Hello, everyone! We're on ToTheMoon. Technological news, sites from the Silicon Valley, the USA and the world. We go out every Sunday, we're twenty-four. Man, that's not real at all. Since we had the best video model, yes, the foundational model of that year, I've never been so popular. Everyone asks me how this will affect the market, what do you think? I'm getting a text today, everybody calls. Order me, my friend, two cottages. It looks like you're such an absolute idiot and you're an operator. The companies I've listed under a big blow. Yeah, so Google is a winner, a big winner for all of this. Only we discussed that, with the arrival of the Trump, half a trillion dollars will be in the US for new investments in AI, as the entire Internet and the entire Silicon Valley exploded in DeepSeek discussions. We'll talk about it. What is this, what is this hard, complicated words? I'd even, uh, maybe, uh, I wouldn't mind at the same time, thought, uh, something came up again. But when you see the fall of NVIDIA's shares almost half a trillion dollars, it becomes very interesting immediately that it happened that it had such a very strong influence.
So, DeepSeek, what do you think? And tell me about this new model that was collected on 10,000, someone writes, really, $50,000, someone's 10,000 maps of NVIDIA. I think NVIDIA died and died right away because the Chinese were the ones who said they were using. I'm in touch. Yeah, I got in touch with that, so I died. They're waiting for sanctions, called, waiting for some NVIDIA problem.
No, like the last edition, I'm very fond of the model. I like design, I like the thinking areas, and then I don't know, there, I'm waking up a university professor, yeah, there, or something, but when I first see thinking, and then I see the answer, I'm a lot. I'm more interested in that. The answer itself may not be as good as the OpenAI, for example, yes? I mean, if you compare the same requests I sent to DeepSeek R1 and O1 the ChatGPT version, the end result of OpenAI is better. Well, for now, on the briefcases I checked.
Yeah, yeah, yeah. I mean O1 20-idellar version. I mean, it's not super-pooper that's advanced versions. I like the final result of this request processing, but I am very happy, first of all, to read the thoughts DeepSeek has added. Yeah, and I repeat, I still don't understand how last time I said it was about marketing, which means it was added to, you know, the implication was more interesting to the viewers. Either it's a real thing that allows for a high score. It's probably gonna make a good deal of this model, but the marketing component is also clearly present.
And marketing, what do you mean
marketing?
Well, look, when you start some prompt at DeepSeek and you put a tie that you're interested in reasoning, yeah, that you need to connect the reasoning, he's starting a process that's in the gate. thinking, that's what he thinks. And he starts thinking and thinking very anthropomorphno. Like the user is asking me to do something like I can do with her? I can get to her on this side. Okay, I'm starting to think, and then he's like, oh, wait, and I think I missed it. And it starts to add something, and the process is going on. It could be a long, long, even on some uncontracted question, you'll have five or six readings, but you're looking at the final result, right? I mean, it's like...
Of course! It's like a man, it's like a man already tells you.
Yeah, yeah, yeah! I'm saying I'm getting a little bit of a hunch. I imagine a student answering some of my questions and--
He's trying.
Yeah. And I feel like I can not only hear his final answer, but also as he approaches that answer. And that gives me great pleasure. Well, first of all, apart from being a pleasure, it's also useful, in my opinion. I mean, when you're studying some kind of topic, you know, impoundly want to figure out how to do something. In one case with OpenAI, you get a specific answer, perhaps well structured in the form of some sort of instruction, as you need to decide. And in another case, you can look at this chain of thought, as you call, yes, that is, the sequence of thoughts, how this answer came, possibly with a very simple, understandable end-up for you, and Then how he got to the final result. I repeat, the final result could be a little worse than the OpenAI, at least on the boxes I checked. But this sequence of thoughts, how the answer came before, how they did, how they did, how they looked at the boxes, what problems they had taken into account, it seems to me a very valuable story. Here. As for your second point on learning there on a small number of GPUs and so on. Well, we're saying from the graduation that it's much easier to repeat than to get it on the start. That's the first. And the second, again, well, we can't believe that there's a parole or a few thousand GPUs, that's all done, right? And the companies that buy it at 100 grand GPU, they're actually spending money in nowhere. And that's why everyone should sell NVIDIA stock or something. Too many strangers there, but I have experience with this model.
