About the leaking code at Anthropic. To date, it is no longer possible to remove such copies from the Internet. The possibility of transporting and importing chat from other models in Google has emerged. The project structure that is now in place is a priori unworking.
I think it's an idea, uh, a start-up.
Hello, everybody! We're on ToTheMoon. Technological news, Silicon Valley sites around the world. I'm still around the world.
We're really around the world today.
Yes, today. I'll be at Creme--
Three, three, three, yeah?
Yes, but I'll be at the Cremnival Valley tomorrow. So, uh, Ilnar, what do you think about the leaking of the Anthropic code? At the beginning, when I was there-- I read one post in X yesterday when, uh, I described the number of reproductions of this code and how, uh, this code was sliding into, uh, blockage, so that, uh, this code is stuck there forever. Although I think it's impossible to remove that number of copies from the Internet to date, there, with people's computers. But someone said, like, "This never leaves, so." And yes, and plus, uh, the descriptions made it feel like X, which means everyone's been drastically cut off the code, uh, and then moved on to, uh, their own code. And then when you start reading about it, you know that, uh, Cloude code isn't there, right? And tell me a little more about this story. I think she's very important, especially, especially she's important. It's super, so it's a company with a new system of soft-walling, applications, approaches to development. There's a artificial intelligence that Anthropic itself has to, uh, close any possibility of copying their own code. They even have a system like that, don't they? From copying your own code. And it's like you're a mistake and a 500-- What's the volume?
I'm not calling the exact size, but as much as the Claude Code reference code, it's online.
Yeah. Tell them what Claude Code is. So for an understanding of the work of these companies. Well, it does matter, doesn't it? It might be a little bit of a change in perception of the market, because it shows, uh, that's a good thing for artificial intelligence
developers.
Well, first of all, uh, totally agree with you. Everything that's on the Internet will never be removed because everyone has been copied, who can. Someone made a trout, you know, there's a response, someone copied themselves. Well, I mean, a lot of people have copied. You say, even in the block, it's all gone. So in no way can Anthropic be able to make that code on people's hands. He's always in huge quantities. Uh, ironic is that Anthropic, they always say that we are the developer of artificial intelligence safely. We're very careful not to make mistakes. They say that the last six months in Anthropic are not written by hands, and they have the entire source code. And the question is, the mistake was created by the man who made the release, or if people don't have the code anymore and do nothing with their hands, it's still artificial intelligence, so that's how. It was, yes, there was such a mistake. I mean, uh, by creating a huge quantity of artificial intelligence products, we're probably putting a lot of potential mistakes in there. And most likely, one of the potential mistakes is that it's come out that the whole company reference code, it's, therefore, out of the open. Claude Code is, well, maybe if not a system-setting product now, but it's very important for anthropic. It is clear that there are a large number of API applications, but a large number of people work through the Claude Code. This is true in programming one of the main products used when AI tools are spoken. I personally know a few very large companies that pay their employees a subscription, there, a Claude Code, and they say that even, well, not that, uh, they recommend us use, but that's right. They strongly recommend the use of Claude Code, saying that it's the best thing that's on the market now. You'll have to use it. There's a way you're gonna pay for it.
They can't use it anymore. There's two and a half billion proceeds. They may not use or, in fact, they can't do anything about this
code.
From the original. Yeah, look, let's figure out what Claude Code is. Claude Code is a shell, an interface that can be easily interacted with, with LLM. In particular, he is trapped under the anthropic Claude. Uh, what's important here? There's some logic inside of him, that's, like, some teams, some extra checks, some logic inside this interface, but the brains still inside LLM. So Claude Code without anthropic Claude is like the same. Yeah, you can put that shell on yourself, you can tie any other LLM, like now a huge number of smart people have done. You can put Qwen, you can get DeepSeek on, you can do something else. So you're gonna have a convenient interface with LLM in the code writing, right? These are all the teams you use when Claude Code-- you work, you have access to it. It works exactly the same way. The only thing that's gonna change your mind now is change. The interface was in the past and the brains were inaccessible. Now, I'm gonna have the whole interface, and that's how all the people who want, who have a little laughs, can put this all on the ground, and I can get it on to those LLM-cam you have, yeah, there, either API or, You'll be able to start with your own. Well, it's like a story on OpenClaude, actually. Here. And accordingly, to continue using this interface with the usual, convenient, well-designed, but with other brains. Because if you use Anthropic's brains, you're gonna have to pay for it anyway. Here. Ah, I wonder if there's a lot of comments that Claude Code says. I mean, if, for example, we take high standards in the code, it's Linux core, yeah, there's some other stuff that's just out there that's gonna make it clear, effective and other. But there's no. Here's a little, uh, pup-pop and a little if we rephrase the standard phrase of the software. Here. Uh, well, whatever it is, it's a separate, uh, confirmation that it doesn't have to be written a straight-up, perfect code to make, like, billions of dollars, yeah, there. If you have a good product, if it's doing all of its tasks, um, it's the task that it's assigned, then it's the money that can be earned without licking the code for the rest of the time. It's another confirmation that some of the start-ups are taking off, yes, there, without a huge pool of programmers, who, like the Olympics, make the code as effective as possible. Here. Uh, well, I guess that's the kind of comments, yeah. The most important thing is that it's an interface and brains. Interface leaks, you can connect to any brain.
And a separate irony, which is exactly what happened to Anthropic.
