Well, internet search in its familiar form will be no more. Not because Google or Yandex will disappear tomorrow, but because the habit itself will disappear — opening the search bar, typing in a question, getting a list of links, reading it all yourself, comparing, keeping everything in your head and hoping the conclusion is right. The next scenario is, of course, much deeper. A person often comes not with a precise task but with chaos. My sales aren't going, I don't understand where to move next.
I don't understand how to build relationships. I need to make a decision. I'm tired of options — what's even going on here. And strong artificial intelligence won't just answer, it will help formulate the real task. What the cause is, what the constraints are, what result is needed, what options exist, where the risk lies, what to check, what to come back to in a week or a month. That, of course, is no longer search and not just a recommendation — it's a new layer of life, of work. A system that helps a person think, hold context and move toward a decision.
For an ordinary person it's a new skill — not asking "find this", but learning to break down your own situation together with artificial intelligence. For business, of course, it's even more important. The customer of the future, it turns out, doesn't just come with a request to buy. They come with a task that artificial intelligence helps them understand. If a business isn't built into that understanding, into that understanding of the task, it may end up invisible. So this episode isn't about the search bar, not about search engines.
This episode is about how the very way of living, working and making decisions on the internet every day is changing.
Hi everyone! We're on the ToTheMoon channel. Tech news, insights from Silicon Valley and from all over the world. And today, of course, we have a really cool special episode. I think it's the biggest, most mainstream topic about true artificial intelligence. What is this going to become? When I first came up with the idea of recording this topic, the system that popped up for me was Perplexity. I don't like talking about Perplexity, I try to focus on the market leaders, and I always go by OpenAI, ChatGPT, Anthropic Claude, xAI Grok and Google Gemini.
That's how my system is set up. Someone will say there's also DeepSeek. I remember there was a time when it was DeepSeek everywhere. And then they'll bring up other Chinese models too. Then they'll list some European models. Like Mistral, the French one. But I say: "Listen, let's concentrate on the real ones, on the really serious systems and on the biggest systems." And people periodically like to bring up Perplexity. Perplexity. Their user volume is strikingly smaller — tens, if not hundreds of times smaller than the big systems'.
Let me remind you that Google Gemini crossed the one billion user mark in its AI search. And that's a very important wake-up call, right. Once again, Google Gemini. They have AI Mode, which they built. It now has more than one billion weekly users. ChatGPT at the moment has roughly a billion weekly users, but all the other systems, obviously, lag behind in numbers — both xAI and Gemini. But these are companies with very serious infrastructure investments, really very serious ones. A story comes to mind — I'll do a separate episode about infrastructure soon.
It's a very important topic — understanding where artificial intelligence is heading and, in particular, the investments of the whole market in general, how it's going to develop. I was watching — the President of the Russian Federation gave a speech recently. He said that Russia is definitely in the artificial intelligence race and has every opportunity to be among the leaders. And he framed the opportunity to lead as: there's talent, there's science. And second, he said, there's electricity, right.
What I found interesting was the realization that he sees artificial intelligence infrastructure only through electricity. I'd like to show everyone and explain that owning gas, oil, electricity — and developing nuclear power — doesn't give you real leadership positions. I mean, it gives you opportunities and gives you a certain foundation, without which many things really are hard to do, in fact — but it doesn't give you the other opportunities. We'll definitely break that down.
So, coming back to Perplexity — whenever people talked to me about Perplexity, their model always absolutely bothered me, and I didn't see a business in it. Even though the company is cool, the owners are great, I even made money on it — on this project. But with all that said, the story is this: it's a company that basically developed a third-party search. I mean, it's a search that uses models, various other models, including OpenAI, Anthropic — it uses different models and gives you information in a certain form.
And then it went on to make a browser, make agents, and that's it. But it doesn't own the models themselves. I mean, the models don't belong to it. It's a kind of external system, some sort of external system. And will it ever be able to compete with Google? On the strength of what? Yes, and that's why we don't see even hundreds of millions of users there. We don't see them and we're unlikely to see them there. Unlikely to see them at all. Most likely, at some point there'll be a decline.