What DeepSeek showed is that, of course, there will be a very strong democratization of artificial intelligence when a man on his computer is not just gonna start ChatGPT, but a ChatGPT who works at his computer. 1,000 times cooler. And that tells us just about-- not about what you think America lost or China won, but what artificial intelligence is incredibly evolving. And we see what the big changes last year have been. We were just making a skate on the most important trends. And of course, we're gonna have a lot of changes in this. What's going on, uh, in the whole process of technology, uh, technology, technology?
Man, that's not real at all. I mean, I've been since when, since we had the best video model, yeah, foundational, I wasn't so popular this April. I was very popular in April, and today I'm popular. Everyone asks me how this will affect the market. What do you think?
It's a dynamics, Sas.
I'm getting a text today, everybody calls. Actually, I'm very popular. Remember me. Well, if it's serious, there's no type of training code, how am I gonna make some precise grades, right? I mean, whatever. I mean, of course, they used the OpenAI data for training. Aaa, of course, the architecture of Mixture of Experts, yes, such high level, as shorts are understood. I'm guessing it's impossible to do it in five million, but in a few dozen millions of millions, it's possible to do so.
You mean dollars, five million dollars, so we can all get the audience.
Yeah, five million won't be allowed, but for a few dozen...
And they write what they did for five million dollars
or what?
Yeah, yeah, yeah. Where within 30 to 100, it's possible to do so. Yaya, I say I support Alexander Weng, that is CEO Scale AI, who said DeepSeek even if they trained on more GPUs, they can't say because there's a problem with expertise, right? So we don't know that. They're lying, not cheating on purpose, not on purpose. This is all we don't know right now. We just realized that... I think we're the key to this, that we've identified that, like Meta, NVIDIA, Perplexity, a, Mistral, Cohere, Writer. So, these companies I've listed are under a big blow, right? So why is NVIDIA in the middle of a hit? Because it's not clear why you're training big models, is it? Well, why? Just. If, on a parole, if it's not so fast commercializing, right? I mean, it's clear that Google will have a lot of GPU because they need to be so angry. They can always redistribute, right? It's still one organization, Gugla, and it's redistributing between internal and external, yes, needs. For training, for inference. So for Google, I think Google will definitely win this. Their strategy is very clear. They're number three now, we've got to be number one. Oracle gave this Stargate Project. Google needs to run, buy GPU. They can train models. Many of your products, many external clients. They'll figure out how to redistribute the GPU, right? That's understandable. OpenAI is not clear now, Antropic. So would VC want to keep supporting these start-ups, especially Mistral and Cohere, to give them money to-- support these crazy valuations, give them a lot of money on GPU? Well, the answer must be no, right? I mean, I think Mistral and Cohere are clearly losers. And Poolside and Magic, that is, all the Gina I companies that have been modeling, who have an assessment, there, billions higher, that's all, I think-
Well, you mean, in this case, you've been raising models that were based on a lot of, uh, hardware-- equipment.
Uh, hardware-- uh, equipment.
Yeah, yeah, yeah, yeah. Well, everyone who's been raising under the model has raised a lot of money.
It's hard to say, because DeepSeek is now, DeepSeek is number one now in AppStore in the USA. So OpenAI is also under attack, I think. Oh, but they're so much, well, as it's clear they're already part of the national project, so they're so fast-tracked, OpenAI, but not s-- well, as substantial, but--
Today they're number one because of 00:07:25. Today, on Monday, they're number one. Oh, now it's clear that Meta is under a big blow because Meta's strategy under the business model has failed with a big crack. Oh, building their infrastructure, they didn't build shit, did they? So nothing's done in three years, but PyTorch was like it was, right? That programming language, right? Python training frambler. Nothing new has been done. Lama's not needed. I mean, it's now getting much more gentle. Their ambitions have become more gentle. NVIDIA is also unclear because if... It became unclear because if they were not, there would be no rabid demanda for modeling, but again, inference, they were more vulnerable to competition. I think that it's all about, yes, I, I agree with a general Twitter reverend that, well, you can't expect AI-reshers with a net worth of $10 to 100 million to grind like Chinese, and looking for new, you know, unlimited training budgets, will find new ways to train models. I mean, I'm generally in agreement with this direction of thought.
Wait, but with AGI, it is clear that what is sought, uh, giants, is not an explanation, that there is no evidence that the creation of AGI will not require a large number of capacity in this case. It's just a short-term task that is now in place. But clearly, to work some artificial intelligence, it needs a huge, hard-wall power, very big.
