I've almost stopped working in Excel in the last six months. In fact, the system is working with memory in the intersection, reviewing some mistakes, revising the code, and so on. And even Anthropic provides examples of big corporations that have introduced it, they use Netflix.
I can sit and say, there's a ceiling height like that, and it's all you've got to do with a 3D model. It's really a mini revolution now.
He did everything, said, "Can you start a session?" I said, "Mogue." I said, "Can you do what the app does?" He said, "Mogue."
SpaceX is all the power that is straight-up, which they have on the Colossus I center, and they're giving Anthropic. It's working here, well, some two options, I guess. Either Ilon's out of the game or he's really disappointed in the team he's got.
Well, it's a political move. Hello, everybody!
We're on ToTheMoon Channel, technological news, Silicon Valley sites all over the world. As Tanya said, today is exactly the world, given the of China. We, Tanya, are you working with me, how much am I with you? Fourteen hours of difference.
Fifteen, 15. I don't even know that this difference exists.
And to me...
It's a little closer.
I got a friend coming in two weeks ago from Shanghai, Shanghai, and he just got... he flew out at 11:00 a.m. and landed at San Francisco at nine. I said, "You got this, it's a plus, you won three hours instead of losing it."
I flew out on Sunday, landed on Tuesday.
Yeah, that's the other way, the other side.
More interesting, huh?
You're getting old. Yeah, you're old. So.
Yeah.
What, why do you want to start tonight? What this week, considering different systems and the information field, we can see a gradual situation where all systems are: someone new makes something, someone catches up, something everywhere. I'll be right back. For example, I read here today that Google will auto-operate all the cars since the twenty-second year, made in the Android environment separate, and there's Gemini. Yeah, I don't know how much he's gonna do, but the whole thing's gonna be fine. So we're going to see him, he's coming forward, right? So, it's like Tesla's got a great job working with Grok directly. Now, as soon as Google settles this issue, it's gonna be in English for the U.S. market, but it's definitely gonna end up elsewhere. But the subject in human life is worthy.
I don't know if it's gonna work as good as Grok's work in Tesla. Yeah? I'm not sure. Parallel here, ChatGPT writes, which means they're fully integrated into Excel and Word, and now ChatGPT is in Excel, Word, and it's kind of not innovation, is it? Because, uh, there's Anthropic there a long time ago. And the big question is, how to use it and whether it should be used directly from Excel ChatGPT or whether the GPT should work with Excel. I mean, I'm curious about your experience as audiences, too, write all the different experiences. I've almost stopped working in Excel in the last six months. Yeah, I mean, I'm doing everything that's hard or I'm making, or I'm creating, like, a model, like Gemini or there in Claude, or ChatGPT. But I'm basically ChatGPT using the version, I'm building all the tools inside. Not much, we know that Anthropic even released an internal, so cool editor, well, we'll give it a probable name as a desk-breaker. And I think that's the world that's gonna move. Although for someone or, for example, for people who have a vital role in Excel's work with standard, aaa, they probably use it inside AI. Again, write, very interesting, yes. You, by the way, Tanya, Ilnar, do you use AI a lot of tools? Like, I have Whoop, an AI bracelet inside or somethingsApp has an AI inside, right? Or there's an AI inside in YouTube, in different sofs, or there's an AI inside. Are you using this AI or mostly AI using directly through the model? Because I'm using it directly, mostly directly through the model. I'm a rare user of AI in different systems.
No, no, I don't mean it.
There's a version in the terminal.
I mean, if you opened it, if you opened, like, I don't know, a photoshop or opened--
Well, I'm, uh, I'm just saying that, uh, work might be a lot of people who think they should use code editors that have different models inside. The same Cursor or Plangs for VSCode, JetBrains or something. But somehow it turns out that I don't even use it. The code is writing pens, and all I don't understand is I'm looking inside the models. Agent development is still in the stuff that I work with, not so far, that it can be directly released. I see the guys use the flames and use Claude in our main. Here. But I'm not going to be around for now. There's nothing else: notion, no Excel, no Photoshop. I just don't use these things.
I'm using, for example, the Inside of the program, including. But first, I'm saying that it's Excel because it's just, you know, a standard form for some people. For example, we often make statements in the architecture of interior design, there are different. I'm, like, if I'm going in some other format than Excel, I'm all just saying, "I can get an Excel plate. And so as many people would be rebuilt for a long time, that there's not an Excel sign, but something else, maybe even more comfortable. But no, they have a sign in their head.