Everything in its time. And it's unlikely these big companies, these big giants, will buy them. They essentially don't need this system, right, because this system is based on other companies. But again, we'll see, right. And they'll definitely sell for some amount of money. There's a cool team there, interesting people and so on. So, the topic — and that's why this topic was very closely tied to that. Because when you start talking about search, people today see artificial intelligence as a system of agents
that, among other things, use Google search or use the search engines of the countries they live in. For example, in particular, the Russian-speaking world, of course, heavily uses Yandex. And that's why I've mentioned Yandex here, mentioned it a lot. If we take a look, Google initially introduced this AI Mode tab as a system where there's a person, they've come to search, come to make a search query. A person comes to Yandex, they've come to search, and supposedly they've been hooked up with artificial intelligence that does some kind of aggregation.
But then, literally a week ago, Google shows up and says: "We have new SEO rules." Two years ago already, when artificial intelligence appeared, I — since I'm involved in many businesses, I'm an advisor in many businesses, an owner in many businesses — I told everyone that now is the time to adjust search queries and adjust what's called search engine optimization. It's going to work differently. And you have the opportunity right now... How will it work differently? Right now there's an opportunity to seed information on the internet about your companies, about yourselves — the information that artificial intelligence will eat up and read.
And artificial intelligence will read that information as, well, important, or correct, or necessary. And a bit later people started listening to this, a bit later they even coined special terms for it — that you need to optimize for artificial intelligence machines. But it turns out they're doing it now — what needed to be done two years ago, not now, right. Thinking that people will still be searching in search engines. After all, if you ask your friends about ChatGPT and, well, about systems like that and similar ones, you'll hear from those people that most of all people use these systems precisely as search, as a way to get some answer — ask a question, get some answer. That is, a kind of information search. What's the weather going to be?
Here I live in California, in Los Altos Hills. Los Altos Hills, so, what's the weather going to be in a week, or tomorrow? So there are these standard types of questions. But if you look deeper, in terms of ChatGPT, in terms of Gemini, Claude — these systems moved past that a long time ago. These systems today let you not ask a question in the search sense, but immediately solve the problem you're facing — the whole problem you're facing, centrally. I mean, you can find out the weather because you're going to some city, or you can, right away, while choosing a trip to some city or organizing that trip, build your requests into the system in such a way that it takes into account what's going to happen with the weather, what your expectations and wishes are, and all the details, all the circumstances.
That is, it folds that into the overall structure of solving the task. And so we run into the main... This applies to agents too. I film a lot about this, but I want to put the emphasis on search once again, because it's an important story of impact for everyone, right.
How in the future all systems will interact with a person, how a person will interact with these systems, and what's actually going to happen. Because, in my understanding, my feeling, the era of search in its current form, the era of advertising in its current form, in terms of contextual advertising, is, of course, going to disappear incredibly fast. Just like a huge number of websites in general, portals that people use today. For example, worldwide, consumption of many sites that deal with providing ordinary information has dropped by tens of times.
For example, in the education world there's a project that completely lost its value after... It was trading at more than ten billion at one point, or around ten billion. It completely vanished from the world, because that project helped kids — well, and all kinds of people, not just kids — answer the questions they'd been set. And I know probably dozens or hundreds of startups where people just want to present information in some prettier way. Do they really think they can handle presenting information compared to ChatGPT, Anthropic and so on?
Again, I'd be interested in your opinion — what you think about this, and which of you are doing related projects like that, and what results you're getting from them. I recently filmed an episode where I talked about my personal opinion that a huge number of apps will disappear, and operating systems will disappear, and serious big solutions will disappear. That, for example, even the coolest program in the world ever created, like Excel — the way that program is used will be transformed.
For example, I hardly ever open Excel anymore to study information or to create information. I mean, I study Excel with ChatGPT, or Claude, and I create Excel with ChatGPT and Claude, right. It's astonishing, but I practically never open files myself in full. By the way, just the other day I filmed an episode about security in the context of artificial intelligence and understanding what data is stored within the model and across all these systems. I strongly recommend you watch that episode.
It will broaden your picture of data safety and of making decisions about interacting with these systems, and, of course, of the opportunities that artificial intelligence opens up for you. And it's important to understand that Google, among all the search engines in the world, has of course flown way ahead. I won't compare it with the Chinese models right now, because China has a somewhat different infrastructure.
There, even the infrastructure in terms of a search engine like Google — it's not that it's absent, essentially, but it works differently, because a lot of people use super apps, right, or ones like WeChat, or such huge complex portal systems as Alibaba, right. I mean, essentially that's also a super app. But the Google engine, of course, is the number one search in the world. Google, of course, isn't trying to completely remove its standard search, because they make insane money on contextual advertising.