Yeah, yeah. Google is definitely a winner. Google is definitely a winner, right? Because Google sees Google as if at the model level, he's a little behind OpenAI. Now everyone knows that the difference is not really that big between these companies, right? That it's possible to get fast, it's probably gonna be fast. Google has many users, training expertise, unique data, external clients. So Google is a winner, a big winner for all of this.
Please write in the comments. I'm very interested in your opinion. Let's do it like a vote. Who thinks? Are the NVIDIA stock going down or growing? And here is the thing about what, and, indeed, if there are more companies to do something easy. And all the giants like Meta, Google, OpenAI, they've already lost the market, and nothing happens. Or vice versa, it's some kind of time story. And the Chinese people have all been fooled and told the truth. And they spent 100 trillion dollars to create DeepSeek. Let's see, let's see. Write it and then we'll make sure we do. The topic, I think, is very fun for discussion. Look, for example, Meta is building this-- a huge number of dates-- and, uh, Manhattan's all over, they're just getting a little bit of a little bit of a wrapped up, right?
Or OpenAI together, that's what we talked about last time, which is supposed to be investing half a trillion dollars in five years. It's just like a start-up discussion, okay? Maybe they'll be five trillions of dollars, maybe less. What's the name coming out of my head? This one, Stargate, yes, Stargate. They're just building infrastructure, general infrastructure and saying they need a huge infrastructure, they need dates, they need-- they need electricity to do all this. I think this is a story that, my personal opinion, history, shows that models are starting on a small number, and, uh, relatively small, yeah? Because 10,000 there or 55,000, uh, 50,000, uh, NVIDIA maps are not that small. When they start on, uh, a little, uh, equipment, it just shows that we're moving, gradually moving into a phase when research is easier and easier. More companies, people, countries will be involved. Well, there's more going to be going to create something very serious. But that's my opinion. I think, in my personal inspection, NVIDIA's shares will return not just to the cost, they'll grow. I think we'll see a very strong growth in the twenty-fifth year. Maybe he's not gonna be like this in the twenty-fourth year because in the twenty-fourth year, NVIDIA grew up a hugely, and that's just... So far, all the models that were coming out, and while we're taking the twenty-fourth year and watching what happened at Google's, what happened at Meta, what happened in X when everyone else made statements and when someone else did. I was doing some open source and so on, it was all somehow, and then everyone caught each other, doing something again, and it was just a little bit of a mess. Then again, the OpenAI appeared, said we had some cool thing, and then we were moving forward. At least once again, I feel it. Let's talk about the next story.
I've been testing a long-standing operator, OpenAI operator. Uh, I thought the case was with what the Openai was wondering what they showed some of the cabs, I don't think I know who they're designed for. So he's sitting, uh, an officer in the OpenAI, saying, "I have a son who likes to eat sushi, not just two years old. I'll order him some food." They go into the open, they go into Instac-- he goes into the p-- uh, delivery, Instacart, and he says, uh, to the operator, "Cup me..." The Operator says, "Come my food." And he's supposed to open up a browser, enters in Instacart or, there, a mobile application opens and orders him food in Instacart for sushi. Case is so weird. I don't know who he's even meant for. Is this a game or something? Yeah, so you're sitting like that, you think, don't you make me a cottage? So. And the operator says, order me some cottage products. It looks like you're such an absolute idiot. You're an operator. Oh, well, you're like, uh, someone's a little bit of a rough story, right? And I imagine this subject will be very serious. Again, my opinion. I'm just here, I think I'll give him a good job. I have a payroll system, and I need a payroll system to monitor certain types of errors that are available every day. And in this payroll, I look manual, I walk in hand. A small, small payment system. I'm the one who gave him this job to open the payroll, enter this payroll, check, check, check, er, c-e, find, er, lock, and send me notices about these locks. Anyway, he opened up, and then, you open a virtual car, you get on that payroll, you go into that payroll system and watch it work. First, the operator didn't work when I switched on another screen. What's going on? Like I don't have two hands, two legs, and I don't have any. I mean, he asked me to work, so he decided to come in with me in the straight. I mean, he's--
Sit, look, Sasha.
Huh? Yeah, sit down, look, yeah. Sit, watch him do it. Sit, watch him do it. Yeah? Here. And, uh, second, of course, he did very unbelievably perform the task. Well, that's totally childish, idiotic, very stupid, performing tasks. I don't know, Ilnar, I've already released someone or Google or Microsoft. You didn't test?
