Anthropic was doing was a development stake, initially to develop a code inside, and what he said he would do Macrohard. And they have a big factory, huge, yeah, that's where Macrohard is written on top of some of the states of the United States. The thing he said was that he would somehow make a competitor Microsoft, yes, what their operating system would be is just proof of the correct Anthropic movement, an incredibly correct movement. Another important story Anthropic said that Claude had become better off with tasks that did not have a clear instruction for each step. Yeah, that's a very important asset now, because if you do some serious project, a very big, scream-- well, a very serious project, that's what it says, for example, to run a part of it. business, the too there will certainly be a huge number of problems. Although there is an example, for example, in San Francisco, that there is a store open from a purely artificial intelligence. But that's it, it's all marketing, it's all faeces, it's all real. But, uh, what Anthropic has, they don't have to say that, but, uh, there's a need for quite some other instructions. The system is starting to get better than what you have to do. Obviously, their rules, their code base, their design requirements, they've been unbelievably trained, what they call, they've fed their own system to create their own system. I'm deeply convinced they have it.
I am deeply convinced that they have a higher priority in using this product inside their company than putting it out. Because their main move is the creation of AGI. Whatever we say, wherever we go, it's still the creation of AGI. And here we are going to see what Anthropic calls the main human skills.
And he says that the main human skills now are to see a big picture, to choose what problems are important, what results to believe and when to stop. And that's the question: what kind of people work in the world, in different companies, corporations, how much do they see this big picture? We often talk on this channel, as important to you viewers, to begin to understand what is a artificial intelligence, and to really deal seriously, really seriously. In doing so, we must understand that most of you work for some company. And the question is, is this company, how does it see a big picture? How much do people in this company see a big picture? The country where you live, the leaders who run this country, how much they see a real big picture, how they realize that this major oil is today, not a oil factory, right? They understand-- and there's no possession of the dollar, yes, today. They understand that there is a laboratory like Anthropic or a laboratory like OpenAI, or a laboratory like Google DeepMind, or a xAI that is built inside. Something that develops itself, develops without a man, develops a supercreation. Will they all do it? Of course not everyone. Well, let's just say, honestly, that someone, well, is guaranteed from these four operators to stay behind, yeah. It seems that they are now more or less the same. And we've stopped even on our own channel telling about some small stuff, who's released a new version of the video-generation or who's released a new version of the image creation, or who's released what kind of images he's got. A new version, I don't know, which is better than mathematical testing or bad-- passes a human test, so-called. The question is how they get into something real innovative, yes. But it's important that the leaders have separated, of course, from the market. Now, most companies are getting really tired of them. And this huge mountain of start-ups that writes parallel, writes a soft, or a large number of already existing operators that seem to be growing up, they're in a very suspended state.
I was just a year ago, setting an example of Nebius and telling them how much they've had a big stock increase. And plus, this is an example of the owner and the founder of the Yandex, leaving-- leaving Russia. And the example is that he's got Nebius' company now standing up a little bit more than ever worth the Yandex or maybe even worth it, yeah. Well, I am deeply convinced that Nebius will cost a lot. It's not my advice to buying something or anything. It's a little assessment of the infrastructure market. Well... that's gonna cost more than $100 billion. I've been talking about it a lot, yeah. The question is, these companies are considered to be very risky in the market. And there's gonna be a company like Salesforce, and it's gonna be less risky than Nebius. So, for me, basically, Salesforce is now a more risky company than Nebius. If we're talking about the development of artificial intelligence. So if artificial intelligence is developed in this stage of self-reconstruction, Nebius will win much more than the salesforce will win. The question of political history is next. How politically can the salesforce agree with these operators inside or with the operator who wins, and will he even be able to agree with them? Or whether these operators need and have no desire to make such decisions as Salesforce, such as CRM decisions. It is clear that the Salesforce for peace is not a matter of conditionality. More serious decisions are important to the world. For example, the Messengers, yes, WhatsApp or Telegram, or important, uh, platforms, there is another type of internal interaction, yes, like Snapchat or like Instagram, or Facebook. Question: What happens next? How will people get information, how will they be together in this information? Because on the one hand, it is clear that OpenAI and Anthropic have not entered the Instagram market at all, for example, or have not entered Facebook. Well, actually, well, as much as they do, they're not really dealing with him now. But if there's a real AGI inside or a real self-study system, then the question is, how hard would it be for them to enter this market? And I see it as uncomplicated.