That's very interesting. What you're talking about the quality of the code. And that's so relative, yeah, actually. And it would seem that Anthropic is the best minds of development, um, of the world. And that's the kind of comments I'm saying, yes, in terms of the quality of the code. It's all very, very relative. I want to go to what Google wants... But this is really important. Here. I'm not, uh, that's not what Anthropic's already fixing the security system, that's not gonna happen again, is it? It's still a serious company with an estimate, and it's probably more than half a million dollars and half a trillion dollars to date, and it's a big, big, e-e-e-e-e-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s-s their infrastructure. Uh, although I-- I don't think he'll help, uh, probably large operators, but it'll help different medium-sized model operators and operators in China.
Yeah, Sas, it could be DeepSeek's code of some kind or Qwen code or something. In fact, they'll just take this interface, a little napkin' to the dock and get out.
Uh, you want to say something? We've been talking about the past-- last time, and now, uh, this update is up. His Google announcing that there was a possibility of transporting and importing chat from other models, uh, in Google. In particular, he describes direct instructions as to how to move, uh, your GPT readings and Claude. Uh, how does that work? Well, first of all, uh, first of all, why would you want to do that? Well, base, why do you need that? I mean, if I used GPT chat for two years, and I have a huge history of my chat and relationship, then why don't I move it to another system, uh, all this data and all--
That whole thing and all history about me, so the system would know who I am. It's important to clarify that, in fact, it doesn't carry all your chat. So you can't turn to your whole story. She's moving information on how you make decisions. The main story is built. And who is it? Even that's a logic of work.
Most people, even who watches the current edition, she's gonna be surprised, why do you have to do that? That's the more I talk to people... My friend was here to Egypt, and a good businessman is totally here, but he opens ChatGPT, but he uses ChatGPT, like a plain version. And there's a ChatGPT and that's it. So you opened up, asked for it, and that's it. I'm telling him, "So, uh, you've made a request, and he's answered you in three seconds. I made a request, so we had to go to one place. He was looking for 18 minutes. " I say, "Do you understand the difference between analysis and quality and errors, and how much is this mistake going to cost you? "Ia, he, uh, at the end of the meeting, he came in and bought himself a version of the Pros. Although we're not much, I'll tell you honestly, we didn't discuss artificial intelligence, practically not screaming-- not touched. It was just a little, uh, little fast--- well, quick comment., but the interesting thing is, people are like that. So people still treat whoop bracelet to pay, $30 a month, and ChatGPT, well, 20 is worth. Okay, I'll pay you a little bit. I mean, it's just like an app to someone who does. Therefore, people may have a transition between migration between systems, it is relatively, well, relatively interesting.
Uh, when I read Google now, uh, I had some doubts because how does it work? Google gives a prompt, and you're putting a promp in ChatGPT. But I know that in ChatGPT when you load a prompt, he's getting you some data. Now, yesterday, uh, we've done a load, uh, uh, I've been loading up at ChatGPT, "What do you know about me, my nature, my decisions, my identity, my personnel, everything." He gave a lot of super-cool. Well, I mean, it's a very detailed description, as my friend said, "No man would have been able to describe you so structurally, well and correctly. So he's been describing it quite clearly. But there's a nuance. First, when he said I was interested in traveling, the only country he had indicated was Egypt.
I mean, and, uh, with the fact that in the context of my last month, I've been asking questions about-- because I've had a lot of displacements, he-- and, uh, there, I've been here, uh, 20 days a week. The fear-- not in his country at all. Of course, a lot of people asked for information. But! I was in Santa Barbara three weeks ago, and he's not talking about Santa Barbara, he's not telling me anything, yeah, there, and he's not asking. Uh, and-- or I at his place, asking about other regions, and he doesn't write about them either. I mean, it's clear that, uh, information about the man he knows, she's on the one hand, he's not bad if you do a lot of chat, he's doing much better now, by the way, than before.
Remember, we raised this question more than a year ago, and, uh, I recommend that all our viewers do this story. Uh, we can-- I can put out my promp, which I gave you, but it was very simple, to be honest. The most important thing is to ask him, there, the nature, the identity, the decision. Who am I, where I am, what I do, everything you know about me. And, uh, do this story, look, write it down, tell us how much you see this story is true. I, by the way, yesterday, discussing with my friend in business, too, I said it was a real topic in terms of getting a job. And, uh-- I told him I would even be in-- in the spot of a lot of companies, sending a prompot that would have to be asked. Or if you were in the OpenAI, uh, someone would be in the OpenAI, then I'd be asking for access to chat rooms if you'd let me. What he said was that it might be illegal in the U.S. And he says it's perfectly easy to do. You come to the company interview and say, "What are you doing? "The man says, "Antropic." "My? Oh, I'm a lot of use. Now, put this prompt in and open on the screen. " So, the man, it's clear that the exclusion of him...
But it's kind of too much, I think.
But look, look, look, look, except...
It's like asking for a psychologist you're working with.
Look, I, look...
... detailed information.
I'd be personal- you know me and you're gonna have to think about my different subjects, right? I would personally have access to OpenAI completely to chat rooms, GPT chat, so he's all chat--
Any man you'd be right now...
No!
...let's say that there would be a record down here so that any sub-signor or, uh, the person who watches it would have studied?
Yeah, first of all, I would love-- give a firm, I'm just saying, OpenAI, because she already owns my data, access to the system, so she can analyze me when she makes a decision, if she'd take it. Decision. If I were to talk about other companies, I wouldn't have given a full system of chat rooms because it would have been a copy of all chat rooms confidentially about me. OpenAI has it. If we were to talk about the downloading of the report on me in real time, and there's any report I'd have downloaded, so I'd download it and show it. Well, I'm here to get a job, and I'm using the system. I'm a man who's not a story. It's very important. If someone here is gonna say something, uh, uh, let me attach my answer yesterday about-- in the PDF, to the full that he's in the chat room, huh? And there, yes...