As far as I understand, that's their main income in terms of advertising. Someone can correct me, write it in the comments, but either way they'll have to do it, because terminals and ordinary command lines like ChatGPT are stepping on their toes. And I, for example, have cut down my Google use not by ten times anymore, probably by a hundred times, right.
And most likely, at some point I'll give it up altogether, as soon as ChatGPT gets interested in making some interesting, nice interface for me and handling my own requests. It's just that they're interested in creating the true artificial intelligence of the future. Same as Anthropic. Google, though, plays in different positions, just like xAI, right. xAI is interested in creating true, serious artificial intelligence in tandem with Starlink, with the satellites, in tandem with Tesla's robots, with their own cars and so on.
I mean, it's a whole huge infrastructure. They're building it, and they're very different. There's OpenAI, Anthropic — those two you can put roughly side by side, right. And even so — I did a separate episode about Anthropic. Anthropic, of course, made a different bet, an incredibly cool one, which is why their valuation is almost — well, of course, less than OpenAI's, but almost the same as theirs, about a trillion dollars. SpaceX has its own story together with xAI, definitely. And Google, they have their own life — an absolutely unique life of their own in the world.
Most likely, it'll be a different company by then, at some point, right? So, the CEO of Google said recently — he was asked a question, they said, look what a poor-quality answer your artificial intelligence gave in search. To which he said: "Well, indeed, the answer could have been better." The problem with search, and with artificial intelligence in search, is that they can give fairly high-quality answers. Although, of course, search has turned into a dumpster. Both Yandex and Google are dumpsters. They can give fairly high-quality answers, but not to what you specifically need, you personally.
That's a very important point. And we see this problem everywhere. When one person needs to quit their job and another shouldn't, one person needs to buy a car and another absolutely must not. I mean, there are a lot of different circumstances that exist in a person's life. And right now, of course, a person goes to search as if for a little tidbit.
Even if you take the notion of a recommendation — recommendation is too small a word, right? I mean, artificial intelligence doesn't just suggest some option, it helps build criteria, a plan, monitoring, come back to the task, work through it. And in education, of course, people love to give examples. That is, it doesn't just recommend a course, it can actually build you a full-fledged learning system today. My daughter goes to school here in Silicon Valley, and, naturally, she's oriented toward college admissions — that's how schools here are set up, they're geared toward admissions, there are good universities here.
Like, my kids are little, and there's Stanford for them to study at, or Berkeley, which is nearby. So there's a certain bar of good universities, and, for example, two years before applying she already has the option to take the SAT, if she's in good shape. And what does it mean to figure out how to pass such a serious exam? It's a very serious task. What does that mean in Google? I have to go in and read through who knows how many articles about it. Now I can upload all the information about her into ChatGPT, write to it that her situation is roughly such and such.
She'd like to try — is it even possible to do this, take this test early, and why do it. Or can it be taken in a year, for example, how often can you take it, where to take it, what the right strategy for taking it might be. And I remember we looked into this question in January or February, and it built a very clear system: what she was supposed to do in the spring, what she should do in the summer, that she takes the first exams, prepares in August, that in the summer she goes to prep courses for a few weeks, that every two weeks she should take a test and get an exact score.
In between those results, prepare in the free Khan Academy system. And when I told her that artificial intelligence — like, Gemini released an SAT prep version, or there are additional courses — she said: "No need for that, there's no point in it." It explained the scoring system, explained the details, and she went in, took the test — I don't remember, in February she had some score, like 1250 or whatever in total. It says: "With this kind of system, overall it's possible to pass. But what do you need?" I say: "She needs to get a score good enough to get into MIT or get into Stanford.
For that you need to have, I don't know, 1700 or 1650, I don't remember what the total score is. To have that, yes, overall there's time to prepare, but you'll have to push hard. Right now you're in the top twenty of people taking these tests in the US. And to score 1650, you'd be in the top one." So this whole system laid it all out completely by itself. That's absolutely not search. I mean, it has nothing to do with search. It was like having a professional, an unbelievably professional consultant.
And it wasn't a professional consultant with some personal opinion of their own — it was like a whole team, a huge team, a big team of different people, whom I can ask with varying probability. What's more, I can always open a new session, right, in ChatGPT or in Anthropic, open a new session and double-check that information, learn it afresh, or ask how people usually act, how people who study at this or that school act, and so on. So, a big difference. For example, if you go to school in Silicon Valley — specifically, she goes to a school that's in the top fifty schools in the US out of twenty thousand, right.