Antropic had it.
Ah, Antropic.
A few months ago, yes.
How much of this story has become real, because I didn't use the Antropic, not the same, and I didn't see how real it was, there, this skies, he was pretty good, easy.
everyone would use that agent, and so on. And the OpenAI somehow shows it so that we're gonna have it in agents now.
Yeah.
Although, in fact, the system looks a little strange at first. Well, I'm picking up the phrases, so I'm a little confused about words. Because, look, you have an operator, first of all, in some browser, well, not in your browser, yeah, but in your OpenAI. You're putting your passwords in there, logs, and so on. Well, let's say it's safe, though it's already got some questions. Then, accordingly, he's communicating with this browser through the squeaky boots, through, uh, there, the card recognition and the other. When we talk about working with the browser, well, come on, it looks really weird. I mean, instead of being directly involved through the code, it's allowed, even API is gone, but you don't mess up pages or anything, and you do scrinches, on the basis of these violinchots, you're making some conclusions. Go on. The route seems to be very wrong in general. It could work with some applications on your computer that no one else has access to except, like, a keyboard, arse and screen. I guess. But, man, how many yuzkeis do you have on this case?
No, we're--
Very little.
They're trying to figure out what's not all API, and it's not likely, API, it's gonna come to all of it. They still want to resolve the question, uh, from the point of view of a simple solution. In theory, I would have preferred. Look, let's just say I gave him access because I knew I had him on the phone. Fuck that.
I'm on my cell phone. Fuck that! I gave him access. Okay, let's get some data. I'm willing, uh, to donate some, uh, some kind of risk for, uh, eff-- very effective, right? Just like we're accessing API, we're also giving someone access. Well, garlic. He turned around, even walked in through the browser. But first, why should I be there? And secondly, the quality of this job. I mean, if it were really smart, it would work there, somehow, and it really worked for a man, but I'd assume that this case could be alive. Even though they've increased now, they have this, uh, 4O Tasks. It's a slip. I mean, uh, it's a pretty good thing. You can test it. That's who's watching us, you can test, see the whole development scenario. They've given it a notice now. So you can put it like that now, you can set up a dish, you have an extra dialogue window, you can do it again, tell him to do some sort of periodicity with-- you're doing some kind of job. Well, it seems like a little bit that it might be, uh, some kind of skilled skies, but it's been working really hard here for now, uh, it's been working really hard. I'm on it. I mean, I'm the-- well, you're right to say that maybe it's a non-working scenario, but I kind of figured out a scenario like that, it worked, and I never got to start it. Ah! That the funniest thing was at the end, I didn't tell you. It's downloading. It decided to export the file from this payment and not to make a scronym. It came in, exported the file, and then it clicked to access the file and wrote, "I don't have access to the file." That was very funny. So it's downloaded this file on a virtual car and then it couldn't open this file to work with him. That's what's going on. I mean, this is a little under-development, and I think it's, uh, absolute under-development. So why put a market in, uh, so stupid, unworkable? And the OpenAI, which, once in the hell, throws these tests out. Or do they want to see the user's behavior? Well, what people do about it, what scenarios they're gonna use, I don't know.
Maybe they want people to see the script.
Sas, it's more like they're just getting more Hippa points than they're getting off the product themselves. So they showed the flames then, huh? There's another version of the plagins there, and then there's this dish coming up, and actually, it's a very strange thing for me. I mean, yeah, on the one hand, I tried to protect them, that it's the difference between the spoon in Telegram, that the crypt is being performed at some point. But if you need some scruples to do on some schedule, well, it's easier to just rest a little while, right? Well, I know that there's not everyone who can rest, but somehow...
Not all of them. No one at all. No one can codify. Yeah.
Sash, but just a little bit of a time-to-be-wrapping for some time-writing-- it's so artificial, I think, a thing. She's really...
Well, simple.
Cut.
Very simple.
Yeah.
Very simple.