Ilnar would never let that happen.
What do I want to add? No, no, no. I'm gonna... I'll explain now.
I can see it right in his
eyes.
What do I want to add? And, Tanya, when I was talking about this kind of work story, I wanted to get it straight. I'd like to put on any request, I'd be honest with you, and I'd be asking him to do it, but in general, if you use that function at work, and it's legal in certain countries, so, I'm sorry that there are requests that are illegal. We don't know. The court in Spain the other day, er, decided that kissing, uh, the hand of a woman without her consent was sexual violence. I don't know, maybe asking, uh, chatting, unloading history is sexual violence, in some country. And that's why I'm not saying, I mean, if it's legal, uh, what would be a prompot? It could be written, removed absolutely things that do not concern or touch certain characteristics, such as some exclusive search requests that people do, there, or an exclusive-- you know? And he's talking about me, he didn't...
China, for example.
That's exclusive.
No, exc-- well, a person may be interested in some things that, uh, can be considered as in society, for example, as a matter of urgency or as a matter of fact, can have, uh-a, medium credibility. Like, he said to me that I was interested in something, something, something, something. I don't know, Egypt. Very funny.
Easter, Sasha.
Very-- yes, very funny.
You've become an east-easterly in the last month.
Yeah. And that's the law-- that's gonna end, right? And, uh, uh, I hope so...
You're not gonna get hired, unfortunately.
Yeah. And then, and then, he says, "You're interested in the garden, there, and so on, yes. I wonder if he's got a special, big area. But a lot of things.
He said, “Tun-Tun, Toun-Tun”.
No, he said, "I'm interested, I'm interested in, there, technological stuff, business." Then-- cars. I liked him saying a few times that I was interested in cars. I sit there and think, what cars do I care? I never understood.
Yeah, well, the man who doesn't have cars in the garage because of the walls.
Come on, come on. Tanya, Tanya, Tanya, I-- our viewers will think of something wrong-- so the point, uh, here I say, like, some, uh, uh, we're on the news, the chat migration and the usefulness of that, yeah., and many of us are all-- each, I think that the man and, well, more than a billion people in five years now realize the hard need for them-- migration between systems, models. But only if we're not one model in the world, right? The need to understand, yes, it will. It's just like we migrate photos, like, or phone contact, and we're migrating. Here. And for the report, write what you think., your opinion and c-- from the perspective of this idea. If anyone uses it, use it in business, describe how it went, how people reacted, how people reacted, what they said, in terms of what happened., and my buddy is saying, ", if a man says, "This is gonna take ten minutes." I say, "We wait, wait, we start, we'll start." And the charais and the charais will be la-- uh, ho-- will be clear.
I hope my team will report on me, and I'm editing--
We can get someone else's phone, go with him.
I'm a ph-- I-- I mean, in fact, in the modern world, the system will understand that you're trying to fake or that it's not credible, or that these requests are incorrect or something. I mean, I still think that if you're sending out every day...
Than, that's what a lot of people do. Why did this story start? This story started with the fact that he employs a lot of people. He said, "Loudy was so much a liar, so it became unrealistic to say. Courses have been offered to tell how to cheat. " And they say, "We catch people, and some people are on--
We catch people, and some people tell us some people tell us that, uh, uh, they say, "You lied." They say, "You're the only one who checked." Here, uh, so given that we're living in the world now, and that's, by the way, not-- well, how much, God, I hired thousands of people, as I saw in the summary of lies. Now, everyone can write the right words, the right indicators, the right parameters, add the details. You can add details and re-examine that there's no detail. Yeah? How many people come to me, they say, "Alexandre, you had this man, he was the most important person, he's got a business up there." I say, "I don't know this man, I don't know." And people don't want to believe that I have a recommendation. They don't want to believe it. I'm telling them, "He didn't work for this company. I was in this company for this time, Director General. I didn't have any directors." Well, if a man was, uh, three-minus three from me, I could forget, there, my name, I could, well, and less two, there, I still had a lot of people, but I'll ask, and yet I have memories, I have memory. Very good, you know, a memory of the name of some of me will still be. What do you think about this, tell me? And tell me if you'd have done that kind of loading. Tanya, you're blocking, but you'd have loaded your chat if you came to work? Not chat, but, uh, I-- that's what the... he wrote, who you are.
Well, um, I guess so. Honestly, yeah.
Of course, yes.
Because, in principle, what am I supposed to hide? How do you call it, I don't make exclusive requests.
I'm doing a separate chat, right?
Yeah, yeah, yeah. You're just making two scouts, one of those in common, two private.
One free and another
feeble.
Yeah. One of those fascinating. No, actually, uh, that sounds radical, though, actually, uh, it doesn't seem to be much different from the intake tests, in my opinion. Remember, sometimes you get to work there, and you get tested there for 100, twenty, 100, thirty questions psychological. And that's-- on the course of this test, there's a psychological profile of yours, and it's decided to take you to work, not to take you to work, you're not going to that position, not you. You're going to take this position. And here in this sense, it's just a cheaper version of this kind of test. I don't know if these tests are popular now, but when the university is done, you come to the company, you're a huge printout, and you sit down, and then you're checking it out. The next day, they tell you, there's a psychotic you're like. Well, you're probably not officially taken from the questionnaire, but it was definitely taken into account in time. There, for the prosthetic, there's more to it, that's what happened all the time. And in this sense, ask your ChatGPT, as your psychological profile, is about this way. You can sit on the questionnaire, then someone will check it out. Either you can, there, through your ChatGPT, get your blindness somehow. Here. The only thing you're saying about everything is lying and everything, I'm gonna add that if it starts to use it as--
No, not everyone's lying. I said it was a lot. Not all of them.