And the fact that the conjectors, I'm allowed to use AI inside the canva. Then another program like SketchUp, which just makes me feel really good every time. It's, uh, it's a 3D simulation option. And I use both inside AI and now I had to load Claude just a week ago because they splintered and Claude made a conjector. When you remember, Ilnar, you and I said it would be great if at some point you could just sit around, dictate models so she could do everything for you. And they've already been doing this. So far, she's a very stunning cococos work, but it's a matter of time, right? I mean, for example, I can sit and say, I have a room five, there, let's say three things, the ceiling height is that. Then drop the picture on top of it, and it's all you're about to figure out a 3D model that I used to draw manually. I mean, in design, it's just something that's really happening right now. I'm not even able to track it.
Actually, this is what you're talking about voice-injection, it's just a recent trend, the last one, there, well, six months, I think it's loud. We've been living on parole there for a long time. But the intensification is going very active. Those who develop with agents and others are already writing, starting to talk. And given that the voice entry is already well and long, it's getting used to. Yeah, and the trend is basically about saying that we should do GPT chat there, Claude, Gemini or anyone else, not write, right? And then, accordingly, just some comment, the truth.
It's funny that design is already active.
It's interesting to know how much design I've always thought about the next thing that Claude himself or GPT chats don't take this function. Because, for example, I got another friend here and decided, so I'm gonna need to paint some of the garage where the tournaments hang out there. And he didn't use, okay, professional, and he's got the tools, and he's at the chat-- in the GPT chat, he took a picture of everything, told me about what he needed the drawing. The GPT teased him as the paint he bought as much as it was. I've done a full-blown, all the tools. We went to the store, he only picked it up, and I had to pay for it. And
I wish GPT had taken it.
Yeah, I didn't take it, but he could have formed a home Depot order by the way.
To complete the home Depot order, by the way, yes, it's a network, a network that is large, at least distributed to the U.S. And make the whole purchase. But I mean, all that happened was not in some kind of side program. And there's a little bit of a plus, because the backup program doesn't have some extra communication where it could say, "Look, what am I supposed to buy, so the kids can paint?" Because with him, my little sons were drawing and painting, and so on. I mean, add some things where the system is limited.
So he used it all, so he's got a jackass out of the system. Here's the code. For me, I've become such a busy user in the past few weeks. And that's how I always wanted to create agents or code without moving into any other system, right? And it's what I'm saying to him to write to me, to confer with my heart sensor. I have a separate sensor one here, I'm studying some rhythms in it. And I once told you that I bought a separate software there, it was expensive, I used it as a mortgage. I thought why would I use this software, even if I paid for it, $500 or $700? We even split up. I wrote what kind of sensor it was and what kind of thing was that, huh? So the sensor is worth $100, it's super professional, unreal, right? So he scans all the time, there, if you wear it, the scanner-- he writes all the heart rate. So I asked ChatGPT to study his data, to conquer his bluetus. He's done everything. I said, "Can you start a session?" He said, "Mogue." I said, "Can you do what the app does?" He said, "Mogue." And then I said, "Can you write me a separate app for Mac?" He said, "Legko." He wrote for Mac. I said, "You can write an interface to him so I can start and keep it. And you're gonna need a report that I usually need in the ChatGPT I did. And here's the report Cubios did. Can you give me that type of report and keep the raw data? And he finally wrote me that pill. I spent a total of 15 minutes, I think maybe 20, because that's all I said "hm," it made me time. And then I had to come over and clicking, because he's been in a mood, right? And I have an app. I've really spent very little. I'm, uh, Ilnar, on camera, I remembered Suzkever, and I think I'm gonna write the same security cameras, because I've got a lot of cameras hanging, I don't know, there's, like, 30 grand. I'll write to the cameras, to the cameras, and I'll put on some sort of video recognition model, so I can recognize certain objects. So the camera will recognize animals there. There's a lot of them here: skunks, coyotes, raccoons, snakes, reindeer and so-- rats, pumas, et cetera.
Cats.
Yeah, yeah, who wants to, yeah, there is. I wanted to-- but he's often wrong, and I wanted to have some of my system. There's a camera, what route she has, what route this animal has been repetited, not repeated how often it's been encountered. It comes, probably, the same animal or other, and so on. To impress, analyse the log, keep the data, well, I'd like to have some of that volume, get onto some other algorithm, get in with other sensors, and so on. And I've got my system written, "You know, those cameras you've got, they've got a closed cloud of your own, and you won't get access to them. Here. But the next time I buy cameras, I already bought the fashion cameras that just had their clouds, and the next time, it turns out I don't care. I can already buy these old, uh, new cameras, but those are professional ones that need to write a sof through API and that I usually do, right? I'm ready to buy that. All right, I'm not gonna be afraid of this anymore because I saw him as close to the sensor and made me an app. I can honestly tell you, it's super cool, right? I came in and said, "Can I have my Whoop on the other side and so on, to collect raw data?" He said, "This protocol is closed." I said, "Okay." But in fact, of course, it's a top pilot, because the data I get, collect and how I launch, that is, I'm going through this uncomfortable soft, and the system actually works for me.