And here's an interesting situation. On the one hand, uh, I want OpenAI to do some fundamental models, development. I mean, well, they say they're gonna have a really cool oc3, right? Well, there's a three-way out there. And this is all around, yeah, the same Canvas, the same crypts they've been doing-
It looks really stupid. I'm c-- it looks... No, look, it looks clear. The only story. You say the following, you say they're making absolutely stupid food, baby, right? That's how the guy who came out says, "I've been working in OpenAI for three months and I've come up with a frontline." It's really like that, right? On one side. On the other hand, you think it might be some sort of strategy to throw such products away, to watch people use, uh, to think what they're use scenarios to move into. AI. I'm just hypothetically suggesting that it might be-- they're beating up with this-- they're watching what new scenarios people might have, what might come up. Because, in fact, they find some kind of use scenarios through it. People didn't know what proms would be writing. Well, obviously, they've gathered a huge amount of, uh, people in the last two years, actually, how people write proms, and how people write queries that people ask about the Internet at the same time.
Nobody knew what they were gonna ask. It's just that people have to figure out how to work with it, what to do, and so on. Here.
It's substantial.
And I wonder how they are and how many people will be drawn to themselves and stretched, right?
One by 30, or what? One to thirty, right?
Yeah, yeah, yeah.
About. Some kind of space difference, some kind of space difference. But you're saying that you're talking about a professional market, you're talking about a B2B market. Right?
Oh, my God. Look, come on. Uh, yeah, it's very narrow, but somehow, if a man, uh, can barely program and wants to make him have some kind of access on the trigger, some LLM request has gone. Not like ChatGPT's time-time, yeah, but any trigger. I saw a man on the camera or something, you got something going on, so it's much easier now through DeepSeek, uh, API to send out a few requests than through OpenAI. And for such a subsoil of OpenAI users, too, perhaps...
It's still a professional use, yes. For the professional use of DeepSeek, can undermine the market for Anthropic, for OpenAI, clearly, there, for Google, professional use can be compromised. For natural persons, however, we'll watch the quality, quality of their system, right? Is that how Sasha says that there's a number one in App Store? It's still a race number, uh, including natural persons, yes, ordinary people in terms of their lives. Because whoever knows who wants to do this, I might even, maybe, and I could programme it, I'll never do it, right? I mean, one for 10,000 people. Uh, just split up so everyone can understand the use scenarios, yeah, that ordinary people won't run this now.
Well, literally the last gun that way. Sasha said that when he talked about Mistral and other companies. They didn't just present the biggest model that was on the level of one and the other, but they did, as they themselves, distilled models that were much more than they called. less volume. Yeah, they're not that good, but they're totally blowing up all the other open source model and Lama and Mistrial, and so on. And so many products are going to be on them now. This is also a very significant change in the market. So, in this regard, the Chinese people, of course, have shaken the market a little bit, and the commercial component, open source. So I wonder what they're gonna show. And as they say, uh, West in the face of OpenAI and Google, Antroika. Can they show something new in principle?
And it's interesting, and if they show, yes, conditionally, and they can achieve results in a few months, as practice shows, will they show? If the Chinese do the same in two months, it's cheaper.
Well, the twenty-fourth year just showed again that, again, it showed that it was cheaper, it didn't always start to cost more. Yeah? I mean, like a company, in terms of, or in terms of, the business of infusion, but we'll see that, see what happens, what happens, what happens, what happens... Who owns DeepSeek? There. There are no corporations in the owners. There's a man who worked, I understand, in the investment, he's got a team, and they've done it with the team together. Sasha, this company, company, does she have a cool owner now or not?
I have one rule about China - never ask who owns, ask what revenue, ask what profil. If you don't ask, you can learn a lot of things. But the point is not to try to figure out the business. That's my only rule about China, which would be... Not if it's kind of serious. Well, I think I can't believe that this is the story that's on the Internet that the cvants have made their own, that the cvants have made their AI a reserver. I can't believe it. That's right. I'm sure there's just a lot of things that don't talk about.
You mean they're such a small team that they could collect a story like that?
No, well, there's probably a lot of things they didn't tell me about. I don't think they're just lying around, but they're not telling. And nobody can check because it's China. That's it. I mean, basically, you know, I got a drink tonight, and I got a company up there... We brought a drink today. I'll tell you what, the company raised two million dollars, and it cloneed up to 100 million revenue in four years, five years. It's like that Cinderella story. They've forgotten to say that they have a private equity that gave them 50 million borrowings. And since, well, they didn't raise it, it's like the loan money they gave back. They had this opportunity. I mean, you have one, you know, people don't cheat so often, they just don't say what they don't want to say. That's it.