Look, if it becomes popular and some good job can only be through it, there's some courses that will teach--
Of course!
Yeah, yeah, yeah. What kind of requests to send, what things to ask.
How much to prepare. I mean, if a man can...
Year of training.
If a person can make 40 requests all year and she can check, for example, if you're using this system. For example, I have a system that can smash the entry parameters and, for example, check me out if I'm lying, not deceive the system. See if I could fool around these chat rooms and make a wrong person, for example, considering that I live there, worked there, my resume, my social profile, and I also have a profile, I have a child, I have a age like this, and so on. And she can take it, so you're gonna have to figure out what you're doing, Ildar. But now, people are making certain identities.
Yes, it is.
Yeah, they can write some articles. And there are people who use it and do it. There's someone who doesn't, but there's someone who does, yeah. A couple of locations in some media. And now you're like a cool man.
It's just that it's so common now.
Yeah.
Well, I put you a prom. Do it, we'll all bet our audience, put our-- our psychological profiles. Ildar, by the way, you said an interesting thing about it. Here, uh, you know, I've been using, uh, polygraph on-- when people get to work for years. And there were businesses where people went, well, thousands of people went through polygraphs. So, 100 per cent of all staff from the person who cleans the place or drives the car to the person who's programming the code or selling or marketing, whatever the person you are, The role, volume, level, all 100 percent of the population were polygraphed. Yeah. Most people, 90% of the cases, people have been quiet. I agree, yes. Twenty--- the only one with more dissent, twenty-five percent of it was there, it's the sales manager. And here's the question: why on the vast mass of people, even the programmers agree? The programmers would seem to be, yes, a very specific, specific reaction. People say, "Well, okay, no problem." I mean, it's clear there's no interest in polygraph asking questions. You won't be asked questions of a personal nature. Not much of a personal question, I don't know, Ildar, have you ever been polygraphed?
Yeah, I did.
I did. Any question of a personal nature you can always say, "I'm sorry, it's not exactly a job, yes, and business is not a business matter." And that subject, uh, she's been distracted a lot. People might be negative or positive, but any serious company checks you along and across. A lot of-- well, so, let's say, not everyone. Different companies are serious and not serious, but are serious, which are not being tested. There's a difference between you and me. Still, just to mention memory, I just gave a memory example, uh, to the fact that memory is now exposing some part of it, and it's pretty good to appreciate the psychological profile. I'm a portrait. Uh, sure, when people's accumulations last years and decades, that's gonna be totally different. And you and I can remember the Black mirror. And the Black mirror is very clear about these boxes, yes, when a man comes to work, he's scanned, scanned his memory, what he's done for the last two or three weeks. And, uh, he's having a big deal, and he knows it's gonna scan and get it, yeah. So this is the same story c-- and when you can't take it out. Now, actually, now, we can clean up. Well, let's just say, now, you don't have to create a person. Now, it's supposed to be a clean-up idea. But the question is, will memory be cleared? Is ChatGPT's memory clear that he understands and knows about you? Someone, by the way, wrote a funny comment at Tazmoon when we told you last time that I went to OpenAI on one question and how, some idiotic thing. I liked the comment from the man who wrote, "so they just agreed to the interview, and also agreed to use all your data in the chat room." Says they-- they didn't need your question, they needed all your data in your chat room. By the way, he might be right because the interviewer disappeared. I don't know what kind of ties I've been gluing.
Here. But I still live in the US.
Yeah, yeah, yeah.
Yeah, about the chat transfer. Uh, I was thinking about that, too. Who needs it? I mean, you want to move, like, yes, there, or I want to move my chat from OpenAI to Google. Even, we're not supposed to be a blind man, are we? That's where they gave the prompt, that's some kind of picture of me, and now we're putting it in Google context. And that's where you take a set of chat, like when you're on your phone to the phone, you and WhatsApp want to move all your chat there, you know, because you haven't had enough time with someone, something else.
Yeah.
So, about the chat room. I've been watching myself for a long time, and the top five to six chat rooms are in this chat-chat story, and I've got a lot of active chat, some of them sometimes fall down, but if they're down, I'm gonna get them if they're special. I don't open it, I never open it. And in fact, they're useful for, uh, ChatGPT in the way of what I'm asking, what we've been talking about or something. But at the time we're discussing some kind of topic, I'm working on one specific chat. I don't turn to the previous one. This is some kind of baggage context that LLM needs. And the question is, you're chatting yourself, basically, there, one year ago, probably not. Either I'm wrong, and right there, you're wondering if you're going to go to the old chat rooms you haven't been in there for more than a month, have you ever come in or not? And the same question for the audience. Is there something you have with a current set, there, five or six chat rooms, and you never fail in your life?
Yeah, that's a good question, by the way. You'll have to write everything in our comments. Well, that's a good question. Tanya, what's going on in this part?
Uh, sorry, Sasha, I had to clean up the dog because I was really snoring. I was afraid it would be on the air.
I'll answer that then you answer, you'll know what.
Yeah, come on.