It seemed to everyone that they would not be able to enter the market and compete with search or compete in the market for some portal systems. For example, portable systems such as Booking, yes, or Airbnb, or portable systems such as Kayak, yes, searching tickets. This system is very convenient for the search for tickets. She doesn't really work in all Russian-speaking space. In Russia, for example, she would not work, yes, or she would not seek directions for some Russian-speaking space, such as Russia and Belarus. But the platform is very interesting. The point is, will there be any platforms for booking tickets of this type? Yeah, they have infrastructure, they have links to all the armoring systems. They have one platform selling another, the same platform sells the third and so on. So they have a huge marketing distribution. But judging by the way the market is now being run by the same Claude or the same GPT chat, I would risk putting these platforms on. But I don't buy or invest in these platforms personally. Yeah, I won't. I'm not gonna look at them. And there's a question. I've seen it recently, if I step aside from artificial intelligence, I've seen it. I like the very brand of designer Ralph Lauren and American clothes. My wife and I were just sitting around saying that the brands are usually the most luxury, they're not Americans. And Ralph Lauren has different lines, including a very expensive line, such a serious level, a level like Louis Vuitton or Dior, or a Brioni. And there are even suits, well, about a plus lesson as the Brioni or Tom Ford. And they've got more money than ever. But it doesn't have a very strong influence on the company's evaluation, because it can influence the company's evaluation, for example, some kind of innovation in artificial intelligence in Booking.com or eBay. But I would be much more. Why? Because these companies, of course, are being evaluated very economically correctly, right? And what's called market. And these are all considered super risky. Although, in its essence, of course, Anthropic OpenAI is a less risky company for me than this whole intermediate layer that is inside. And Ralph Lauren is less risky than Booking or more than eBay. But that's my personal opinion. There are, by the way, three scenarios that Anthropic translates into human language that is possible.
First, they say that the trend is generally slowing down and models have reached the ceiling, and they help but cannot choose an important scientific goal. And then, accordingly, society has very much time to adapt. The second scenario is the labs are accelerating. So artificial intelligence makes most of the performance, and people set targets and check. And for example, a small team can do a huge organization's work. It's the most likely scenario ever that can be considered, of course, at this point of time. And I can see that. I see it personally, even by my efficiency, which I have. And there's a third scenario. The scenario was considered and considered. And Elon Musk in this scenario, even negative in this scenario, with about 25 per cent, twenty per cent of the chances were that there was a complete self-defense. So artificial intelligence is the one that develops the next version of itself, and then the main limitation is the computing power, energy, security checks, etc. Yeah, that's certainly not the current development, but the possible scenario. Many people, many scientists say that this is all nonsense, that artificial intelligence is not at a stage of development, that it's a little bit different. And I remember two years ago, a huge number of people talking about artificial intelligence. It's just, it means some lazy and a system of generation next, next word, yes, probability of the next word or the next part of the word. That was a lot of stuff, I think, and it's been produced by different, different information, scientific articles, publications and everything. But we see today that the way models work is quite different. Not all of you, of course, see it. I'll tell you again that there's a big difference between a free version of the top systems, between the pay-off version of the top systems and the expensive version of the top systems. There's a huge, big difference.
System. There's a huge, big difference. There's a big difference between the-- in the country where you're launching even a expensive system or a pay-for-charge system, I'm not talking about a free system, there's a big difference between ChatGPT in Europe, European. Not all countries, but as far as it works in Eastern Europe, how it works in the States, yes, or how it works in China. There's a big difference. You're the one who should understand it for yourself. And it would be cool if you tried or had access to some more serious systems. Why do I recommend that many people use and do something with ChatGPT Codex or Anthropic Claude, just to feel what's going on. I mean, just feel it. We'll take off a few briefcases in the near future, where we can show the Codex and, uh, and what it does in simple things, so that you can perhaps understand this scenario better.
Actually, with regard to the trends that are in the Silicon Valley, I want to highlight a few trends that are related to the subject that we are dealing with today. First, that's what you can say, the lab's become a product, yes. The product used to be chat-bot or was a model. That's all the comparisons of the endless ones that were on this subject. Now the product becomes the laboratory cycle itself. That's what I said, there's an idea, a code, an experiment, then there's a test of this experiment and a new model, yeah. And the question is, who speeds this cycle, he's got an incredible advantage, yes. So the second, uh, oope-- well, the very narrow topic for the Silicon Valley is that it is now not just mind, but the length of the task, yes. It's important that time artificial intelligence can work on its own. Uh, and it's more understandable than the abstract level of intelligence. And how much self-employment is a question, there, minutes, hours or days. And now, just yesterday, I was sitting with my editor, and she's got, uh, one of the editors, plus the ChatGPT version, not Pro. I say, "How long do you have the longest request?" And she says, "I have the longest request, well, there, three minutes or four minutes." I said, "Well, look, just this subject I told you about was a 20-minute request. This was done in other capacities, earmarked capacity and still in parallel, uh, processes. So we were telling you that, uh, Anthropic released the Dynamic Workflow system. These are dynamic flows that are launched simultaneously to accomplish one task. The system itself creates many parallel flows for one task.
