The social platform X has already blocked forty-two thousand accounts that were operating automatically through chatbots. And this is an example of a much bigger problem. Artificial intelligence allows one person to do the work of ten. But how do we now distinguish an extremely effective person from a bot? Companies themselves encourage people to automate work, and then those same companies decide at what point automation becomes a violation. If artificial intelligence wrote a text, chose who should receive it, and you only reviewed it and pressed the button, did you do it or did artificial intelligence do it?
And if one person uses AI to perform the work of ten people, is that efficiency or bot-like behavior? Could the tool that gives you an advantage today become the reason your account is blocked tomorrow? That is what we are going to discuss today on ToTheMoon. An incredible number of people now use artificial intelligence. They automate all kinds of things with it and apply it in their professional work, at home, and everywhere else. People build different systems without understanding the consequences that may follow.
Of course, not everyone in the Russian-speaking community knows Twitter well or uses the network, but it is enormously widespread in the United States and in many countries around the world. It is also a very unusual platform in terms of what Elon Musk has done with it. X is closely connected with xAI and with a range of technologies. One of the company’s executives recently made an interesting statement. He said: “On our platform, we identified forty-two thousand accounts that were using chatbots to publish replies automatically.” What does that mean?
It is actually a very common practice. Huge numbers of bloggers and other people explain how to automate your work. Take LinkedIn, for example. I consider LinkedIn a garbage dump. When someone writes to me there, I do not even read the message, because LinkedIn has turned into a genuine garbage dump of automation: automatic messages after someone adds you as a connection, then another message a day later, two days later, three days later. It is just endless mass spam—completely insane spam.
So what were these forty-two thousand X accounts doing? I will explain the problem first, and then tell you what X executive Nikita Bier did with the accounts, because that part matters. These were a lot of accounts. People had configured systems that entered different accounts, automatically liked posts, automatically left messages, or published various material. We often see the same thing on YouTube and other platforms. You look at a comment and get the sense that it was not written by a real person.
I have seen a great deal of this on Telegram as well: you can tell immediately that a real person is not doing it. That creates a very interesting dilemma. What exactly are forty-two thousand such accounts? Nikita Bier blocked them. He removed them completely from the system. These accounts will almost certainly never be restored or allowed to remain on the platform. And that introduces a very interesting issue. X uses language saying that autonomous communication between artificial intelligence and users, with no human in the loop, conflicts with the platform’s mission.
Nikita wrote that the platform is meant to show the real pulse of humanity. In other words, real people are supposed to be there. That is very important. At the same time, there is an interesting contradiction. I know that, during a private conversation at one meeting, Elon Musk told a close friend of
mine that he could imagine content on platforms like this being generated automatically by the platform itself. You look at that and start thinking. There is one possible future in which you enter a social network and the platform itself generates the content. Today’s discussion is not really about social networks, of course, but the distinction matters. There is one boundary where content is created only by the social platform, and another where content is created by people—actual people—rather than exclusively by the platform.
We already have special services where, for example, music must be created only by humans and artificial intelligence is not allowed. There are services where AI is permitted, services where it is already used in enormous quantities, services where AI-generated content must be labeled, and services where no label is required. But what does “using artificial intelligence” even mean? What does it mean for a person to have reviewed or not reviewed something? Suppose I write someone an email with the help of artificial intelligence.
I provide all the requirements at the beginning, the system formulates the message, I copy it, and I send it. How do we determine AI’s degree of participation? Even the nature of my instructions can vary. Perhaps my instruction was only two words. Perhaps I simply said, “Write a reply to this email.” Perhaps I said, “Reply in my style.” Or perhaps I said, “Reply to this email and resolve these specific issues.” Who draws the line? It is impossible to draw it cleanly—absolutely impossible.
That is why this precedent of deleting accounts seems extraordinarily risky to me. Let me explain why. Nikita Bier published a post on X about these forty-two thousand accounts, and that is how I saw the story. The post does not mean, of course, that a user cannot ask ChatGPT, Claude, or Grok to draft a reply. X presents the main problem as a situation in which an account’s software independently searches for users’ posts, decides whom to contact, generates comments, publishes them without human review, and repeats the process at large scale.
But what if a person personally repeats the same process at large scale? What if I hire someone to do it on my behalf? What if I am simply a fanatical person who does it myself? Should I definitely be blocked? What if I am just extremely efficient? We now live in a world where efficiency is very hard to interpret. In the past, a company hired people and said: “Everyone should make roughly the same fifty calls per day,” or “Everyone should write roughly the same amount of text.” I have always disliked that approach categorically.
I have never liked averaging people. I have always said that human beings differ from one another in fundamental and enormous ways. You can hire a web designer who produces in one day what another person would take three months to produce. I have had people like that more than once. It does not mean there are many of them, but they exist. These are people who produce an extraordinary result. In principle, such a person should be paid as much as fifty other designers, but the world does not work that way.
With the arrival of artificial intelligence—especially in the United States—we have started to see completely different compensation. There have been cases in which people working with AI were reportedly paid ten million dollars a year or even fifty million dollars a year. Why? Because the assumption is that this person may create something unique or bring a special capability, and that a company has to pay that amount in order not to lose the game. Zuckerberg, for example, followed that logic.
Yet, looking at the past year and a half, Meta appears to be losing anyway. How do you identify the unique person? What actually makes someone unique? And will that person bring those unique skills into this particular role at this particular company? Artificial intelligence widens this gap even further. On the one hand, there is supposedly a problem because some automation exists. On the other hand, how is that automation different from a person who can do something extraordinarily quickly?