Illar, Ilnar, look, you're asking for the current ChatGPT or Claude interface now, or Gemini, uh, or Grok, and do you need to move these thousands of chat rooms I have thousands to go to another? So, you're gonna have to move the chat I asked you a year ago? Look, in the current interface, the stream of the interface, I'm gonna express the word "pogan" interface, you know, clearly, it's a super uncomfortable interface for the
work, tuck.
C, that's the data. Of course Ilnar, I don't need him. Even in the projects, the old tasks I've ever asked are the problem. Not much, for example, because I use GPT chat, I sometimes open other systems, I use GPT chat, I'm looking. The search is dirty and bad. And finding something when you don't know the truth, I don't know, the name of a read or a true chain of words, it's very hard. I mean, you can't, you don't have a search you opened. He says, "What are you looking for?" I said, "Look, I was asking about the baby and the age and food. Can I find somewhere?" And he's really looking at all the chat, finding all the chat rooms where the kids were, and finding all the chat. Not just doing, there, stupid s-- chatting on the same words and synonyms, and so on. Search. And plus this tape is wrong. I mean, uh, I wonder why I'd like to move them and not lose them? I would imagine that in the future there would be a system and a model that could, Illnar, what I've described now, pack. And when I ask the system for something next time, like in a business, defined, in a tax system, for example, asking about a question or corporate contracts, I'll ask you a question, and she'll say, "I'll look up your whole story about these corporate episodes. I see that you've already acquired these forms in these jurisdictions, and you've had such a deal, like, here agencies, that kind of deal." Ta-ta-ta-ta-ta will raise and tell me, "We're going to keep going on this subject or not?" I say, "Of course, yes! "This is what I was studying, "then, this is the deal." She'll tell me, "Get new ones, for example, to get me that you've been up or up or use it." And keep going.
Or just an example, still a home-based one, right? I have a garden, and I've been at the GPT chat for years. And I make every time I explain my flowers to him, towns. I tell him every time I have dozens of ways to protect flowers or feed them. I've got to take pictures of them every time. I can't just say, "Look, take all the flowers and recommendations from all the colors, and what protection and recommendations are. And tell me, I see this design, I still have to do what, and I'm the same every time, "however I need to put millions of this means, that means--
Have you tried to create a project? Sad.
Look...
Sently speaking, the garden project.
Look, I have a lot of projects. There's no garden, by the way, Tanya. I think I should set up a project. I just brought it here. You're saying, "Sash, you've got so many gardens, so do it." But I'm a garden-- I can't do all the projects because it's a weird interface again, it's a weird thing. Plus, even when you're in the project, you're wearing a lot of chat. Ilnar said it was fun right now. You're five or six, like some chat you use, which you know where they are, in what projects or not in projects, you're in. All that went away is because the names are stupid, because you have this sideline incomprehensible, invisible, you can't-- it's not a system for normal work. They're wasting money and creating Pulse, and then he's gonna die and shut it down, right? I think he's already dead. Here. Maybe I'll stay, I don't know. They're-- they're basically the system they tried to do, which shows you how your life is, but it shows, uh, some kind of news kit-- well, some kind of shit, right? Here. Unworking, unworking nonsense, which is likely to be dead. It's like they-- they put on a mobile device the model configuration, and on the web-based model configuration, they're all flying, they're changing the wrong ones. On the mobile attach-- on the desctop app, it works differently. And you sit there and you think, "What time can I do?" Well, how many can you be glowing? That's why they're like the Moon. Here. Ah, Tanya, what's going on with your chat?
C, uh, viewers, too-
Illar, I understand correctly that you're not-- and you don't have to move all the chat rooms, but the key, basically, that you're up there, right? What was there two years ago, maybe, and, uh, and don't have to break the past, right?
It's a month. Like, if you haven't used chat for a month, you haven't been there, he's been out there quite a bit. You need him? You're looking at him? Or is that how you could get rid of him?
You know, uh, well, here, I think it's kind of a human thing to do. I need to, I can sometimes go up on some subject, and then I can't even forget about it for a long time, then I'm back in my mouth, this chat is already, this or my project, I don't know how many projects I have. Really, uh, maybe tens. And that helps me. And I remember where, uh, searching and so on. Uh, I mean, I-I'd really, uh, I'm just saying, uh, it's important that, uh, there's a story, there are different questions, uh, and that-- but again, like Sasha says, sometimes I'm asking him something, uh, To restore, yes, what we've already studied together. Uh, and it's starting. He's got some totally unexpected things that I've forgotten about, and some super-price thing is, uh, forget or mix. There may be a age, like the same child, confused or starting to give me advice. I'm asking about my daughter, who's been there for a year and two months, and he's starting to give me advice for my pregnancy. Well, well, how would some of your inconsistencies go? So, yeah, I guess it doesn't make sense to move all the chats so far.
It's gonna be a total pot, but in time, I hope that this system will be much better and more built up.
Yeah, look, soon, uh, digital nomad will play new paints. These are the athletes that came out there. And, in fact, life is placed in two suitcases and a laptop. And you're out there, moving from one country to another and so on. And there's a digital nomadem to call people who, there, are switching from the system to the system and don't take any chat with them, yes, whatever.
Yeah, I'm starting a new life.
I came here and I left.
Yeah. I don't want to remember anything about the past. That's why it's over. No... no... chat integration, no new identity. I'm getting ready for a new job.
A year later, in OpenAI, we need to grow a new identity.
Yeah.