This is the first time I've ever introduced this topic, I think in the models for users, for the usual, Elon Musk, when the Grok flagship was, and they released the first version, which is very expensive, $200 then, $250, I think she was worth it. And, uh, this model has them, they say you have four different kinds of sciences, right? And they even put them out, wrote, and you're running, like, one, two, three, four, parallel. It was always very interesting to see. You didn't know what was going on in the background. It is clear that modern models now have parallel flows. But the question is, what are the modern models? What systems? I'm not sure ChatGPT has versions Pro running parallel to the flows, yes, 5.5 Pro. But in Anthropic, I can definitely see that-- in Codex, I guess that's partly what I see, too. In his examples, yes. You're here as audiences, write what you see on your examples, that everything depends on the systems you write. But in the A-- Cl- in Antr-- in Claude Code, I can definitely see the flow balancing, but it does work weird. So the system says I've run ten systems down there, and I'll be back. And then she's, like, disappearing, and I don't know where she is, yes. The interesting subject for the Silicon Valley, which is actually now, is that the test becomes a new market, yes. It's when artificial intelligence generates more code, ideas, experiments, and the most valuable is not the generation, but the verification. Different tests, audits, security, there, evaluation tests that happen, and this is the constant observation. What do you mean, surveillance? When the system has the ability to witness what happens in parallel. I'm about to use three systems like this in developing serious systems. I mean, I work at the same time at ChatGPT 5 Pro, 5.5 Pro, as analyzing systems, repositories, uh, pros and evaluations. In parallel, there is Codex that makes different analyses, development, and parallel there is Aicnthrop Claude Code, which is still analysing different architectures, and it also takes different actions. And in fact, I'm always rechecking these systems on the one hand, although this is a well-known subject that doesn't always have to be rechecked, but like me, I do independent audits. I mean, for example, I have Anthropic, uh, Claude Code, it's Opus 4.8, a million context window, a maximum version. He works for me, I'd say it's what's in it, what's called the Silicon Valley, like, uh, surveillance, yeah. When the system keeps the system constantly in line with the other system being developed. And that's a fun story. I mean, if there was a system, if ChatGPT today or Claude, uh, but they can't do it, they could always monitor what's going on in my life, why, for example, I need to, There's something interesting about, uh, the offline of the device, some equipment like robots or some sensors, or some microphones. That would be very cool because they could really take and take decisions in parallel with different themes. I mean, if I'm gonna do something about Anthropic, ChatGPT would have gotten in, said, "Look, look, I don't know, he's sick or he's got to go to a certain place with the kids. I mean, if they were these systems were working, uh, working in parallel. I'm a little late for my piano class, and I see that I've already been depressed, late, not getting a graduation. And the system would know, for example, that she would see that I'd just postponed the class, and she would have told my teacher in a stash that I'm late, yeah. There's no problem for my teacher here to wait. He and I have a series of arrangements, and I would have a huge plus that I'm not distracted. But for that, the system should be a little more manipulative of my life. I don't mind. I mean, there's a monitoring zone where I'm totally okay with it if she does. You're gonna share, by the way, what systems are you against to put her in? I see I'm not saying anything about the cameras you have at night, in there, in the bathroom or in the shower. You know, these jokes can start, right? Or someone's gonna say something about the financial system. Okay, let's not get in there. Although your banks have access to them anyway. But still write even to systems like whatsApp or Telegram, you'd really have access to them. Oh, I brought you three tracks in the Silicon Valley. I want to bring you the fourth thing that a man moves above the chain of all decisions. Less manual implementation is taking place, more choices of goals, criteria of success and responsibility, yes. On the one hand, it's important for your work where you are. I think we should clean this whole story, which concerns junior specialists. Or that some younger people will leave, they'll be some smartest thing. I think the smartest people are just saying that they're gonna be smart. But it's some kind of abstraction. Everybody stays, right? But it's important to probably choose your education or even understand your own career or business that you want to do, yes, or your life. What do you plan from your family inside? The fifth trend I want to highlight today is that all the world's calculations that are going on today, that's all I said infrastructure, energy is the dependence on the emergence of some kind of thing. A super-mouth ejiai, yes, that the calculation is getting, so in the quads, we'll call politics. And because if the pace of progress starts to depend on the computing power, what is called compute in the professional slang is what they call envi-- they start to depend on the power of the data centers, then the race. artificial intelligence, of course, it's becoming infrastructural, geopolitical. And it's not worth mentioning all the world leaders who don't even know it, yes. And I've given you a recent example when people think that if they have a lot of energy, they'll be very fast in their artificial intelligence. And the sixth trend you're looking for right now is to draw attention to it. It's not even a trend, yes. There's no sign of six training days of the Silicon Valley, but look at this point and this position. What if someone says, uh, needs to slow down, and, uh, it's like a pause, then there's a question about who sees speed, who checks who's entitled, uh, this is the brakes that's gonna be pulled and stopped? And how do you make sure that others don't work and do anything, huh? And the slowing story is really important. She's incredible to me. Over the past three years, I am personally very pleased that I have slowed down in a vast number of my decisions concerning technology, the infrastructure of the artificial intelligence market.