I complete a colossal number of tasks in a single day, and to me that feels ordinary. When I ask many other people to do certain things, I sometimes think: why is this person not doing it? What is going on? It is so elementary. But that is not the right way to reason, because people have different capabilities, different strengths, and different orientations toward different kinds of tasks. At the same time, we should not pretend that one person simply does more in one area while another always compensates by doing more somewhere else.
That is not how it works. If we look at final effectiveness, some people create ten million units of value and other people create one hundred. Effectiveness really can differ enormously. X’s rules contain another very interesting statement: the fact that a human mechanically presses the Publish button does not guarantee safety. X says that large-scale, repetitive, irrelevant, or intrusive replies may be considered spam regardless of who pressed the button.
That is strange. Imagine our ToTheMoon channel. People leave comments. Members of my team review them, and I also review them periodically. I may go through fifty or seventy comments and write “Thank you,” ask “What are you doing here?”, leave a like, a heart, or an emoji—in other words, produce quite repetitive replies. Am I a bot in that case? No. I expressed some reaction to a message a person left for me. Moreover, YouTube itself offers automatically generated replies to comments.
That is strange too, but it still offers them. What exactly are you trying to help with? Are you trying to help people manage a channel? Look at the contradiction. On the one hand, I am apparently not supposed to automate everything, because the platform may deny me access to features or stop promoting my videos. On the other hand, the platform itself is trying to turn the person into a robot by offering all these automatic replies. X actively promotes Grok. You open X and the company proudly invites you to use it: summarize this post for me; summarize the video someone shared; tell me directly in the feed whether this person was telling the truth.
The platform itself is encouraging me to use artificial intelligence. So where is the line? If I write a reply and ask Apple’s tools to rewrite it with better vocabulary, improve the Russian or English, or simply check the spelling, what am I allowed to do and what am I not allowed to do? There is a contradiction between what an X executive says and what Elon Musk says. X’s official automation rules, updated in April 2026, still permit bots that generate replies with artificial intelligence, but only with X’s prior, explicit written permission.
Automated replies are also allowed under a set of conditions: the user must contact the bot or clearly request a response; the user must know in advance that the response will be automated; only one automated reply may be sent per request; there must be a simple way to opt out; and the reply must relate to the user’s original request. Automatically searching for other people’s posts by keyword and replying without being asked is prohibited. Brands and publishers must also obtain separate approval from X for such automated campaigns.
So the official position appears to be this: artificial intelligence is acceptable as an assistant to a person, but artificial intelligence acting as an independent participant in the social network is forbidden without special permission from X. Yet that also means some companies have received special permission. So is it allowed or not? If AI prepares the text, a person reads it and publishes it personally, the risk of being blocked is supposedly low. But is it actually low or not?
If a program monitors the feed and automatically comments on relevant posts, the risk is high. Why should that behavior be prohibited in principle? Suppose a real person is sitting on the other side—someone hired to do exactly that, someone with a financial interest in doing it. Is that also prohibited? Am I not allowed to hire a person who does this twenty-four hours a day and pay them a large amount of money? What is the difference? There are people who are motivated to behave this way.
I see people who comment constantly. I do not have gigantic social networks, but my YouTube channels still have around two hundred thousand subscribers, and I have some tens of thousands of followers on Instagram. I can see that some people comment or send messages specifically to draw attention to themselves. The same is visible on X. Are those people engaging in manipulation or not? One of the main questions I am examining with you today does not really belong only to
social networks. It concerns the tools that these same companies—OpenAI, Anthropic, and others—are now offering. It also concerns actions recommended by well-known people around the world. ToTheMoon is just one channel. We are fairly large in the Russian-speaking market and probably produce one of the largest volumes of content measured by audience views. But the broader question is this: a famous person recommends that you take some action. Could that action ultimately cause you to violate a law or a platform rule and result in everything being blocked—your channels, your automations, or your systems?
Could it create a serious problem for you? How are you supposed to understand the actual rules of this game? The automation of social networks is a mature, deliberately built process. Bots certainly exist, but notice that the old bots have not disappeared. X still discovers enormous networks of mass accounts engaged in bot traffic. There are huge numbers of bot-traffic accounts on X, Instagram, and Facebook. Many YouTube channels—even channels by prominent creators that many of you watch—have enormous numbers of inflated views and fake subscribers.
You can often see it easily: a channel has two million subscribers, but only fifty thousand people watch a video. Or perhaps the numbers are not inflated and YouTube simply decided to reduce that channel’s reach. What is actually happening inside these accounts? Whom do social networks promote and whom do they suppress? Whom do they show and whom do they hide? Who can you trust today? You can find more videos like this on our channel. I also want to add that the only way to help a channel grow on YouTube is to subscribe if you have not already done so, leave comments, press Like, and share the content with friends.
These are simple actions for you. You can see that we publish a great deal of material. Someone recently asked, “When do you sleep if you produce so much content?” I have many different projects, tasks, and businesses in which I am involved. My life is not small; I have four children. We have been building ToTheMoon for more than two years. The channel is funded with my own money and resources from my private businesses, which develop and create different kinds of content in different fields.
We have editors, designers, video editors, and other people on the team, and of course we use a large number of services. We plan only to increase the amount of content on the channel. I want to develop it very seriously and bring you a great deal of useful, interesting, strong information. What we need from you is support that helps worthwhile content continue to appear. I am always very grateful for that kind of support. Thank you very much. We already have an active audience and there are always comments, but I encourage you to leave even more so that, by some incomprehensible miracle, we can keep up with YouTube’s unique algorithms.
Have a good day, everyone.