No, so why show that it really affected growth? What's the point? It's basically what I think Meta says they have a billion users. Well, I got it, I came and said that you have everyone who has whatsApp and Facebook, Instagram, all Meta users, well, or some small percentage of these people. That's it. You've got all the users. How do they use it? Well, I mean, the users of Lama. Who's gonna launch it? Who uses how often asks what kind of requests are being asked, huh? What length of requests are asked and what complexity of requests are being asked? That, it must be a big question. And that's the kind of statistics that I'm getting.
I'm asking you a question. And that kind of statistic, that's how no one else has, nobody sees it, and nobody really understands it, and nobody really knows it. Mmm. Aah, de-- but everybody, it's gonna be time. You know how we can go back to this company. Today, uh, what's the end of February, huh?
And I'll see what happens at the end of the year. We'll be making a final issue and remembering the main trends of the year. We're just at the beginning, by the way, how much? Two weeks ago, we had a graduation or three weeks ago, the main trends of the twenty-fifth year, based on the twenty-fourth year. Who wasn't watching, look, please. It's just a very good hit in the trend. And, um, we, by the way, have talked a lot of things out there that two leaders in the world are clearly drawing out country--- China and the United States. And how China has grown up so much over the twenty-fourth year, and continues to attack the five USs in the twenty-fifth year. It's just that it's really going to be a little crazy, right? Here, and especially here, Trump said about $500 billion. Right here, just right away, China hit him, China hit him sideways, so DeepSeek, yeah. At the same time, almost simultaneously.
Yeah. I wonder if this is the gap between, uh, the way Trump said, and then China actually collapsed the markets. There was an anon in India that they wanted to build the biggest, uh, data processing center. They have investments, well, they have a lot to date, and they also have investments in billions. Not 500 billion, of course, but as in another region, but as a consequence of this story, it's also beginning to invest. So the rich people on this side, after Trump, started looking more active, started abnormalizing.
And China is getting in here, yes. How much is it? On half arillion, you say, yes, Sashi, NVIDIA's capitalization fell for today's...
Well, there's a plus-minus, yeah. There's a place to watch what happens next. What happened is that. It's just what it's got to do, right? It's connected, uh, if someone thought-- well, maybe it's related to something-- but the fall's substantial, yeah. Maybe someone connects what, um, impulse, shares cost 100, 40, there's five, one hundred forty seven, now it's worth 18. But at the time of the release, something might change, yes, we'll see how much it'll be. She's already below three trillions now. When the graduation started recording the graduation, it cost a little over three trillions. She's worth two and nine trillion. Can you imagine, yes, what numbers are? Very interesting numbers are under discussion. We're gonna sit, "Shit, so, like, a hundred billion, so the company's got it." But that's a number, it's just a drawing. The guy who sells now may lose. Maybe he made it because he bought it differently. Here. Someone's gonna buy it now, and it means, right, well, he's got a plus, for example, it's gonna be. I mean, it's hard to say. I think that now, uh... What's this about? This is because it's supposed to be less necessary to build equipment. So if again, he-- they built on NVIDIA, then there's a lot more team going to do it because we need less investment. If you need more investment, more people can do it, more can do it. They also buy these-- they will also start buying NVIDIA equipment. I mean, a little weird. That's weird. Or something more serious can be done. So if they say you can build a good model for a small number, what if you increase the number? You can build something better, you can give people better services. I mean, you're the one who said that about one, like, the result was better, right? Or the system is slowing down. I mean, I have, like, a pay-as-you-go version of the Pro, there, two hundred dollars a month, uh, and I still see that it takes a lot of time. I'm waiting. So in the 20-dollar version, they can write you, "You don't have any more, more requests. Yes, the limit is over." What if I want to get some huge resources inside? So this is all a very specific, specific subject, right? I'm probably just asking questions. It's a model like that, that was a year ago or six months ago, it sort of solves. Well, we're living on progress. I'm just saying something else. I can see that I've already started to analyze, ask, do some other stuff. So I'm not a corporation, am I? Which, really, yeah, if you look at the B2B market, most companies, that's most, I know a lot of different spectrum companies, practically nobody really does, uh, new ones. Models. Well, deep, no one really analyses data even from the CRM system perspective. I'm saying, well, well, seriously. The whole analysis is very often just like this kindergarten or often a hip and used a little. So people use it a little bit.
No, not above the professional level. Well, I'm not talking about people who are very smart and super-smart.
Thank you very much, everybody. We're on ToTheMoon Channel, technological news and sites from the Silicon Valley. We're going out on Sunday. I'll see you in a week. Bye, everybody.