It's also connected to the GPT chat. I mean, here. I'm just like an addition again.
Very interesting, it's like Excel. Excel is number one, there's a peace saft, right? But you said Tanya, by the way, that people are asking for reports in Excel. I don't work in Excel, but I'm sending people Excel reports that are made in ChatGPT. Anyway, it's all about it, because ChatGPT is today--
I'm doing this now, too. I mean, I hate Excel.
Yeah, yeah. And he's making you the top pilot of the table, right? I mean, he's doing this to be a super quality. He beats them on the sheets, he'll unbear... and he'll make it normal. That's important, right? So ChatGPT does what most people don't know what it is. So people have a very big problem with data normalization. He's making you a different cutting, and you can check their calves, see. I mean, that, uh, level is different, that's, in terms of creation. I wonder if Excel will stay. I wonder if our viewers will write about this part in general, even about an extracel or something. What you want to do is go on. So, look, there's a very interesting subject. Everyone's been stepping up their memory. And there's a, uh, more technical story that Anthropic introduced dreaming and regime, I don't know, Elnar, have you been in or out? It's--
No, I, yeah, I saw the news that Claude could see dreams, yeah. But...
Yeah, but it's kind of a dream. Actually, it's a story that isn't dreaming, right? It's basically a system in the middle of the day, and, uh, it works with memory, renegotiates some mistakes, redefines the code, and so on. And even Anthropic provides examples of big corporations that have introduced it, they use it. There's a Netflix doing something, someone else, huh? But the point is, this system is parallel, while you're not working with it, doing internal work with your memory or through a review of something.
Well, you and I know perfectly that we're not moving on the subject of "response." We're moving in a new time of work, including a memory work. And here's ChatGPT you-- you're going to be renovating. Today, the day before yesterday, I get a little bit of a little bit of a little bit of a new job, "I'm sure I want to try a new job." I don't know what kind of a job I've been working for. He's in English from the top. It's ChatGPT, it's OpenAI. I have the whole system everywhere, so it says I have to answer Russian, right? I mean, I read English easily, but my chat in Russian. And he's texting me. He writes the English request and the request is as follows. He says, "So please, uh, based on what we said I should do on this weekend?" He writes this question in English, first of all, and I'm also given an answer in English, which is what I'm supposed to do on this weekend? It's weird, right? Because I have all the machines written in Russian. Look at the quality of the GPT and memory. So what did he offer me? First of all, he offered me. Well, I'm a super-steamer of GPT, huh? I mean, I have a lot of chat rooms, projects and everything. He beat me up first, took out the weather that's gonna be on the weekend. Here. And he wrote, "Because I know you." And then he writes business, there, e-waisering, public performances, spiritual practices, family, lithstyle. And you live in Bay Eria, it's south of San Francisco, where the Ring Valley is. I'll do you a high quality, and a high quality reboot you for the weekend, so... And I'm telling you, Saturday morning, you have to land something. And I'm suggesting you go to the park, so you go to school. There's a falloly service, falloly, like, parks like that are alone. Well, Tanya, you know what this is, right? And naturally, where the house is, the house, the flowers, the gardens, the gardens. And he's offered me a study to go from nine to twelve volunteers, volunteers, so some team is gonna do something. And then, and then, we're still about five of those same nonsense, and he offered me the same idea. I don't even want to read anymore. Then I was offered to go to Stanford, so the art center and then, uh, somewhere, huh? And he's, uh, doing it on me, doesn't know I'm going to Stanford on Sunday to Stanford for a piano concert, like, right? Or he doesn't know what four kids are, a garden and a bunch of stuff. And I have a dog on Saturday and a fucking wife, and I have a friend in my guests. And I'm gonna do whatever I need.
I'm going, I'm running, so I'm going in the morning and the day.
I'll make a volunteer.
Yeah. And tonight, he offered me a symphony orchestra to Stanford, so... Here. And that means I'm three, right here, and that's what I-- I'm gonna die if I go to three of those activities, so it's a day. It's totally stupid, stupid memory reminds me when the new chat started, what was it called? Remember where he made that one for a week? I don't know if she's still working or not working.
Pulse.