If you go to this school, you have your own process. If you go to a different school, you have your own process. If you go to a school outside the US, you have yet another process of your own. And then — why do you even need to take this test at all? Are you even planning to apply anywhere? Or, for example, there was another story that also involved her. I'm just giving you examples that will let us sort out the whole concept of search today. When we were figuring out what she should study, which directions and tracks to take.
And I remember the system asks me — this was in January: "Tell me, which field does she want more?" I say: "Listen, she wants to be a mathematician at the same time. At the same time she wants to become someone who works in fashion, as a designer, or sews something, or manufactures, and to do psychology. At the same time she wants to do music." It says: "Well, look, when the situation is undefined, then you need to take such-and-such a track." Yes, there was a question of whether she had to take physics at school, advanced physics.
It said: "Then you definitely need to take advanced physics." That was very interesting from a knowledge standpoint. And it explained why that choice was made. Obviously, I asked follow-up questions about it. I mean, there's no way to compare this with any articles, with any outside people. And essentially, of course, if we're being honest, people who use artificial intelligence this seriously, especially those who already know how to build their own AI native — as it's called — systems, or software created with the help of artificial intelligence.
These aren't agents, not some agents — these are independent, different things. And using different services independently — then, of course, a person is freed from search, they don't need search. I mean, what do I need Google for? Right now I use Google to look up, by a certain keyword, what the seats look like on a plane. On a plane, a Boeing, flying from Mexico City to Paris. I needed to look that up, for example, the day before yesterday. I opened it in Google, although I probably could have already told ChatGPT: "Show me photos of these planes." But I'm just used to making that kind of query in Google and opening the images.
And because there are lots of images, I figure it out quickly. Although, essentially, the interface in ChatGPT really already handles it for you. What else do I need it for? I practically don't need it anymore. To check where my sites stand in the search results, SEO-wise. But once again, Google put out a big overview. I won't devote a huge amount of time to that today, telling you about it, but they've put out a really big overview in terms of SEO — it's going to change. Very important. It doesn't mean that the SEO requirements that existed before will disappear.
But the point is, you have to understand that it won't be SEO, right — it'll be a situation in which search engines won't exist at all.
For example, there's an interesting statement from Google's vice president of search, right, that is, the search person — Elizabeth Reid. She says — literally just recently, a couple of days ago, she said that Google calls this a new era of AI search and the biggest update to the search bar in twenty-five-plus years. So, an important detail: the new bar helps formulate questions, keeps context as the dialogue continues. But, you understand, these are phrases and statements about absolutely nothing, right?
Because it's something neutral and something very political. I mean, you see, systems like Google and like Yandex — they really couldn't change the search system, the search bar, in twenty years; they were afraid to modernize it, they made some micro, micro-targeted changes. But a person no longer needs context saved as the dialogue continues. Today there's Claude at Anthropic, or ChatGPT, which stores thousands of your sites, your chats and a huge number of files about you.
They know information about you down to the last detail. They know your behavior patterns, the style of answers for you. Why would I need dialogue continuation, right? But what else are they supposed to say? Or here's Shivani Mohan, the vice president in charge of Data Science at Google, who wrote that AI Mode has exceeded one billion users. By the way, here they wrote that it's monthly users, and the average query that users ask in this AI Mode — that's the artificial intelligence mode inside Google search, right — in the US is three times longer than a regular one, right.
And that planning queries grew eighty percent faster than the overall volume of AI Mode. But it's obvious that today artificial intelligence is adopted in the world only very relatively. And even when we talk about ChatGPT's billion users — which is still more stable than Google's billion users, because those are search users — still, most of those people are free ChatGPT users. These are users who, well, ask some micro-queries and use ChatGPT as a kind of formality. And even if it seems like they're using something complex, they're using it as a formality. Among the people around me — my circle is very diverse.
There are businesspeople, politicians, doctors, economists. There are people who are seventy years old, thirty, forty, anyone you like, right. Marketers, whoever, right. I can say that usually one in ten uses artificial intelligence systems more or less decently. All the rest are grandpas when it comes to search, right. But that's how it is.
And, for example, in parallel, Demis Hassabis — a very well-known person, the head of DeepMind at Google, and a person who's genuinely developing true artificial intelligence. And he describes Gemini, in this case, as a future world model that understands — the way it does text — how to build plans and act through a device. That's not search, it's an attempt to build a universal assistant that understands the world and a person's situation. So that's, of course, something completely different.