Look, it's got to be strong, yeah, so it's gonna be like, you know, making it all go away, right? And not all people are really ready. I was just thinking about removing ninety percent of the contacts I have on my phone because I'm sick of seeing this endless number of birthdays in Telegram, yeah, that you can't turn off. It's just a completely complete function. Or on top of people, they put their stories, and I, um, a-a-a-- many-- that could cause, uh, weirdness, but I don't know how many contacts I have, there are, like, ten, tens of thousands or thousands, I don't know. Well, a lot. And you know, I was sitting here at the hotel last night, uh, waiting for something. I opened up, well, I got this idea, I opened and started, so I'm gonna delete it. I see there are numbers that are recorded, uh-huh, when they were still dialing without a code, just like a six-digit number in Belarus, right? So, with the fact that I have , the phone came up, um, two thousand second year, the cell phone, that was a long time ago. Yes, two thousand two years, that's a long time ago.
So how many contacts have these people need to know-- these people, they change their phones, these people, they change their phones. I'd like to add one story to the projects. The project structure that is now in place is a priori unworking. I mean, she's not gonna work in the model. It's some kind of hole that was made, perhaps for corporate history, to sell B2B. Why is she unworking? Because if you-- well, it's logically made by its own. Here, Tanya, you said you had a very, very many chat, and you're building it there, to make it for the garden. I'm, like, moving, there, I've been organizing a trip, I've had a project, and I'm adding my questions to this project. And there are chat rooms, I'm talking about them. But when I moved into the country, even on this trip, the country becomes the main activity here very much. I have so many chat rooms I'm not asking them in the projects, I'm starting to ask them out of the project. In fact, the system should be the one to set all the lines. It's not a project job, she, she has to spread the shadows, she has to combine them in terms of logic. Logic-- they-- she has to be in the sense of why unite? Because I'm sometimes in Tuzumun, I have a project like Tuzumuna, right? I'm in Tuzumun's project asking, I'm first of all, every time I see something interesting in a week or something, I have to wait all the time for him to move into the project. There's not much of a thing like that: when you're saying "to get in the project," you need to wait because if you turn it off, it might not add. There's a vibration of the phone, right? And you're kind of adding and waiting four seconds. But that's a total thing! First, it's a code design that's totally deficient, right? And second, it's a total idiot. I mean, you're sitting, I'm waiting because I know he can't move. And, uh, here's the question, the moment is that there are things I'm asking about AI that are not just Tuzumun, but AI, which are, for example, my business, and-
It is, for example, my business and one or the other, or AI is a family or AI, which is a garden. So what do I do? I need to-- what do I have to put the project in? This structure, therefore, should not be so operational in terms of modern data accounting models, modern systems. You have to connect the system yourself. And it's a difficult task, yes, well, totally uncomplicated. This is a recollection of Max's graduation, which tells us what the company culture is, what they do, and we realize we can fight or talk long. But the OpenAI company is doing AGI. They, by the way, when they turned Sora off, they wrote at their place, they wrote, uh, a description of what they are now focusing on in terms of development. Now, I'm here. Oh, that's what they're doing, what they're laughing at. They wrote: in coding, in, uh, in the code, on AI, the software tools and the code writing are shifting the trick. I'm really interested, so I'm gonna say, and in coding, they had a trick shifted to something else? Enterprise is business, products for companies that know more money, contracts and business benefits. I wonder if you had something else to do with enterprise. And in AGI, super app, the big goal is to build one strong universal AI product that can do many different tasks in one place rather than keep many separate road areas. So, look, when they were writing about the video, they thought the video was too expensive and not the top priority. Where's the usual people? That's what it says. So they have coding that's carrying money because coding is now a huge use, and Anthropic has become a point of view for them. And you can see that. They're the first line to write coding. But obviously, Anthropic became a testament to them. This is the code red, which was so on. So, a company with the largest number of users in artificial intelligence in the world, announces the code red about the company and says it again. Sam Altman said, "You have users, like in Texas," yeah, there. She's announcing, so that's a story. While Anthropic was really targeted, it was always very strong and treated people who were, uh, with the fto that were developing. And then they're re-educated, saying, "For us coding is important and beyond enterprise." And enterprise is a story. Now, in the Silicon Valley, I'm in contact with different start-ups or businessmen who talk about their business, and all the investors are always saying that it's important that enterprise bring in business. The profits, so they brought the money. Although I'm actually looking at who gets the money and how I get the money. And you see, everything is very different, especially in AI. By fact. But there's a device like that, so... And draw attention to the fact that enterprise is not to help people, but to enterprise, yes, it is more understandable that money is where money and benefits are available to companies. And that's getting a problem, right? We've been talking about this in the past graduation. And what, and what, did they do to companies in the world, have been a real benefit, if they, uh, have directly benefited from the current money.
Here. But our channel is aimed at helping people.
Look, you know, me, Sasha, what's the idea? Look, my ChatGPT and there, let's say, your or Illnar is like three completely different identities, right? If there's this chat migration. And, like, you know, you have a co-sponsoring session, right? For a pretty good time, it's a very serious money. If I could, well, that's a thought on the edge of madness, but still. If a man had asked you to, uh, copy your chat, re-examine it, and he'd be bringing it back, and then he's getting a new one, uh, there, I don't know what it is, the same ChatGPT or anyone else would give him answers. based on your logic, the right thing you've learned. There's a thing in this, too. Well, it's probably a certain time--
It would be fun if he gave it to a man, but he would have, uh, shown, and how would I ask a question or something?
Yeah, as much as you ask, what would you--
I'd think I'd ask what I'm asking and what I'm interested in. That's a very interesting idea. I think it's an idea, uh, a start-up.