Yeah, pulse. She's working, she's not working. Someone's using it or not, someone's using it or not. I don't think anyone's using it. They've done some bullshit again, though there might be something interesting to show you about memory. I mean, it's like the system's gone down, you know, it's not smart. They're now default, five and five installs. They've had a decline in hallucinations, and she's got to answer faster. So even a free version of the system works much better. It's just a top pilot. And it's just a little bit of a little bit of a memory. They also say they're gonna start talking about it.
They're gonna start telling me where they're getting their memories. So they have a new function if you meet it. It's a very useful story, but I think it's not gonna be a very good time at OpenAI right now. They say that you have many memory sets. I'll remind you that there is a predetermined memory, there is where you write, there is a memory you said: to keep certain data in the chat room. There's a memory where you're personalized. There are chat rooms you wrote. And ChatGPT understands that it's used and used, by the way, by her own good. I don't know. I see now I'm starting a new chat, he gives me examples of my chat. I mean, it's a good thing to use. And now ChatGPT has started a new thing, and you can see it, and on the basis of what memory it was based on. So you can understand what memory it was. I don't know how this works. Cool, not cool, because I have a plant that doesn't give me any cuts or abbreviations without explanation. So any abbreviation that comes into chatting should be explained, well, because the abbreviations in the world are very large. And so when you have a ChatGPT, you get them different forever. So any I-- ChatGPT doesn't care about my memory, he gives me abbreviations, abbreviations on the abbreviation, and then you sit, read. According to the USCMEAP review, that's what the statistics are. What are you doing with this? Here. The memory will be developed in a clear way. And perhaps Anthropic will show some more cool things.
Maybe not. Since Anthropic released Dreaming, we're waiting for this Dreaming and Gemini and ChatGPT. Yeah, they'll show up, they'll be clear. By the way, I was writing in the code-- Sam Altman, "We've expanded all the limits again, and we've got a Fast regime." Elnar, do you use the code or not?
Yeah.
But the Fast regime is five and five extra hai from me, right? Five and five extra hai. But it's probably about the extra hai regime, right? Not in the mode, I think you'll be there, you're getting low, medium and high. I'm still-- I'm telling you honestly, I still don't understand why they're making a garage like this different models and names, why they have an app on Mac different from the code. And I think it could have been combined. That would be really cool. And it seems that Codex Cloud has been working more or less. Although it's hard for me to use it for now, I'm still using Codex Offline. I mean, this Codex, I remember, is the OpenAI agent environment that is installed on the computer, where you have the ability to, in fact, not program. So she's doing more like a setup for you, creating, creating, creating, like, some tool in that nu--- sure, that's the code, right? Or write a program. But we shouldn't be interested in this writing the program because I'd be like a GPT chat, which does something and ends up with results. And he has a lot of steps he's done. I'm just so happy here. Yeah, there's a file he created a bunch of files. And I gave an example last time that it was like... how easy it was to connect to the CRM system and unload the data. By the way, he recognized all the calls very easily. And that's, of course, the top manoeuvre, how easy it was to do. Now, a bunch of handless programmers, analysts who often don't answer me for a long time, I'm not going to be able to turn them back. So, here. And I'll be treating people who use it because I don't want to write myself. Who has any examples of these simple aplikes, too, tell me, because I think it's a very useful subject for a normal life. And as one of my friends said, who came to me, that I finally-- I got the opportunity to realize my dreams. I've had a lot of this in my work. I've been writing a lot of technical assignments in my life, and I've often been unable to let anything out myself. I mean, I have some idea, I let it go, I can't go through, and I have to go to the software, designers, the editors, and so on. And now I have a lot of things to do, to do it myself, although I can honestly say I don't want to do it.
I'm fighting hard. But when staff and colleagues and companies are boycotting the use of new tools, or lacking knowledge, mind, some kind of lookout, they're gonna have to do it. If someone connects to the cameras, tell me, how interesting they are, it's very interesting. Such cabs, especially video recognition and home-based, are not programming. I mean, I'm more interested in the case here. I think it's important in Tuzmuni. It's the case where people do it, well, as a very simple interface, very light. That's how my friend draws a wall in his normal life. So he didn't know what color he was gonna pick, didn't know where to buy it, didn't know what kind of volume to buy, what other instruments to buy, what professional sequence it was supposed to be. And yet, when you're all set, everything's done, and you're still drawing a picture, and then you're gonna check it out. That's very interesting, of course. By the way, ChatGPT was on Tanya's store. He took a picture of the palitre while I was waiting for my money, and he took pictures of the paint palitra. I think he's gonna buy the most expensive paints that mix in the store. They're ready, right? And...
Everybody mixes in the store, Sasha.
Yeah?
All the paints mix.