And, by the way, an interesting point: a huge number of people already — Sam Altman, for one, has spoken a lot about reasoning models, that is, models that can reason. What are we seeing now? That even OpenAI is removing the toggles for model selection, where it says the model reasons. Because today all decent modern models reason incredibly well. Just yesterday — well, as of the filming of this episode — Claude, that's Anthropic, released an update, Claude Code Opus 4.8. They have 4.8 Max.
And within these updates a feature appeared that lets you split things up and work on code, or on your whole system, in parallel threads — in hundreds of parallel threads at once, developing something for you simultaneously. Can you call that an agent? Not at all. Can you call it ordinary search that reasons? Of course not. This is the creation of something on a global scale. It's like you, as a person, over the course of your day, are simultaneously in the middle of planning events.
Like right now I'm filming a ToTheMoon video, and before that I was doing something health-related, before that I was pumping up the tires, before that I was putting the top up on the convertible. I need to go, because I need to help a buddy get through his driving test. And before that I ordered some groceries, and before that I drove into town, took a walk, and before that I held a meeting with my editor, talked with my wife, with the kids, with the dog, fed the dog in the morning, woke up, and so on.
I mean, half a day has gone by for me, or even less, right, and there's a huge number of events that have happened. Programmed some of those systems, reacted in terms of investments, on the investment market, talked with my partners. I mean, a huge number of different actions. Bought something, ordered something, a lot of different actions happened. And that's exactly what a large number of simultaneous threads is. Can any kind of search even be compared to that? No. I mean, to google — the concept of "googling" is essentially disappearing. And we'll get some new concept.
We don't have it yet, right. And these systems won't exist. And as I said at the very beginning, whoever thinks that electricity will save artificial intelligence, will help develop it — it won't. And whoever thinks that since they have search engines, those will be able to develop something — they won't develop it and won't make it. Obviously, their own systems will be figured out and built locally. And, for example, Yandex says, in any case, that, for example, Alice AI in search lets you solve complex tasks,
hold dialogues, work with documents, images, form structured answers with sources, and so on. But the thing is, even that — that's already the past. It's like parsing PDFs, or your documents, or analyzing ZIP archives. That's the past. Or the number of documents the system can process. Even the size of the context window is the past, right. Because you can — well, at some point you'll be able to, in any case — upload as many files as you want. The question is whether you have access to such systems.
It's like, if I, for example, am limited to a one-million-token context window in Claude, or in Codex, then obviously the programmers at Codex and Claude aren't limited to a million-token context window, right. Well, it'd be strange to think they were. Note, by the way, I might film a separate episode about this, in terms of strategy. I said from the very beginning that every company has its own different strategy. The most important thing in the world is understanding what a company's strategy is, right, what its strategy is in terms of what it does.
If you take xAI, you'll see that they're inevitably about to enter the world of software development. This complex of theirs, where they're doing Macrohard, and where they just had this thing where they said they're giving people the ability to create agents and are starting to program within their own system. You'll see them burst into the software development market. And why software development? Because imagine you have some kind of watch or some kind of pendant, or, well, there's the phone a person carries around, or there's some robot, or some wall-mounted device.
The main thing will be: how do you start communicating with that device? Why would I even go into some search, or go into some other system? The question is whether the device will support that communication. That's a separate matter, whether it will support it. But search as such, of course, doesn't exist there. Moving on — we'll keep going in a moment — be sure to write what you think about this. A vote would be really interesting. Who do you think will win, who will pull it off, which systems might manage it, which systems won't be able to do it.
Or does someone think that search will continue to exist and won't disappear at all, in this sense. Once again, I urge everyone to understand — obviously, Google — I'm in no way saying that Google will disappear, it definitely won't, right, it's a serious company. I'm saying that Google search as such will go away. And it's an interesting topic for all of you to look at, from the standpoint of the role you currently hold, especially if you're connected to this. If you work in marketing, or you're in sales, you manage projects, or you're an entrepreneur, you have your own business, or you're building some IT product, or you're investing.
Even if you're just an ordinary person. I mean, to understand that artificial intelligence isn't about asking for a fact, it's a way to make sense of life, work, studies, money, your choices, right. To approach these systems with a different mindset, to understand how they work, so you don't end up at the tail end with all this later, but instead adapt this world for yourself, adapt this world into your own life. And here on the channel, of course, there'll be people who want to do exactly that. And of course, we'll test all this in the near future together with you and see what happens.
Share your thoughts, your comments, and see you in new videos on our channel.