I mean, on the one hand, I'm very much about him what's called, I'm gonna drown and tell you, and we're really a lot of things just saying.
We're very much simply keeping information about artificial intelligence. It's not even about Tuzmoon, yes. We've got him, not so small, but generally, aaa, where I'm talking, where I'm at, where I'm acting as an advisor or just talking to someone on the one hand. On the other hand, I've put myself on a very strong pause on many episodes. I realized that the number, like start-ups and ideas that are being done now, is pointless. Uh, the number of start-ups that will die and grow bigger and bigger. Because I was initially bidding on companies like OpenAI, they'd be very serious. And you can re-examine my releases. I said, a half a year ago, I'm sure companies like OpenAI would be worth trillion dollars. And we're here for that. Though it seemed like it was funny for the market. Is that what you think about the evaluation, by the way? I know that we, uh, our audience, it's not about the investment market, but I have a feeling that the OpenAI, HyAI, Anthropic, Google is a tentrillion company. Although, well, it's clear that it's gonna take time. And first, it's probably gonna be some four, five, six, seven, eight, uh, nine trillion. Google, I see that plan has passed. These companies are in the near future, uh, they'll catch up with him. Although Google has, I thought I was telling you in the past graduation, he has an unrealistically strong advantage. Oh, well, it's very serious to get out of here, yeah.
I'm talking anthropic today, by the way. In the case of OpenAI, OpenAI, he also talks not just about chat-bots, but about artificial intelligence that accelerates artificial intelligence research. Look, it's clear that the goal of OpenAI is that artificial intelligence is, uh, this is ASI, there, or AGI. I mean, it was just that, you know, that human, it means that human- super-- super-crazy, yeah, it was. But at the same time, all they say in the market is that it's a very old, obsolete story like that two years ago. And I think they'd better tell them more about their central position because whatever they're relying on the proceeds, the market will win from someone who's actually got those billions. users, yes. Google DeepMind, uh, he, on the one hand, says that they write a huge number of codes using artificial intelligence, that artificial intelligence means that algorithms are offered there, checking them, improving them. Over the cycle. But it's not a complete self-improvement, and at least it's not even about what Anthropic is saying. I think they're all gonna start talking about it and talk about it. Google will be much more complicated than OpenAI Anthropic, in terms of the beast, yes. This is Apple doing a new update here. The Apple conference was held the other day, they updated and told them they had artificial intelligence pedits. I'm not interested in listening, yes, because when they show up on the front stage and say, "Look, so Siri will be here to tell us how the football world will be going." Well, that's really funny. Well, it's just, it's just funny, yeah. And they show that the pictures are there. Uh, is that a question, why are they doing this? Because I think it should be their shares. But it's clear that their shares won't fall because it's a corporation that's got a huge amount of funds. But still. Once again today's thought I want to tell you that we're not gonna be measuring models in the near future, we're not gonna measure chat-bots, we're not gonna measure agents, we're not gonna take these different tests off, but we're gonna take a look at them different ones. Let's measure-- that's what's called not the IQ model, yes, but we're gonna measure the speed of these laboratories. I mean, how much of these labs, uh, are created by the same artificial intelligence. It's an important parallel line to understand the AI native world or AI native soft. I'll see you in our nearest releases. Don't forget to sign and leave your comments.