The man was-- huh?
It's a base, bro.
Yeah, man was twice as much as he planned in the first place as GPT was planning on, by the way. Here. And when there was one mistake, when the context window was overwhelmed, he took a picture of the palitra in the store and said to pick up the colors in that palitra. And he died here. Well, of course. Although I think that's the way the palitres are being chosen in the right picture, the idea is to work. No, it shouldn't.
No. You're just a knife right now. That's the only thing you've ever said. I don't believe GPT chatting can even do with the best development, because he has, uh-
Why?
He's limited to the iPhone camera, for example, right? And she's far from human eyes. I mean, I never even choose a color on a computer, I have to see it alive, make it look good and look at different things in this particular room.
That's a famous subject, isn't it? Yeah, that's a famous subject. Like a monitor, like a camera, and so on.
It'll be just completely different colors.
I thought if he told GPT chat, what color is paint. Look, this brand will give the brand number in the store. Plus, he takes a picture, plus he looks like a shot. The GPT was drawing him. He used some color code, he'll go and check on that code. Still, the palette of the flowers of professional producers, it's a digital code, digital, and it's supposed to be a match.
Tanya, I thought so.
You know, I have, like, a special sensor that cost a lot of money that uses these, I suspect so much, Benjamin Moore, you're probably talking about some kind of price in America. It's a lot of expensive paint, but it's a sensor special, which needs to be brought to the wall with certain lighting, it's allowed. He'll tell you this color code you can come with. But that's all so relative, too, because it's gonna depend on the way the lights went on that wall. And I still have more faith in my eyes than-base, of course, for most users, this is a really great subject. And as I'm not tired of saying that design is more accessible to many people, but it's gonna be sky and earth for me. What GPT showed, what you're gonna end up in your garage, I'll see that's totally different.
If he gave the code. I have one question, but as a friend came to me, I changed his colors. I showed him that he was wrong in color on his own, on his own, on my eye. I found a mistake right away. That was it. It was real at the store, by the way.
You see, yes.
Look, look. If we're trying to get a job, Tanya, GPT chat to paint the wall, he's picking-- I'm telling him I'm gonna have to buy some paint for it. And for example, I'm willing to buy some sort of color or a certain price, small, medium and so on. He's coming right away. I don't think he should give me a picture of the hair, and then there's no paint like that, or they're worth a million dollars. He's going there, found a palitre, and he's showing this palitre and he says that this palette will be a little different. But you have to buy these codes, and that's gonna help you as much as possible. I think that was the task that was made then. Then my friend wouldn't have to pick a paint in the store in the eye if he had set the target for GPT chat. Because I would have set the target for GPT chat now, thereby minimizing the risk of choosing further. At another point, GPT said, for example, you could come and if you wanted to have fun, I gave you some idea what the drawing should be, but you're all gamma colors. And what's blue and blue, dark blue, blue, blue, blue. There's fifty blues out there, you take your own.
It's far more than fifty, believe me.
Yeah, yeah, yeah, yeah. I'm talking on parole, right?
Yeah, yeah, yeah. I'm talking on parole, right? Again, it's probably a brand-dependent thing. I mean, I can come to the store and, like, I want to buy some cheap paint, and I won't have that kind of verbal, right? Or she won't look like that.
Right?
You know, you can even get a piece of paint from a cue-- that's pulling from the wall, coming over, giving, and they'll make you the same. So there's no limit in the edge-- in the color, you can get any color. But I don't believe GPT's even gonna do that in the near future. What a human eye designer can do.
By the way, the human eye is a diaper-human eye. That's why it's losing to date, it's considered losing cars with a lidar, losing, for example, the technologies Tesla did. I mean, Ilon Mack once said, "We're not gonna have a lidar." Everyone was laughing, saying it's impossible, impossible to do. And I read one, there, one technical post on this subject, which happens when the system scans three dozen lidar spaces and then separate cameras, three dozen spaces with cameras. creates a more probable mistake than learning to use cameras, right? And while I see the result here, it's not in other states, oh, not in other countries, I see the result of Tesla, which still drives everywhere, anywhere. Her autopilot works pretty well. I see my type of modern autopilot, which is made by lidar cards, then sewn into the road car, and how tight it works compared to Tesla's working. And it was a story that there was no reason to have a man's eye and work through a camera, not through scanning, uh, space and objects, remoteness of these facilities, and so on. Here. So let's see what happens, of course, that's what happens. But these cabs, they're very interesting, and they learn more or less rightly or wrongly to ask questions again. Although you and I understand that we must come to a system where the system will be recommending and advising you from the beginning that it is not just a stitch to paint the wall, but then say, "Look, if you have a wall, You're not getting enough money, you're not gonna have a choice of paint, like you can mix, you're only gonna be able to buy. And so we have to figure out what colors are, and let's see those flowers how this fantasy looks like you." Yeah?