It's real, by the way, I think you have a lot of followers, they'd... No, I'm not a start-uper. I don't have enough nerves for that.
That's why I'm telling the audience. That's why I'm telling the audience.
Well, I think the demand would be.
I don't see much, look, I don't see much, I don't see much, very rarely, the idea where I can call it a start-up, because I usually say, "I'm laughing," it's, uh, throw it out-
Oh, that's the best compliment. Well, tonight was a good night.
I have a lot of businessmen who know each other, who have a lot of money. They make the valley start-ups. I think they're not gonna make it, yes, never. So, here. And this...
So our audience can take it. I'm not gonna hurt you.
It's a very-- it's an interesting idea from the point of view... But again, look. Oh, look, look what's going to happen now, like, like, what kind of audience, what kind of a problem is now. Here we are, I have IT, you know, IT business, and yesterday, with my partner, there, with mine, that when I'm in the process of establishing any AI systems, people through the API for companies, corporations, for, for a normal person, They use cheap models. That's important. They use cheap models. That's incredible. So they don't use a model that thinks 17 minutes, which restaurant is better. What was the question?
There Lesha chose, "Let me have good restaurants with a view in Cairo." Here!
Tripadvisor is so first.
He tells me the name, I'm hitting him. He tells me the name, I hit him, and the restaurant doesn't exist. Well, some Google request took some sort of article, and I wrote, "Please, please, Arab restaurants in Cairo. Make me a comparison table for four, five, six. Tell me how far they're from this hotel. And so we can figure it out from zero to 100. The restaurant of the Lebanese kitchen at the Four Seasons Seventy. And he started doing it. He's been doing me a long time. What did he do? He studied four hundred and forty-eight sources. I remember now, four hundred and forty-eight sources. He left-- I think it wasn't 17 minutes, it must have been more demanding. He's gone... Well, a lot of stuff, a lot of stuff. But when he gave us the result, when he gave us, we used this list further, and he's been on the list for days. But when we got the result, it was very interesting because before that, thinking heavy, that is, the maximum reasoning model, gave me some other restaurants. Because once I've identified seventy, look, she had to look for more, she wanted to find the system more, and all ten were higher than those seventy. And before that, this Four Seasons restaurant was number one. Look, before that, she had one. That's the difference. And here, when people, uh, mobile applications or web, or, in their own corporations, use API, they, they don't make a request for that quality and don't load it into Pro theory because in Pro The version of the request from-- just a flat price, there's a token size, right? Then back out, the size of the tokens, I guess time, I don't know. Well, I-- well, there's probably time to go back, uh, in Pro. There must be time because, uh, Pro's theory often doesn't answer that big answer, but maybe it's like eighty minutes, right? It depends on what you gave her in the front and told her, uh, what you'd do. And there's not even a question for the Tokens. Because if you told her to study and analyse, there's at least 15 thousand sources, what's going to happen to her?
Illnar, there's time at Pr-- you don't know, Pro version of API, uh, time counts or not?
I mean, you know, because you're talking about a model, which means she's a token genetic, and the token velocity is limited. I mean, anyway, she's been thinking something long, and she's got to get a lot of tukens. Or to think a lot.
Yeah, yeah. But these intermediate tokens, they count?
Yeah, sure.
So, the intermediate reasoning, is it counting?
They can be considered different rates, there, or something, clashed, not clashed. I'm not gonna tell you the details.
But it's clearly taken into account.
But it's taken into account, yes.
Yeah, it's clearly taken into account.
Yeah, yeah, yeah.
Because otherwise, you can make a short request, "Let me analyze eight hundred thousand sources and get a type-- and give me one answer, right? Jacobs. It's so readable, it's not gonna work.
Yeah, sure.
It won't work. Here. So-- yeah, that's a pity. So, uh, this is, uh, this is a super important subject. And, Tanya, your request that you say, his negotiability-- that's the whole world, as I've seen these starttaps that try to create, copy, er, Buddhism, so create an e-I, copy, You, uh, set up a sales manager, so you can get a therapist. They all use cheap models. And what you're saying now is a serious model, serious, serious stuff, a serious study. What the model gave me isn't a prank. She's got something quickly analyzed and handed out. She's the one who's been in contact with me, accumulating some data, analysing, analysing. Last night, by the way, who's-- who's technically skilled, you can analyze, read what you've been looking at in Cloud Code, right? There's a story in there that
In the code, right? There's a story where there's a memory she's referring to right there, and there's one she's in there, I don't know, that's the right word, you know, like, uh, asynchronous, right, yeah, that's what it's like? Or from-- or al--- well, uh, when it's parallel, parallel to you in the process, there's still a building that's still under study, and it's a parallel study. I mean, one-- I asked, for example, she had some memory of me, she did, and then, like we were talking about flowers, she starts looking at my flowers, which, for example, We were parallel. That's totally different. And lifting old chat and looking for flowers and doing everything, and breaking this set of information is what Google is, including trying to do to analyze your core so that it can analyze your accumulated core. Your whole story is all, uh, to be able to get these fragments up and down at some point in time, maybe fast on this fragment when you've already got the data, like, He's been tagging around dozens of thousands of words or topics, and he knows, oh, there's 250 chat rooms that have high weight to the garden. We're gonna put them in the context of this request. And that's why, by the way, Tanya, it's just that it might work, or it's too big a system. Not enough, I have to be created. I mean, he was talking about me yesterday, if I had told him, I told my friend yesterday that I had, well, had a stockpiling, hundreds, maybe, like, a thousand different