I think it would be fun to put the projector in there. With a GPT chat, brush the picture, projector it over there, project it and make it look like this. Just as parents are interested...
You know, Ilnar, you're
You can see it in kind.
also--
The first, first potential problem I've seen, we've already been speaking with flowers. And the second potential problem, I think, with the size and proportions in the end, which can be solved by the projector. Because GPT chats, that the same Nana Banana is very sinful, they're actually just taking it from somewhere, I don't even know the logic that's in the design. They always don't match reality. That's why Sketch Up, for example, it was removed through Cloud when you were injecting, there's a way to regulate the size. This function is not in the GPT's chat of the pictures, nor in Nana Banana, you can't. You're trying to explain to him, do twice as much, do twice as you can. And it's kind of a really stunning thing. So I wonder how this drawing will end up looking. I don't know how difficult it is for you, Sas, but I think it's the second potential problem that might arise.
What he gave, look, he gave a screwdriver, I don't know how right it is, a slo-word, what is the right name.
Well, elevation, huh?
With the scale. Yeah, it's a scale condition, I don't know, ten centimeters, there, or inch, or something. And it showed in terms of the real length, width, height and everything that he had originally given. We gave him a width and a height. He's all very structural, structural and clear, and he's shown what he looks like. I mean, according to the plan, it's okay. As far as he's inside the width and heights, he's got the right thing to know about the objects that were on the wall because there were different tourists, there were rock climbing steps. As far as he clearly did, it's another question, right? And maybe he could see them as you speak correctly, not--
Model, yes
I can't count. But I think if he's doing this again, he's not doing it in Nana Banana, but giving it through a tool that, you know, through a common chat or a common environment, special, some kind of module, The first time you're gonna describe it in the original data is correct, the most correct reference data. I think there's a chance to get a better picture. But recently, for example, I'm in GPT chat room asking me to move couches over there, and in the same room, I had a wish, there, something to change, and he's in my hair. I asked him again, he's hanging again.
He's been doing it for a long time.
He didn't move anything like that. Yeah, and I realized I didn't really understand, I'm not really dealing with it. I have no discretion, I'm not a professional man here. I mean, I like something visually, but how do I create this final picture? I can't make it, and I don't want to do it now. And I don't even know what the GPT chat-checking needs to do to me. Because what GPT did to me, I can do it myself to date. Yeah, I mean, here. And all the rest of this is I have to go to Tana Pyrocha. Well, that's right.
Yay!
I don't have another option. Well, I can't treat you forever, so that's it.
Why not?
And accordingly, and accordingly, I don't trust the system yet. But I think that a lot of people in the world, they need
some kind of simple design decision, and they'll get a simple solution.
Yeah, 100 percent.
What kind of theme do you want to go to? C-- so fast, quick information has been provided, statistics have appeared, how many people in China use different models. I still decided to give up the latest current data on the number of users that are currently available in different systems. I think it's a very useful, well, useful design for perception. So the first story is in China, it's supposed to be six hundred million users and all of China. And I will recall that Chinese models give a sum of five hundred or six hundred million users. But it's also users, if you put everyone off different systems, there are people who cross. It is believed that around two and a half billion people are being used in the world. But I don't really understand how it's done. I think it's an abstract number. It still remains that OpenAI has nine hundred million weekly users, one and two billion monthly, like. And Gemini has a million monthly, nine hundred, nine hundred or eight hundred million monthly. There's one interesting aspect about Grok that I spoke a few times and yet I'm speaking because Elon Musk and SpaceX. Grok is part of XAI. X--- this is SpaceX, right? It's a Grok. Can you call SpaceX, Elnar, right? And...
In fact, yes, it's a related company.
In fact, yes. Still, it can be the most expensive 100-- well, it's the start-up in the world, right? Because if he comes to the IPO, he'll be assessed there about two trillions. We've talked about it. And, well, they have the most important infrastructure, they have very serious data centers, which even some corporations don't even have at all. And for example, they now have an Anthropic agreement that Anthropic will use their power, including. And with SpaceX--
That's the most important news. Yeah, that's the most important thing I think this week.
This is a big topic. Oopa-- why did I come in with Grok like that? Grok remembered by number of users. So, I wonder if Grok, he might not be a great confession in the world because he only has sixty million users counted, but active. But Grok is in the top of three systems in the US, and yet inside the US is an active system, it's very much developing, yes. And I wouldn't even look at the users again, her HAI connection, yet this connection to SpaceX and the connection to SpaceX infrastructure, and there's still a connection to Tesla. But as much as Tesla is another corporation.
I mean, SpaceX is separate Tesla, right? And they're so dense, though inside the infra--- they're kind of connected and re--usable technologies inside. But what Anthropic has concluded with SpaceX is very interesting, especially when the courts are now in between, what is the name of this court? This is a trial between Elon Musk v. Sam Altman and OpenAI. Although the defendants are in a series, Greg is Sam Altman's partner. There were, there were Mira Murati, an ex-Federal AI. There's Eli Suqewer, this man Elon Musk attracted, and then Suzkever himself didn't want to let Elon Musk get a single office at OpenAI. There. There's a big story there. Microsoft in defendants.
Microsoft in defendants, yes, Microsoft in defendants and in seriousness, they have decided to enrich themselves in the subject of a charitable organization. The court is that he may be subject to special promulgation separately, perhaps not as it goes. The court is that the essence is that Elon Musk says that the original OpenAI was created as a non-profit organization, a charitable organization that will develop the system for the benefit of society. She was, by the way. These are people who have disconnected from Gogle for-- because Google is a corporate structure that was approaching to do, well, we're calling it Gemini now. Then it wasn't Gemini, but it was AGI, yes, and, and it was artificial intelligence. And OpenAI was separated to do why the name is OpenAI, open artificial intelligence, to make it for people. Yeah. And the main subpoena is that Ilon Mac says there would be no company without me. I'm attracting the main people, I'm attracting the key contracts and partners, such as Microsoft and NVIDIA. I gave a lot of money, there, tens of billions of dollars, so you broke everything you could. You owe me $150 billion. By the way, the interesting thing is, Sam Altman wrote to Elon Musk a week ago, "Sam, you wanted to make a commercial company, and we offered you a share of this company, and you turned your share off." And he wrote, "You're still on it." Here. It looks like a bribe. No? And, here. Well, there's a whole series of things against Mask. Because they say that Mask called the first GPT chat with a trash app and a schoolboy, and that's what's gonna do better. And the Mask wanted to make the system commercial. The mask wanted to run it as a single man. The mask wanted to join Tesla. And actually, he's doing XAI now and he wants to compete with Open and Open-ended. So, really, SpaceX is in the Anthropic as Anthropic, look, it's a company that's further connected from OpenAI. Next, so Dario's off the OpenAI. Right, Elnar? Was that the same chain? Yeah. And creates Anthropic. A company that already costs almost a trillion dollars. And less time has passed. So, the OpenAI has been in the midst of 2015, and Anthropic has been in the middle of a few years. And the company is already worth a little more money in the round than the OpenAI. And they said that the OpenAI is not for people, but we're for people, we're for security, and we're about safety. And including that safety and security in artificial intelligence were dealt with with with Max Grigoriev. And the Anthropic is disconnected. And here's Anthropic with Elon Musk. Elnar, do you want to add spaceX on this part, Anthropic? Yeah. Because, like you said, it's a big deal.
The whole market is how long we've been talking about the last year and a half, fighting for chips, everyone missing chips, prices grow, memory is worth it. Everybody's fighting for their chips. They impose sanctions against China to prevent him from being chipped, so he doesn't get it or anything. And then the news is, SpaceX is all the power that is clear, and it's on the date-centre Colossus-1, and it's giving Anthropic.
Well, of course not free, but for use. So we're not gonna use that as SpaceX, and we're gonna give all our $20,000 NVIDIA chips to another company. And it is here that this resource, which everyone is fighting for, is needed to create new models, train, compete, everything higher and higher. Ilon's just giving it away. It's working here, well, some two options, I guess. Either Ilon has outplayed all the people and like, don't play this competition, don't pay millions of dollars to super-coolers, scientists, researchers, and so on, but you just need to get this resource now, until he's gone. Because it's been years, and these cards are not as valuable as they stand now. And then it seems that he's out of the game and he just earned money on it, and everyone else has been scattering their resources. Either he was really disappointed in the team he's got. There was a news recurrently that he gathered here, then he kicked everyone out again, then something else. So there's a constant reset on the team and a failure to meet the results that the team is doing, and just put a cross on all these research projects. Because we recently discussed that they and Kursor have interrelationships, yes, that's why they have a partnership for the sixty billions that they have the opportunity to buy out and close some of the programming.
Ilon already said that the Coursor isn't worth so much that it's not a very good system and it's not written well. As I recall, the last were data. Maybe something changed.
Well, GPT's the first one that's written to a schoolboy.