lectures, Well, hundreds, a lot of different articles and materials. It's tens of thousands of pages. And there I tell you, uh, I'm telling you, I'm giving some content because I have a YouTube-canal, and there's a lot of stuff, yeah, groups are driving, and, uh, but old, like, data that were ten years ago, No. I mean, if he could also put this data into context and that I'm telling you to create some kind of design and in it, Tanya, how to think about it, but he can't now, he's got it. context window is limited. When I'm on the migration side, Google gives up to five hygabytes to download. And ChatGPT, for example, now gives hygigabytes, I think if I'm not wrong or there before the hygigabaytha is down-- I don't know how much, by the way, I'm just, uh, just, there was a couple weeks ago, I remember, Outlook. I couldn't download my archive, yes, in Outlook, old, young. Google's giving me five gigabytes in the event of migration. I have a question, by the way: how will the prompt give it to you? Why five gigabytes? Maybe this is the subject when I can be right about myself. And you know what I'm talking about, that's what these files are about me, right? And he'll start studying and downloading it into a separate layer. By the way, this Google migration chat is not just, uh, it's not just a... it's just one chat, it's a separate special area that counts as a memory of you. So, Tanya, the idea is awesome. Ah, but even now, if people start doing it or do it, they need to spend a lot of money, uh, to give people answers. That's how my friends say, "Sasha, you're standing for OpenAI very expensive. Your requests, the endless analysis of hundreds of thousands of calls, so checks and everything, they're very, they're very expensive." So if you check, um, check out some queries, they're worth... Uh, Ilnar, by the way, if you ever have, uh, you're sitting a lot, you're gonna notice how much you're gonna need, uh, that's what you're looking at... Doesn't mean you need to pay that money. Maybe he's a little expensive, there's $10, 20 dollars, or 30, you'll check. It's interesting to see just what price policy is for a difficult request in Pro.
Because everyone's talking about prices, there, about tokens, about some microchid. I was once telling you how much time it was worth... the voice of this, uh, voice-making thing is one thing. How much are these complex systems worth? Sam Althman, guys, two weeks ago, at some speech, he said, "We're going to pay for power." He said about this year ago that you'd have to pay $20,000 to ask. You know what he's calling now? So a year ago, he said there would be systems when you're $20,000, for example, paying, and you're doing some business business business. He said, "They'll be systems when you pay billions of dollars to solve one question. Millions of dollars." And that, and that's very, uh, very important to take his subject. It's very important. There are questions. Remember what happened? Question: War to start or not to start, huh? So, uh, Grok says, "Yeah," "Is it Anthropic or who says, "No," there, ChatGPT or Gemini says, "Well, let's think we're gonna doubt it." ChatGPT said, "No." These are episodes of this type, yes, when you're studying, studying a lot of different possibilities, and thousands of pages of reports are issued to different units, there, companies and the rest. So you're saying I'm buying this company or not buying it there, for... Or a boo-- if you buy, what time? It's $20 million or $100 million. And you say, in answering this question, we'll spend $100,000. And he's on it, he's on his way. So, uh, bra-- Sam Althman keeps talking about it, right? So we're guaranteed to go to the ma-- to another system of analysis. And, Tanya, this is an interesting thing. I mean, if you say I'd like to get this co-teaching you just told me, and maybe you should have a ChatGPT, ChatGPT, like, not a little, like, a little bit of a pimp. I know, $30,000. He's gonna take my data, everything will make it all up and pack it to you in some system.
Maybe it'll make a separate start-up.
I see it more like this, personal artificial intelligence or personal ChatGPT, or something. Because the context window, at least now and in the near future, especially in an effective context window, does not take all the information that would be needed to get a similar identity. It's more like, you know, uh, that's how it's... Well, for example, you can build a genetic genealist tree. You pay special people, they go to the archives, they find information, all records and stuff in these archives. And there, they collect your genetic tree. It's worth a lot of money. It's likely that there will be a similar payment for ChatGPT, a start-up or someone else. And that's a Drevo, uh, information about you, digitizing everything that you're saying, there's five gigabyte Outlook, uh, tens of thousands of phone numbers, lectures, show where you went from, Who you learned that these people do that. Well, that's some information that's going on. It's not in the context window after that. I mean, it's a conditional, cheap way to do it, there, RAS-Found, yeah, so that there, at every request, the system could find some kind of piece of information, take it to its own context from this huge, huge, volume and, accordingly, answer. But that's not it. Rather, it's gonna be a fainting, uh, model, big model, and it's very expensive, under, uh, the information you've been gathered about. And then we'll be there, ChatGPT with the shades, there, Sasha Volchek, with the shades, there, Tanie, there, with the shades, there's a claw-some man. Uh, on current money, it's worth, uh, crazy money, because you need to--
Yeah.
...to intune the model, to make sure you didn't break anything else and so on. Well, that's right, yeah. I'm afraid to call some numbers now, but it's closer, there's hundreds of thousands of dollars. But there's a model that will produce a result to a particular person or speak like that person. What's this got? Is that the ChatGPT is, what's it like? Five four. You've taught this personal intelligence under a specific model. Here comes GPT, there's five, six or something. And you've got to-- you have to pay all that money back for the pre-learning and stuff. I mean, it's gonna be, but it won't be mass, it's gonna cost, well, very substantial money. What about this person, what about that chat? Maybe Tanya, your idea is, you can rent out, chase or sell, or whatever. I mean, like,
Dana, rent Tanya Pyrozkov.
Here. It's never gonna work, it's not rented. Thank you very much, everybody. I'll see you in a week.