Kursor wanted it, and Kursor wanted to give a round at 60 million billion, and this round failed. And then Ilon was told the news was about to buy 50 bucks. And Ilon goes out and says, "There, the one, the one, the one. So, Kursor is something bullshit. And they kind of fell down and offered them a little more than ten. I'm gonna have to see what happened in these last few weeks. A week ago, there was a subject like that.
Well, the point is, Ilon's design seems to be either puzzling or something. Because the resources that are needed to learn new fundamental models, he's giving another company, in fact, a competing company.
Or it's a political move.
I see what money is.
Something I think Ilon Mack's gonna be doing a little bit of a pause.
And there's another story. Maybe Anthropic just got so strong that he could just come in, stumble his finger, say, I need those resources, how much money do you need? And toss any amount that's needed. Because they're getting capitalization so fast. I remember a few months ago we were discussing that they were four times less than OpenAI, and they were far from there. And here we are discussing that they are already under the trillion, so they're attracted to this assessment to new rounds. And it looks like they can stumble into any object and say, all we have to do is, let's deal with this. And then they found just two hundred thousand NVIDIA cards, they're starting to use them. Anyway, this news is a little shocking.
It's a very big move from the Anthropic perspective to SpaceX. I guess we just don't really understand the tricky Mask game, but I was really surprised when I saw the news.
Well, they're all mixed up. Still, the computer is not enough in this case. There was an update that OpenAI is required now a year, sixty, this year will be $60 billion in compute, and that in the next five years, it will cost six hundred and fifty billion. I'm gonna pay you a dollar for computing. And you and I should understand that this is the amount of these calculations, it can't decrease, it can only increase. People get used to it in a way that's unreal and they're getting more and more into it, and they're getting used more.
I mean, it's clear that people are getting cheap in there now. There are models worth $20,000, $200, there's a code of laundry. It is clear that all this will be limited at some point in time, but still the resources of these capacities will increase. I mean, he can't grow up, right? And just Goldman Sachs said that the growth of Capex was capital. In fact, let's do it. In a simple chain, you could say on AI and compute power, it would grow to over $7 trillion dollars over the next five years. Well, I'm telling you, this is a very serious, serious amount of money, investment, company value. And the report looks pretty realistic. Well, planning is pretty realistic, because we see what money is being spent now. What are the other companies gonna do about it? Like DeepSeek was now, attracted to a round, he's fifty billion, yes, attracted, Elnar, me?
I think so.
Is he fifty, fifty, fifty, or 50? Fifty-seven-sized. He's fifty-six.
I think it's an assessment.
Estimated fifty. Yeah, it was a little bit of a little bit of a special news. At first, it was like 50, then fifty-seven. The media write what they want, right? And it's not the money that you know, it wasn't even around, is it? With that Capex that's pouring in America. And at the same time, it was news that NVIDIA's top charges in China were more than a million dollars. I mean, the price is just dried up in a crazy way, because it's impossible to buy. Well, you're getting less access. And logically, no one has access to toptops, right? And I think the line is worth a man saying you have a cycle. It's earning, there's a few years to use the previous generation's fee there. New futures will be used sooner. I mean, we all know that Tesla, well, SpaceX in this case, they're making their own payments, and they're gonna be running them off, then they're gonna be using them in NVIDIA for a whole series of years. I mean, it's still a future, right? The future.
About seven trillions that Goldman Sachs. It's just funny. One of the first episodes we've got on the canal.
In one of the first episodes we have on the canal, we discussed that Elon is driving around the world, gathering seven trillions to build chip factories and so on, attracts money.
And you still thought how many budgets, uh, Republic of Belarus, to a total. And here we are, like, the same number. Goldman Sachs is talking about seven trillions now. It's just a little bit of a numerical magic, funny.
I wanted to add that I was surprised here because I am in China talking to various large producers, suppliers and interested in many of them whether they use AI. Naturally, as I expected, they're very active in using it, but none of them use Chinese LLM. No man told me that. Most of the time, everyone uses Claude, not even ChatGPT. And they're very active in the infiltration of agents. What they told me they were doing with agents, I kind of lost a thread at some point, because they're getting some unreal stuff.
Well, that's a great news for them about, uh, SpaceX interaction with Aicic. Because right after that, they've taken off the limits somewhere, somewhere, lifted the limits on Claude's use of the subscription code because they've cut the limits quite hard for the last few months. And after this news, these limites went up again. So, therefore, for use in Claude Desktop's agent systems, or somewhere else, it's getting much easier to do.
Write what topics you still want to hear on our channel, and I'll see you in a week. ToTheMoon Canal.